Advanced Fluorescence Microscopy
Peter J. Verveer Editor
Methods and Protocols
Methods in Molecular Biology 1251
M E T H O D S I N M O L E C U L A R B I O L O G Y
Series EditorJohn M. Walker
School of Life SciencesUniversity of Hertfordshire
Hat fi eld, Hertfordshire, AL10 9AB, UK
For further volumes: http://www.springer.com/series/7651
Advanced Fluorescence Microscopy
Methods and Protocols
Edited by
Peter J. VerveerDepartment of Systemic Cell Biology, Max Planck Institute of Molecular Physiology,
Dortmund, Germany
ISSN 1064-3745 ISSN 1940-6029 (electronic)ISBN 978-1-4939-2079-2 ISBN 978-1-4939-2080-8 (eBook) DOI 10.1007/978-1-4939-2080-8 Springer New York Heidelberg Dordrecht London
Library of Congress Control Number: 2014953904
© Springer Science+Business Media New York 2015 This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of the material is concerned, specifi cally the rights of translation, reprinting, reuse of illustrations, recitation, broadcasting, reproduction on microfi lms or in any other physical way, and transmission or information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or hereafter developed. Exempted from this legal reservation are brief excerpts in connection with reviews or scholarly analysis or material supplied specifi cally for the purpose of being entered and executed on a computer system, for exclusive use by the purchaser of the work. Duplication of this publication or parts thereof is permitted only under the provisions of the Copyright Law of the Publisher’s location, in its current version, and permission for use must always be obtained from Springer. Permissions for use may be obtained through RightsLink at the Copyright Clearance Center. Violations are liable to prosecution under the respective Copyright Law. The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication does not imply, even in the absence of a specifi c statement, that such names are exempt from the relevant protective laws and regulations and therefore free for general use. While the advice and information in this book are believed to be true and accurate at the date of publication, neither the authors nor the editors nor the publisher can accept any legal responsibility for any errors or omissions that may be made. The publisher makes no warranty, express or implied, with respect to the material contained herein.
Printed on acid-free paper
Humana Press is a brand of SpringerSpringer is part of Springer Science+Business Media (www.springer.com)
Editor Peter J. Verveer Department of Systemic Cell Biology Max Planck Institute of Molecular Physiology Dortmund , Germany
www.springer.com
v
The fl uorescence light microscope has become a common tool in the life sciences, with broad applications in many fi elds of biology. Recent decades have seen a rapid development in the fi eld, with the introduction of many new techniques to visualize and quantify biologi-cal samples at a molecular level.
This book aims to provide an overview of the advanced methods in fl uorescence micros-copy that have found application in biology, or that are promising to become essential tools in the future. Each chapter focuses on a different method and attempts to provide a practi-cal guide for application in biological systems.
With this book we attempted to cover a broad range of advanced fl uorescence micros-copy methods. In several cases, the instrumentation that is needed might not be easily avail-able to the biologist, due to their novelty. In these cases, we have attempted to provide instructions for building such equipment, along with protocols for their application.
For some time now, fl uorescence microscopy has been a standard tool for the molecular biologist in such fi elds as cell biology, neurobiology, and development biology. As they become more readily available, the new tools described here will be of tremendous use to those same scientists. We hope that this book will help them to apply these methods in the biological systems that they are interested in.
Dortmund, Germany Peter J. Verveer
Pref ace
vii
Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
1 The Physical Basis of Total Internal Reflection Fluorescence (TIRF) Microscopy and Its Cellular Applications. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Natalie S. Poulter , William T. E. Pitkeathly , Philip J. Smith , and Joshua Z. Rappoport
2 Two-Photon Excitation Microscopy and Its Applications in Neuroscience . . . . 25 Ricardo Mostany , Amaya Miquelajauregui , Matthew Shtrahman , and Carlos Portera-Cailliau
3 Live Spheroid Formation Recorded with Light Sheet-Based Fluorescence Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 Francesco Pampaloni , Roli Richa , Nariman Ansari , and Ernst H. K. Stelzer
4 Fluorescence Microscopy-Based RNA Interference Screening . . . . . . . . . . . . . 59 Manuel Gunkel , Nina Beil , Jürgen Beneke , Jürgen Reymann , and Holger Erfle
5 Fluorescence Resonance Energy Transfer Microscopy (FRET). . . . . . . . . . . . . 67 Katarzyna M. Kedziora and Kees Jalink
6 Localizing Protein–Protein Interactions in Living Cells Using Fluorescence Lifetime Imaging Microscopy. . . . . . . . . . . . . . . . . . . . . . 83 Yuansheng Sun and Ammasi Periasamy
7 Analysis of Biomolecular Dynamics by FRAP and Computer Simulation . . . . . 109 Bart Geverts , Martin E. van Royen , and Adriaan B. Houtsmuller
8 Fluorescence Correlation Spectroscopy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 Mark A. Hink
9 Homo-FRET Imaging Highlights the Nanoscale Organization of Cell Surface Molecules . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151 Suvrajit Saha , Riya Raghupathy , and Satyajit Mayor
10 Practical Structured Illumination Microscopy . . . . . . . . . . . . . . . . . . . . . . . . . 175 E. Hesper Rego and Lin Shao
11 4Pi Microscopy of the Nuclear Pore Complex . . . . . . . . . . . . . . . . . . . . . . . . . 193 Martin Kahms , Jana Hüve , and Reiner Peters
12 Application of STED Microscopy to Cell Biology Questions . . . . . . . . . . . . . . 213 Natalia H. Revelo and Silvio O. Rizzoli
13 Three-Dimensional Photoactivated Localization Microscopy with Genetically Expressed Probes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231 Kelsey Temprine , Andrew G. York , and Hari Shroff
Contents
viii
14 Direct Stochastic Optical Reconstruction Microscopy (dSTORM) . . . . . . . . . . 263 Ulrike Endesfelder and Mike Heilemann
15 Optogenetics: Optical Control of a Photoactivatable Rac in Living Cells . . . . . 277 Taofei Yin and Yi I. Wu
Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
Contents
ix
NARIMAN ANSARI • Physical Biology Group, Buchmann Institute for Molecular Life Sciences (BMLS) , Goethe Universität Frankfurt am Main , Frankfurt am Main , Germany
NINA BEIL • BioQuant, ViroQuant-CellNetworks RNAi Screening Facility , Ruprecht-Karls- Universitat Heidelberg , Heidelberg , Germany
JÜRGEN BENEKE • BioQuant, ViroQuant-CellNetworks RNAi Screening Facility , Ruprecht-Karls-Universitat Heidelberg , Heidelberg , Germany
ULRIKE ENDESFELDER • Institute of Physical and Theoretical Chemistry , Goethe University , Frankfurt am Main , Germany
HOLGER ERFLE • BioQuant, ViroQuant-CellNetworks RNAi Screening Facility , Ruprecht-Karls-Universitat Heidelberg , Heidelberg , Germany
BART GEVERTS • Department of Pathology, Josephine Nefkens Institute, Erasmus Optical Imaging Centre , Erasmus MC , Rotterdam , The Netherlands
MANUEL GUNKEL • BioQuant, ViroQuant-CellNetworks RNAi Screening Facility , Ruprecht-Karls-Universitat Heidelberg , Heidelberg , Germany
MIKE HEILEMANN • Institute of Physical and Theoretical Chemistry , Goethe University , Frankfurt am Main , Germany
MARK A. HINK • Department Molecular Cytology, van Leeuwenhoek Centre for Advanced Microscopy (LCAM) , University of Amsterdam , Amsterdam , The Netherlands
ADRIAAN B. HOUTSMULLER • Department of Pathology, Josephine Nefkens Institute, Erasmus Optical Imaging Centre , Erasmus MC , Rotterdam , The Netherlands
JANA HÜVE • Laboratory of Mass Spectrometry and Gaseous Ion Chemistry , The Rockefeller University , New York , NY , USA
KEES JALINK • Cell Biophysics Group, Department of Cell Biology B5 , The Netherlands Cancer Institute , Amsterdam , The Netherlands
MARTIN KAHMS • Laboratory of Mass Spectrometry and Gaseous Ion Chemistry , The Rockefeller University , New York , NY , USA
KATARZYNA M. KEDZIORA • Cell Biophysics Group, Department of Cell Biology B5 , The Netherlands Cancer Institute , Amsterdam , The Netherlands
SATYAJIT MAYOR • National Centre for Biological Science , Bangalore , India AMAYA MIQUELAJAUREGUI • Department of Neurology , David Geffen School of Medicine
at UCLA , Los Angeles , CA , USA RICARDO MOSTANY • Department of Neurology , David Geffen School of Medicine at UCLA ,
Los Angeles , CA , USA FRANCESCO PAMPALONI • Physical Biology Group, Buchmann Institute for Molecular Life
Sciences (BMLS) , Goethe Universität Frankfurt am Main , Frankfurt am Main , Germany AMMASI PERIASAMY • W.M. Keck Center for Cellular Imaging, Biology , University of Virginia ,
Charlottesville , VA , USA REINER PETERS • Laboratory of Mass Spectrometry and Gaseous Ion Chemistry ,
The Rockefeller University , New York , NY , USA WILLIAM T. E. PITKEATHLY • Physical Sciences of Imaging for the Biomedical Sciences (PSIBS)
Doctoral Training Centre , University of Birmingham , Edgbaston, Birmingham , UK
Contributors
x
CARLOS PORTERA-CAILLIAU • Department of Neurology , David Geffen School of Medicine at UCLA , Los Angeles , CA , USA ; Department of Neurobiology , David Geffen School of Medicine at UCLA , Los Angeles , CA , USA
NATALIE S. POULTER • School of Biosciences , University of Birmingham , Edgbaston, Birmingham , UK
RIYA RAGHUPATHY • National Centre for Biological Science , Bangalore , India ; Shanmugha Arts Science Technology and Research Academy , Thanjavur , India
JOSHUA Z. RAPPOPORT • School of Biosciences , University of Birmingham , Edgbaston, Birmingham , UK
E. HESPER REGO • Department of Immunology and Infectious Diseases , Harvard School of Public Health , Boston , MA , USA
NATALIA H. REVELO • STED Microscopy Group, European Neuroscience Institute , Deutsche Forschungsgemeinschaft Center for Molecular Physiology of the Brain/Excellence Cluster , Göttingen , Germany ; International Max Planck Research School for Neurosciences , Göttingen , Germany
JÜRGEN REYMANN • BioQuant, ViroQuant-CellNetworks RNAi Screening Facility , Ruprecht-Karls-Universitat Heidelberg , Heidelberg , Germany
ROLI RICHA • Physical Biology Group, Buchmann Institute for Molecular Life Sciences (BMLS) , Goethe Universität Frankfurt am Main , Frankfurt am Main , Germany
SILVIO O. RIZZOLI • STED Microscopy Group, European Neuroscience Institute , Deutsche Forschungsgemeinschaft Center for Molecular Physiology of the Brain/Excellence Cluster , Göttingen , Germany ; Department of Neuro- and Sensory Physiology , University of Göttingen , Göttingen , Germany
MARTIN E. VAN ROYEN • Department of Pathology, Josephine Nefkens Institute, Erasmus Optical Imaging Centre , Erasmus MC , Rotterdam , The Netherlands
SUVRAJIT SAHA • National Centre for Biological Science , Bangalore , India LIN SHAO • Howard Hughes Medical Institute , Ashburn , VA , USA HARI SHROFF • Section on High Resolution Optical Imaging, National Institute of
Biomedical Imaging and Bioengineering , National Institutes of Health , Bethesda , MD , USA
MATTHEW SHTRAHMAN • Department of Neurology , David Geffen School of Medicine at UCLA , Los Angeles , CA , USA
PHILIP J. SMITH • Physical Sciences of Imaging for the Biomedical Sciences (PSIBS) Doctoral Training Centre , University of Birmingham , Edgbaston, Birmingham , UK
ERNST H. K. STELZER • Physical Biology Group, Buchmann Institute for Molecular Life Sciences (BMLS) , Goethe Universität Frankfurt am Main , Frankfurt am Main , Germany
YUANSHENG SUN • W.M. Keck Center for Cellular Imaging, Biology , University of Virginia , Charlottesville , VA , USA
KELSEY TEMPRINE • Section on High Resolution Optical Imaging, National Institute of Biomedical Imaging and Bioengineering , National Institutes of Health , Bethesda , MD , USA
YI I. WU • Robert D. Berlin Center for Cell Analysis and Modeling , University of Connecticut Health Center , Farmington , CT , USA ; Department of Genetics and Developmental Biology , University of Connecticut Health Center , Farmington , CT , USA
Contents
xi
TAOFEI YIN • Robert D. Berlin Center for Cell Analysis and Modeling , University of Connecticut Health Center , Farmington , CT , USA ; Department of Genetics and Developmental Biology , University of Connecticut Health Center , Farmington , CT , USA
ANDREW G. YORK • Section on High Resolution Optical Imaging, National Institute of Biomedical Imaging and Bioengineering , National Institutes of Health , Bethesda , MD , USA
Contents
1
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_1, © Springer Science+Business Media New York 2015
Chapter 1
The Physical Basis of Total Internal Reflection Fluorescence (TIRF) Microscopy and Its Cellular Applications
Natalie S. Poulter*, William T.E. Pitkeathly*, Philip J. Smith, and Joshua Z. Rappoport
Abstract
Total internal reflection fluorescence (TIRF) microscopy has gained popularity in recent years among cell biologists due to its ability to clearly visualize events that occur at the adherent plasma membrane of cells. TIRF microscopy systems are now commercially available from nearly all microscope suppliers. This review aims to give the reader an introduction to the physical basis of TIRF and considerations that need to be made when purchasing a commercial system. We explain how TIRF can be combined with other micros-copy modalities and describe how to use TIRF to study processes such as endocytosis, exocytosis, and focal adhesion dynamics. Finally, we provide a step-by-step guide to imaging and analyzing focal adhesion dynamics in a migrating cell using TIRF microscopy.
Key words Total internal reflection fluorescence (TIRF), Endocytosis, Exocytosis, Focal adhesions
1 Introduction
Total internal reflection fluorescence (TIRF) microscopy, also known as evanescent wave/field microscopy, is a technique that uses a special mode of sample illumination to exclusively excite fluorophores that are within approximately 100 nm of the cover glass/sample interface. This mode of illumination is based on an evanescent field which is produced when light rays are totally inter-nally reflected at the cover glass/substrate interface. The evanes-cent field does not propagate deeper into the sample; therefore, images are not contaminated with fluorescence from out-of-focus planes. As a result, TIRF microscopy provides a means of imaging fluorophores near the cover glass/substrate interface with a high signal-to-noise ratio, thus making it the current gold standard
*Authors contributed equally.
2
technique for studying events that occur at the adherent plasma membrane of cells [1]. The recent rise in the popularity of this microscopy modality in the field of cell biology has been due, in part, to the work of Daniel Axelrod. Axelrod has been key to the development and promotion of this technique and has written various reviews on the subject [2, 3]. The current availability of “off-the- shelf” objective-based TIRF systems from all major microscope companies has also facilitated its use in the cell biology laboratory, as user-friendly operation makes TIRF microscopy accessible to the general scientific population rather than confining it to those with significant amounts of optical know-how.
When imaging cultured cells, TIRF microscopy has the advan-tage over confocal and epifluorescence microscopy of providing high-resolution, high-contrast images of the region closest to the glass coverslip (Fig. 1) making it perfect for studying processes such as endocytosis [4–7], exocytosis [8–10], cell adhesion [11–14], and cytoskeletal dynamics [15–18].
Although there now exists numerous commercially available TIRF systems, there are still several considerations that need to be taken into account before purchasing a system to ensure that you meet the requirements of your experiment and achieve the desired image quality.
Most commercially available TIRF microscopes use a “through-the- objective” optical configuration (we describe this and its alter-native below). One of the most important considerations for this type of setup is the choice of the objective lens used. High numeri-cal aperture (NA) lenses are always desirable for high-resolution optical microscopy, as they ultimately determine the “resolving power” of a system due to Abbe theory of imaging [19]. It is gen-erally agreed in the literature that the physical minimum for an objective-based TIRF setup is 1.40 NA, but in practice, 1.45 NA is usually the smallest used. The reasons for this are described in detail in the “objective-based TIRF” section. Usually the cost of an objective lens increases with the numerical aperture (NA) of the lens. Therefore, choosing the correct objective for your applica-tion, while considering cost implications, is a significant factor when designing your TIRF system.
Another import element that will affect the image quality is the type of camera chosen. Factors influencing camera performance are described in detail in Millis [20], but the main points are sum-marized here. Monochromatic charge-coupled devices (CCDs) are generally preferable over color-based CCDs. Cameras come with different bit depths; a higher bit depth means that a greater num-ber of gray levels can be detected, increasing dynamic range of intensity measured. For example, a 12-bit camera in our system gives 212 = 4,096 gray levels. If imaging speed is an important factor in your research, an electron-multiplying CCD (EM-CCD) may be more appropriate. EM-CCDs are much more sensitive to low light and therefore allow exposure times to be reduced, allowing
Natalie S. Poulter et al.
3
Fig. 1 TIRF versus epifluorescence imaging. A fixed MDA-MB-231 breast cancer cell labeled for alpha-tubulin (a, b) and phospho-paxillin (c, d). TIRF imaging (a, c) results in a much higher signal-to-noise ratio for both structures compared to epifluorescence imaging (b, d) which has a hazy appearance. Scale bar: 10 μm. (g, h) Show the illumination schematics for TIRF (g) and epifluorescence (h)
TIRF Microscopy for Cell Biology
4
faster imaging and also reducing photobleaching of the sample. The physical pixel size of EM-CCDs is typically larger than those in standard CCDs, and hence, the resolution of the resulting images is lower. To compensate for this, higher magnification objectives, e.g., 150×, may have to be used to ensure that optical images are sampled above the Nyquist criteria. EM-CCDs are also very expensive and are easily damaged by relatively intense light signals or prolonged low-intensity light signals.
Many applications of TIRF require imaging of more than one color, for example, imaging dynamics of a receptor where the receptor (EGFR) is labeled in green and a vesicle marker (clathrin) is in red [21]. In many other instances, simultaneous acquisition of the two colors would be beneficial as a lot of processes occur very rapidly and switching between different wavelengths might mean that any colocalization may not be detected, when it really does occur. If rapid imaging of two color samples is necessary for your application, it may be worth investing in an emission beam splitter that allows simultaneous imaging of the two colors. This separates the emitted light into two separate beams (one per color) that are each projected onto one half of the camera chip. Although the two colors are imaged simultaneously, there is the drawback that the area that can be imaged is half of that which can be obtained via sequential imaging. Therefore, whether to perform simultaneous or sequential two-color imaging will depend on what is more important to your application.
Once you have your system up and running, the next thing you need to think about is the design of your experiments. Although it is true that an image is worth a thousand words, a TIRF image of a cell without proper quantitative analysis is just a pretty picture. When planning experiments, it is necessary to think about ways to get the most data out of your image and to ensure that you really are measuring the processes that you want to study. In this chapter, we will briefly outline the physical basis of TIRF microscopy, describing the two main different optical configura-tions, prism- and objective-based, and also introduce another spe-cialist TIRF technique called variable angle TIRF microscopy (VA-TIRFM). We will discuss the benefits of combining TIRF with other microscope modalities and then delve into some of the biological applications of TIRF microscopy and include important tips and tricks for setting up experiments and analyzing the imag-ing data.
Every advantage of using TIRF microscopy emanates from the physical properties of the illumination method that it employs. Therefore, an understanding of the fundamental physical princi-ples behind the formation of TIRF images is absolutely necessary to reap the full benefits of this powerful microscopy technique. This knowledge is also essential when trying to interpret and
1.1 The Physical Basis of TIRF
Natalie S. Poulter et al.
5
extract quantitative information from the images obtained through TIRF, as factors such as fluorophore position within the illumina-tion field have a significant influence on the apparent intensity of its fluorescence. This section aims to provide the reader with some of the key physical concepts behind TIRF microscopy as well as important considerations for choosing the appropriate TIRF con-figuration for various biological applications.
TIRF microscopy relies on what is known as an evanescent field, which serves to exclusively illuminate a thin plane just above the imaging surface, i.e., a glass coverslip. The evanescent field is produced when light rays are totally internally reflected at the interface between the imaging surface and an aqueous medium. Below is a short explanation of the fundamentals of total internal reflection for dielectric media (media that does not conduct electricity; most applications of TIRF in cell biology use dielectric media).
The refractive index, n, of an optical medium tells us how electromagnetic waves, in this case visible light, propagate through it, relative to how it propagates through a perfect vacuum. It is defined as n = c/v, where c is the velocity of light in a vacuum and v the velocity of light in the medium. When light rays propagating through one medium (e.g., glass) meet an interface of another medium (e.g., air, water, cytosol) which has a different refractive index, the subsequent direction of the light rays is changed depending on the angle at which the light meets this interface (Fig. 2). Some of the light rays may be reflected from the interface and some may be transmitted into the second medium (Fig. 2a). If the refractive indices of both media are known, n1 and n2, as well as the angle of incidence, θ1, then Snell’s law gives us the angle at which light rays are transmitted and/or reflected from this interface, θ2 Eq. 1.
n n1 1 2 2sin sinq q= (1)
Fig. 2 Ray diagrams for the case of total internal reflection. This figure shows how the direction of light rays changes when they propagate through one medium (gray ) and encounter an interface of a second medium (white ), such that they experience an abrupt decrease in refractive index. (a) The angle of incidence is less than the critical angle. (b) The angle of incidence is equal to the critical angle. (c) The angle of incidence is greater than the critical angle
TIRF Microscopy for Cell Biology
6
Total internal reflection occurs when the refractive index of the first medium is greater than that of the second, n1 > n2, and the angle of incidence is greater than what is known as the “critical” angle, θc. When the angle of incidence is equal to the critical angle, θ1 = θc, light rays emerge into the second medium and propagate tangentially along the direction of the interface, or θ2 = 90° (Fig. 2b). So Snell’s law gives us: n1 sin θc = n2 sin 90°, sin 90° = 1, and therefore, θc = sin− 1(n2/n1). For angles of incidence greater than θc, light rays are totally reflected back into the first medium (Fig. 2c).
As we have just seen, Snell’s law describes which direction light rays leave a boundary between two media for a given angle of inci-dence, but it tells us nothing about the proportions of the light reflected and transmitted from such a boundary; Fresnel’s equa-tions must be used in for this. When Fresnel’s equations are solved for the case of total internal reflection, they show that although the propagation of the light rays is totally reflected, there is still an electric component of light which crosses the boundary and, hence, still a field in the second medium. This field is known as the evanes-cent field, and its intensity decays exponentially with perpendicular distance from the boundary:
I I z dz = -( )0 exp / (2)
where Iz is the field intensity at the distance, z, from the interface, d is the decay constant of the field, and I0 is the intensity of the field at the interface (z = 0).
The mathematics behind this is beyond the scope for this article, but interested readers are referred to texts [19, 22] or other good optics text books. In addition, Fresnel’s equations demonstrate that there are also implications for the amplitude of an evanescent field for different polarizations of the incident light [2]. However, apart from some specialist experiments [23, 24], the effects of polarization are negligible.
The penetration depth of the evanescent field for a particular TIRF configuration is critical. Local areas of high refractive index within a sample will convert the evanescent field into scattered propagating light which will in turn contaminate images. For that reason, it is often desirable for a TIRF system to have the capability to vary the penetration depth of the evanescent field. The equation below shows what determines the penetration depth of the evanes-cent field:
dn n
=-
l
p q0
12
1 224 sin
(3)
1.2 Penetration Depth
Natalie S. Poulter et al.
7
This is perhaps one of the most important equations when designing TIRF-based experiments as there are severe technical/economic implications on the ability to adjust the penetration depth. It shows that the penetration depth is dependent on the wavelength of the excitation beam, λ0; the angle of incidence of the excitation beam, θ1; and the refractive indices of the two media: n1 and n2. In practice, we do not usually have much control on the refractive index of a biological sample (n2), and we are usually con-fined to specific excitation wavelengths (λ0) of a specific fluoro-phore once in a sample. We cannot vary the refractive index of the imaging surface once intact during an experiment (n1). Therefore, experimentally we can only control the penetration depth through variation of the angle of incidence. However, as described below, with many commercially available TIRF objectives, the ability to vary the angle of incident light and thus penetration depth can be limited.
When TIRF microscopy was first applied in the field of cell biology, the experimental setup consisted of a few basic components added to an inverted fluorescence microscope [25]. Since then, many dif-ferent optical configurations have been designed and presented in publications, most of which fall into two categories, either prism- based or through-the-objective-based TIRF systems.
The key to producing an evanescent field at the imaging surface/sample interface is the delivery of light rays to the interface at the correct angle of incidence. In order to get a “pure” evanescent field, all light rays must meet the interface with the same angle of incidence. A pure evanescent field is one which follows a single exponential profile Eq. 2 with a given decay constant, d. If the incident beam of light meets the interface with a spread of angles, then a field consisting of multiple evanescent fields is created, each with different penetration depths; thus, we no longer follow the profile of a single exponential. Therefore, the excitation beams must be well collimated, and the imaging surface should be as flat as possible. With the aim of imaging cell substrates, glass prisms were optically coupled to glass coverslips, to which cells were adhered, to ensure that these two conditions were met. Hence, the origin of the term “prism-based” TIRF. There are several different configurations for prism-based TIRF microscopes, the first intro-duced into cell biology studies by Axelrod in 1981 [25].
Although through-the-objective systems are becoming increas-ingly popular, there are numerous benefits of using prism-based systems. Some of these benefits include:
1. Prism-based systems are typically cost-effective as they can be constructed with “off-the-shelf” components and “home built” in a laboratory/department workshop. Thus, they are
1.3 TIRF Configurations
1.3.1 Prism-Based TIRF
TIRF Microscopy for Cell Biology
8
not confined to any particular manufacturer or microscope system.
2. They usually achieve a “purer” evanescent field due to the fact that it is easy to get well-collimated light to the reflective boundary (i.e., minimal scattered light).
3. The resultant evanescent field covers the full field of view, again due to ease of focusing collimated light to the sample interface.
4. Prism-based systems can be designed and built to suit a specific purpose.
Some of the main disadvantages of prism-based systems include:
1. It is difficult to change the angle of incidence (although for simple prism-based setups, a variable angle design is demon-strated in Stock et al. [26]).
2. It is more complicated to set up each experiment as they require the imaging surface (typically a glass coverslip) to be optically coupled to the prism using substances such as glycerol.
As the potential of TIRF was realized, and its use in cellular imag-ing increased, various developments were made to the technology which led to the design of the “through-the-objective”-based sys-tems. The main advantage of this design is that the excitation light is delivered by the objective lens like most other microscopy modalities, therefore making it easier to perform experiments on a high-throughput basis.
The excitation beam is directed off-axis in the back focal plane of the objective and along the periphery of a high NA lens such that it emerges from the front aperture at a sufficiently high angle to achieve total internal reflection at the coverslip/sample inter-face. In objective-based TIRF systems, high NA objectives are a necessity because of the high angle of incidences required for achieving total internal reflection at the coverslip/sample interface. High NA provides the freedom of a larger range of incident angles and hence variation of penetration depth; therefore, most com-mercially available lenses are 1.45 and 1.49, but higher NA lenses are available. Olympus has a 100×, 1.65 NA objective which is capable of producing a very pure evanescent field at low penetra-tion depths. The drawback of this lens is that it requires a special immersion oil (which is volatile) and expensive sapphire high refractive index coverslips to match. Fish’s review has a good dis-cussion on the NA of an objective and the range of above critical angles [27].
As stated above, in order to get a “pure” evanescent field, the incident light rays need to all meet the coverslip/sample interface at the same angle of incidence. The challenge with the objective- based systems is maintaining a well-collimated beam of light through the special confinements of the margins of the objective lens.
1.3.2 Objective-Based TIRF
Natalie S. Poulter et al.
9
There are many advantages of using an objective-based TIRF system, including:
1. It is easier to use than a prism-based system as it is not restricted by the position of a prism in the setup and does not require the imaging surface to be optically coupled to a prism with a liquid such as glycerol.
2. The user is free to use different off-the-shelf imaging surfaces, i.e., microscope slides, coverslips, and glass-bottom dishes, providing that they have an appropriate refractive index.
3. Light paths can be changed at a “flick of a switch” to image with different wavelengths or imaging modalities.
4. Useful for multipoint, multi-dish, or slide experiments as the excitation and emission beams come from the same side of the objective.
Essentially, all commercially available TIRF microscopes employ objective-based TIRF. These generally allow rapid change of TIRF angle and switching between excitation beams. Some even have simultaneous dual-wavelength or multiwavelength imaging capability by exciting with two or more different wavelengths at the same time and chromatically separating the light emitted from the sample using a beam splitter. Simultaneous wavelength can be achieved in two ways: (1) Use a beam splitter combined with a right-angle prism, so that the separated light paths are projected in adjacent positions on the same CCD chip. (2) A separate CCD could be used for each light path coming from the beam splitter. The former has the obvious disadvantage of sacrificing the field of view for each channel, and the latter is considerably more expan-sive as it requires additional cameras.
Although through-objective TIRF systems do offer many advantages over the prism-based setups, there are also a few disadvantages:
1. The evanescent waves produced by the through-the-objective- based systems are usually less “pure” than prism-based systems. This is because more rays are scattered as the excitation beam is guided through the back focal plane of the objective than in the prism-based systems.
2. Due to the obscure path the light has to take to leave the objective at the correct angle, the resultant diameter of the beam leaving the objective is very small, typically smaller than the field of view of the microscope. Consequently, the width of the evanescent field is also less than the field of view of the microscope, and hence the image intensity is seen to drop off at the edges of images.
TIRF Microscopy for Cell Biology
10
3. The maximum angle of incidence achievable with through-the- objective systems is restricted by the NA of the objective, and therefore, smaller penetration depths are achievable with prism- based systems.
4. Purpose-built TIRF objectives are specially engineered so typically cost several thousand pounds.
Earlier we discussed the factors that influence the penetration depth of the evanescent field and gave the equation which relates them. We also point out that the only one of these we have any real control over during an experiment is the angle of incidence of the laser beam. Altering this gives us control of the penetration depth of the evanescent field providing the option of illuminating as deep or as shallow into the sample as we wish (Fig. 3), within the con-straints of the system (realistically between 100 and 400 nm although 60–600 nm is theoretically possible depending on the specific setup). The technique of VA-TIRF makes possible the measurement of submicroscopic z-distances, with slices up to 10 times thinner than those achieved with a confocal scanning micro-scope. This allows a much finer level of detail to be resolved in in vitro studies, for example, determining distances between fluo-rescent probes and cell membranes [28] or the ability to map cell- substrate topology [26]. There are several ways to implement such a technique. The most common models incorporate a single or series of rotating mirrors [26, 28], though setups with an acousto- optic modulator and telecentric lens optics with wavelength- dependant deflection angles and varying image planes are also
1.3.3 Variable Angle TIRF Microscopy (VA-TIRFM)
Fig. 3 Variable Angle TIRF. This figure shows how the penetration depth of the evanescent wave varies with incident angle of the incoming laser. The resultant images are shown beneath
Natalie S. Poulter et al.
11
relatively common [29, 30]. As long as the optics contain a correction for movement of the exciting laser beam in the x – y plane when the angle of incidence is changed, then variable angle TIRF is possible on nearly any system.
Images obtained through this method can be interpreted in a number of different ways. In cell biological applications, qualita-tive measurements, when the detection of membrane-bound fluo-rophores is differentiated from those located in the cytosol alone, are possible as are quantitative observations based on relative levels of fluorophore excitation [31]. As of yet, we are not aware of reports of a 3D image stack being created from the images, but the theoretical framework is there.
The addition of a TIRF system to a microscope leaves its optical configuration relatively unchanged; therefore, the sample is still accessible by other microscopy light paths. Most commonly, TIRF is combined with epifluorescence microscopy, due to the fact that they are both widefield techniques and use the same collection optics (filters, dichroics, CCD cameras) which makes it easy to switch between them. Epi-illumination can be achieved in one of the following two ways: (1) through use of an independent epi- light source such as a standard arc lamp light source and (2) by changing the angle of incidence from above the critical angle for TIRF to a subcritical angle such that light is no longer reflected and hence propagates into the sample. The former method is the one more commonly seen in the literature [32–34] as the mechanical switching between each excitation light path can be performed very rapidly (typically a hundreds of milliseconds, some as fast as tens of milliseconds). The latter is less popular due to the speed at which most TIRF systems can “detune” the TIRF beam and then accu-rately “tune” it again. However, now commercially available micro-scopes are being produced which can rapidly and accurately change the angle of incidence of the excitation beam for this purpose.
Combining TIRF with epifluorescence has its obvious limitations, the main being the large depth field of the epi-illumination. In order to overcome this limitation, it can be combined with confo-cal microscopy because of its optical sectioning capability. Images from confocal microscopes represent a much thinner cross section than epifluorescence images (~1μm compared to ~10μm) because the effective volume of the point spread function (PSF) of confocal is much smaller than that of epi. In conjunction with the sharp PSF, a pin hole is placed before the detector which prevents out-of- focus light rays from reaching the detector. Conventional point- scanning confocal microscopes use a pair of oscillating mirrors to scan the focal point of the excitation light across planar sections through the sample. Any fluorescence from the focal point is then detected using a photon multiplier tube, converted to a digital
1.4 TIRF and Other Microscopy Techniques Combined
1.4.1 Epi–TIRF
1.4.2 Confocal–TIRF
TIRF Microscopy for Cell Biology
12
signal, and the image is recorded serially by a computer. Because images are constructed in this point-by-point manner, the acquisi-tion time per image can be significantly greater than that of the CCD-based acquisition of epi or TIRF (e.g., ~1 s compared to ~100 ms for a 1,000 × 1,000 pixel image). As a result, it would not be possible to perform some of the same experiments which com-bine epi with TIRF as the distribution of fluorophores within the evanescent field can change significantly in the time it takes to acquire a confocal image. To overcome this limitation, fast confo-cal systems such as a resonance scanner or spinning disk configura-tions can be used which are capable of acquiring a 512 × 512 two-dimensional image in ~100 ms (some spinning disk systems can acquire images in less than 20 ms). Resonance scanner systems use high-temporal-resolution photon multiplier tubes and form images in the same way conventional confocal microscopes do. To facilitate colocalization studies with TIRF and resonance scanner images, their data sets must be co-registered as they will generally differ by a rotation, scale, and position within the image. A method for automated co-registration is presented in Pitkeathly et al. [35].
Spinning disk confocal microscopes use an array of micro lenses on a fast rotating disk which is used to simultaneously illu-minate multiple points in the sample rather than scanning a single point across it. Images are recorded using a high-grade CCD, and in some setups, the TIRF images are acquired with the same CCD. If the same CCD is used for both light paths, then the images should already be naturally registered. However, this con-figuration will potentially require significant switch times to alter the light paths.
For all dual modality experiments, the limiting factor is the time it takes to switch from one modality to the other, especially in live-cell experiments. In the case of the TIRF–resonance scanner confocal combination, the switch over can take a few seconds as light paths have to be mechanically switched over. This may pro-hibit the use of this combination for some experiments.
As seen in Eq. 3, the penetration depth is also dependent on wave-length. Hence, when performing simultaneous multiwavelength experiments, this needs to be taken into consideration. If two differ-ent excitation wavelengths of light are being used simultaneously, and delivered at the same angle of incidence, then the evanescent field corresponding to the light with the longer wavelength will penetrate deeper into the sample. Moreover, some experiments may require that the penetration depth of both wavelengths of light used to be precisely the same. In this case, the angle of incidence of light of each excitation beam will have to be different. Commercially available microscopes such as the Olympus cellTIRF microscope are able to control the angle of incidence of each excitation beam independently.
1.4.3 Dual/Multicolor Experiments
Natalie S. Poulter et al.
13
Some experiments may require precise knowledge of the penetra-tion depth and the profile of the evanescent field. Although the theoretical profile of the evanescent field is described as an expo-nential function for a given angle of incidence, in practice, this is often not the case. Not all light rays arrive at the same angle of incidence because of scattering on route to the objective lens; therefore, the field is a mixture of exponential functions. Mattheyses and Axelrod developed a method for determining these experi-mentally [36], and they found that a function consisting of two exponentials accurately described the profile of the field in their through-the-objective-based apparatus.
Endocytosis is the process by which a cell internalizes membrane- bound or extracellular material and there are several endocytic pathways in operation in cells [37]. The most well defined of these is clathrin-mediated endocytosis (CME). Clathrin-coated pits (CCPs) form at the plasma membrane before they are pinched off as vesicles inside the cell through the action of the GTPase dyna-min [38, 39]. Fluorescently tagged clathrin and dynamin con-structs have been developed and validated for use in live-cell imaging studies [40, 41]. The unique properties of TIRF micros-copy make it ideal for studying endocytic events [4, 42, 43]. When cells express fluorescently labeled clathrin, the CCPs appear as spots at the cell surface. As clathrin-coated vesicles (CCVs) form, they are still seen at the plasma membrane, but as they are pinched off and moved into the cell, and therefore out of the TIRF field, they seem to “disappear” (Fig. 4). If you want to ascertain if your protein of interest is really undergoing endocytosis, there are vari-ous criteria that need to be met [44]. Firstly, the putative endocytic event must be seen to disappear over successive frames and not move laterally out of the field of view. To ensure that this can be detected, you must make sure that you are imaging frequently enough to capture the endocytic process. We tend to image with “no delay.” This means that if the exposure time was set to 400 ms, then the camera will take an image every ~400 ms. This mode of camera capture is called streaming and occurs through continuous “frame grabbing” from the CCD chip. Secondly, appropriate con-trols are needed to verify that a spot disappearing from the TIRF field is due to an endocytic event and not photobleaching.
To ensure that events of endocytosis are not actually misidenti-fied occurrences of photobleaching, it is important to quantify the intensity of other neighboring spots in the field of view (Fig. 4b, c). If the intensity of the spot of interest decreases over the period of analysis but a nearby spot shows no change in intensity, then you can be fairly sure that photobleaching is not playing a role in spot disappearance. Another way of checking that a spot is being endo-cytosed is if it disappears from the TIRF image but is still visible in an epifluorescence image, indicating that it has moved further than
1.4.4 Other Considerations
1.5 Biological Applications of TIRF
1.5.1 Endocytosis
TIRF Microscopy for Cell Biology
14
100 nm into the cell. This has been used in various studies [6, 45, 46], but as the TIRF and epi images are generally not taken at the same time, care needs to be taken in subsequent analysis. Additionally, if your protein of interest is undergoing endocytosis, then it should colocalize with other markers of endocytosis and its internalization/disappearance from the TIRF field should be coin-cident with the disappearance of these independent endocytic markers. As mentioned previously, fluorescently tagged clathrin and dynamin constructs have been developed and these can be used in combination with your protein of interest to see if they colocalize and disappear together. Dynamin is also involved in some clathrin-independent endocytic pathways, so it may be a good marker to use if your protein is not internalized via CME. However, in our opin-ion, the gold standard for scoring a bona fide event of endocytosis depends upon the observation that an extracellular soluble ligand which has bound to a receptor on the cell surface internalizes at the same time, place, and rate as your intracellular endocytic markers. In one example, we incubated cells in transferrin (Tf), the ligand for the Tf receptor, which is internalized by CME, and simultaneously imaged clathrin and Tf entering the cell [5]. The co-disappearance of the endocytic marker within the cell (clathrin) with a soluble ligand placed outside the cell (Tf) clearly demonstrates that endo-cytosis was occurring in these cells.
Fig. 4 Quantifying endocytosis. A clathrin-coated pit labeled with clathrin-dsRed (spot A) is seen to disappear from the TIRF field over time (a). A neighboring CCP (spot B) does not disappear and serves as a control for photobleaching. Quantification of the fluorescence intensity of both spots is shown in the graphs below. (b) shows the total intensity of spot A decreases over time and (c) shows the intensity of spots A and B taken across the dotted line shown in (a). The intensity of spot B remains unchanged throughout the imaging so the disappearance of spot A is due to endocytosis and not photobleaching. This research was originally published in Biochemical Journal, J.Z. Rappoport, focusing on clathrin-mediated endocytosis. Biochemical Journal. 2008; 412: 415–423 © the Biochemical Society
Natalie S. Poulter et al.
15
Exocytosis occurs when an intracellular vesicle fuses with the plasma membrane, releasing its contents to the extracellular envi-ronment [47]. Exocytosis is important for many processes such as insertion of particular receptors at defined areas of the plasma membrane, release of neurotransmitters at the pre-synaptic termi-nal, and removal of waste products from the cell. TIRF microscopy is ideal for studying exocytosis as the evanescent wave decays over a distance that is roughly the same thickness as a secretory vesicle [48], so as the vesicles move towards the plasma membrane, they are seen to get brighter. Again, certain criteria must be met to ensure that a true exocytic event is occurring as there have been instances where dyes that accumulate in vesicles previously used to show exocytic events (e.g., acridine orange (AO)) have not been appropriately employed. The “flash” of increased intensity previ-ously thought to depict fusion of an AO-labeled vesicle with the membrane actually can be due to rupture of the vesicle carrying the dye triggered by photodamage caused through imaging [48, 49]. As a result, an integral membrane protein is deemed to be bet-ter for visualizing exocytosis as its diffusion in the membrane is slower (2D, rather than 3D). Thus, it is both easier to visualize and to quantify a membrane protein moving laterally within a mem-brane rather than a soluble protein diffusing in solution. Figure 5 shows the exocytosis of a GFP-tagged transferrin receptor (TfR- GFP) in HeLa cells. The vesicle carrying the receptor can be seen to get brighter as it nears the plasma membrane and moves deeper into the evanescent wave. In this instance, the GFP tag is intralu-minal, and as GFP fluorescence is partially quenched at low pH, when fusion begins and the acidic vesicle interior is neutralized by the extracellular medium, the fluorescence intensity is observed to peak at the onset of fusion. The fluorescence intensity then starts to decline as the TfR-GFP spreads out into the plasma membrane, increasing the width of the intensity profile as this lateral diffusion continues.
Strict quantitative criteria have been set out by the laboratory of Sanford Simon to explicitly demonstrate the delivery of membrane- associated cargo to the cell surface via exocytosis [48, 50], and this should be applied to confirm a suspected exocytic event. These criteria are based upon the fact that while the total intensity of the cargo from within the fusing vesicle will not change immediately post-fusion, the area of that signal will. However, it must be emphasized that the speed of acquisition is critical for imaging exocytosis. From Fig. 5, you can see that this exocytic event is completed within just over 1 s so imaging needs to be as rapid as possible. Simultaneous dual-color imaging is also useful when studying exocytosis. To confirm that your protein is involved in exocytosis, it is good practice to co-express it with known mark-ers of exocytosis such as the soluble protein neuropeptide Y (NPY) and the membrane protein vesicular stomatitis virus glycoprotein
1.5.2 Exocytosis
TIRF Microscopy for Cell Biology
16
(VSVG), tagged with a different color to your protein of interest. This type of analysis was carried out by Grigoriev et al. [10] in a study looking at the role of Rab6 in exocytosis. You can then use TIRF to quantify colocalization and time of appearance and fusion with the membrane of the two proteins to confirm whether exocy-tosis is occurring. As exocytosis is a rapid process, sequential acqui-sition of the two colors would not be appropriate in this instance as the events would be completed before the filters had time to move into place and any colocalization would be missed.
Focal adhesions (FAs) are large, multiprotein complexes that act to link the actin cytoskeleton of cells to the extracellular matrix through heterodimeric transmembrane proteins called integrins [51, 52]. Regulation of these structures is important for cell migra-tion; new focal contacts form at the front of migrating cells, and these can either be quickly turned over or can mature into more stable FAs that allow the cell to transmit traction forces (through
1.5.3 Focal Adhesion Dynamics
Fig. 5 Analyzing exocytosis by TIRF. (a) A TIRF timelapse showing a vesicle carrying transferrin receptor (TfR)-GFP fusing with the plasma membrane. (b) Quantification of the GFP intensity across the dotted line shown in (a) over time. As the vesicle fuses with the plasma membrane, the fluorescence intensity peaks and then declines as the receptor diffuses within the plane of the membrane
Natalie S. Poulter et al.
17
the actin cytoskeleton) to the ECM and vice versa [53]. FAs need to be disassembled at the back of the cell so the rear can be retracted and the cell can move forward. As FAs form at the adherent cell membrane, they are best visualized through the use of TIRF microscopy to eliminate unwanted signal that might obscure their pattern and position in the cell. Any treatment that affects cell migration, be it through genetic means or application of a pharma-ceutical agent, could be acting on the formation or disassembly of FAs in the cell and this is often a topic of investigation. For exam-ple, inhibiting CME in fibroblast cells resulted in a reduction in cell migration and prevented efficient disassembly of FAs in these cells [54]. The effect of a treatment on FAs can be assessed by measuring the intensity, the size, and the position of the FAs in the cell over time. However, using the intensity and size of a structure in a TIRF image is perhaps not the best measurement because other factors can affect them. The number of fluorophores in a structure will affect the intensity, as will the position of the fluoro-phore in the z-direction. Objects closer to the coverslip appear brighter and bigger than those further away, and due to the sensi-tivity of TIRF, even nanometer differences in vertical position can cause noticeable changes in intensity. Thus, FA turnover (the time it takes for an FA to form and then disassemble) is generally the most direct measure of FA dynamics. Figure 6 shows a FA labeled
Fig. 6 Focal adhesion turnover. (a) A focal adhesion (FA) labeled with paxillin-RFP forming, maturing, and disas-sembling in an MDCK migrating at a wound edge seen via TIRF microscopy. (b) Quantification of the total fluo-rescence intensity of the FA over time. The region being measured is shown as a red ROI in the first image in (a)
TIRF Microscopy for Cell Biology
18
with paxillin-RFP in an MDCK migrating in a wound healing assay. A FA can be manually identified and a region of interest (ROI) drawn around it. Information such as area and intensity can be taken from the ROI for each time point and plotted in a graph to show how the FA increases in size and intensity over time before decreasing as it disassembles. The absolute size and intensity can-not be determined due to the reasons detailed above, but they give an indication of any relative differences over time, and in combina-tion with the time scale, comparisons between treatments can be made. It should be noted though that slower moving cells may have decreased FA turnover rates; thus, perturbations that affect FA dynamics may represent indirect effects.
Manual analysis of FA turnover is laborious and subject to bias as the FA chosen to be analyzed is often the biggest and easiest to identify. Thus, these FAs may not be representative of the popula-tion. As such, automated FA detection programs have been devel-oped that allow the identification, tracking, and analysis of various FA properties over time, and one from the Gomez lab has been made publically available at http://gomezlab.bme.unc.edu/tools [11]. An additional advantage of automated tracking over manual identification is that a large number of FAs can be identified and analyzed, increasing confidence in the results seen. If an automated program does not work on your data sets, it is important to manu-ally analyze as many FAs as possible within each cell. In the next section of this chapter, we will provide a step-by-step guide to imaging focal adhesion dynamics on a Nikon TIRF system using the breast cancer MDA-MB-231 cells migrating in a scratch wound assay as an example.
2 Materials
1. MDA-MB-231 cells from the Health Protection Agency Cell Culture Collection.
2. Complete medium: DMEM (Lonza), 10 % fetal calf serum (FCS) (Labtech International), 1 % penicillin/streptomycin (Gibco).
3. 1× trypsin (Gibco). 4. Lipofectamine 2000 (Invitrogen). 5. Paxillin-RFP construct (courtesy of Dr. Maddy Parsons, King’s
College London). 6. Cell imaging medium (CIM): Hanks balanced salt solution
(HBSS; Sigma) dissolved into 10 mM HEPES at pH 7.4, with FCS at 5 % (V/V).
7. Neubauer cell counting chamber. 8. 35 mm glass-bottomed imaging dishes (MatTek Corporation,
USA).
2.1 Biological
Natalie S. Poulter et al.
19
1. Nikon TIRF system based on a Nikon Eclipse Ti inverted microscope, utilizing illumination through the microscope objective (CFL Plan Apo 60× NA 1.49, Nikon).
2. Microscope incubator (Okolab S.r.l) set at 37 °C for live-cell imaging.
3. Green diode 561 nm laser. 4. 12-bit iXon 1 M EM-CCD camera. 5. Nikon NIS-Elements Advanced Research v3.2 software.
3 Methods
1. Trypsinise a T75 flask of MDA-MB-231 cells and count the number of cells per 1 ml using a counting chamber. Seed 6 × 105 cells onto each 35 mm glass bottom dish and incubate in complete medium at 37 °C, 5 % CO2 for 24 h.
2. On day 2, transiently transfect cells with paxillin-RFP using Lipofectamine 2000 according to the manufacturer’s instruc-tions using a ratio of 4 μg DNA to 10 μl Lipofectamine 2000 per 35 mm dish of cells. Remove transfection medium after 3 h to prevent cell death, replace with fresh complete medium, and leave the cells in the incubator to express the construct for 24 h.
3. On day 3, wound the confluent layer of cells using a pipette tip, remove the medium, and replace with CIM that has been pre-warmed to 37 °C (see Note 1). Before using the microscope, observing laser safety guidelines, ensure that the TIRF laser is aligned and focused for the objective you are going to use.
4. Place the dishes of cells onto the microscope stage within a microscope incubator that has been pre-warmed to 37 °C for at least several hours before imaging (see Note 2). Ensure that the glass bottom of the dish is clean and it is completely level on the microscope stage (see Note 3).
5. Using a 60× TIRF oil objective (NA 1.45) in bright-field mode, focus the cells using the course focus and then fine- tune using the extra-fine focus setting before turning on the perfect focus system (PFS).
6. Search for transfected cells at the wound edge using the 563 nm laser (for RFP) at a “detuned” angle (see Note 4).
7. Once a transfected cell has been identified gently, adjust the laser angle until TIRF has been reached (see Note 5).
8. Set the exposure time for the image; ensure that no pixels are saturated but structures are bright enough to be seen. Typical exposure times are 200–500 ms.
9. Once all the cells to be imaged have been identified and their XY positions saved, set up a timelapse (see Note 6). For focal
2.2 Hardware
TIRF Microscopy for Cell Biology
20
adhesion (FA) turnover assays, we generally image the cell every 1 min for 1 h, but this can be adjusted depending on the speed the cells are moving (i.e., faster cells need imaging more often and vice versa). Timelapse imaging files are saved auto-matically on the NIS-Elements software as .nd2 files.
10. Use the timelapse files to get quantitative information on FA dynamics. The ROI tool can be used to select FAs and measure their fluorescence intensity over time, indicating their assembly and disassembly rates. Other parameters such as area and num-ber of FAs can also be determined in the NIS- Elements AR software.
4 Notes
1. If your microscope setup does not have CO2 capabilities, then buffering the pH of the cell environment using CIM is a good alternative. Alternatively, if CO2 is available, you can image cells in DMEM in a 5 % CO2 atmosphere containing the appro-priate supplements but without phenol red.
2. To minimize stage drift during the course of a timelapse exper-iment, it is important to ensure that the microscope incubator, and therefore the microscope components, is up to the required temperature (usually 37 °C). It helps to have the temperature probe immersed in water in the incubator as this gives a better indicator of the temperature in the dish being imaged and is also a more constant temperature so the heater will not be switching on and blowing on your samples every time there is a slight dip in temperature.
3. Any dirt or liquid on the coverslip area of the dish can hinder good TIRF imaging. Clean the coverslip with ethanol and ensure it is dry before imaging. Also make sure the imaging dish is flat and there is enough oil. If you see a circular interfer-ence pattern when imaging, remove the dish, clean the oil from objective and the dish, and apply fresh oil and ensure dish is replaced so that the coverslip is flat and try again (see Mattheyses et al. [1] for more detailed troubleshooting information).
4. Detuning the TIRF involves changing the angle of the laser so that the cells can be visualized via propagated light, in an epifluorescence- like state. This helps you to see transfected cells and also identify the edge of the wound, which is very dif-ficult in TIRF mode.
5. TIRF is when there is no “haze” visible around the cell; gener-ally, the cell gets dimmer as the laser angle is moved towards TIRF and then it suddenly gets brighter just before TIRF is reached. You know when you are in TIRF when all the out-of- focus light disappears and when you focus up and down only
Natalie S. Poulter et al.
21
the structures you are looking at go in and out of focus; no other fluorescent molecules deeper in the cell come into focus (i.e., only a single focal plane can be visualized). There are sev-eral websites that provide useful information on TIRF micros-copy, along with interactive tutorials and example images, for example:http://www.microscopyu.com/articles/fluorescence/tirf/tirfintro.htmlhttp://www.olympusmicro.com/primer/techniques/fluores-cence/tirf/tirfhome.html.
6. If your system has an automated XY stage and allows XY posi-tions to be saved (i.e., allowing more than one cell to be imaged in a timelapse), make sure that the number of positions being imaged fits into the time interval between images or else you will not be imaging as frequently as you think you are.
Acknowledgments
The authors would like to acknowledge funding through BBSRC Project grant BB/H002308/1. WTEP and PJS are funded through the Physical Sciences of Imaging for the Biomedical Sciences (PSIBS) Doctoral Training Centre, and NSP is funded through British Heart Foundation New Horizons grant NH/11/6/29061. The TIRF microscope used in this research was obtained through Birmingham Science City Translational Medicine Clinical Research and Infrastructure Trials Platform, with support from Advantage West Midlands (AWM).
References
1. Mattheyses AL, Simon SM, Rappoport JZ (2010) Imaging with total internal reflection fluorescence microscopy for the cell biologist. J Cell Sci 123:3621–3628
2. Axelrod D (2001) Total internal reflection flu-orescence microscopy in cell biology. Traffic 2:764–774
3. Axelrod D (2008) Total internal reflection flu-orescence microscopy. Methods Cell Biol 89:169–221
4. Rappoport JZ, Simon SM (2003) Real-time analysis of clathrin-mediated endocytosis dur-ing cell migration. J Cell Sci 116:847–855
5. Rappoport JZ, Benmerah A, Simon SM (2005) Analysis of the AP-2 adaptor complex and cargo during clathrin-mediated endocytosis. Traffic 6:539–547
6. Merrifield CJ, Perrais D, Zenisek D (2005) Coupling between clathrin-coated-pit invagi-nation, cortactin recruitment, and membrane scission observed in live cells. Cell 121:593–606
7. Merrifield CJ, Feldman ME, Wan L et al (2002) Imaging actin and dynamin recruit-ment during invagination of single clathrin- coated pits. Nat Cell Biol 4:691–698
8. Fix M, Melia TJ, Jaiswal JK et al (2004) Imaging single membrane fusion events medi-ated by SNARE proteins. Proc Natl Acad Sci U S A 101:7311–7316
9. Akopova I, Tatur S, Grygorczyk M et al (2011) Imaging exocytosis of ATP-containing vesicles with TIRF microscopy in lung epithelial A549 cells. Purinerg Signal 8:59–70
TIRF Microscopy for Cell Biology
22
10. Grigoriev I, Splinter D, Keijzer N et al (2007) Rab6 regulates transport and targeting of exo-cytotic carriers. Dev Cell 13:305–314
11. Berginski ME, Vitriol EA, Hahn KM et al (2011) High-resolution quantification of focal adhesion spatiotemporal dynamics in living cells. PLoS One 6:e22025
12. Partridge MA, Marcantonio EE (2006) Initiation of attachment and generation of mature focal adhesions by integrin-containing filopodia in cell spreading. Mol Biol Cell 17:4237–4248
13. Fletcher SJ, Poulter NS, Haining EJ et al (2011) Clathrin‐mediated endocytosis regu-lates occludin, and not focal adhesion, distribu-tion during epithelial wound healing. Biol Cell 104:238–256
14. Lock FE, Ryan KR, Poulter NS et al (2012) Differential regulation of adhesion complex turnover by ROCK1 and ROCK2. PLoS One 7:e31423
15. Manneville JB (2006) Use of TIRF microscopy to visualize actin and microtubules in migrat-ing cells. Meth Enzymol 406:520–532
16. Webb RL, Rozov O, Watkins SC et al (2009) Using total internal reflection fluorescence (TIRF) microscopy to visualize cortical actin and microtubules in the drosophila syncytial embryo. Dev Dynam 238:2622–2632
17. Dixit R, Ross JL (2010) Studying plus-end tracking at single molecule resolution using TIRF microscopy. Method Cell Biol 95:543–554
18. Gardner MK, Charlebois BD, Jánosi IM et al (2011) Rapid microtubule self-assembly kinet-ics. Cell 146:582–592
19. Graham-Smith SF, King TA (2000) Optics and photonics: an introduction. Wiley, Chichester, UK
20. Millis BA (2012) Evanescent-wave field imag-ing: an introduction to total internal reflection fluorescence microscopy. Methods Mol Biol 823:295–309
21. Rappoport JZ, Simon SM (2009) Endocytic trafficking of activated EGFR is AP-2 depen-dent and occurs through preformed clathrin spots. J Cell Sci 122:1301–1305
22. Born M, Wolf E, Bhatia AB (1999) Principles of optics: electromagnetic theory of propaga-tion, interference and diffraction of light. Cambridge Univ Press, Cambridge, UK
23. Anantharam A, Onoa B, Edwards RH et al (2010) Localized topological changes of the plasma membrane upon exocytosis visualized by polarized TIRFM. J Cell Biol 188: 415–428
24. Sund SE, Swanson JA, Axelrod D (1999) Cell membrane orientation visualized by polarized total internal reflection fluorescence. Biophys J 77:2266–2283
25. Axelrod D (1981) Cell-substrate contacts illu-minated by total internal reflection fluores-cence. J Cell Biol 89:141–145
26. Stock K, Sailer R, Strauss W et al (2003) Variable‐angle total internal reflection fluores-cence microscopy (VA‐TIRFM): realization and application of a compact illumination device. J Microsc 211:19–29
27. Fish KN (2009) Total internal reflection fluo-rescence (TIRF) microscopy. Curr Protoc Cytom Chapter 12, Unit 12.18
28. Olveczky BP, Periasamy N, Verkman A (1997) Mapping fluorophore distributions in three dimensions by quantitative multiple angle-total internal reflection fluorescence microscopy. Biophys J 73:2836–2847
29. Oheim M, Loerke D, Chow RH et al (1999) Evanescent-wave microscopy: a new tool to gain insight into the control of transmitter release. Philos T Roy Soc B 354:307–318
30. Rohrbach A (2000) Observing secretory gran-ules with a multiangle evanescent wave micro-scope. Biophys J 78:2641–2654
31. Loerke D, Stühmer W, Oheim M (2002) Quantifying axial secretory-granule motion with variable-angle evanescent-field excitation. J Neurosci Methods 119:65–73
32. Weisswange I, Bretschneider T, Anderson KI (2005) The leading edge is a lipid diffusion barrier. J Cell Sci 118:4375–4380
33. Millán J, Hewlett L, Glyn M et al (2006) Lymphocyte transcellular migration occurs through recruitment of endothelial ICAM-1 to caveola-and F-actin-rich domains. Nat Cell Biol 8:113–123
34. Saffarian S, Kirchhausen T (2008) Differential evanescence nanometry: live-cell fluorescence measurements with 10-nm axial resolution on the plasma membrane. Biophys J 94: 2333–2342
35. Pitkeathly WT, Poulter NS, Claridge E et al (2011) Auto-align – multi-modality fluores-cence microscopy image co-registration. Traffic 13:204–217
36. Mattheyses AL, Axelrod D (2006) Direct measurement of the evanescent field profile produced by objective-based total internal reflection fluorescence. J Biomed Opt 11: 014006
37. Doherty GJ, McMahon HT (2009) Mechanisms of endocytosis. Annu Rev Biochem 78:857–902
Natalie S. Poulter et al.
23
38. Damke H (1996) Dynamin and receptor- mediated endocytosis. FEBS Lett 389:48–51
39. Roth MG (2005) Clathrin-mediated endocy-tosis before fluorescent proteins. Nat Rev Mol Cell Biol 7:63–68
40. Gaidarov I, Santini F, Warren RA et al (1999) Spatial control of coated-pit dynamics in living cells. Nat Cell Biol 1:1–7
41. Cao H, Garcia F, McNiven MA (1998) Differential distribution of dynamin isoforms in mammalian cells. Mol Biol Cell 9:2595–2609
42. Rappoport JZ, Taha BW, Lemeer S et al (2003) The AP-2 complex is excluded from the dynamic population of plasma membrane- associated clathrin. J Biol Chem 278: 47357–47360
43. Yarar D, Waterman-Storer CM, Schmid SL (2005) A dynamic actin cytoskeleton functions at multiple stages of clathrin-mediated endocy-tosis. Mol Biol Cell 16:964–975
44. Rappoport JZ (2008) Focusing on clathrin- mediated endocytosis. Biochem J 412:415–423
45. Soulet F, Yarar D, Leonard M et al (2005) SNX9 regulates dynamin assembly and is required for efficient clathrin-mediated endo-cytosis. Mol Biol Cell 16:2058–2067
46. Lee D, Wu X, Eisenberg E et al (2006) Recruitment dynamics of GAK and auxilin to clathrin-coated pits during endocytosis. J Cell Sci 119:3502–3512
47. Jahn R, Südhof TC (1999) Membrane fusion and exocytosis. Annu Rev Biochem 68: 863–911
48. Simon SM (2009) Partial internal reflections on total internal reflection fluorescent micros-copy. Trends Cell Biol 19:661–668
49. Jaiswal JK, Fix M, Takano T et al (2007) Resolving vesicle fusion from lysis to monitor calcium-triggered lysosomal exocytosis in astrocytes. Proc Natl Acad Sci U S A 104: 14151–14156
50. Schmoranzer J, Goulian M, Axelrod D et al (2000) Imaging constitutive exocytosis with total internal reflection fluorescence micros-copy. J Cell Biol 149:23–32
51. Burridge K, Fath K, Kelly T et al (1988) Focal adhesions: transmembrane junctions between the extracellular matrix and the cytoskeleton. Annu Rev Cell Biol 4:487–525
52. Burridge K, Chrzanowska-Wodnicka M (1996) Focal adhesions, contractility, and signaling. Annu Rev Cell Dev Biol 12:463–519
53. Webb DJ, Parsons JT, Horwitz AF (2002) Adhesion assembly, disassembly and turnover in migrating cells-over and over and over again. Nat Cell Biol 4:E97–E100
54. Ezratty EJ, Bertaux C, Marcantonio EE et al (2009) Clathrin mediates integrin endocytosis for focal adhesion disassembly in migrating cells. J Cell Biol 187:733–747
TIRF Microscopy for Cell Biology
25
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology, vol. 1251, DOI 10.1007/978-1-4939-2080-8_2, © Springer Science+Business Media New York 2015
Chapter 2
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
Ricardo Mostany, Amaya Miquelajauregui, Matthew Shtrahman, and Carlos Portera-Cailliau
Abstract
Two-photon excitation (2PE) overcomes many challenges in fluorescence microscopy. Compared to confocal microscopy, 2PE microscopy improves depth penetration, owing to the longer excitation wavelength required and to the ability to collect scattered emission photons as a useful signal. It also minimizes photo-damage because lower energy photons are used and because fluorescence is confined to the geometrical focus of the laser spot. 2PE is therefore ideal for high-resolution, deep-tissue, time-lapse imaging of dynamic processes in cell biology. Here, we provide examples of important applications of 2PE for in vivo imaging of neuronal structure and signals; we also describe how it can be combined with optogenetics or photolysis of caged molecules to simultaneously probe and control neuronal activity.
Key words 2-Photon, Axonal bouton, Calcium imaging, Channelrhodopsin, Confocal microscopy, Dendritic spine, Electroporation, Green fluorescent protein, Optogenetics, Oregon Green BAPTA, Photomultiplier tube, Synaptic plasticity, Turnover, ScanImage, Uncaging
1 Introduction
Fluorescence microscopy has emerged as the preferred tool for studying the structure and dynamics of biological systems both in vivo and in vitro. In recent decades, we have witnessed unprec-edented technological advances in molecular and cellular imag-ing, which was fueled in large part by the discovery of fluorescent molecules that can be used to image cellular structure and the dynamics of organelles or even single proteins [1]. The sensitivity of fluorescence detection for these molecules (e.g., fluorescent proteins from a variety of aquatic invertebrate species such as jelly-fish, synthetic fluorescent dyes, and quantum dots) is exquisite [2]. Chemists and molecular biologists have also modified fluorescent proteins such that individual proteins can be used to monitor changes in intracellular calcium, pH, protein–protein interactions,
1.1 Fluorescence Microscopy
26
or the function of single enzymes such as kinases or the ubiquitin proteasome system [1].
In parallel with these discoveries, rapid advances in fluores-cence microscopy techniques have led to creative new ways to study biological processes [3]. For example, confocal microscopy allowed scientists to visualize these fluorescent molecules with improved spatial resolution, fluorescence lifetime imaging micros-copy (FLIM) and Forster resonance energy transfer (FRET) led them to study protein–protein interactions and fluorescence recov-ery after photobleaching (FRAP) made it possible to investigate protein turnover and trafficking, to name a few. In addition, the diffraction resolution limit of light microscopy was recently broken with novel methods, such as stimulated emission depletion (STED), bringing us to a new era of super-resolution microscopy [4–6].
Besides the discovery of fluorescent proteins, perhaps the most sig-nificant advance in bio-imaging of the last 25 years was the inven-tion, in 1990, of two-photon excitation (2PE) microscopy [7]. Since then, 2PE has become widely popular for in vivo imaging of neuronal structure because of its superior depth penetration and reduced photobleaching compared to confocal or epifluorescence microscopy [8]. Imaging deeper with 2PE is achieved by means of excitation with an infrared laser and optics that collect much of the scattered emitted light, while maintaining excellent spatial resolu-tion. An added bonus is that 2PE greatly decreases photodamage (such as photobleaching and phototoxicity) and is therefore more compatible with experiments requiring prolonged continuous tis-sue illumination (e.g., calcium imaging). Both advantages make 2PE ideally suited to chronic in vivo imaging in the intact brain.
Single-photon excitation requires the absorption of a high energy photon to excite an orbital electron of a fluorophore to a vibrationally and electronically excited state (excitation; Fig. 1a). When this electron relaxes to its ground state, it emits a photon of light of a different wavelength than the excitation photon (fluores-cence emission). In 2PE this process is achieved by the quasi- simultaneous absorption of two photons: the first one excites the electron to a virtual intermediate state, and the second one com-pletes its excitation to reach the final excited state.
The main shortcoming of single-photon excitation is its inef-ficient fluorescence excitation: because light absorption occurs throughout much of the specimen, a pinhole in front of the detec-tor is required in confocal microscopy to reject fluorescent photons emanating from outside the focus (Fig. 1b). Unfortunately, the pinhole similarly rejects in-focus photons that subsequently scatter, and as a result only unscattered photons contribute to the signal. This inefficiency demands high laser power for imaging, which cre-ates unwanted photodamage [9] and limits imaging depth to the free mean path of visible light (≤100 μm in biological tissue [10]).
1.2 Two-Photon Excitation Microscopy
Ricardo Mostany et al.
27
Fig. 1 Principles of 2PE microscopy. (a) Jablonski energy diagram showing the electron excitation process in single- (left ) and two-photon (right ) fluorescence microscopy. Single-photon excitation (left ) requires the absorption of a high energy photon to excite an orbital electron of a fluorophore to a vibrationally and electroni-cally excited state (excitation). When this electron relaxes to its ground state, it emits a photon of light of a different wavelength than the excitation photon (fluorescence emission). In 2-photon excitation (2PE; right ), this process is achieved by the quasi-simultaneous absorption of two photons; the first one excites the elec-tron to a virtual intermediate state, and the second one completes the excitation to reach the electronic excited state. (b) Confocal versus 2PE microscope systems. In confocal systems (left ), the presence of a pinhole right before the photodetector rejects the photons emitted from outside the focus (e.g., photon #6), as well as those scattered on their way to the PMT (e.g., photon #4). Only unscattered photons coming from the focal plane are able to pass through the pinhole (e.g., photon #5) and contribute to the signal. In 2PE systems (right ) there is no need for a pinhole since photons contributing to the signal come only from the geometrical focus of the excitation spot (e.g., photon #5), even if they were scattered on their path to the PMT (e.g., photon #4)
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
28
In 2PE, two lower energy photons are absorbed simultaneously to excite the fluorophore. Because of the steep dependence of absorption rate on photon concentration (light intensity), fluores-cence is confined to the geometrical focus of the laser excitation spot, which provides inherent optical sectioning. A pinhole is not required because all emitted photons (regardless of how much they scatter on their path to the detector) convey a useful signal (Fig. 1b). Greater tissue penetration is possible in 2PE (compared to confocal microscopy) thanks to the longer excitation wavelength used and to the ability to collect scattered emission photons as a useful signal. These properties of 2PE also limit photodamage since the absorption is confined to a tiny focal volume, and fewer excitation events are required to achieve the same signal due to the improved collection efficiency. In addition, because 2PE micros-copy requires photons of lower energy than those used in one- photon excitation, this reduces photodamage further.
2 Materials
Although commercial systems are available for multiphoton microscopy, some users may prefer to custom build their 2PE microscope (Fig. 2). Because the hardware and software needed for laser beam scanning and data acquisition in 2PE and confocal microscopy are so similar, commercial confocal systems can some-times be converted into a two-photon microscope [11]. For fur-ther reading on configuring a 2PE system, please refer to the following resources [3, 8, 9, 12].
Mode-locked lasers are well matched to the requirements for efficient 2PE (short pulse width of 50–100 fs and high repetition rates of ~100 MHz). In particular, tunable femtosecond Ti:Al2O3 (titanium:sapphire) lasers are commonly used in most laboratories because the spectral range (690–1,050 nm) is sufficient to excite a wide variety of available fluorophores. Fixed wavelength mode- locked lasers in the 1,000–1,250 nm range (e.g., Nd:YLF, Yb:KYW, or Cr:forsterite) can also be used, for example, to image red-shifted fluorescent proteins [13], and they are more affordable than tun-able lasers. Longer wavelengths of ~1,300 nm can be achieved when an optical parametric oscillator (OPO) is coupled to a stan-dard Ti:Al2O3 laser; this configuration may be desirable to achieve deeper tissue penetration [14] or multicolor imaging [12]. Others have reached record depths of imaging using regenerative amplifiers as the excitation source, which lower the laser repetition rate and thereby increase the yield of nonlinear optical processes [10, 15].
In addition to the laser, the main components in a standard 2PE microscope are a beam expander, a fast shutter, scanning mirrors (closed loop or resonant), photomultiplier tubes (PMT), and the different optical components (e.g., scan and tube lenses, reflecting and dichroic mirrors, objective lens), needed to guide the laser beam to the sample and the emitted light to the PMTs (Fig. 2).
2.1 Instrumentation: Laser and Microscope
Ricardo Mostany et al.
29
For custom microscopes several image acquisition and image pro-cessing tools have been developed:
1. ScanImage [16] is an image acquisition application for con-trolling laser scanning microscopes. It was developed in MATLAB (The MathWorks, Inc.) and is currently used by >150 laboratories worldwide (https://openwiki.janelia.org/wiki/display/ephus/ScanImage). This versatile software uses simple graphical user interfaces to operate the scan mirrors
2.2 Image Acquisition and Image Processing Tools
Fig. 2 Components of a 2PE microscope. The basic components of a 2PE micro-scope include the following elements: the two-photon laser (Ti:sapphire laser), a beam expander (the goal is to fill the back aperture of the objective lens), a half-wave plate combined with a polarizing cube to control and measure the laser power, a fast shutter to control exposure, scanning mirrors (vertical and horizon-tal, in this example they are closed-loop galvanometer mirrors), a scan lens, a tube lens, a dichromatic mirror, the objective, a collection lens, the photomulti-plier tubes, and a CPU to integrate and control the image acquisition process
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
30
(scan speed, zoom, rotation) and define the properties of the images to be acquired (e.g., number of slices in a stack, step size), as well as settings to carry out uncaging experiments.
2. Another helpful application is ImageJ (http://rsbweb.nih.gov/ij/) and some of its plug-in collections (www.macbiophotonics.ca/imagej; http://www.uhnresearch.ca/facilities/wcif/imagej/), which are used for image processing and analysis.
3. Neurolucida (MicroBrightField, Inc.) and its confocal image stack module is a commercial software that can be used to trace and analyze complete axonal and dendritic arbors of neurons within image stacks acquired with 2PE and extract useful infor-mation (e.g., length, branching nodes, branch order, complex-ity index, etc.) to characterize the neuron imaged. Standard tools for analysis of calcium imaging data (see Subheading 3.4) are still being developed [17], and individual labs use custom routines usually written in MATLAB.
3 Methods
Both synthetic dyes and fluorescent proteins are available to image the structure of neurons and glia in the brain (e.g., tracing the dendrites and axons of particular neuronal types) or to record neu-ronal activity with calcium imaging. Commonly used synthetic dyes for imaging neuronal structure with 2PE include Lucifer yel-low, Alexa fluor, DiI, and fluorescein. Intravascular injection of fluorescent dextrans permits the study of blood flow dynamics at the level of capillaries in the superficial layers of the cortex [18–21]. In addition, methoxy X-O4 is a fluorescent compound that crosses the blood-brain barrier and binds to amyloid plaques in mouse models of Alzheimer disease [22]. Synthetic calcium indica-tor dyes include Fluo-4, Fura-2, and OGB-1. Suforhodamine 101, a red dye that labels glia [23], is commonly used in calcium imag-ing experiments to distinguish neurons from glia.
A wide array of fluorescent proteins also exists over a rainbow palette of colors that allow an increasing number of applications [24, 25]. For example, EGFP, YFP, mCherry, and td-Tomato are frequently used for imaging neuronal structure with 2PE, while TNXXL, YC3.60, and GCaMP6 have been developed as geneti-cally encoded fluorescent calcium indicator proteins [26–29] (see Note 1). For in vivo imaging of brain cells, specific cell subsets can be labeled with fluorescent proteins through an array of genetic methods (alone or combined). Below we discuss briefly the most commonly used methods for labeling cells for imaging with 2PE:
1. Filling cells with synthetic dyes: For live imaging of neural structure with 2PE microscopy, cells can be labeled with fluo-rescent dyes using Diolistics [30], by direct injection of dyes
3.1 Making the Brain Visible for 2PE Microscopy: Fluorescent Labeling of Cells and Tissues
Ricardo Mostany et al.
31
into the brain or by filling them with a particular dye during whole-cell recordings [31]. Similarly, synthetic calcium dyes can be introduced directly into cells or electroporated using a patch pipette [32]. Alternatively, the acetoxy-methyl ester (AM) variants of dyes like Fluo-4 or OGB-1 can be bulk-loaded into the brain, which allows for recording the activity of large ensembles of neurons [33–35].
2. Transgenic and gene targeting (see Note 2): Mice engineered to express fluorescent proteins under the control of specific pro-moters (e.g., thy1 promoter; [36]) have been used for chronic imaging neuronal structure in vivo for a decade [37, 38]. The use of conditional gene expression (e.g., Cre-loxP, Flp-FRT) and inducible systems (e.g., Tet-ON/Tet-OFF) provides fur-ther spatial and temporal control of gene expression [39].
3. Transfection methods: In the context of live cell imaging with 2PE, nucleic acids (DNA or shRNA constructs) are most com-monly introduced in neurons using plasmid electroporation or viral infection. These methods can be used in Cre mice for conditional gene activation to obtain higher labeling specificity [39]. For in vivo electroporation, the delivery of a plasmid into neurons of living mice is achieved by injecting a DNA solution into the brain (or into a single cell with patch-clamp electro-physiology) followed by short electric pulses that permeabilize cell membranes temporarily and allow plasmid entry [40]. For in utero electroporation, the DNA is injected into the cerebral ventricles in mouse embryos in order to target subpopulations of neuron precursors (e.g., layer 2/3 pyramidal neurons of the cerebral cortex) [41–43]. Recombinant viruses (e.g., adenovi-ruses, lentiviruses, herpes viruses) are increasingly used for gene delivery into neurons, typically via stereotaxic intracranial injections targeted to a particular brain region [44–46].
Two different surgical preparations, glass-covered cranial window and thinned skull, have been developed to get optical access to the brain and image structure and functionality of labeled neurons [47–50]. The cranial window preparation requires the removal of a piece of skull (leaving the dura intact), and then the craniotomy is covered with a tiny glass coverslip (Fig. 3). The thinned skull preparation requires mechanical thinning of the superficial layers of the skull, preserving a thin layer of the bone that allows imaging through it. There is also a variant of the thinned skull approach, in which the skull is thinned, polished, and ultimately reinforced with a layer of cyanocrylate glue and a cover glass [51]. With any of these methods, one can achieve sufficient spatial resolution with 2PE microscopy to detect individual dendritic spines or axonal boutons.
3.2 Cranial Window for In Vivo 2PE Microscopy
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
32
The main advantage of the glass-covered cranial window method is that it allows for multiple imaging sessions to be con-ducted for longitudinal imaging of the same neuronal or glial pro-cesses within large fields of view. The cranial window approach (but not the thin skull preparation) also makes it possible to per-form certain manipulations of the brain before sealing the crani-otomy with a cover glass (e.g., electrode implantation, direct intracortical pharmacology, bolus loading of calcium dyes, viral injections). In addition, craniotomies are necessary for implanting gradient refractive index lenses (GRIN) in experiments that require imaging of deep brain structures [52] (see Note 3). The main drawback of the cranial window method is that it is more techni-cally demanding as only the best preparations remain optically transparent for weeks or months, and even the slightest perturba-tion of the dura mater will cause a quick worsening of the quality of the imaging. Because the skull is never breached with the thinned skull technique, the incidence of infection and inflammation is minimal, and the success rate is much higher [53]. One downside of transcranial imaging is that skull thinning, which has to be repeated before every imaging session, can only been done a lim-ited number of times (<3), or else the quality of imaging deterio-rates. In this regard, the polished and reinforced preparation avoids the bone inflammation that results from repeated thinning [51].
Fig. 3 Cranial window surgery for in vivo imaging with 2PE microscopy. The location of the cranial window is selected based on anatomical landmarks or functional imaging (1). In this case, the window was placed over the left barrel cortex. Under anesthesia, the skull is exposed, and a circular portion of the bone is gently carved with a pneumatic drill (2). Next, the bone flap is removed with small forceps (3) taking care not to damage the underlying meninges and vasculature (4). A glass coverslip is gently placed over the craniotomy (5). The edges of the glass window are sealed with cyanoacrylate glue and dental cement (6). A small well is also made around the window with dental acrylic to accommodate the objective lens and a drop of water for imaging. A titanium bar is then embedded in the dental acrylic (7), which can later be used to attach the mouse on to the microscope stage. The cortical vasculature can be seen under the microscope objective (7). Please refer to Mostany R and C Portera-Cailliau (2008) for a video of the procedure
Ricardo Mostany et al.
33
The structure and function of different cell types in the brain that express fluorescent dyes or proteins can be imaged in vivo using 2PE microscopy allowing the study of dynamic anatomical and functional changes as a result of learning and memory or in response to sensory inputs from the environment, or how they compensate for or degenerate in disease. The extraordinary reso-lution of the images acquired with 2PE microscopy makes it pos-sible to observe tiny neuronal structures, such as spines and axonal boutons (~1 μm in diameter), and even quantify the turn-over and trafficking of synaptic proteins (e.g., PSD95, Ras) within these structures [13, 54]. In the last decade, several studies that imaged synaptic structure in vivo with 2PE were able to record changes in synapses that were either associated with sensory experi-ence and learning motor tasks (reviewed in [55, 56]), or triggered by stroke [18]. Chronic 2PE microscopy in vivo can also be used to image neuroglial [57] and neurovascular [51] interactions, blood flow dynamics [21, 19], and amyloid plaques [58, 59], among others. By imaging neurons in the intact brain of living mice, such longitudinal, high-resolution imaging studies of den-dritic and axonal segments have provided valuable information about dynamic aspects of synapses that could not previously be examined using histological studies in fixed tissue (Fig. 4).
3.3 Imaging Neuronal Structure with 2PE Microscopy
Fig. 4 Chronic high-resolution imaging of dendritic spines with in vivo 2PE microscopy. High-resolution images of dendritic spines acquired with in vivo 2PE microscopy before and after stroke. The day of imaging is shown in the lower right-hand corner. Shown is an apical dendritic segment from a layer 5 pyramidal neuron in peri- infarct cortex before (green ) and after (red ) unilateral permanent middle cerebral artery occlu-sion (MCAO) in mice. All images are maximum intensity projections (4–7 slices, 1.5 μm apart). A few exam-ples of always present spines (yellow arrowheads ), gained spines (green arrowheads ), and lost spines (red arrowheads ) are shown. Blue asterisks at +4 d post MCAO denote transient dendritic swelling after stroke. See Mostany et al. [18] for details
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
34
When neurons fire action potentials, the intracellular concentra-tion of calcium ions rises. One can record neuronal activity opti-cally by using fluorescent dyes (or proteins) that respond to binding of calcium by changing their spectral properties (e.g., their fluores-cent intensity increases or decreases, or their excitation/emission spectra change). Calcium imaging using 2PE microscopy is an ideal tool for interrogating large ensembles of neurons in the intact brain [32] because it offers advantages over traditional single unit recordings using microelectrodes. First, with calcium imaging one can record signals from hundreds (potentially thousands) of neu-rons simultaneously. When combined with mouse genetics to label individual subpopulations of neurons, one can also be certain of which cell types the calcium signals are being recorded from. Second, calcium imaging is less invasive, and circuits can be recorded without penetrating electrodes that might disrupt nor-mal activity.
For calcium imaging, one can use 2PE microscopy to record calcium transients in single dendritic spines [60, 61] or to monitor the spontaneous or evoked activity of large ensembles of neurons simultaneously with single cell resolution [62–64] (Fig. 5).
Unfortunately, two-photon calcium imaging suffers from important drawbacks, such as poor temporal resolution and low signal-to-noise ratio [17, 65]. Newer generation genetically encoded calcium indicators with improved signal-to-noise will overcome
3.4 Recording Neuronal Signals with 2PE Microscopy
3.4.1 Calcium Imaging
Fig. 5 Calcium imaging of neuronal ensemble activity with 2PE microscopy. (a) Typical field of view of layer 2/3 neurons (green) and glia (yellow) stained with the calcium indicator dye Oregon Green BAPTA-1AM and imaged with in vivo 2PE microscopy. Sulforhodamine 101 (a red dye) was used to stain glia. The image is a single frame (~120 μm below the dura) in a representative calcium imaging movie (3 min, 3.9 frames per second) from a 14-day-old mouse. (b) Calcium traces showing the relative changes in fluorescence intensity over the baseline fluorescence (ΔF/F) of 5 different layer 2/3 neurons from a representative calcium movie. The upward deflections represent spiking events within those neurons
Ricardo Mostany et al.
35
some of these limitations. For example, although synthetic indicators are superior for detecting single spikes, some of the newer geneti-cally encoded indicators (YC3.60, GCaMP6f) may detect action potentials quite reliably [66]. Importantly, recent developments in faster scanning (e.g., acousto-optical deflectors), parallelization of 2PE (e.g., multifocal multiphoton microscopy), and improved pho-todetectors suggest that over the next decade optical probing of neural activity with calcium imaging could eventually be an excellent alternative to electrophysiology, offering a less invasive approach to record action potential firing in large ensembles of identifiable neu-rons in three dimensions (see Subheading 4, step 2).
Another way to optically probe neuronal activity is to image changes in membrane potential. For this purpose several voltage- sensing proteins (e.g., FLaSh, VSFP2, SPARC, Flare, Opto-patch) have been designed that work well with 2PE microscopy and could in theory be useful to monitor the activity of thousands of individual neurons simultaneously [67–69]. These sensors are usually FRET based and can report both subthreshold changes in membrane potential and spiking activity of neurons. Unfortunately, most current approaches for voltage sensing with genetically encoded or synthetic indicators have poor spatial res-olution and very low signal-to- noise ratio, requiring the averag-ing of many stimuli to detect responses.
Another advantage of 2PE microscopy is that it can easily be com-bined with optogenetics and be used to uncage neurotransmitters for temporally and spatially precise manipulation and probing of neuronal activity within intact neural circuits.
This technique enables researchers to silence or stimulate geneti-cally specified classes of neurons (or other electrically excitable cells) with exquisite temporal precision using light-sensitive mole-cules [70, 71]. For example, channelrhodopsin-2 (ChR2) is a light-activated cation channel that depolarizes neurons, whereas halorhodopsin (NphR) and archaerhodopsin-3 (Arch) are a chloride channel and an outward proton pump, respectively, that enable almost complete silencing of neurons. Using optogenetic tools, cell type-specific and minimally invasive photostimulation has revealed causal relationships between activity of neuronal pop-ulations and animal behavior. In addition to evoked behaviors, electrophysiological recordings (e.g., patch-clamp or “optrodes”) [72] or calcium imaging [73] can be used as the readout of neuro-nal activity. Thus, by combining 2PE microscopy with calcium dyes or voltage-sensitive dyes, subsets of neurons expressing opto-genetic sensors could be specifically modulated with light to reveal and mimic patterns of connectivity with phenomenal temporal and spatial precision [74]. In the future, as improved red-shifted
3.4.2 Voltage-Sensing Proteins
3.5 Combining 2PE Microscopy with Optogenetics, Glutamate Uncaging
3.5.1 Optogenetics
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
36
calcium indicator dyes become available, it will be easier to combine optogenetics with two-photon calcium imaging. Although optoge-netic manipulations are by enlarge done with single-photon excita-tion, 2PE of ChR2 is possible in vivo [75], and this option may be preferable when precise stimulation of individual synapses or por-tions of circuits is desired.
The photorelease of caged, biochemically inert effector molecules (e.g., nucleotide, neurotransmitter, second messenger) can be used to control molecular interactions in cells [76]. With two-photon uncaging, the absorption energy is used to release the caged mol-ecule, typically a neurotransmitter, from its protective chemical group, in order to mimic normal synaptic activity. The photoacti-vation of caged neurotransmitters has achieved synapse-specific resolution with the use of 2PE in brain tissue. By combining 2PE and uncaging (e.g., of MNI-glutamate), postsynaptic receptors of individual spines can be stimulated with excellent temporal and spatial resolution while structural and activity changes are assessed. These techniques have been successfully applied to the study of postsynaptic signaling and local circuit mapping at the level of indi-vidual dendritic spines [77–79]. An exciting recent advance was the successful stimulation of identified dendritic spines in vivo, in adult mice, using two-photon uncaging [80].
The invention of 2PE microscopy has revolutionized fluorescence imaging over the last 2 decades, thanks to the vision of physicists, chemists, engineers, and biologists working towards a common goal. Over the next few years, we will continue to witness unprec-edented advances that lead to additional improvements in this powerful tool. In particular, developments that will enhance our ability to image deeper in the brain and with better temporal reso-lution in awake behaving mice will be especially useful.
Light scattering of both the near-infrared excitation wavelengths used in 2PE microscopy and of the emitted fluorescence imposes limits on how deep one can image [81]. As a result, in vivo cal-cium imaging with conventional 2PE has been restricted to the superficial layers of the neocortex. A variety of approaches have been established to increase the depth range for in vivo calcium imaging, including the use of high numerical aperture objectives and longer wavelengths (e.g., with fixed-wavelength laser sources or OPOs), using adaptive optics and wavefront optimization to correct for spherical aberrations and lensing effects [82, 83] or implementing ultrashort-pulsed regenerative amplifiers [15]. An alternative approach to deeper imaging is to improve on the col-lection of scattered fluorescent emissions that are not transmitted through the objective and would otherwise be a wasted signal, for example, by using a ring of optical fibers around the objective [84]. A different approach entirely is to use microendoscopes to reach
3.5.2 Two-Photon Uncaging
3.6 Future Directions
3.6.1 Deeper and Brighter Imaging
Ricardo Mostany et al.
37
deep brain structures, such as the hippocampus [52, 85]. In addition, signal-to- noise ratios will improve as better fluorophores become available. For calcium imaging this is a critical issue, because when faster scanning is implemented (see below), the pixel dwell time of the laser will be reduced, resulting in less signal from the indicator. Major efforts are underway to design better genetically encoded calcium indicators to solve these problems [28].
A major problem with calcium imaging is that it is hard to relate changes in fluorescence of the indicator molecule to neuronal spik-ing both in terms of determining the exact numbers of action potentials and especially the timing between them. This is due to limitations in the signal-to-noise ratio of most indicators (see above) and in the speed of acquisition. Recent solutions have been put forward to improve image acquisition speed. For example, spatio-temporal multiplexing of multibeam scanning with 2PE can improve temporal resolution and make it possible to record neu-rons at different depths simultaneously [86]. Alternatively, one can record only useful signals from cell bodies using targeted path scan-ning [87] or acousto-optical deflectors [88, 89]. Lastly, there is considerable effort in the field to shape the excitation beam to mini-mize or eliminate scanning, in order to rapidly excite large numbers of cells in parallel without significantly compromising the pixel dwell time. These novel geometries include the use of Bessel beams [90] and light sheets [91] to achieve thin 2PE planes perpendicular to the emission path or the implementation of spatial light modula-tors that shape the laser beam into an arbitrary light pattern, which allows for simultaneous imaging (or uncaging) at different locations [92] (see Note 4). Further developments in years to come will no doubt bring us closer to our goal of being able to record, with mil-lisecond precision, the firing of thousands of neurons, distributed over a volume of brain tissue in behaving animals.
An important goal of neuroscientists is to link changes in the structure and function of neuronal circuits to changes in behavior. It is therefore critically important to study these phenomena in the intact behaving animal. In the case of calcium imaging of neuronal activity, this has been accomplished by recording in head-restrained mice while they navigate virtual environments [93] or by miniatur-izing 2PE microscopes into portable devices that allow investiga-tors to image freely moving animals [84, 94].
4 Notes
1. GECIs come in two varieties: single-fluorophore sensors and sensors involving FRET between two proteins. A good example of single-fluorophore GECIs is the GCaMP family of sensors, which is composed of a circularly permuted EGFP molecule
3.6.2 Faster Imaging
3.6.3 2PE Microscopy in Behaving Animals
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
38
that is flanked on one side by the calcium-binding protein calmodulin and on the other by the calmodulin-binding pep-tide M13. FRET-based sensors (e.g., the yellow cameleon dye YC 3.60) require a nonradiative energy transfer between an excited donor fluorophore (e.g., enhanced cyan fluorescent protein) and an acceptor fluorophore (e.g., circularly permuted Venus protein). The two fluorophores are connected by a linker sequence that is composed of a calcium-binding protein (e.g., calmodulin-M13, troponin); upon calcium binding, a conformational change in the linker protein brings the distance between the donor and acceptor fluorescent proteins to less than 10 nm, which is necessary for FRET to occur. This is reflected by a decrease in the fluorescence intensity of the donor protein (blue) and an increase in the acceptor fluores-cence (green), and the calcium signal is therefore expressed as the ratio of the two.
2. Additional information on specific gene expression patterns in available mouse lines can be found in Allen Brain Atlas (http://www.alleninstitute.org), Genepaint (http://www.genepaint.org), GENSAT (http://www.gensat.org), or The Jackson Laboratory (http://jaxmice.jax.org).
3. The microendoscopy approach to image deep structures relies on the use of GRIN lenses, which use internal variations in the refractive index (as opposed to curved refractive surfaces of conventional lenses) to guide light. In essence, GRIN lenses act as an optical relay that projects the scanning pattern of the 2PE microscope to a focal plane deep inside the tissue sample. A wide array of microendoscopes is available with varying physical length, optical working distance and numerical apertures, and field of views. An important shortcoming of microendoscopy is that it is associated with some degree of tissue damage because it relies on the insertion of a GRIN lens into the brain.
4. Unfortunately, because these light-sheet approaches are cur-rently limited in their depth penetration, it will be challenging to adapt them for live imaging in the intact rodent brain.
Acknowledgments This work was supported by the Stein Oppenheimer Endowment Award and by grants from the US National Institutes of Health (5R01HD054453 from the Eunice Kennedy Shriver National Institute of Child Health and Human Development and 5RC1NS068093 from the National Institute of Neurological Disorders and Stroke).
Ricardo Mostany et al.
39
References
1. Miyawaki A (2011) Proteins on the move: insights gained from fluorescent protein tech-nologies. Nat Rev Mol Cell Biol 12:656–668
2. Eigen M, Rigler R (1994) Sorting single mol-ecules: application to diagnostics and evolu-tionary biotechnology. Proc Natl Acad Sci U S A 91:5740–5747
3. Wilt BA, Burns LD, Wei Ho ET et al (2009) Advances in light microscopy for neuroscience. Annu Rev Neurosci 32:435–506
4. Hell SW, Wichmann J (1994) Breaking the diffraction resolution limit by stimulated emis-sion: stimulated-emission-depletion fluores-cence microscopy. Opt Lett 19:780–782
5. Hell SW (2003) Toward fluorescence nanos-copy. Nat Biotechnol 21:1347–1355
6. Sigrist SJ, Sabatini BL (2012) Optical super- resolution microscopy in neurobiology. Curr Opin Neurobiol 22:86–93
7. Denk W, Strickler JH, Webb WW (1990) Two- photon laser scanning fluorescence microscopy. Science 248:73–76
8. Svoboda K, Yasuda R (2006) Principles of two- photon excitation microscopy and its applica-tions to neuroscience. Neuron 50:823–839
9. Denk W, Svoboda K (1997) Photon upman-ship: why multiphoton imaging is more than a gimmick. Neuron 18:351–357
10. Theer P, Hasan MT, Denk W (2003) Two- photon imaging to a depth of 1000 microm in living brains by use of a Ti:Al2O3 regenerative amplifier. Opt Lett 28:1022–1024
11. Nikolenko V, Nemet B, Yuste R (2003) A two- photon and second-harmonic microscope. Methods 30:3–15
12. Entenberg D, Wyckoff J, Gligorijevic B et al (2011) Setup and use of a two-laser multipho-ton microscope for multichannel intravital fluo-rescence imaging. Nat Protoc 6:1500–1520
13. Gray NW, Weimer RM, Bureau I et al (2006) Rapid redistribution of synaptic PSD-95 in the neocortex in vivo. PLoS Biol 4:e370
14. Kobat D, Durst ME, Nishimura N et al (2009) Deep tissue multiphoton microscopy using longer wavelength excitation. Nat Protoc 17: 13354–13364
15. Mittmann W, Wallace DJ, Czubayko U et al (2011) Two-photon calcium imaging of evoked activity from L5 somatosensory neu-rons in vivo. Nat Neurosci 14:1089–1093
16. Pologruto TA, Sabatini BL, Svoboda K (2003) ScanImage: flexible software for operating laser scanning microscopes. Biomed Eng Online 2:13
17. Grewe BF, Helmchen F (2009) Optical probing of neuronal ensemble activity. Curr Opin Neurobiol 19(5):520–529
18. Mostany R, Chowdhury TG, Johnston DG et al (2010) Local hemodynamics dictate long- term dendritic plasticity in peri-infarct cortex. J Neurosci 30:14116–14126
19. Mostany R, Portera-Cailliau C (2008) A method for 2-photon imaging of blood flow in the neo-cortex through a cranial window. J Vis Exp. (12). pii: 678. doi:10.3791/678. PMID: 19066563
20. Schaffer CB, Friedman B, Nishimura N et al (2006) Two-photon imaging of cortical sur-face microvessels reveals a robust redistribution in blood flow after vascular occlusion. PLoS Biol 4:e22
21. Kleinfeld D, Mitra PP, Helmchen F et al (1998) Fluctuations and stimulus-induced changes in blood flow observed in individual capillaries in layers 2 through 4 of rat neocortex. Proc Natl Acad Sci U S A 95:15741–15746
22. Klunk WE, Bacskai BJ, Mathis CA et al (2002) Imaging Abeta plaques in living transgenic mice with multiphoton microscopy and methoxy-X04, a systemically administered Congo red derivative. J Neuropathol Exp Neurol 61: 797–805
23. Nimmerjahn A, Kirchhoff F, Kerr JN et al (2004) Sulforhodamine 101 as a specific marker of astroglia in the neocortex in vivo. Nat Methods 1:31–37
24. Shaner NC, Steinbach PA, Tsien RY (2005) A guide to choosing fluorescent proteins. Nat Methods 2:905–909
25. Chudakov DM, Matz MV, Lukyanov S et al (2010) Fluorescent proteins and their applica-tions in imaging living cells and tissues. Physiol Rev 90:1103–1163
26. Lutcke H, Murayama M, Hahn T et al (2010) Optical recording of neuronal activity with a genetically-encoded calcium indicator in anes-thetized and freely moving mice. Front Neural Circ 4:9
27. Tian L, Hires SA, Mao T et al (2009) Imaging neural activity in worms, flies and mice with improved GCaMP calcium indicators. Nat Methods 6:875–881
28. Looger LL, Griesbeck O (2012) Genetically encoded neural activity indicators. Curr Opin Neurobiol 22:18–23
29. Mank M, Santos AF, Direnberger S et al (2008) A genetically encoded calcium indicator for
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
40
chronic in vivo two-photon imaging. Nat Methods 5:805–811
30. Gan WB, Grutzendler J, Wong WT et al (2000) Multicolor “DiOlistic” labeling of the nervous system using lipophilic dye combinations. Neuron 27:219–225
31. Portera-Cailliau C, Pan DT, Yuste R (2003) Activity-regulated dynamic behavior of early dendritic protrusions: evidence for different types of dendritic filopodia. J Neurosci 23: 7129–7142
32. Grienberger C, Konnerth A (2012) Imaging calcium in neurons. Neuron 73:862–885
33. Garaschuk O, Milos R, Konnerth A (2006) Targeted bulk-loading of fluorescent indicators for two-photon brain imaging in vivo. Nat Protoc 1:380–386
34. Golshani P, Portera-Cailliau C (2008) In vivo 2-photon calcium imaging in layer 2/3 of mice. J Vis, Exp
35. MacLean J, Yuste R (2005) Imaging action potentials with calcium indicators: practical guide. In: Yuste R, Konnerth A (eds) Imaging neurons a laboratory manual, 2nd edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp 351–355
36. Feng G, Mellor RH, Bernstein M et al (2000) Imaging neuronal subsets in transgenic mice expressing multiple spectral variants of GFP. Neuron 28:41–51
37. Grutzendler J, Kasthuri N, Gan WB (2002) Long-term dendritic spine stability in the adult cortex. Nature 420:812–816
38. Trachtenberg JT, Chen BE, Knott GW et al (2002) Long-term in vivo imaging of experience- dependent synaptic plasticity in adult cortex. Nature 420:788–794
39. Jefferis GS, Livet J (2012) Sparse and combina-torial neuron labelling. Curr Opin Neurobiol 22:101–110
40. Judkewitz B, Rizzi M, Kitamura K et al (2009) Targeted single-cell electroporation of mamma-lian neurons in vivo. Nat Protoc 4:862–869
41. Saito T, Nakatsuji N (2001) Efficient gene transfer into the embryonic mouse brain using in vivo electroporation. Dev Biol 240:237–246
42. Dixit R, Lu F, Cantrup R et al (2011) Efficient gene delivery into multiple CNS territories using in utero electroporation. J Vis Exp. (52). pii: 2957. doi:10.3791/2957. PMID: 21730943
43. Cruz-Martin A, Crespo M, Portera-Cailliau C (2010) Delayed stabilization of dendritic spines in fragile X mice. J Neurosci 30:7793–7803
44. Atasoy D, Aponte Y, Su HH et al (2008) A FLEX switch targets Channelrhodopsin-2 to
multiple cell types for imaging and long-range circuit mapping. J Neurosci 28:7025–7030
45. Lowery RL, Majewska AK (2010) Intracranial injection of adeno-associated viral vectors. J Vis Exp. (45). pii: 2140. doi: 10.3791/2140. PMID: 21113119
46. Schultz BR, Chamberlain JS (2008) Recombinant adeno-associated virus transduc-tion and integration. Mol Ther 16:1189–1199
47. Holtmaat A, Bonhoeffer T, Chow DK et al (2009) Long-term, high-resolution imaging in the mouse neocortex through a chronic cranial window. Nat Protoc 4:1128–1144
48. Mostany R, Portera-Cailliau C (2008) A crani-otomy surgery procedure for chronic brain imaging. J Vis Exp. (12). pii: 680. doi:10.3791/680. PMID: 19066562
49. Cruz-Martin A, Portera-Cailliau C (2010) In vivo imaging of axonal and dendritic structures in developing cortex. In: Sharpe J, Wong R (eds) Imaging in developmental biology: a lab-oratory manual, 1st edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY, pp 513–522
50. Yang G, Pan F, Parkhurst CN et al (2010) Thinned-skull cranial window technique for long-term imaging of the cortex in live mice. Nat Protoc 5:201–208
51. Drew PJ, Shih AY, Driscoll JD et al (2010) Chronic optical access through a polished and reinforced thinned skull. Nat Methods 7: 981–984
52. Jung JC, Mehta AD, Aksay E et al (2004) In vivo mammalian brain imaging using one- and two-photon fluorescence microendoscopy. J Neurophysiol 92:3121–3133
53. Xu HT, Pan F, Yang G et al (2007) Choice of cranial window type for in vivo imaging affects dendritic spine turnover in the cortex. Nat Neurosci 10:549–551
54. Yasuda R, Harvey CD, Zhong H et al (2006) Supersensitive Ras activation in dendrites and spines revealed by two-photon fluorescence lifetime imaging. Nat Neurosci 9:283–291
55. Holtmaat A, Svoboda K (2009) Experience- dependent structural synaptic plasticity in the mammalian brain. Nat Rev Neurosci 10: 647–658
56. Yu X, Zuo Y (2011) Spine plasticity in the motor cortex. Curr Opin Neurobiol 21:169–174
57. Nimmerjahn A, Kirchhoff F, Helmchen F (2005) Resting microglial cells are highly dynamic surveillants of brain parenchyma in vivo. Science 308:1314–1318
58. Spires TL, Meyer-Luehmann M, Stern EA et al (2005) Dendritic spine abnormalities in amyloid precursor protein transgenic mice demonstrated
Ricardo Mostany et al.
41
by gene transfer and intravital multiphoton microscopy. J Neurosci 25:7278–7287
59. Tsai J, Grutzendler J, Duff K et al (2004) Fibrillar amyloid deposition leads to local syn-aptic abnormalities and breakage of neuronal branches. Nat Neurosci 7:1181–1183
60. Sabatini BL, Svoboda K (2000) Analysis of cal-cium channels in single spines using optical fluctuation analysis. Nature 408:589–593
61. Chen X, Leischner U, Rochefort NL et al (2011) Functional mapping of single spines in cortical neurons in vivo. Nature 475:501–505
62. Cossart R, Aronov D, Yuste R (2003) Attractor dynamics of network UP states in the neocor-tex. Nature 423:283–288
63. Kerr JN, Greenberg D, Helmchen F (2005) Imaging input and output of neocortical net-works in vivo. Proc Natl Acad Sci U S A 102: 14063–14068
64. Ohki K, Chung S, Ch'ng YH et al (2005) Functional imaging with cellular resolution reveals precise micro-architecture in visual cor-tex. Nature 433:597–603
65. Gobel W, Helmchen F (2007) In vivo calcium imaging of neural network function. Physiology 22:358–365
66. Hendel T, Mank M, Schnell B et al (2008) Fluorescence changes of genetic calcium indi-cators and OGB-1 correlated with neural activ-ity and calcium in vivo and in vitro. J Neurosci 28:7399–7411
67. Perron A, Mutoh H, Akemann W et al (2009) Second and third generation voltage-sensitive fluorescent proteins for monitoring membrane potential. Front Neural Circ 2:5
68. Ahrens KF, Heider B, Lee H et al (2012) Two- photon scanning microscopy of in vivo sensory responses of cortical neurons genetically encoded with a fluorescent voltage sensor in rat. Front Neural Circ 6:15
69. Chanda B, Blunck R, Faria LC et al (2005) A hybrid approach to measuring electrical activity in genetically specified neurons. Nat Neurosci 8:1619–1626
70. Fenno L, Yizhar O, Deisseroth K (2011) The development and application of optogenetics. Annu Rev Neurosci 34:389–412
71. Bernstein JG, Garrity PA, Boyden ES (2012) Optogenetics and thermogenetics: technolo-gies for controlling the activity of targeted cells within intact neural circuits. Curr Opin Neurobiol 22:61–71
72. Deisseroth K (2011) Optogenetics. Nat Methods 8:26–29
73. Zhang YP, Oertner TG (2007) Optical induc-tion of synaptic plasticity using a light-sensitive channel. Nat Methods 4:139–141
74. Peron S, Svoboda K (2011) From cudgel to scalpel: toward precise neural control with optogenetics. Nat Methods 8:30–34
75. Papagiakoumou E, Anselmi F, Begue A et al (2010) Scanless two-photon excitation of channelrhodopsin-2. Nat Methods 7:848–854
76. Ellis-Davies GC (2009) Basics of photoactiva-tion. Cold Spring Harb Protoc. pdb top55
77. Matsuzaki M, Ellis-Davies GC, Nemoto T et al (2001) Dendritic spine geometry is criti-cal for AMPA receptor expression in hippo-campal CA1 pyramidal neurons. Nat Neurosci 4:1086–1092
78. Sobczyk A, Scheuss V, Svoboda K (2005) NMDA receptor subunit-dependent [Ca2+] signaling in individual hippocampal dendritic spines. J Neurosci 25:6037–6046
79. Ashby MC, Isaac JT (2011) Maturation of a recurrent excitatory neocortical circuit by experience-dependent unsilencing of newly formed dendritic spines. Neuron 70:510–521
80. Noguchi J, Nagaoka A, Watanabe S et al (2011) In vivo two-photon uncaging of glutamate revealing the structure-function relationships of dendritic spines in the neocortex of adult mice. J Physiol 589:2447–2457
81. Oheim M, Beaurepaire E, Chaigneau E et al (2001) Two-photon microscopy in brain tis-sue: parameters influencing the imaging depth. J Neurosci Methods 111:29–37
82. Rueckel M, Mack-Bucher JA, Denk W (2006) Adaptive wavefront correction in two-photon microscopy using coherence-gated wavefront sensing. Proc Natl Acad Sci U S A 103: 17137–17142
83. Booth MJ (2007) Adaptive optics in micros-copy. Philos Transact A Math Phys Eng Sci 365:2829–2843
84. Engelbrecht CJ, Gobel W, Helmchen F (2009) Enhanced fluorescence signal in nonlinear microscopy through supplementary fiber-optic light collection. Opt Express 17:6421–6435
85. Flusberg BA, Cocker ED, Piyawattanametha W et al (2005) Fiber-optic fluorescence imaging. Nat Methods 2:941–950
86. Cheng A, Goncalves JT, Golshani P et al (2011) Simultaneous two-photon calcium imaging at different depths with spatiotempo-ral multiplexing. Nat Methods 8:139–142
87. Lillis KP, Eng A, White JA et al (2008) Two- photon imaging of spatially extended neuro-nal network dynamics with high temporal
Two-Photon Excitation Microscopy and Its Applications in Neuroscience
42
resolution. J Neurosci Methods 172: 178–184
88. Grewe BF, Langer D, Kasper H et al (2010) High-speed in vivo calcium imaging reveals neuronal network activity with near- millisecond precision. Nat Methods 7:399–405
89. Duemani Reddy G, Kelleher K, Fink R et al (2008) Three-dimensional random access multiphoton microscopy for functional imag-ing of neuronal activity. Nat Neurosci 11: 713–720
90. Planchon TA, Gao L, Milkie DE et al (2011) Rapid three-dimensional isotropic imaging of living cells using Bessel beam plane illumina-tion. Nat Methods 8:417–423
91. Truong TV, Supatto W, Koos DS et al (2011) Deep and fast live imaging with two-photon scanned light-sheet microscopy. Nat Methods 8:757–760
92. Nikolenko V, Watson BO, Araya R et al (2008) SLM microscopy: scanless two-photon imaging and photostimulation with spatial light modu-lators. Front Neural Circ 2:5
93. Dombeck DA, Harvey CD, Tian L et al (2010) Functional imaging of hippocampal place cells at cellular resolution during virtual navigation. Nat Neurosci 13:1433–1440
94. Kerr JN, Nimmerjahn A (2012) Functional imaging in freely moving animals. Curr Opin Neurobiol 22:45–53
Ricardo Mostany et al.
43
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology, vol. 1251, DOI 10.1007/978-1-4939-2080-8_3, © Springer Science+Business Media New York 2015
Chapter 3
Live Spheroid Formation Recorded with Light Sheet-Based Fluorescence Microscopy
Francesco Pampaloni , Roli Richa , Nariman Ansari , and Ernst H. K. Stelzer
Abstract
We provide a detailed protocol for a three-dimensional long-term live imaging of cellular spheroids with light sheet-based fl uorescence microscopy. The protocol allows the recording of all phases of spheroid formation in three dimensions, including cell proliferation, aggregation, and compaction. We employ the human hepatic cell line HepaRG transfected with the fusion protein H2B-GFP, i.e., a fl uorescing histone. The protocol allows monitoring the effect of drugs or toxicants.
Key words Light sheet-based fl uorescence microscopy , LSFM , SPIM , Three-dimensional cell cultures , Live-cell assay , Tumor spheroids , Spheroid formation , HepaRG cells , Live-cell imaging
1 Introduction
Data gathered during long-term fl uorescence imaging provides the basis for the analysis of spatiotemporal processes in single cells, tis-sues, and whole organisms [ 1 ]. Three-dimensional imaging is required to investigate cellular processes under close-to-natural conditions, especially in three-dimensionally organized cell cul-tures. The issues of phototoxicity, photobleaching, high recording speed, and three-dimensional imaging capability have been assessed by employing spinning-disk confocal fl uorescence microscopy [ 2 ], heavily optimized wide-fi eld fl uorescence microscopy with subse-quent deconvolution (OMX) [ 3 ], and light sheet-based fl uores-cence microscopy (LSFM) [ 4 , 5 ]. Among these three microscopies, LSFM (Fig. 1 ) is particularly gentle towards living samples. The light sheet-based illumination irradiates the specimen with an extremely low energy of about 2 μJ at 488 nm in the illumination plane [ 6 , 7 ]. Since LSFM takes advantage of modern cameras, a very high stack recording speed is possible. While in our fi rst imple-mentation of a Digital Scanned Light Sheet Microscope (DSLM), a speed of six planes/second was achieved, where each plane consisted of 2,048 × 2,048 pixels [ 6 ], an entire stack of 100 or
44
more images should now become available every second. Another important advantage of LSFM is the high dynamic range, which supports very complex image processing. With LSFM, long-term fl uorescence imaging of developing Arabidopsis thaliana lateral roots [ 8 ], fruit fl y ( Drosophila melanogaster ) embryos [ 5 , 9 ], and zebra fi sh ( Danio rerio ) embryos for up to 72 h has been per-formed without impairing the development [ 6 ]. The high speed allows one to follow, e.g., the mitosis and migration of more than 10,000 cells in a zebra fi sh embryo [ 6 ].
LSFM is also extremely well suited to record the behavior of live mammalian cells for long periods of time [ 10 ] and is the tool of choice for imaging three-dimensional cell cultures, such as cel-lular spheroids. The latter become more and more popular in basic research [ 11 ], for drug screening [ 11 – 15 ], and for personalized medicine [ 16 ]. Most immortalized and tumor cell lines, primary cells, as well as stem cells can form spheroids by spontaneous aggregation on nonadhesive substrate [ 17 ]. The spheroid forma-tion consists of aggregation, delay, and compaction phases. The entire process takes 3–7 days [ 18 ]. Previous work has shown that LSFM is particularly well suited to image tumor cell spheroids in three dimensions at high resolution [ 19 ] and with live cells [ 20 ].
Fig. 1 Light sheet-based fl uorescence microscopy (LSFM). ( a ) Setup of a single plane illumination microscope (aka SPIM) [ 5 ]). ( b ) Principles of LSFM imaging. Left : a single plane in the specimen is illuminated by a light sheet. Only the plane that is observed is also illuminated, resulting in lower photobleaching and lower photo-toxicity. Center : by moving the specimen through the stationary light sheet, a three-dimensional stack of images is recorded. Right : by rotating the specimen multiple-view image stacks are obtained. Combining multiple different views of the specimen increases the resolution along the z -axis. ( c ) Close-up of a specimen chamber, showing both the illumination and detection objective lenses oriented at a 90° angle with respect to each other
Francesco Pampaloni et al.
45
We show how to perform long-term live imaging by combin-ing the LSFM with a perfusion chamber. Agarose beakers are employed to allow the cell aggregation in a confi ned environment. As an example, we record the complete formation of a spheroid over a time period of 6 days. We use the recently developed hepatic human cell line HepaRG, which is well suited for drug and toxicity screenings [ 21 ].
2 Materials
1. Cell growth medium : Add 50 ml FBS (100 % fetal bovine serum), 5 ml penicillin/streptomycin, and 5 ml 200 mM L - glutamine to 445 ml DMEM (1× Dulbecco’s modifi ed eagle medium phenol-free). Mix 500 ml of medium. Store at 4 °C.
2. Cell dissociation : StemPro Accutase Cell Dissociation Reagent (Life technologies, #A1110501).
3. Agarose aliquots : Prepare a 1 % solution of high-melting point agarose (Sigma, A9539) in PBS. Aliquot the agarose solution in 2 ml tubes. Store at 4 °C.
4. Transduction reagent : BacMam viral particles containing a his-tone 2B-GFP expression cassette (Invitrogen C10594).
1. 25 cm 2 tissue culture fl ask (e.g., from greiner bio-one, www.greinerbioone.com ).
2. Hemocytometer (e.g., Neubauer hemocytometer).
HepaRG, terminally differentiated hepatic cells derived from a human hepatic progenitor cell line that retains many characteristics of primary human hepatocytes (e.g., from Life Technologies, HPRGC10).
Custom templates, see further in the text.
1. Sharp angled-tip precision forceps (e.g., Excelta SKU 50-SA, http://www.excelta.com ).
2. 1 ml syringes. 3. 0.55 × 25 mm (24 G × 1″) hypodermic needles. 4. Custom LSFM mounting system for the agarose beaker (details
further in the text).
1. Various suitable LSFM implementations are described in details in [ 5 , 22 , 23 ] (SPIM), (6) (DSLM), and [ 8 ] (monolithic DSLM, mDSLM). A commercial LSFM is available from Zeiss (Lightsheet Z.1, http://microscopy.zeiss.com/microscopy/en_de/products/imaging-sys tems/l ightsheet-z-1.html#Introduction ).
2.1 Chemicals and Reagents
2.2 Cell Culture Equipment
2.3 Cell Line
2.4 Templates for the “Agarose Beakers”
2.5 Further Equipment
2.6 Imaging
Light Sheet-Based Fluorescence Microscopy
46
2. Long working water-dipping distance objective lenses (e.g., Carl Zeiss W N-Achroplan 10×/0.3 NA).
1. A compact cell incubator (e.g., Galaxy ® 14S CO 2 incubator, New Brunswick http://eshop.eppendorfna.com/products/New_Brunswick_Galaxy_14S_CO 2 _Incubator ).
2. Heated hose for tubing system (e.g., series WSKW from Winkler http://en.winkler.eu ).
3. Microprocessor-controlled table-top temperature controller (e.g., WRT2000X, from Winkler http://en.winkler.eu ).
4. A 4 meter-long gas-permeable silicon tubing (e.g., from Reichelt Chemietechnik, Germany, 1 mm internal diameter, 2 mm external diameter, http://www.rct-online.de ).
5. Gas-impermeable rubber tubing (e.g., from Reichelt Chemietechnik, Germany—EPDM/PP pharmaceutical tub-ing, 1.6 mm internal diameter, 4.8 mm external diameter, http://www.rct-online.de ).
6. Peristaltic pump (e.g., REGLO digital from Ismatec, Germany, http://www.ismatec.de/de_d/pumpen/s_reglo/reglo_digi-tal.htm ).
7. Two-stop tubing for the peristaltic pump (e.g., 1.30 mm two- stop PharMed BPT tubing, Ismatec SC0328).
8. Autoclavable Luer-Lok connectors, mini tubing olive connec-tors (inner Ø 2 and 1.5 mm), and quick-disconnect coupling/nipple systems with valve (e.g., THOMAFLUID-POM with 1.6 mm nozzle). All this part can be purchased, e.g., from Reich elt Chemietechnik, Germany, http://www.rct-online.de ).
Image processing software (e.g., Fiji, a variant of ImageJ, http://fi ji.sc ).
3 Methods
1. Rapidly thaw the frozen HepaRG cells (within 5 min) in a 37 °C water bath. Dilute the thawed cells in a 5 ml pre-warmed growth medium. Centrifuge cells for 4 min at 300 × g . Resuspend the pellet in 5 ml pre-warmed cell growth medium. Plate cells at high density in a 25 cm 2 cell culture fl asks to speed up recovery. Change medium after 24 h. Propagate cells at 90–100 % confl uence. Detach cells with 500 ml StemPro Accutase Cell Dissociation Reagent, and resuspend cells in 4.5 ml cell growth medium.
2.7 Temperature and Gas Control System for Time- Lapse Live Imaging
2.8 Software for Image Acquisition and Image Processing
3.1 Thawing of the HepaRG Cell and Transduction with the Histone 2B-GFP Nuclear Marker
Francesco Pampaloni et al.
47
2. Determine the concentration of cells in the medium with a hemocytometer (e.g., Neubauer hemocytometer).
3. Expand HepaRG cells as monolayer on a 75 cm 2 tissue culture fl ask, until they reach a 70 % confl uence.
4. Add the BacMam transduction reagent with a histone 2B-GFP expression cassette to the plated cell, by pipetting the viral par-ticles directly in the media at the density recommended by the company data sheet. Baculoviral vectors for fl uorescent label-ing of the cell nuclei have a relatively low cellular toxicity [ 24 ]. After 24 h of incubation with the BacMam particles, cells are detached by employing StemPro Accutase and the cell suspen-sion is diluted to 2 × 10 3 cells in 25 μl (the volume of the mea-suring chamber mounted in the LSFM). The transduction effi ciency approaches 90 % (Fig. 2 ).
5. Perform a control experiment by forming HepaRG spheroids in a suitable U-well plate (e.g., the HydroCell Surface™ 96-well plate). The original cell suspension is diluted in growth medium, and 100 μl is transferred in each well. An 8-channel pipette can be employed to reduce the pipetting time. Incubate the multiwell plate under standard cell culture conditions. Spheroid formation is completed after 2–6 days (Fig. 3 ).
1. A caster can be made from an aluminum tube and an alumi-num plunger (Fig. 4a, b ). Employ aluminum or stainless steel for both the tube and the plunger to support the rapid cooling of the agarose gel. Use a conically shaped plunger tip to sim-plify the aggregation of the cells in the center of the beaker (Fig. 4a, b , see Note 1 ).
2. Autoclave the caster. 3. Prepare a 1 % solution of high-melting point agarose in PBS
and aliquot it in 1.5 ml tubes.
3.2 Agarose Beaker Molding
Fig. 2 HepaRG cells transduced with nuclear marker H2B-GFP. Left : transmitted light image. Right: fl uores-cence image. Objective lens: CZ Achroplan 10×/0.3. Ex/Em: 488/520 nm. Microscope Zeiss Axiovert 40 CFL
Light Sheet-Based Fluorescence Microscopy
48
Fig. 3 Aggregation of a HepaRG spheroid for 32 h. 2,000 cells/well expressing H2B-GFP were seeded into a 96-well plate coated with an agarose layer. The cells do not adhere to the agarose layer and form compact spheroids within days (aka liquid overlay method). Transmitted light and fl uorescence images are shown. Objective lens: Nikon CFI Plan Fluor 10×/0.30, WD 16 mm. Ex/Em: 488 nm/515 nm. Time-lapse interval: 30 min. Microscope: Nikon Eclipse confocal fl uorescence microscope
Fig. 4 Schematic representation of the template and the production of agarose beakers. ( a ) The template consists of a stainless steel barrel and a plunger. ( b ) The assembled template. ( c ) The liquid 1 % high-melting agarose is sucked into the template by pulling the plunger
Francesco Pampaloni et al.
49
4. Place few aliquots of the 1 % agarose solution in a heating block at 95 °C for 15 min in order to completely melt the agarose.
5. Working under the laminar fl ow sterile hood, immerse the caster in one 1.5 ml tube with the liquid agarose solution. Suck the agarose into the template by pulling up the plunger (Fig. 4c ). Place the caster at 4 °C for 5 min to allow a fast hard-ening of the agarose gel.
6. After agarose hardening, push the plunger out of the template. Separate the agarose beaker from the plunger by completely immersing it in PBS and by gently pushing the beaker out with a sharp angled tip precision forceps (Fig. 5 ).
7. Store the agarose beakers at 4 °C in PBS with 1 % penicillin/streptomycin.
1. Work under the laminar fl ow sterile hood. 2. With sharp angled tips forceps, place one agarose beaker verti-
cally inside a petri dish. Place the beaker in contact with the wall of the petri dish in order to have a stable support during the cell seeding (Fig. 5d ).
3.3 Cell Seeding in the Agarose Beaker
Fig. 5 The photographs illustrate the extraction procedure of the agarose beaker from the stainless-steel template. ( a ) The agarose beaker is still on the template. ( b ) The agarose beaker is immersed into sterile PBS and gently pushed out of the template with a sharp-tip angled forceps. ( c ) The agarose beaker is carefully manipulated with the forceps, and ( d ) the beaker is placed vertically in an empty petri dish in order to fi ll it with the cell suspension
Light Sheet-Based Fluorescence Microscopy
50
3. Remove any residual storage PBS from within the beaker with a 1 ml syringe connected to a 0.55 × 25 mm needle (24G × 1″).
4. Slowly inject in the beaker the cell suspension by employing a 1 ml syringe with a 0.55 × 25 mm needle (24G × 1″). Avoid the formation of air bubbles in the cell suspension during injection ( see Note 3 ).
5. Insert the beaker vertically into a 1.5 ml tube, and centrifuge it at 300 × g for 5 min. The centrifugation allows for a rapid seed-ing of the cells at the beaker’s bottom.
1. Most of the LSFM chambers have a large volume of about 10 ml. Thus, a perfusion system is necessary in order to main-tain a constant pH during long-term time-lapse experiments.
2. Provide the LSFM perfusion chamber with an inlet hole at the bottom of the chamber and with an outlet hole at the top of the chamber (Fig. 6 , inset).
3. Connect mini tubing olive connectors to the inlet and the outlet.
4. Autoclave the chamber.
1. Gas exchange tubing : wrap the 4 m-long gas-permeable thin silicon tube around a stable aluminum cylinder forming a tight spiral (Figs. 6 and 7 ). The long and thin silicon tube allows fast equilibration of the fl owing media with the outside gas envi-ronment. By placing the silicon spiral in a CO 2 incubator set to 5 % CO 2 , the pH of the cell culture media can be stably main-tained at 7.4.
3.4 Preparation of the LSFM Perfusion Chamber
3.5 Assembly of the Inlet and Outlet Perfusion Tubing
Fig. 6 Schematic representation of the assembled cell culture media perfusion system. Inset : a specimen chamber equipped with inlet and outlet connectors
Francesco Pampaloni et al.
51
2. Inlet tubing : connect with Luer-Lok connectors both ends of the silicon tube spiral to about 1 m-long gas-impermeable tube. The gas-impermeable tubing ensures that no gas exchange between the perfusion media and the environment outside the CO 2 incubator occurs. Connect one end to the two-stop tubing for the peristaltic pump, as well as the exten-sion tubing that will be connected with the perfusion chamber. Use quick- disconnect valve coupling nipples to connect the tube with the perfusion chamber. Connect the tube at the other end with the growth media supply bottle (Fig. 6 , see Note 2 ).
3. Outlet tubing : connect with Luer-Lok connectors one approxi-mately 1 m-long gas-impermeable tube to the two-stop tubing for the peristaltic pump. Connect the extension tubing that will be linked with the perfusion chamber. Use a quick-discon-nect valve coupling nipple to connect the tube with the perfu-sion chamber (Fig. 6 ).
4. Autoclave both the inlet and the outlet tubing.
1. Prepare a custom-made LSFM holder for the agarose beaker similar to the one shown in Fig. 7 . The bottom part of the holder must fi t the agarose beaker. The material of the holder must be autoclavable (e.g., POM, see Note 4 ). Notice the sharp-tip pin used to stably anchor the beaker to the holder (Fig. 7a, b ). The upper part of the holder must be designed to
3.6 LSFM Holder for the Agarose Beaker
Fig. 7 Mounting the agarose beaker that contains the seeded cells with the specimen holder. ( a , b ) The bottom part of the specimen holder is fi rst immersed in sterile media or PBS. The capillary force facilitates the inser-tion of the agarose beaker in the holder. A sharp-tip pin prevents it from dropping down. ( c ) The photograph shows the specimen inside the LSFM chamber
Light Sheet-Based Fluorescence Microscopy
52
ensure a stable connection with the LSFM xyz -stage once inserted in the LSFM chamber (Fig. 7c ).
2. In the laminar fl ow sterile hood, vertically position the agarose beaker containing the cell suspension in one petri dish (see e.g., Fig. 5d ).
3. Wet the bottom part of the beaker holder with sterile PBS or medium. This facilitates the insertion of the beaker into the holder.
4. Insert the agarose beaker into the holder. The easiest proce-dure is keeping the beaker vertically in a petri dish and approaching the holder from the top. The wet bottom part of the holder will “suck” the agarose beaker facilitating the connection.
5. Insert with forceps the pin laterally through the holes of the holder, and punch the agarose beaker so that it will not drop down during the experiment.
1. A commercially available tube heating system is a simple solution to control the temperature in the LSFM perfusion chamber.
2. Insert the inlet tube into the tube heating system as shown in Fig. 8 .
3. Connect the inlet and outlet tubing to the LSFM perfusion chamber as shown in Subheading 3.4 and Fig. 6 .
4. The fi nal medium temperature at equilibrium depends also on the fl ow speed. Perform a test in order to fi nd the optimum fl ow speed in order to reach a medium temperature of 37 °C.
1. Sterilize the objective lens of the LSFM with ethanol. 2. Connect the autoclaved LSFM chamber to the objective lens. 3. Pipette at the bottom of the empty chamber 1 ml of sterile sili-
con oil. A thin layer of silicon oil prevents bacterial contamina-tion by effectively isolating the media in the chamber from the outside environment.
4. Cover the chamber with a petri dish lid in order to maintain the inside sterile.
5. Place the gas exchange silicon spiral in the CO 2 incubator (Fig. 9 ). Both ends of the spirals are provided with Luer-Lok connectors and go outside the incubator through the standard cable aperture (visible in Fig. 9 , inset).
6. Assemble the inlet and outlet tubing system in a sterile hood as illustrated in Subheading 3.5 and in Fig. 6 .
7. Put the growth media supply bottle on ice, in order to prevent bacterial contamination of the media over several days.
8. Insert the fi nal part of the inlet tubing in the heating system as illustrated in Subheading 3.7 and in Fig. 8 .
3.7 Assembly and Testing of the Temperature-Controlled Heating System
3.8 Assembly of the Whole System and Start of the Time Lapse
Francesco Pampaloni et al.
53
9. Connect the inlet and outlet tubing to the LSFM chamber by employing a quick-disconnect valve coupling nipple, and start both the peristaltic pump and the heating system in order to fi ll up the chamber with growth media at 37 °C.
10. Once the chamber is fi lled up and the media temperature is stable, insert the mounted agarose beaker prepared as illus-trated in Subheading 3.6 and Fig. 7 .
11. Start the time lapse.
We observed the formation of a HepaRG spheroid from 2,000 seeded cells with a Carl Zeiss W N-Achroplan 10×/0.3 objective lens. Figure 10 shows the entire process in both transmitted light (Fig. 10a ) and fl uorescence contrast (shown as a three-dimensional isosurface rendering, Fig. 10b ). We recorded in total 600 three- dimensional stacks with two channels (transmitted light and fl uorescence) with an interval of 15 min between two stacks.
3.9 Results
Fig. 8 Assembly of the temperature-controlled heating system. The fi nal part of the inlet tube is inserted through the heating system, which warms the perfused medium
Fig. 9 Photograph of the gas exchange spiral placed inside a small CO 2 incubator ( a ) ensuring a constant pH 7.4 of the perfusion medium. A 5 % CO 2 concentration is set in the incubator. ( b , c ) The silicon tube spiral wrapped around an aluminum cylinder
Light Sheet-Based Fluorescence Microscopy
54
Each three-dimensional stack contains 135 single frames spaced 2.6 μm along the z -axis (81,000 frames within 150 h). The pro-gressive aggregation and compaction of the HepaRG cells pro-ceeds for approximately 60 h. From 60 h onwards, spheroid compaction occurs, as inferred from the volume reduction of the spheroid shown in Fig. 10b . We counted the mitotic events during the spheroid formation. The three-dimensional stacks were visually inspected plane by plane in their entirety and over time with ImageJ
Fig. 10 Formation of a HepaRG spheroid during a period of 6 days. 2,000 cells were seeded into the 1 % agarose beaker. ( a ) The transmitted light frames were recorded during the fl uorescence time-lapse imaging with the TC-LSFM. ( b ) Three-dimensional surface rendering of the corresponding fl uorescence stacks. The fl uorescence fusion protein employed for nuclei visualization was H2B-GFP. The rendering of the fl uorescence data was performed with the ImageJ plugin “3D view.” Scale bar 100 μm. For both transmitted light and three- dimensional fl uorescence imaging the same objective lens, Zeiss W N-Achroplan 10×/0.3 was employed
Francesco Pampaloni et al.
55
and its Bio-importer plug-in. Figure 11 provides representative examples of the detected mitotic events. Due to the excellent signal- to-noise ratio of the LSFM stacks over the whole recording period, this procedure can be automated. In total, we recorded 90 cell divisions, which is suffi cient to provide a representative picture of the cell divisions profi le in the spheroid.
The results of the analysis are plotted in Fig. 12 . The data shows that the majority of the mitotic events occur prior to the onset of the compaction phase, which takes place 60 h after seed-ing the cells. The cell divisions decrease during the compaction phase of the spheroid. Mitosis occurs primarily at the periphery of
Fig. 11 Representative mitosis events recorded during the HepaRG spheroid formation. Each pair of frames represents a mitotic event, showing the nucleus before and after mitosis at a time interval of 15 min. The cells express the fusion protein H2B-GFP. Objective lens: CZ N-Achroplan 10×/0.3
Fig. 12 Analysis of mitotic events during the formation of the HepaRG cells spheroid. The temporal axis is color coded according to the color bar. ( a ) Maximum projection of the mitotic events on the xz plane. As shown in the diagram, the majority of the mitotic events occur during the fi rst 50 h. The blue and red dashed circles represent the approximate shape and position of the aggregating cells during the initial phase of the spheroid formation ( blue ) and in the fi nally formed spheroid. Please note the majority of the mitotic events that occur in the outer cell layer of the forming spheroids. ( b ) Maximum projection on the yz plane. c ) Complete xyzt repre-sentation of mitotic events during the spheroid’s formation
Light Sheet-Based Fluorescence Microscopy
56
the spheroid rather than in the core, as apparent from the three- dimensional data plotted in Fig. 10a .
These results show that an LSFM equipped with a perfusion system can be used to monitor the complete process of spheroid formation in three dimensions for a period of at least 6 days.
4 Notes
1. The caster tip can have other shapes than a conical one. Possible shapes are fl at or truncated cone. In order to have a U-shaped bottom, just produce an agarose open tube with the caster, and then close one end with a droplet of agarose. The surface tension will induce a U-shaped bottom.
2. It is convenient to pre-fi ll the inlet tubing with medium before connecting it with the LSFM perfusion chamber. This avoids air bubbling and the formation of foam in the chamber once the peristaltic pump is started.
3. The fi lling of the agarose beaker with the cell suspension can be alternatively conducted with thin glass capillaries connected to a normal plastic hypodermic syringe with a 20 cm-long fl ex-ible silicon tube. This allows a very precise control over the fl ow. In this way, the cell suspension can be dispensed without the formation of air bubbles.
4. The LSFM holder for the agarose beakers as well as the perfu-sion chamber should be made of an autoclavable plastic mate-rial, e.g., POM. Metals should in general be avoided due to potential cellular toxicity.
Acknowledgments
The authors thank the Deutsche Forschungsgemeinschaft (DFG) and the German Federal Ministry of Education and Research (BMBF) for fi nancial support (project ProMEBS). We thank Berit Langer for her outstanding support.
References
1. Baker M (2010) Cellular imaging: taking a long, hard look. Nature 466:1137–1140
2. Homem CCF, Reichardt I, Berger C et al (2013) Long-term live cell imaging and auto-mated 4D analysis of drosophila neuroblast lin-eages. PloS One 8(11):e79588. doi: 10.1371/journal.pone.0079588
3. Carlton PM, Boulanger J, Kervrann C et al (2010) Fast live simultaneous multiwavelength
four-dimensional optical microscopy. Proc Natl Acad Sci U S A 107:16016–16022
4. Huisken J, Stainier DY (2009) Selective plane illumination microscopy techniques in devel-opmental biology. Development 136:1963–1975
5. Huisken J, Swoger J, Del Bene F et al (2004) Optical sectioning deep inside live embryos by selective plane illumination microscopy. Science 305:1007–1009
Francesco Pampaloni et al.
57
6. Keller PJ, Schmidt AD, Wittbrodt J et al (2008) Reconstruction of zebrafi sh early embryonic development by scanned light sheet microscopy. Science 322:1065–1069
7. Keller PJ, Stelzer EH (2008) Quantitative in vivo imaging of entire embryos with digital scanned laser light sheet fl uorescence micros-copy. Curr Opin Neurobiol 18:624–632
8. Maizel A, von Wangenheim D, Federici F et al (2011) High-resolution live imaging of plant growth in near physiological bright conditions using light sheet fl uorescence microscopy. Plant J 68:377–385
9. Keller PJ, Schmidt AD, Santella A et al (2010) Fast, high-contrast imaging of animal devel-opment with scanned light sheet-based structured- illumination microscopy. Nat Methods 7:637–642
10. Pampaloni F, Kroschewski R, Berge U et al (2014) Tissue-Culture Light Sheet Fluorescence Microscopy (TC-LSFM) allows long-term imaging of three-dimensional cell cultures under controlled conditions. Integrative Biology. doi: 10.1039/C4IB00121D
11. Harma V, Virtanen J, Makela R et al (2010) A comprehensive panel of three-dimensional mod-els for studies of prostate cancer growth, invasion and drug responses. PLoS One 5:e10431
12. Friedrich J, Seidel C, Ebner R et al (2009) Spheroid-based drug screen: considerations and practical approach. Nat Protoc 4:309–324
13. Matsuda M, Shiratori S (2011) Correlation of antithrombogenicity and heat treatment for layer-by-layer self-assembled polyelectrolyte fi lms. Langmuir 27:4271–4277
14. Takano S, Tian W, Matsuda M et al (2011) Detection of IDH1 mutation in human glio-mas: comparison of immunohistochemistry and sequencing. Brain Tumor Pathol 28:115–123
15. Lee MY, Kumar RA, Sukumaran SM et al (2008) Three-dimensional cellular microarray
for high-throughput toxicology assays. Proc Natl Acad Sci U S A 105:59–63
16. Goto A, Hoshino M, Matsuda M et al (2011) Phosphorylation of STEF/Tiam2 by protein kinase A is critical for Rac1 activation and neurite outgrowth in dibutyryl cAMP-treated PC12D cells. Mol Biol Cell 22:1780–1790
17. Kelm JM, Timmins NE, Brown CJ et al (2003) Method for generation of homogeneous multi-cellular tumor spheroids applicable to a wide variety of cell types. Biotechnol Bioeng 83:173–180
18. Lin RZ, Chou LF, Chien CC et al (2006) Dynamic analysis of hepatoma spheroid forma-tion: roles of E-cadherin and beta1-integrin. Cell Tissue Res 324:411–422
19. Verveer PJ, Swoger J, Pampaloni F et al (2007) High-resolution three-dimensional imaging of large specimens with light sheet-based micros-copy. Nat Methods 4:311–313
20. Lorenzo C, Frongia C, Jorand R et al (2011) Live cell division dynamics monitoring in 3D large spheroid tumor models using light sheet microscopy. Cell Div 6:22
21. Guillouzo A, Corlu A, Aninat C et al (2007) The human hepatoma HepaRG cells: a highly differentiated model for studies of liver metab-olism and toxicity of xenobiotics. Chem Biol Interact 168:66–73
22. Greger K, Swoger J, Stelzer EH (2007) Basic building units and properties of a fl uorescence single plane illumination microscope. Rev Sci Instrum 78:023705
23. Huisken J, Stainier DY (2007) Even fl uores-cence excitation by multidirectional selective plane illumination microscopy (mSPIM). Opt Lett 32:2608–2610
24. Zeng J, Du J, Lin J et al (2009) High-effi ciency transient transduction of human embryonic stem cell-derived neurons with baculoviral vec-tors. Mol Ther 17:1585–1593
Light Sheet-Based Fluorescence Microscopy
59
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology, vol. 1251, DOI 10.1007/978-1-4939-2080-8_4, © Springer Science+Business Media New York 2015
Chapter 4
Fluorescence Microscopy-Based RNA Interference Screening
Manuel Gunkel , Nina Beil , Jürgen Beneke , Jürgen Reymann , and Holger Erfl e
Abstract
Using RNAi interference (RNAi), it is possible to study the effect of specifi c gene knockdowns in mam-malian cells. In this protocol we present the automated preparation of “ready to transfect” multiwell plates and cell arrays, on which cells can be grown which are then reversely transfected with one type of siRNA in every individual well or spot. Additionally, different microscope types for screening approaches are com-pared and considerations about the information workfl ow are made.
Key words High-throughput screening , High content screening , Automated sample preparation , Automated microscopy , Data mining
1 Introduction
Microscopy-based RNA interference (RNAi) screening has become a powerful tool to elucidate gene function in mammalian cells in a previously inconceivable manner [ 1 ]. In this chapter we describe the workfl ow of the ViroQuant-CellNetworks RNAi screening facility in BioQuant, Heidelberg University, from automated sample prepara-tion to screening microscopy [ 2 ], supported by sophisticated data mining tools [ 3 , 4 ].
The ideal specifi c gene knockdown experiment by RNAi allows addressing controlled gene silencing independent of any stochastic gene expression and cell cycle effects. This generates an unambigu-ous cellular phenotype for the assay of interest. Unfortunately, this does not refl ect the reality of RNAi experimentation, especially not in high-throughput RNAi workfl ows. Uncontrolled knockdown effi ciencies in combination with stochastic gene expression and cell cycle effects, just to mention a few, cause the occurrence of popula-tions of phenotypes after gene knockdown [ 5 ]. To increase the information content of RNAi knockdown experiments, single-cell
60
information should be collected and considered for evaluation [ 6 ]. Besides that, many biological assays require an image resolution suffi cient to resolve subcellular structures or to obtain information on physical interactions between cellular components. Many RNAi screens taking cellular population into consideration have been published, and the need for microscopy-based image acquisition is steadily growing.
As different processes of interest within a biological system occur at defi ned points in space and time, different microscopic observation and investigation methods have to be considered. Nowadays, advanced and fully automated data acquisition (DAQ) techniques are needed for performing large-scale screening assays. The techniques are optimized in terms of throughput (number of acquired samples per time) or information content, e.g., resolution and live cell imaging ( see Table 2 for a small overview of screening microscopes).
Besides assay preparation and imaging, a central aspect is data mining, i.e., information workfl ows that not only provide analysis tools but also contain assay-specifi c information for a proper docu-mentation and interpretation of the analysis results itself. Ideally, such an information pipeline covers the entire screening workfl ow containing library information, plate layouts resulting from assay preparation and dedicated image, statistics and bioinformatics results, as well as the visualization of the data.
Here, we give an introduction to high-throughput sample preparation for microscopy-based RNAi screens and considerations for the proper choice of screening microscopes for a certain resolution demand.
2 Materials
1. siRNA oligonucleotides (Ambion). 2. OptiMEM + Glutamax (Gibco). 3. Sucrose (USB). 4. Gelatin (Sigma-Aldrich). 5. Lipofectamin 2000 (Invitrogen). 6. Fibronectin, human (Sigma-Aldrich). 7. 384-Well low-volume plates (Nunc). 8. 384 Deep well plates (Eppendorf). 9. 384-Well image plates (BD Falcon). 10. Lab-Tek chambered cover glass (Nunc). 11. Sterile fi lter, 0.45 μm (Millipore).
Manuel Gunkel et al.
61
3 Methods
For RNAi screening, “ready to use” solid-phase reverse transfec-tion can be prepared in 96- and 384-well plates as well as on cell arrays [ 7 ]. The right choice depends on many factors ( see Table 1 ).
1. Prepare an siRNA stock solution by dissolving lyophilized siR-NAs with MilliQ water to a fi nal concentration of 30 μM.
2. Transfer 3 μl of OptiMEM, containing 0.4 M sucrose, to each well of a 384-well low-volume plate.
3. Add 3.5 μl of Lipofectamin 2000 to each well of the 384-well low-volume plate and mix thoroughly.
4. Add 5 μl of the respective siRNA stock solution to each well of the 384-well low-volume plate and mix thoroughly.
5. Incubate for 20 min at room temperature. 6. Add 7.25 μl of a 0.2 % gelatin solution containing 1 × 10 − ² %
fi bronectin to each well of the 384-well low-volume plate and mix thoroughly.
See also Note 1 .
1. Option 1: Preparation of 96-well plates “ready to transfect”: Dilute 18.75 μl of the source siRNA transfection solution with
468.75 μl MilliQ water and dispense 25 μl per well of a 96-well plate.
Option 2: Preparation of 384-well plates “ready to transfect”: Dilute 18.75 μl of the source siRNA transfection solution with
187.5 μl MilliQ water and dispense 5 μl per well of a 384-well plate.
Option 3: Preparation of high-density cell arrays “ready to transfect”:
3.1 Preparation of the Source siRNA Transfection Solution for Solid-Phase Transfection
3.2 Preparation of “Ready to Transfect” Plates
Table 1
Advantages and drawbacks of different plate types
Plate type Advantage Drawback Proposed application
96-Well plate
Large number of cells per well, area ~30 mm 2
Low number of wells
Validation screen with preselected low number of targets
384- Well plate
Medium number of wells – Genome-wide screen
Cell arrays Full assay on a single plate. No need for plate handling. Cell seeding and antibody staining simultaneously on whole plate
Relatively small area (~0.13 mm) per spot, ~50–300 cells
Genome-wide screen with expensive reagents or rare patient cells
Fluorescence Microscopy-Based RNA Interference Screening
62
The siRNA transfection solution is not further diluted and printed directly onto the cell array by a full contact printer ( see Note 2 ). No further steps are needed!
2. Dry the 384-well plates for 2.5 h at medium drying force in the Speed Vac ( see Note 3 ).
Various types of microscopes can be used for screening approaches. Crucial factors for the choice of the right system are the imaging speed and the quality of the readout. First we want to give a com-parison between three different fl uorescence-based screening sys-tems, a wide-fi eld (Olympus, IX81), a spinning disc confocal (PerkinElmer, Opera), and a confocal (Leica, SP5) microscope. On each system, the same screen was performed. A 96-well plate (ibidi) was used on which a layout with 12 wells ( see Fig. 1 ) was repeated 3 times, one repetition for each microscope screen. In each well, multiple subpositions were acquired until the total recorded area per well covered about 1,240 × 930 μm 2 . At each subposition a z-stack with 24 steps (z-step width 2 μm) was recorded for three different color channels. The imaging parameters of the different microscopes are compared in Table 2 .
3.3 Comparison of Different Screening Microscopes
Fig. 1 Layout of the plate and readout example. Four wells are used as controls where no siRNA was applied ( dark gray , top and bottom row ), in four wells a non- silencing siRNA was applied ( light gray , left column ), and in four wells siRNA for Kif11 knockdown was transfected ( red , right column ), which resulted in pro-metaphase arrest of the cells. The relative number of cells in prometaphase is counted and the relative amount compared to all imaged cells is determined
Manuel Gunkel et al.
63
Table 2
Imaging parameter for three different microscope systems
Imaging parameter Wide fi eld (Olympus IX81)
Spinning disc confocal (PerkinElmer Opera)
Confocal (Leica TCS SP5)
Objective 20× air, NA = 0.75 20× water, NA = 0.7 40× oil, NA = 1.25
Pixel size 322.5 nm 330 nm 331 nm
Z-step 2 μm 2 μm 2 μm
Number of planes 24 24 24
Field of view 433 × 330 μm 2 439 × 330 μm 2 338 × 338 μm 2
Number of fi elds 3 × 3 3 × 3 4 × 3
Acquired area 1,233 × 930 μm 2 1,239 × 930 μm 2 1,238 × 938 μm 2
Channel acquisition Sequential Sequential Sequential
Acquisition time per frame (RGB, in ms)
300, 100, 100 1,000, 40, 40 293, 293, 293
Total acquisition time 51 min 59 min 8 h 13 min
Possible readouts can be manifold; some examples would be the siRNA knockdown-induced up-/downregulation of a certain protein which can be measured via fl uorescence labeling and inten-sity analyses, morphological changes of fl uorescently labeled cellu-lar compartments (phenotype detection), or increase/decrease of cellular density. In Fig. 1 , a phenotype readout example for the sample screen is shown. Here the phenotypes were determined by a workfl ow setup using the workfl ow and data mining manager KNIME ( www.knime.org ) with additional image processing (KNIP, tech.knime.org/community/image-processing) and WEKA [ 8 ] plug-ins, which performed three steps: image segmentation for cell detection, feature extraction for the individual cells, and feature clustering by an x-means algorithm [ 9 ]. Further details concerning data analysis using KNIME can be found on our webpage ( http://www.bioquant.uni-heidelberg.de/technology- platforms/viro-quant-cellnetworks-rnai-screening-facility/data-analysis.html ).
Further possible imaging methods spanning low-resolution (high-throughput) to high-resolution (low-throughput) screening are listed in Table 3 . The numbers (based on a reference sample with 384 imaging positions) refl ect only a small subset of possible screening applications. The choice of the screening microscope to be used depends strongly on the assay and the information content which is needed. For instance, in the case that cell-based intensity expression profi les have to be analyzed, very fast 2D wide-fi eld
Fluorescence Microscopy-Based RNA Interference Screening
64
imaging with low resolution is suffi cient. In contrast, if 3D (structural) information of small objects is targeted, confocal imag-ing is the method of choice.
A clear information workfl ow is mandatory in microscopy-based screening approaches. Here we will describe a typical information workfl ow for RNAi screening.
On the one hand, a typical data mining workfl ow contains assay-specifi c metadata like siRNA library and plate layouts includ-ing dedicated transformation protocols which are needed in order to translate the siRNA library plate layout (as given by the manu-facturer) to the assay-specifi c plate layout. On the other hand, the information workfl ow contains data processing tools and provides analysis results that are based on the plate layout as integrated into the data pipeline upstream of the information workfl ow: The results from image processing are correlated with the correspond-ing siRNA approach during statistical data processing. Based on this, bioinformatics tools gain information by network and path-way reconstruction which fi nally leads to a siRNA correlated inter-pretation of the entire assay. For that purpose, advanced information workfl ows consist of different analysis tools and a dynamic data-base featuring continuous interaction with analysis results and cor-responding update of the database, respectively ( see Fig. 2 ).
4 Notes
1. The preparation can be done manually or by an automated liquid handler (e.g., “MICROLAB STAR” by Hamilton).
2. In our case either a 96 pin spotter “ChipWriter Pro” by Bio- Rad or a 384 pin spotter developed by Graffi nity Pharmaceuticals.
3. The drying time depends on the amount of plates and the number of fi lled wells. (e.g., drying 4 384 plates takes approxi-mately 20 min).
3.4 Information Workfl ow for Screening Microscopy
Table 3
Overview of different screening microscopes in terms of speed and applicability
Imaging type Data Resolution Speed Example approach
Wide fi eld, e.g., Olympus IX81 2D 10×, NA = 0.4 ~3 h Intensity profi les Phenotypic penetration
Confocal (spinning disc), e.g., PerkinElmer Opera
3D 60×, NA = 1.4 ~10 h Colocalization Structural information
Confocal (laser scanning), e.g., Leica SP5
3D 63×, NA = 1.4 ~20 h Colocalization Structural information
Manuel Gunkel et al.
65
Acknowledgment
We like to thank Michael Berthold, FB Informatik und Informationswissenschaften, Uni Konstanz, and the whole KNIP- Team for image analysis support. The ViroQuant-CellNetworks RNAi screening facility is supported by the Federal Ministry of Education and Research (BMBF) and funded by the FORSYS pro-gram ViroQuant (Project 0313923) and by CellNetworks-Cluster of Excellence (EXC81).
References
Fig. 2 Illustration of a screening microscopy information workfl ow. Ideally, the integrated database which fi nally contains the overall results of the screen is synchronized in a further step with external screening data-bases in order to search for correlations and overlaps
1. Neumann B, Walter T, Hériché J-K et al (2010) Phenotypic profi ling of the human genome by time-lapse microscopy reveals cell division genes. Nature 464:721–727
2. Liebel U, Starkuviene V, Erfl e H et al (2003) A microscope-based screening platform for large- scale functional protein analysis in intact cells. FEBS Lett 20:394–398
3. Hamecher J, Rieß T, Bertini E (2011) A versa-tile framework for the analysis of high- throughput screening data. In: Koeppl H, Aćimović J, Kesseli J, Mäki-Marttunen T, Larjo A, Yli-Harja O (eds) Eighth international work-shop on computational systems biology (WCSB). Tampere University of Technology, Tampere, pp 57–60
Fluorescence Microscopy-Based RNA Interference Screening
66
4. Carpenter AE, Jones TR, Lamprecht MR et al (2006) Cell profi ler: image analysis software for identifying and quantifying cell phenotypes. Genome Biol 7:R100
5. Knapp B, Rebhan I, Kumar A, Matula P et al (2011) Normalizing for individual cell popula-tion context in the analysis of high-content cellular screens. BMC Bioinformatics 12:485
6. Sacher R, Stergiou L, Pelkmans L (2008) Lessons from genetics: interpreting complex phenotypes in RNAi screens. Cur Opin Cell Biol 20:483–489
7. Erfl e H, Neumann B, Liebel U et al (2007) Reverse transfection on cell arrays for high content screening microscopy. Nat Protoc 2:392–399
8. Pelleg D, Moore A (2000) X-means: extending k-means with effi cient estimation of the number of clusters. In: ICML '00 proceedings of the seventeenth international conference on machine learning. Morgan Kaufmann Publishers Inc. 727–734
9. Hall M, Frank E, Holmes G et al (2009) The WEKA data mining software. ACM SIGKDD Explor Newslett 11:10
Manuel Gunkel et al.
67
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology, vol. 1251, DOI 10.1007/978-1-4939-2080-8_5, © Springer Science+Business Media New York 2015
Chapter 5
Fluorescence Resonance Energy Transfer Microscopy (FRET)
Katarzyna M. Kedziora and Kees Jalink
Abstract
FRET (Förster Resonance Energy Transfer) microscopy breaks the resolution limit of light to let us investigate the conformation and function of proteins within living cells. Intensity-based methods are the most popular and direct approach to detect FRET. Among them, detection of sensitized emission signals and ratio imaging of specially designed FRET sensors are routinely used in modern cell biology laborato-ries. In this chapter, we provide protocols for both these techniques. We guide the reader through the mathematical corrections necessary to calculate the sensitized emission image. We illustrate this approach with an example of studying the interaction of nexin (SNX1) proteins. In the ratio FRET protocol, we focus on monitoring changes in cellular concentration of cAMP with an EPAC-based FRET sensor.
Key words FRET, FilterFRET, Ratio imaging, Nexin, cAMP, EPAC
1 Introduction
Functional imaging allows us to look at protein-protein interac-tions and protein conformational changes within living cells. One of the most powerful functional imaging techniques is FRET microscopy [1, 2]. Specially designed FRET probes can spy on the activity of proteins (e.g., caspase cleavage assays [3]) or the con-centration of small second messengers within a cell—Ca2+ [4], cAMP [5], PIP2 [6], etc. FRET microscopy provides answers to an exponentially increasing number of biological questions [7], as it gives insight into details two orders of magnitude smaller (1–10 nm) than the resolution available in classical fluorescence microscopy (~250 nm).
FRET (Förster Resonance Energy Transfer) is the physical phenomenon of radiationless transfer of energy between two molecules—the energy donor and acceptor [8]. When two fluoro-phores are engaged in energy transfer, excitation of the donor molecules results in emission of fluorescence characteristic of the acceptor, rather than the donor (sensitized emission).
68
The FRET efficiency (E) is defined as the fraction of the energy absorbed by the donor molecules that is transferred to acceptor molecules. For a given pair of fluorophores, this efficiency depends on their distance but also on their orientation and environment. Changes in FRET efficiency are usually greatest when fluorophores are within a distance 1–10nm from each other (spatial resolution available for FRET). Yet distance and orientation are often impos-sible to uncouple. Therefore, FRET efficiency responds to the spa-tial distribution of fluorophores, but it is not a simple “molecular ruler” in the complex settings of a living cell.
Several techniques were developed to study FRET efficiency. Perhaps the most intuitive approaches are based on studying changes in the intensity of fluorescence emission of the donor and acceptor, including simple ratio imaging, detection of sensitized emission, and “filter FRET” [9–11]. Ratio imaging suffices in many cases to detect FRET. This is very convenient as we can draw biologically valid conclusions from this kind of FRET experiment without applying the correction procedures necessary to quantita-tively detect sensitized emission. Both of these techniques are described in the current chapter.
To understand the idea of such experiments, consider an experiment in which two molecules of interest are fused to donor and acceptor fluorophores, respectively. In this experiment, fluo-rescent signal can be collected in three different channels, defined by excitation and detection parameters optimized for donor or acceptor molecules (Fig. 1b, c):
●● d—donor excitation (λdexc) and donor emission (λd
em).●● s—donor excitation (λd
exc) and acceptor emission (λaem).
●● a—acceptor excitation (λaexc) and acceptor emission (λa
em).
The fluorophores in this thought experiment are designed to be perfect for FRET experiments (Fig. 1a). A huge overlap between the emission spectrum of the donor and the absorption spectrum of the acceptor maximizes the possible efficiency of FRET. Yet there is neither cross-excitation between the fluorophores nor leak- through of fluorescence emission of one of them into the channel optimum for its partner.
The signal of the non-interacting molecules in this experiment can be detected directly in channel d (Fig. 1b continuous line) and
Fig. 1 (continued) (c, perfect FRETpair; f, real-world fluorophores). (g) Table summarizing signals present in the channels of a FRET experiment. Note that in case of a perfect FRET pair only diagonal elements are nonzero. Non-diagonal signals are present only in case of leak-through (signals marked with *), cross-excitation (signals marked with **), or both (signals marked with ***). Donor molecules are essentially not cross-excited at acceptor excitation wavelength (IaD − S and IaS signals are grayed out). Equations to the right of the table introduce parameters linking the signals, originating from the same population of molecules but detected in different channels
Katarzyna M. Kedziora and Kees Jalink
69
b
c
da
e
f
g�������������
populationsof molecules
detectionchannels d s a
ID−S IdD−S IsD−S
∗ IaD−S∗∗∗ −→ IsD−S = βIdD−S
IS IdS∗ Is
S IaS ∗∗ −→ IdS = δIsS
IA IdA∗∗∗ IsA ∗∗ Ia
A −→ IdA = αIaA IsA = γIaA
Fig. 1 Signals detected by ratio imaging and FilterFRET. (a–c) Spectra of a perfect FRET pair, i.e., there is no leak-through and cross-excitation between the fluorophores. (d–f) Spectra of real-world fluorophores—the CFP and YFP pair. Note that there is considerable leak-through of the emission of each fluorophore into the detection range of its partner. Acceptor molecules can be cross-excited at the donor excitation wavelength, but the excitation of donor molecules with an acceptor excitation wavelength is always negligible. (a, d) Excitation and emission spectra of a FRET pair with optimized excitation wavelengths (λd
exc and λaexc) and detection ranges
(λdem and λa
em) indicated (a, perfect FRET pair; b, real-world fluorophores). (b, e) Signal detected in d and s channels with FRET (dotted line) or without (continuous line). Note that signals in d and s channels are detected after exciting with donor excitation wavelength (λd
exc), (b, perfect FRET pair; e, real-world fluorophores). (c, f) Signal detected in channel a (with an excitation optimized for acceptor molecules—λa
exc) is not affected by FRET
FRET Microscopy
70
channel a (Fig. 1c). If molecules of interest can interact with each other, we expect FRET between their fluorophores. The signal in channel d is diminished (ID − S) as now the donors that are in a com-plex transfer part of their absorbed energy to the acceptor mole-cule, rather than emitting it directly as photons. Signal of acceptors excited by FRET (sensitized emission, IS) appears in channel s. The signal detected in channel a does not change as the presence of FRET does not interfere with the signal of (directly) excited acceptor molecules—IA (Fig. 1c). Thus, we have introduced three detection channels as well as three populations of molecules. This setup is summarized in the table (Fig. 1g). In case of a perfect FRET pair, only diagonal elements of the table are nonzero—i.e., the signal from different populations of molecules can be detected only in the channels optimized for them. It is the signal of sensi-tized emission detected in channel s that provides the information on the presence and magnitude of sensitized emission and as a consequence about the presence of interaction between the studied molecules.
Unfortunately, in the experiments performed with real-world fluorophores, cross-excitation and leak-through usually complicate matters (Fig. 1d). The signal in channel s is no longer a measure-ment of pure sensitized emission. With (Fig. 1e dotted line) or with-out FRET (Fig. 1e continuous line), it always contains a leak- through signal from donor molecules and a signal from cross- excited accep-tor molecules. In this situation, some signal from any of the three populations of molecules can be detected in all three detection chan-nels (Fig. 1g). For example, Hence the signal in channel d (first column in the table Fig. 1g) is the sum of the signal from excited non-fretting donor molecules (ID − S
d), signal of sensitized emission leaking back into the donor channel (ISd), and signal of cross- excited acceptor molecules also leaking back into the donor channel (IAd).
In most experiments, cross-excitation of donor molecules is very small and can be neglected. Hence the signal in channel a originates only from the population of acceptor molecules—IAa (ID − S
a and ISa appear gray in the table in Fig. 1g). Nevertheless, the measurements obtained from the channels d, s, and a do not suf-fice to calculate the value of the sensitized emission signal (ISs) because we have to determine seven different intensities (non-gray in the table) from just three measurements. Note however that in any specific row in the table (Fig. 1g), all the intensities originate from a single population of molecules. Therefore, a set of param-eters (α, β, γ, and δ) that relate them to each other can be intro-duced (see equations to the right of the table in Fig. 1g). These parameters depend solely on the specific settings of the microscope (filters, spectral sensitivity, etc.) and on the spectral characteristics of donor (β) or acceptor (α, γ, and δ) molecules. Therefore, they can be calculated based on measurements of pure populations of
Katarzyna M. Kedziora and Kees Jalink
71
donors or acceptors. For example, β just represents the fraction of leak- through of donors in the s channel.
On wide-field fluorescence microscopes, these factors are fixed for a given fluorophore. However, the parameters of the excitation and detection pathways in a confocal microscope can be tuned independently. Moreover, the intrinsic instability of laser intensities may easily amount to 5 %, even in modern instruments, making it often impossible to recreate exactly identical settings between the experiments. As a consequence, it is best to collect the data on the magnitude of the correction factors at the same time as performing the FRET experiment.
Based on the measurement of three channels together with four introduced parameters, it is finally possible to solve the system of seven equations to calculate the value of sensitized emission, ISs:
ISs =
- - -( )-
s d ab g abbd1
It is crucial to understand what kind of information is provided by the sensitized emission signal. By definition, this signal is a mea-surement of the amount of energy being transferred. It follows the spatial distribution of interactions between proteins or their activ-ity and depends on the number of fluorophores available to inter-act with each other. Note that because sensitized emission is sensitive to the parameters of both excitation and emission, it is not possible to quantitatively compare two experiments as long as they are not acquired in exactly the same conditions.
FRET efficiency, on the other hand, is fully quantitative and can thus be compared between different experiments. Whether it is necessary to calculate FRET efficiency depends on the nature of the biological question that underlies the experiments. To calculate FRET efficiency, the measured sensitized emission has to be nor-malized, e.g., by dividing it by the intensity of the (unquenched) donor signal (Ed), signal of acceptors (Ea), or a combination of the two of them (see Note 1). Such normalizations are never straight-forward. For example, to determine Ed, channels d and s are used that share the same excitation parameters but differ in sensitivity to donors and acceptors, respectively. On the other hand, the chan-nels a and s that are needed to calculate Ea may have the same detections settings but do not share the same excitation intensities. Thus, quantitative determination of FRET efficiency requires an additional calibration, the details of which are beyond the scope of this chapter but may be found in literature [11].
The first protocol in the method section of this chapter describes a confocal experiment designed to study the interactions between nexin proteins (SNX1) by measuring FRET. Overexpression of a fusion construct (SNX with fluorescent proteins) enables the detection of signal mainly from tubulovesicular endosomal
FRET Microscopy
72
membranes and from the cytoplasm [12]. Therefore, confocal microscopy was chosen to visualize the fine tubular structures by optical sectioning. In this experiment, we make use of reference cells, expressing only one fluorophore, to calculate the values of the correction parameters.
The second protocol in the method section shows how to measure the intracellular concentration of cAMP with an EPAC- based sensor. For this single-polypeptide FRET sensor, ratio imaging is the method of choice. Ratio imaging, i.e., simply recording the ratio of the d and s images, is the easiest and most popular technique of FRET microscopy. The ratio image is insensi-tive to fluctuations in excitation light intensity and it normalizes signal differences caused by cell morphology (e.g., differences in cell height) and in expression level. If the stoichiometry of donor and acceptor fluorophores is fixed, quantitative data can be obtained simply by performing an end-point calibration. This can, for example, be the addition of forskolin together with IBMX which drives cAMP levels so high that they saturate the response of the FRET sensor.
2 Materials
1. Cell culture media.
(a) HEK293 (Embryonal Kidney Cells, American Type Culture Collection), N1E-115 (mouse neuroblastoma), and HeLa (cervical cancer) were cultured in DMEM (Gibco) sup-plemented with 8 % FCS (Sigma) and 1 % antibiotics (penicil-lin and streptomycin), in a 5 % CO2, 37 °C, humidified incubator.
(b) Cells were imaged in DMEM-F12 medium (Gibco). (c) Transfection reagent—PEI (Polyethylenimine,
Polysciences) was dissolved in ethanol to the concentration of 1.5 mg/ml and stored in a glass container in −20 °C.
2. Agonists were added from 1,000-fold concentrated stocks:
(a) Adrenalin (Sigma). (b) Forskolin (Santa Cruz Biotechnology). (c) IBMX (Sigma).
3. Coverslips (Thermo Scientific, 24 mm, #1) were sterilized in 70 % ethanol. Each coverslip was put in a separate well of a 6-well plate and left in a laminar flow to dry.
4. Plasmids.
(a) SNX1-CFP and SNX1-YFP constructs were a kind gift of Professor Peter Cullen, University of Bristol, England.
(b) TEPACVV construct was described previously [13].
2.1 Cell Culture
Katarzyna M. Kedziora and Kees Jalink
73
1. A confocal microscope capable of exciting CFP (e.g., with a 442 nm HeCd laser line or a 458 nm Argon laser line) and YFP (e.g., a 514 nm Argon laser line). Readout of the signal should be simultaneous with two PMT detectors in the ranges 470–510 and 520–600 nm (slight deviations from these ranges work too). A high-NA immersion objective chromatically cor-rected for CFP/YFP detection should be used.
2. Image analysis software that can calculate images represented with floating point numbers (e.g., the freeware package ImageJ [14] or Fiji [15]).
1. Any inverted fluorescent microscope that can excite CFP (in a range 430–470 nm) and collect donor (470–510 nm) and acceptor (520–600 nm) emission. If a wide-field system is used, it would be beneficial to use an image-splitting device to acquire the images of donor and acceptor simultaneously.
2. Image acquisition software capable of calculating and plotting the ratio between CFP and YFP emission in a selected region of interest (ROI) during an experiment.
The imaging systems were equipped with a thermostated chamber for live cell imaging (37 °C) with controlled atmosphere of 5 % CO2 and elevated humidity. If not available, a HEPES-buffered system should be used.
3 Methods
Below we provide two protocols for preparing and performing intensity-based FRET experiments. The first protocol gives detailed information about the use of sensitized emission (FilterFRET) to study interactions of proteins. The second one deals with ratio FRET imaging of dynamic changes of small molecules inside the cells.
1. Select a pair of fluorescent proteins that constitute a good FRET pair. Their spectral properties have to be compatible with the sample to be studied and the image acquisition system (see Notes 2 and 3).
To perform the example experiment, the classical FRET pair CFP (Cyan Fluorescent Protein) and YFP (Yellow Fluorescent Protein) has been chosen.
2. Prepare fusion constructs of the proteins of interest with fluorescent proteins. Test extensively whether tagging with fluorescent protein may affect the localization or function of the proteins (see Note 4).
For this example two constructs SNX1-CFP and SNX1-YFP have been prepared.
2.2 Microscopy and Image Analysis
2.2.1 Sensitized Emission
2.2.2 Ratiometric FRET
2.2.3 Temperature Control
3.1 Sensitized Emission/FilterFRET
3.1.1 Planning of the Experiment
FRET Microscopy
74
3. Prepare reference cells, i.e., cells that express either donor or acceptor fluorophores alone. Make sure that their signal matches the signal from the cells of interest (see Note 5). It is convenient to choose the reference constructs so that the ref-erence cells can be easily discriminated from the FRET cells (see Note 6).
Reference cells used to perform the example experiment were HEK cells stably expressing CFP or YFP (HEK-CFP and HEK-YFP cell lines).
4. Make sure that the imaging setup is prepared for live cell imag-ing (see Note 7).
5. Optimize the image acquisition: optimal alignment (see Note 8), minimized chromatic aberration (see Note 9), and flatness of field (see Note 10) are extremely important in a FRET experiment.
1. Day1: Plate cells of interest on glass coverslips (see Note 11). In the example, experiment N115 cells were seeded at
30,000 cells/well in 6-well plates containing sterilized 24 mm coverslips.
2. Day 2: Transfect the cells with the FRET constructs. Transfection protocol is cell-type specific. It may be necessary to refresh the cell culture medium ~12 h after transfection.
N115 cells were transfected with SNX1-CFP and SNX1-YFP constructs using PEI transfection reagent (1 μg of each DNA + 4.5 μg of PEI for each well).
3. Day 3 (morning): Seed reference cells on the coverslips with the cells of interest (see Note 12).
HEK-CFP and HEK-YFP cells growing in a culture flask were trypsinized and seeded onto the coverslips in the 6-well plate at 200,000 cells/well. Allow them minimally 2 h to attach and spread on the surface.
4. Day 3 (afternoon)/day 4: Transfer a coverslip with cells to a chamber suitable for live cell imaging (see Note 13). Check that the imaging chamber does not leak (see Note 14) and place it on the microscope. Allow the temperature to equili-brate before starting measurements to avoid focus drift.
1. Find a healthy FRET cell that expresses both constructs. The image must also contain at least one donor and one acceptor reference cell.
2. Focus and zoom to optimize the visibility of details in the FRET cell.
3. Acquire three images in channels d (donor excitation and donor emission), s (donor excitation and acceptor emission),
3.1.2 Preparation of a FilterFRET Experiment
3.1.3 FilterFRET Experiment
Katarzyna M. Kedziora and Kees Jalink
75
and a (acceptor excitation and acceptor emission)—see intro-duction. Take care that the acquired images are of high quality (see Notes 15 and 16) and make sure no motion artifacts are present (see Note 17).
Images of N115 cells expressing SNX1-CFP and SNX1-YFP constructs together with reference cells are presented in Fig. 2a–c.
If time-lapse imaging is required, repeat the acquisition of d, s, and a images with the desired frequency.
1. Calculate and subtract the value of background in the d, s, and a images. Background should be calculated as the mean value of the signal within a ROI that contains no cells (see Note 18).
2. Optionally, apply preprocessing procedures. It depends on the biological question and the quality of the acquired images what kind of procedures will be beneficial (see Note 19).
In the example experiment, d, s, and a images were subjected to smoothing using a 3 × 3 averaging filter.
3. Calculate the values of the correction parameters:
(a) Measure the values of the signal of an acceptor reference cell (HEK-YFP) in three channels d, s, and a (da, sa, and aa).
(b) Calculate the values of the α, γ, and δ correction parameters:
a =
d
aa
a
d =
d
sa
a
g =
s
aa
a
(c) Measure the values of the signal of a donor reference cell (HEK-CFP) in two channels d and s (dd, sd).
(d) Calculate the value of the ß correction parameter:
b =
s
dd
d
(e) Calculate the value of sensitized emission signal according to
ISs =
- - -( )-
s d ab g abbd1
,
as derived in the introduction.
3.1.4 Image Processing of FilterFRET Data
FRET Microscopy
76
Fig. 2 Sensitized emission and ratio FRET. (a–f) Sensitized emission. N1E-115 cells expressing SNX1-CFP and SNX1-YFP were seeded together with reference cells (HEK-CFP and HEK-YFP). Images were collected in three channels (d, s, and a). Based on these data, the values of correction parameters were α = 0.01, β = 0.56, γ = 0.23, δ = 0.04, and sensitized emission (IsS) was calculated. Scale bar represents 10 μm. (a) Image col-lected in d channel. (b) Image collected in a channel. (c) Overlay of d and a channels showing FRET cell as well
Katarzyna M. Kedziora and Kees Jalink
77
1. Select a FRET sensor to visualize changes in the concentration of a molecule of interest. The TEPACVV sensor is used in this example to follow the dynamic behavior of the cytosolic cAMP concentration [13].
2. Select an imaging setup (confocal or wide field—see Notes 20 and 21) and make sure that it is prepared for live cell imaging (see Note 7).
3. Find acquisition software that enables simple image analysis during the experiment. In fast dynamic experiments, it is ben-eficial to have immediate feedback on the result, enabling us to adapt the experiment accordingly.
1. Day 1: Plate cells of interest on glass coverslips. In the example experiment, HeLa cells were seeded at
50,000 cells/well in 6-well plates containing sterile coverslips. 2. Day 2: Transfect the cells with the FRET sensor constructs
according to common laboratory practice. HeLa cells were transfected with the TEPACVV construct using
PEI transfection reagent (1 μg of DNA + 3 μg of PEI for each well).
3. Day 3: Transfer a coverslip with cells to a chamber suitable for live cell imaging (see Note 13). Check carefully for fluid leaks (see Note 14) and place it on the microscope. Allow the tem-perature of the sample to equilibrate before starting measure-ments to avoid the main source of drift.
1. Find a healthy cell expressing the sensor construct at moderate level (see Note 22).
2. Focus, zoom in, and set the parameters for the acquisition. Often, the signal from an entire cell is averaged in this kind of experiments. Find a compromise between spatial resolution, signal-to-noise ratio, and bleaching/phototoxic effects.
3. Acquire d (donor excitation and donor emission) and s (donor excitation and acceptor emission) images.
4. Using proper acquisition software, draw a region of interest around the cell or group of cells. This region is used during the experiment to calculate d over s ratio (Fig. 2g).
3.2 Ratio FRET
3.2.1 Planning of a Ratio FRET Experiment
3.2.2 Preparation of a Ratio FRET Experiment
3.2.3 Experiment
Fig. 2 (continued) as donor only and acceptor only cell. (d) Image collected in s channel. (e) Calculated sig-nal of sensitized emission (I sS), showing that resonance energy transfer only takes place in tubular structures. (f, g) Ratio FRET. (f) Example images of HeLa cells expressing TEPACVV construct before and after stimulation with forskolin and IBMX. (g) Graph shows changes in signal in (d) CFP and (s) YFP channels during the experi-ment. The ratio CFP/YFP follows the changes in cAMP concentration after stimulation of the cells
FRET Microscopy
78
5. Set a proper sampling rate and duration of the experiment (see Note 23), and start acquisition of a baseline (see Note 24).
6. Perform a mock stimulation (negative control) with vehicle (see Note 25).
7. Add a proper stimulus to the cells (see Note 26). In the example, cells were stimulated with 60 nM of adrenalin
(Fig. 2f, g). 8. When the response to the stimulus is complete, add additional
stimuli or perform an end-point calibration. This calibration doubles as a positive control for the experiment.
In the example, cells were stimulated with a mixture of for-skolin (5 μM) and IBMX (5 μM) to raise cytosolic cAMP levels maximally (Fig. 2g).
4 Notes
1. The different normalizations Ed and Ea provide fully quantita-tive yet different information on FRET efficiency. To under-stand this idea, imagine an experiment in which the acceptors highly outnumber the donor molecules. If in this experiment all the available donor molecules are in complex with acceptors and the efficiency of energy transfer within a pair is 100 %, the calculated value of Ed will be 100 %, but the value of Ea will be very low.
2. The choice of a proper FRET pair may play a significant role in the success or failure of the experiment. Obviously, it has to match the available imaging system. One should also take into account fluorophore brightness and resistance to bleaching, mat-uration time, possible photochromic changes, and propensity to create dimers. A full discussion on choosing proper fluorophores can be found in a range of excellent publications [16, 17].
3. If possible, the fluorophores should be the only source of signal during the FRET experiment. Therefore, it is important that the level of autofluorescence of the cells is negligible within that spectral range.
4. All considerations of fusion proteins, like N or C terminal tag-ging or proper folding, apply to the FRET constructs. Moreover, the interaction of the studied proteins should bring the fluorophores in close proximity. Therefore, proper design of the linker region is very important [18, 19].
5. Reference cells are to be imaged simultaneously with the cells of interest. Due to the limited dynamic range of microscopic images, it is important that fluorescence intensity of both cell types is approximately similar.
Katarzyna M. Kedziora and Kees Jalink
79
6. Reference cells serve only as a source of pure signal from donors or acceptors. These can be transiently transfected cells (select cells that can be easily and efficiently transfected—such as HEK293T cells) or stable cell lines. For convenience, the distribution of fluorophores should allow easy distinction between reference cells and the cells of interest. Moreover, the reference signal should be present exactly in the plane of inter-est in confocal imaging—for example, if the experiment is designed to study the FRET signal within a cell nucleus, a ref-erence signal localized on the ventral membrane is of little use.
7. Depending on the duration of the experiment (time lapse ver-sus single measurement), it is important to keep cells healthy and avoid stressing them during imaging. Discussion on proper imaging media can be found in literature [20].
8. Proper alignment of the optical system is very important because misalignment may cause spatial shift between the images, as well as necessitate very high excitation power in order to acquire images of good S/N ratio.
9. Chromatic aberration of the setup should be minimized as much as possible. The use of objectives specially designed to decrease the chromatic shift in the spectral range of choice is highly recommended.
10. It is important to ensure highly homogenous illumination within the field of view. If the flatness of the field is not satisfac-tory, shading correction procedures should be applied to the acquired images [21].
11. Plating density depends on the specific cell type and on the selected protocol for transfection. However, care should be taken that on the day of experiment there is enough space on the coverslip for the reference cells to adhere.
12. The number of reference cells of each type should exceed the number of cells of interest 3–5 times. This abundance of refer-ence cells ensures that most FRET cells are accompanied by at least one of each reference cell types. The specific number of reference cells needed is cell-type specific, but in our hands, seeding at ~20 % density gives good results.
13. Transfer of a coverslip should be fast to prevent cells from dry-ing. Sharpened forceps make it easy to pick up the coverslip from the 6-well plate. Add imaging medium immediately after transfer.
14. Any leakage of an imaging medium will deteriorate image quality and may be harmful to the microscope. Clean the bot-tom of the coverslip in the imaging chamber and dry it with a paper tissue. Any leakage can be easily found by checking for moist spots (which will appear dark) on a dry tissue pressed to the bottom.
FRET Microscopy
80
15. For imaging, use the full dynamic range available to maximize the amount of collected information. However, overexposed pixels should definitely be avoided as they introduce bias in the calculated results.
16. It is critical to acquire images with maximum signal-to-noise ratio. To introduce necessary corrections, several calculations have to be performed with the images. This inevitably leads to the propagation of noise. If the quality of acquired images is not satisfactory, the calculated results may become very noisy and difficult to interpret. Acquisition of good quality images may require averaging and/or accumulating individual images. For time lapse, balance between the quality of the images and the well-being of the cells.
17. d and s images can easily be acquired simultaneously on a con-focal setup, but to acquire the a image, a change in excitation light is necessary. To avoid possible movement artifacts (e.g., when imaging fast-moving vesicles), it is highly advisable to acquire d, s, and a images by changing the excitation light between lines and not frames on a confocal system.
18. Make sure that the region of interest selected to calculate the value of the background does not contain any very dim cells. This may be easier to check by using a LUT (look-up table) that emphasizes differences in dim regions.
19. Standard procedures include smoothing (to reduce noise) and thresholding (to exclude some regions from analysis). Exclusion of some regions may help in the interpretation of calculated results; for example, in background regions noise may be pro-nounced. Overexposed pixels must also be excluded from the analysis, as they do not provide useful information. However, one has to check whether such operations introduce bias in the results.
20. Confocal microscopy is of course preferred if small spatial details are to be resolved (e.g., if a FRET sensor localizes to some small structures within a cell). Wide-field microscopes are usually faster and may be less detrimental to the cells. In case very fast readout (ms or sub-ms temporal resolution) is required, consider using a microscope that does not create an image but rather collects all photons together on two PMTs.
21. On a confocal microscope, it is straightforward to acquire d and s images simultaneously. On a wide-field microscope, one has to change emission filters and acquire these two images sequentially. Besides the danger of introducing motion arti-facts, this also necessitates double exposure and thus more photobleaching. Therefore, it is advisable to use a beam split-ter device that enables simultaneous detection of both chan-nels side by side on a single CCD [11].
Katarzyna M. Kedziora and Kees Jalink
81
22. Very high expression levels of fluorescent constructs may affect cell morphology and deteriorate the FRET response as well as cause saturation. On the other hand, a low signal-to-noise ratio may become a problem in low expressing cells.
23. Optimize the timing of the experiment as needed: the faster the biological process, the higher the necessary time resolu-tion, but this will also affect bleaching and phototoxic effects.
24. Make sure that the baseline is long enough to give you a good indication of the variability and drift in the signal. This depends on the signal-to-noise ratio, stability of the system, and bio-logical fluctuations.
25. Make sure to properly mix the medium after addition of stim-uli to avoid gradients of concentration (pipetting up and down ~5 times with a P200 pipet in a 2 ml imaging dish should pro-vide proper mixing). Mix the medium slowly and never point the pipet tip directly at the imaged cells in order not to squirt them away.
26. Stimulants are most often added from concentrated stocks. Take up the required volume of stimulus solution in a P20 pipet. Next, gently remove the yellow tip with the stimulus solution from the P20 and attach it to a P200 pipet. Make sure that the stimulus is not ejected in this step by keeping the P200 pipet piston slightly pressed; release it slowly while attaching the tip. The underpressure created in this way will prevent stimulus from spilling from the tip. Add the stimulus by mix-ing the sample medium several times with the P200 pipet (see Note 25).
References
1. Ciruela F, Vilardaga JP, Fernandez-Dueñas F (2010) Lighting up multiprotein complexes: lessons from GPCR oligomerization. Trends Biotechnol 2:138–147
2. Brunger AT, Strop P, Vrljic V et al (2011) Three dimensional molecular modeling with single molecule FRET. J Struct Biol 173:497–505
3. Xu X, Gerard ALV, Huang BCB et al (1998) Detection of programmed cell death using flu-orescence energy transfer. Nucleic Acids Res 26:2034–2035
4. Nagai T, Sawano A, Park ES et al (2001) Circularly permuted green fluorescent proteins engineered to sense Ca2+. Proc Natl Acad Sci U S A 98:3197–3202
5. Ponsioen B, Zhao J, Riedl J et al (2004) Detecting cAMP-induced Epac activation by
fluorescence resonance energy transfer: Epac as a novel cAMP indicator. EMBO Rep 5: 1176–1180
6. Hertel F, Switalski A, Mintert-Jancke E et al (2011) A genetically encoded tool kit for manipulating and monitoring membrane phos-phatidylinositol 4,5-bisphosphate in intact cells. PLOS One 6:e20855
7. Sun Y, Wallrabe H, Seo S et al (2011) FRET microscopy in 2010: the legacy of Theodor Forster on the 100th anniversary of his birth. Chem Phys Chem 12:462–474
8. Forster T (1946) Energiewanderung und Fluoreszenz. Naturwissenschaften 6:166–175
9. Gordon G, Berry G, Liang XH et al (1998) Quantitative fluorescence resonance energy transfer measurements using fluorescence microscopy. Biophys J 74:2702–2713
FRET Microscopy
82
10. Wouters FS, Verveer PJ, Bastiaens PIH (2001) Imaging biochemistry inside cells. Trends Cell Biol 11:203–211
11. Jalink K, van Rheenen J (2009) FilterFRET: quantitative imaging of sensitized emission. Fret and Flim Techniques, vol. 33, 1st edn. Elsevier B.V., pp. 289–349
12. Zhong Q, Lazar CS, Tronchere H et al (2002) Endosomal localization and function of sort-ing nexin 1. Proc Natl Acad Sci USA 99: 6767–6772
13. Klarenbeek JB, Goedhart J, Hink MA (2011) A mTurquoise-based cAMP sensor for both FLIM and ratiometric read-out has improved dynamic range. PLOS One 6:e19170
14. Schneider CA, Rasband WS, Eliceiri KW (2012) NIH Image to ImageJ: 25 years of image analysis. Nat Methods 9:671–675
15. Schindelin J, Arganda-Carreras I, Frise E et al (2012) Fiji: an open-source platform for biological- image analysis. Nat Methods 9: 676–682
16. Shaner NC, Lin MZ, McKeown MR et al (2008) Improving the photostability of bright monomeric orange and red fluorescent pro-teins. Nat Methods 5:545–551
17. Stepanenko OV, Stepanenko OV, Shcherbakova DM et al (2011) Modern fluorescent proteins: from chromophore formation to novel intra-cellular applications. BioTechniques 51:313–318
18. Komatsu N, Aoki K, Yamada M et al (2011) Development of an optimized backbone of FRET biosensors for kinases and GTPases. Mol Biol Cell 22:4647–4656
19. Chen H, Puhl HL, Koushik SV et al (2006) Measurement of FRET efficiency and ratio of donor to acceptor concentration in living cells. Biophys J 91:39–41
20. Goldman RD, Spector DL (2005) Live cell imaging: a laboratory manual. Cold Spring Harbor Laboratory, Cold Spring Harbor, NY
21. Tomazevic D, Likar B (2002) Comparative evaluation of retrospective shading correction methods. J Microsc 208:212–223
Katarzyna M. Kedziora and Kees Jalink
83
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_6, © Springer Science+Business Media New York 2015
Chapter 6
Localizing Protein–Protein Interactions in Living Cells Using Fluorescence Lifetime Imaging Microscopy
Yuansheng Sun and Ammasi Periasamy
Abstract
In the past decade, advances in fluorescence lifetime imaging have extensively applied in the life sciences, from fundamental biological investigations to advanced clinical diagnosis. Fluorescence lifetime imaging microscopy (FLIM) is now routinely used in the biological sciences to monitor dynamic signaling events inside living cells, e.g., Protein–Protein interactions. In this chapter, we describe the calibration of both time-correlated single-photon counting (TCSPC) and frequency domain (FD) FLIM systems and the acquisition and analysis of FLIM-FRET data for investigating Protein–Protein interactions in living cells.
Key words Fluorescence lifetime imaging microscopy (FLIM), Förster resonance energy transfer (FRET), FLIM-FRET, Time-domain FLIM, Time-correlated single-photon counting (TCSPC) FLIM, Frequency-domain FLIM, Protein–protein interactions
1 Introduction
Fluorescence lifetime is the average time a molecule spends in the excited state before returning to the ground state, typically with the emission of a photon. The natural fluorescence lifetime of a fluorophore (in the absence of non-radiative processes) is an intrin-sic property of the fluorophore. The natural fluorescence lifetimes of widely used fluorescent probes in cellular imaging such as fluo-rescent proteins and organic dyes are typically within 10 ns. The first nanosecond lifetime measurements using optical microscopy were made in 1959 [1, 2]. Since then, numerous FLIM method-ologies have evolved for various biological and clinical applications [3, 4]. Although FLIM techniques can be challenging for biolo-gists without a physics background, they provide an unprecedented level of information about the functions, dynamics, and interac-tions of proteins in living cells under physiological conditions at high temporal and spatial resolutions.
The fluorescence lifetime of a fluorescent molecule carries information about its local microenvironment and can be very
1.1 Overview of FLIM and Its Applications in the Biological Sciences
84
sensitive to certain environmental changes. Therefore, cellular responses to events such as changes in temperature, pH, and ion concentrations (e.g., calcium) can be measured accurately using FLIM. For example, high-speed widefield FLIM was employed to measure the change in calcium concentration in live cells using calcium sensitive probes (e.g., Oregon Green) [5–7]. Two-photon excitation (TPE) FLIM was used to map the response of rigor cross-bridges to stretching of the myosin essential light chain in skeletal muscle fibers by probing the microenvironment of the interface region of the myosin lever arm domain with Coumarin [8]. FLIM was implemented to investigate different conforma-tional changes of the presenilin 1 (PS1) protein that is associated with Alzheimer’s disease (AD) providing further understanding of the AD diagnosis [9]. FLIM techniques were also applied in plant biology. Eckert et al. [10] used the widefield single-photon count-ing FLIM technique to investigate the fluorescence dynamics of the chlorophyll d-containing cyanobacteria Acaryochloris marina.
FLIM has been employed to investigate a number of human diseases using endogenous autofluorescent molecules in human cells and tissues and shows great promise in several clinical applica-tions. Nicotinamide adenine dinucleotide (NAD+) is a coenzyme found in all living cells and carries electrons from one reaction to another through redox reactions in metabolism. When NAD+ accepts electrons from other molecules, it forms NADH which is highly fluorescent with peak absorption and emission maxima at 340 and 460 nm, respectively [11]. NADH can be imaged with TPE microscopy [12], serving as a convenient noninvasive fluores-cent probe of the cellular metabolic state. A new window for can-cer diagnosis was opened up using TPE FLIM to detect the free (shorter lifetime) to bound (longer lifetime) NADH ratio to moni-tor cellular metabolic states [13, 14]. Monitoring the autofluores-cence from human skin, the FLIM methodology was applied to distinguish between basal cell carcinomas and surrounding unin-volved skin [15].
One of the major FLIM applications is to measure FRET, which is the non-radiative energy transfer from an excited molecule (the donor) to another nearby molecule (the acceptor), via a long-range dipole–dipole coupling mechanism. The most basic concepts of FRET are described by [16–19]:
E R R r= +( )0
606 6/
(1)
R n J J
f f d
f d06 2 4 1 6
4
0 211= ´ ´ ´ ´{ } =( ) ( )
( )-
¥
¥
ò
ò. ,
/k
l l l
lQYD A
D A
D
el
ll
(2)
1.2 FLIM and Förster Resonance Energy Transfer (FRET)
Yuansheng Sun and Ammasi Periasamy
85
As shown in Eq. 1, the efficiency of energy transfer (E) from the donor to the acceptor is dependent on the inverse of the sixth power of the distance (r) separating them, subject to the Förster distance “R0,” at which half of the excited-stated energy of the donor is transferred to the acceptor (E = 50 %). Determination of R0 (in Ångstrom) for a FRET pair is given by Eq. 2, where κ2 is the orientation factor between the donor emission and acceptor absorption dipoles, ranging from 0 to 4; n is the medium refractive index; QYD is the donor quantum yield; J is the degree of the donor emission-acceptor absorption spectral overlap; εA is the extinction coefficient of the acceptor at its peak absorption wave-length; and fD(λ) and fA(λ) are the normalized donor emission and acceptor absorption spectra, respectively. For the majority of FRET pairs, the R0 values are on the scale of a few nanometers. Thus, FRET is usually limited to distances less than about 10 nm. Due to the dependence on the sixth power of the donor-acceptor distance, E is very sensitive to the distance change around R0 at the sub- nanometer scale (Fig. 1). Therefore, measuring FRET provides a sensitive tool for investigating a variety of phenomena that pro-duce changes in molecular proximity [20–24].
Using FLIM, FRET events can be identified by measuring the reduction in the donor lifetime that results from quenching in the presence of an acceptor, and E can be estimated from the donor
Fig. 1 FRET efficiencies (E) versus the separation distance between the donor and the acceptor (r), based on Eq. 1, for a Förster distance (R0) of 4, 5, or 6 nm, at which E is 50 %. Due to the dependence on the sixth power of the donor-acceptor distance, E is very sensitive to the distance change around R0—as an example highlighted for the R0 of 5 nm, the E values fall sharply for the distance increase from 3 nm to 7 nm
Localizing Protein-Protein Interactions Using FLIM
86
lifetimes determined in the absence (τD—unquenched lifetime) and the presence (τDA—quenched lifetime) of the acceptor, as shown by Eq. 3 [11].
E = -1
ttDA
D (3)
Since only donor signals are measured for determining E in FLIM- FRET, the method does not usually require the corrections for spectral bleedthrough that are necessary for intensity-based FRET measurements of the sensitized emission from the acceptor [20–24]. The fluorescence lifetime is insensitive to changes in fluorophore concentration, excitation light intensity, or light scattering but sen-sitive to environmental changes—these facts make FLIM- FRET an accurate method for FRET measurements. The FLIM- FRET method has been routinely applied in many laboratories including ours for studying Protein–Protein interactions and investigation of signaling events in a variety of biological systems [9, 25–46].
FLIM techniques are generally subdivided into time-domain (TD) and frequency-domain (FD) methods. The basic physics that underlie the two methods is essentially identical, since they are finite Fourier transforms of each other. In TD FLIM, a fluoro-phore is excited by a pulsed light source, which is synchronized to high-speed detectors and electronics; its fluorescence decay profile is directly measured in a number of (at least two) sequential time bins after (and relative to) the excitation pulse (time zero); its fluo-rescence lifetime can be estimated by analyzing the recorded decay profile (Fig. 2a). In FD FLIM, a modulated light source is used to
1.3 Overview of FLIM Techniques
Fig. 2 The fundamental principles of time-domain (TD) and frequency-domain (FD) FLIM measurements. (a) TD FLIM usually employs a pulsed light source that is synchronized to high-speed detectors and electronics to measure the emission (Em) signals at different time bins relative to the excitation (Ex) pulse (time zero), directly producing a fluorescence decay profile; the fluorescence lifetime is estimated by analyzing the recorded decay profile. (b) FD FLIM typically uses modulated light sources and detectors and measures the phase shift Φ and the amplitude attenuation M {M = (f/F0)/(e/E0)} of the emission relative to the excitation, which are then ana-lyzed to estimate the fluorescence lifetime
Yuansheng Sun and Ammasi Periasamy
87
excite a fluorophore, whose emission signals are measured by a detector modulated either at the same modulation frequency of the excitation (homodyne) or at a slightly (a few hundred to thou-sand hertz) different frequency (heterodyne); analyzing the phase shift(s) and amplitude attenuation(s) of the emission relative to the excitation makes it possible to extract the fluorescence lifetime of the fluorophore (Fig. 2b). The repetition rate of the excitation source used in TD FLIM or the fundamental modulation fre-quency of the excitation source used in FD FLIM needs to be chosen according to the fluorescence lifetime to be measured. The rate or frequency for measuring nanosecond lifetimes in most biological applications should be on the megahertz scale, e.g., 10–100 MHz.
Many TD and FD FLIM techniques were developed and dem-onstrated for various biological applications. Fortunately, most of them are now commercially available, e.g., from Becker and Hickl (BH, www.becker-hickl.de), Picoquant (www.picoquant.com), ISS (www.iss.com), Lambert Instruments (www.lambert-instruments.com), and Intelligent Imaging Innovations (www.intelligent- imaging.com). These systems can be stand alone or integrated with existing widefield or confocal or multiphoton (MP) microscopy systems. Each FLIM technique has its own strengths and limita-tions and choosing the most suitable one is really dependent upon the biological investigation at hand. Several FLIM techniques are compared in Table 1.
In the majority of FLIM measurements, a fluorescent decay is mea-sured from a population of fluorescent molecules. The intrinsic decay kinetics f(t), represented by the numbers of emitted photons as a function of time relative to the excitation, can be modeled by Eq. 4:
f t f a t
n
N
n n( ) = -( ){ }=å0
1
exp /t
(4)
where f0 is the number of emitted photons at time zero and N (N ≥ 1) is the number of different fluorescent species—each one is associated with a distinct fluorescence lifetime (τn), and its percentage over the number of total fluorescent molecules is often termed as the pre-exponential factor (an). For a multi-exponential case (N > 1), the apparent lifetime (τa) calculated by Eq. 5 is often used to indicate the mean lifetime of all fluorescent species.
t ta = ( ) =
= =å ån
N
n nn
N
na a1 1
1,
(5)
A variety of computer algorithms have been developed for FLIM data analysis [3, 4, 11]. Some are associated with specific FLIM data acquisition methods. For example, the rapid lifetime
1.4 Overview of FLIM Data Analysis
1.4.1 The General Fluorescent Decay Kinetics Model
Localizing Protein-Protein Interactions Using FLIM
Table 1 Comparison of several commonly used FLIM techniques
FLIM detection Typical implementation
Time- correlated single- photon counting (TCSPC)
TD FLIM
TCPSC [47] is typically implemented on laser scanning microscopes using a pulsed single- or multiphoton laser, a fast PMT or APD detector (timing jitter <300 ps), and TCSPC electronics [5, 30, 48, 49]. The TCSPC device synchronizes the detector and the scanning clock to the excitation pulse and records the arrival time and spatial information for each detected photon. Accumulating photons for thousands or millions of excitation pulses, a “photon counts” histogram (fluorescence decay profile) for estimating the fluorescence lifetime can be built at each pixel (or voxel) of a 2D (or 3D) image. A multichannel PMT module has also been employed in TCSPC FLIM to obtain spectrally resolved lifetimes [34, 50]
Gated image- intensified camera
TD FLIM
The gating-camera FLIM is typically implemented on widefield and spinning-disk confocal microscopes using a single-photon pulsed laser, a gated image-intensified camera, and gating-control electronics. The gating-camera can be operated at a superfast speed to detect photons within a time (gating) window for a few hundred picoseconds to a few milliseconds relative to the excitation pulse [5, 27]. A number of images are acquired in sequential gating windows to estimate the lifetimes. Extracting single-component lifetimes requires collecting two gated images at minimum, which may only take a few seconds [6, 51]. Due to the fast data acquisition speed, the gating-camera FLIM was implemented in a high- content screening platform [45, 52]
Streak cameraTD FLIM
A streak-camera system, consisting of a streak scope and a fast CCD camera, can be operated to transform the temporal profile of a light pulse into a spatial profile on a detector by causing a time-varying deflection of the light across the width of the detector [53]. In streak-camera FLIM, each line of an image consists of the time resolved information for a pixel location along the X-axis and a series of images are acquired for the Y-axis. The streak-camera FLIM uses a pulsed laser that is synchronized to a streak-camera system and was implemented on MP laser scanning microscopes [29, 31, 53]. Using a spectrograph device, the streak- camera FLIM was also demonstrated to obtain spectrally resolved lifetimes [31]
Heterodyne digitalFD FLIM
In traditional heterodyne FD FLIM, both the light source and the detector are modulated, but at a slightly different frequency, e.g., a few hundred hertz. The recently developed digital FD FLIM employs a modulated pulsed excitation source (one- or multiphoton laser), but does not modulate the detector at all, making the technique much more simpler to implement in laser scanning confocal microscopes [54]. In digital FD FLIM, the detector (PMT or APD) is working in a manner like photon counting, and all operations including the generation of the light modulation frequency, the generation of the cross-correlation sampling frequency, and the assignment of the time of arrival of a photon to a bin are digital, allowing multifrequency measurements (for extracting multicomponent lifetimes) done simultaneously and thus greatly improving photon efficiency and data acquisition speed [46, 54]
Homodyne FD FLIM using an image- intensified camera
The camera-based FD FLIM is typically implemented in widefield or spinning-disk confocal microscopes, using an LED or diode laser excitation light source and an image-intensified camera, both modulated at the same frequency (homodyne) [55–57]. This FLIM method is very fast since only several images need to be acquired to obtain one lifetime image and this may only take only a few seconds [58]
89
determination (RLD) algorithm was developed to analyze time-resolved images acquired by a gated image-intensified camera [5, 27, 59]. Some FLIM data analysis strategies vary depending upon the particular biological models. For example, global analysis [60–63] can significantly improve the signal-to-noise ratio (SNR) of fluorescence lifetime images, but the method requires the assumption that identical fluorescence relaxation parameters per-tain to the pixels grouped together for analysis, which may not be valid in many experimental systems.
Fitting analysis is commonly used in both TD and FD FLIM data analyses [11, 64]. In TCSPC FLIM, a measured decay g(t) is typi-cally represented as the convolution of the intrinsic decay f(t) (Eq. 4) and the instrument response function (IRF) of the FLIM system plus noise n(t), as shown in Eq. 6:
g t f t n t( ) = ´ ( ) + ( )IRF (6)
Given the IRF and some a priori parameters of the intrinsic decay, e.g., the number of exponential components (N in Eq. 4), an iterative fitting (deconvolution) procedure is usually applied to estimate the values of each lifetime component (τn in Eq. 4) and the corresponding pre-exponential factor (an in Eq. 4). At each iterative step k, an estimated gk(t) is calculated from the convolu-tion of the IRF and the intrinsic decay modeled by the estimated lifetimes and pre-exponential factors; the estimated gk(t) is then compared to the measured g(t) to produce the standard weighted least squares (χ2), which will be used as a criterion to evaluate the fitting significance and to determine whether or not to stop the iterative fitting procedure. The value of χ2, indicating a good fit for an appropriate model and a random noise distribution, should be close to 1 predicted by Poisson statistics with enough data points for fitting [11, 64].
In FD FLIM, the measured data at each pixel is usually composed of both phase delay (Φ) and amplitude modulation ratio (M), as shown in Fig. 2. The lifetime of a single fluorescent species (N = 1 in Eq. 4) can be directly calculated from either the phase (τΦ—the phase lifetime) or the modulation by Eq. 8 (τM—the modulation lifetime).
t wF F= tan / (7)
t wM = - ( )1 2M M/ (8)
where ω is the modulation frequency. A difference between τΦ and τM is an indication that the data does not follow a mono- exponential time course. For a multi-exponential case (N > 1 in Eq. 4), the phase and modulation values are often measured at several
1.4.2 Fitting Analysis of TCSPC FLIM Data
1.4.3 Model-Free Phasor Plot Analysis of FD FLIM Data
Localizing Protein-Protein Interactions Using FLIM
90
m odulation frequencies. From low to high frequencies, the phase delays increase from 0 to 90°, while the modulation ratios decrease from 1 to 0. To estimate the values of each lifetime component (τn in Eq. 4) and the associated pre-exponential factor (an in Eq. 4), the weighted least squares numerical approach can be used to fit both phase and modulation values measured at different frequen-cies, given some priori parameters of the exponential model, e.g., the number of exponential components (N in Eq. 4). Similar to TCSPC FLIM data fitting analysis, the calculated χ2 and residuals provide an evaluation for the fitted results.
Other than the fitting analysis, a graphical method for FLIM data presentation and analysis was developed. This method, named as polar plot [65], AB plot [66, 67], or phasor plot [68, 69] by different groups, is essentially based on the transformation of FD FLIM data defined by Eq. 9.
G M S Mw w w w w w= ( ) = ( )cos , sinF F (9)
Given a modulation frequency ω, each measured FD FLIM data point (Φω and Mω) can be directly located in a 2D plot using Gω as the horizontal axis and Sω as the vertical axis (Eq. 9). For a single- lifetime species, the phase (Eq. 7) and modulation (Eq. 8) lifetimes are equal, from which a relationship between the Gω and Sω coor-dinates shown in Eq. 10 can be derived.
S Gw w2 2
0 5 0 25+ -( ) =. . (10)
This Gω − Sω relationship is represented as a semicircle curve center-ing at (Gω = 0.5, Sω = 0) with a radius of 0.5 in the phasor plot. The semicircle curve indicates the lifetime trajectory with decreasing lifetime from left to right, where (1, 0) indicates lifetimes near zero to (0, 0) being infinite lifetime. A point falling on the semicircle will have only a single lifetime, while a point that falls inside the semicircle will have multiple lifetime components.
Since no fitting is involved, the phasor plot approach is inde-pendent of any underlying physical model. By plotting raw data in the phasor plot, one can distinguish between single- and multicomponent lifetimes, identify the lifetime value of a single-lifetime species on the semicircle, and also visualize the relative lifetime changes between complex lifetime distributions. More conveniently, many data distributions measured from different samples can be directly compared in a same phasor plot. In FLIM-FRET measurements, FRET can be identified by compar-ing the phasor plot distributions of the donor fluorophores in the donor-alone versus the “donor + acceptor” specimens. Several analytical tools were developed based on the polar, AB, or phasor plot approach for multicomponent analysis without nonlinear fitting [66, 67] and determining the FRET efficiency [57, 68], e.g., the “FRET Trajectory” functions in the SimFCS
Yuansheng Sun and Ammasi Periasamy
91
software (www.lfd.uci.edu/globals) [68]. It should be noted that these methods can be also employed to analyze raw TCSPC FLIM data after being transformed into the frequency domain through the digital Fourier transform—such a function is pro-vided by the SimFCS software [68].
A typical FLIM-FRET experiment involves three steps: 1. Calibrate the FLIM system with fluorescence lifetime
standards. 2. Acquire FLIM data from biological specimens, typically includ-
ing specimens that only have the donor and from specimens that contain both the donor and the acceptor.
3. Analyze the data to extract the lifetime information and inter-pret the analyzed data to illustrate the biological activity or other experimental objectives.
Here, we describe the procedures that are necessary to use both a TCSPC FLIM method and a FD FLIM method to monitor Protein–Protein interactions in live specimens based on FRET. Both synthetic and real biological models are used for FLIM-FRET demonstrations in live cells.
2 Materials
The basic principle of TCSPC FLIM is described in Fig. 2a and Table 1.
1. The TPE TCSPC FLIM system is implemented on a Bio-Rad Radiance 2100 confocal/MP microscope system using the MP system configuration, which is operated by the Bio-Rad LaserSharp 2000 software (www.zeiss.com/micro). The Bio- Rad scanning unit is attached to a Nikon TE300 inverted microscope.
2. The TCSPC module is a BH SPC-150 computer board (www.becker-hickl.de).
3. The FLIM detector is a BH PMC-100-0 PMT, which has a response time of approximately 150 ps [30, 46] and is attached to the non-descanning side port of the microscope.
4. The TPE laser is a Coherent 10 W Verdi pumped tunable (700–1,000 nm) mode-locked ultrafast (pulse width <150 fs) pulsed (repetition rate of 78 MHz) laser (www.coherent.com).
5. A remotely controlled filter wheel containing several emission filters is placed before the FLIM detector. The 480/40 nm (for Coumarin 6 and Cerulean) and 525/50 nm (for Fluorescein) emission filters (www.chroma.com) were used.
6. A Nikon 60X/1.2NA water objective was used for all TCSPC FLIM measurements.
1.5 Outline of a FLIM-FRET Protocol
2.1 TPE TCSPC FLIM System
Localizing Protein-Protein Interactions Using FLIM
92
7. The TCSPC module and FLIM detector are controlled by the BH SPCM software.
8. The acquired TCSPC FLIM data is analyzed using the BH SPCImage software.
The IRF can be estimated from the acquired decay data itself. However, it is preferable to directly measure the IRF experimen-tally, e.g., by recording scattered excitation light. For visible light excitation, a strongly scattering specimen such as nondairy coffee creamer is conventionally used to record the IRF [5]. For infrared light excitation, the IRF can be measured using a sample (such as urea crystal) that yields second-harmonic generation (SHG) sig-nals [70]. SHG is a nonlinear process that delivers a signal at one- half of the excitation wavelength.
1. A thin layer of urea crystal sandwiched between a glass cover-slip and a glass slide was used to measure the IRF of the TPE TCPSC FLIM system.
The basic principle of digital FD FLIM is described in Fig. 2b and Table 1. We use the ISS ALBA confocal digital FD FLIM/FCS system (www.iss.com).
1. The ISS ALBA FastFLIM system is attached to a Nikon TiE300 inverted microscope.
2. The excitation source is a Fianium SC450-6 pulsed supercon-tinuum laser module (www.fianium.com), which covers the excitation wavelengths ranging from 460 to 2,200 nm and has a repetition rate of 60 MHz and a pulse width of 6 ps.
3. The system is equipped with two identical PerkinElmer SPCM-AQRH- 15 avalanche photodiode (APD) detectors (www.perkinelmer.com).
4. Various combinations of excitation, emission filters, and dichroic mirrors can be configured through the system software. The 445/20 nm (for Coumarin 6, Fluorescein and Cerulean) and 560/25 nm (for Rose Bengal) excitation filters and the 480/40 nm (for Coumarin 6 and Cerulean), 531/40 nm (for Coumarin 6 and Fluorescein), and 630/92 nm (for Rose Bengal) emission filters (www.semrock.com) were used.
5. A Nikon 60X/1.2NA water objective was used for all FD FLIM measurements.
6. The phase shifts and modulation attenuations of the emission relative to the excitation at three frequencies (60, 180, 300 MHz) are simultaneously measured at each pixel of a fluo-rescence lifetime image.
7. The ISS VistaVision software is used for data acquisition and analysis.
2.2 Measurement of the TCSPC FLIM Instrument Response Function (IRF)
2.3 Confocal FD FLIM System
Yuansheng Sun and Ammasi Periasamy
93
Prior to any scientific investigation, a FLIM system must be cali-brated with standard fluorophores of known lifetimes. Many of them have been used for calibrating FLIM systems, and a few com-monly used ones are given in Table 2. The following standards were used for calibrating our FLIM systems at room temperature (~22 °C).
1. Coumarin 6 dissolved in ethanol (~2.5 ns). 2. Fluorescein dissolved in Tris buffer at pH 7.5 (~4.05 ns). 3. Rose Bengal dissolved in water (~0.12 ns). 4. Rose Bengal dissolved in methanol (~0.54 ns). 5. Rose Bengal dissolved in 2-proponal (~0.98 ns). 6. Rose Bengal dissolved in methanol (~2.57 ns).
It is desirable to employ FRET reference standards in addition to positive and negative controls to verify the FLIM-FRET results. A comparative method to determine the accuracy of FRET measure-ments was developed by the Vogel laboratory (NIH/NIAAA) [74, 75]. The approach uses “standards” in the form of genetic constructs encoding fusions between donor and acceptor fluores-cent proteins separated by defined amino acid (aa) linker sequences. A series of FRET-standard constructs were generated through encoding Cerulean [76] and Venus [77], directly coupled by either
2.4 Fluorescence Lifetime Standards for FLIM System Calibration
2.5 FRET Standards for Verification of a FLIM-FRET Approach
Table 2 List of several commonly used fluorescence lifetime standard fluorophores
FluorophoreAbsorption peak a Emission peaka Solvent Lifetimeb Reference
Coumarin 6 ~460 nm ~505 nm Ethanol ~2.5 ns [46]
HPTS ~454 nm ~511 nm Phosphate buffer (pH 7.5)
~5.3 ns [46]
Fluorescein ~494 nm ~521 nm Phosphate or Tris buffer (pH > 7)
4–4.1 ns [54, 71]
Rhodamine B ~542 nm ~565 nm Water ~1.7 ns [54]Methanol ~2.5 ns [72]Ethanol ~3.1 ns [5, 55, 73]
Rose Bengal ~560 nm ~571 nm Water ~0.12 ns [5, 53, 73]Methanol ~0.54 nsEthanol ~0.74 ns2-propanol ~0.98 nsAcetonitrile ~2.38 nsAcetone ~2.57 ns
aThe absorption and emission peaks may vary upon the solventbAll lifetimes given were measured at room temperature (~22 °C)
Localizing Protein-Protein Interactions Using FLIM
94
a 5, 17, or 32 aa linker—named as C5V, C17V, and C32V, corre-spondingly [75]. In addition, a construct of a low FRET efficiency as a negative control was also made using a 229 aa linker and is named CTV, where Cerulean and Venus are separated by a TRAF domain [74]. These plasmids are now available at www.addgene.org/Steven_Vogel. Here, three FRET-standard constructs expressed in live mouse pituitary GHFT1 cells (transfected by FuGENE 6) are used to verify both TCSPC and FD FLIM-FRET approaches.
1. Cells alone without any transfection. 2. Cells transfected with Venus alone. 3. Cells transfected with Cerulean alone (donor-alone control). 4. Cells transfected with the C5V construct. 5. Cells transfected with the C32V construct. 6. Cells transfected with the CTV construct.
The example biological model for demonstration is the basic region-leucine zipper (bZip) domain of the CCAAT/enhancer binding protein alpha (C/EBPα) transcription factor. The bZip family proteins form obligate dimers through their leucine zipper domains, which position the basic region residues for binding to specific DNA elements. Immunocytochemical staining of differen-tiated mouse adipocyte cells showed that endogenous C/EBPα was preferentially bound to satellite DNA-repeat sequences located in regions of centromeric heterochromatin [78, 79]. When the C/EBPα bZip domain is expressed as a fusion to a fluorescent protein in cells of mouse origin, such as the pituitary GHFT1 cells used here, it is localized to the well-defined regions of centromeric het-erochromatin in the cell nucleus [80]. FRET microscopy has been demonstrated to be a perfect tool for detecting the homo- dimerization of C/EBPα in living cells [24, 32, 33, 38, 46]. A FRET system for this biological model can be built by fusing the C/EBPα bZip domain to two fluorescent proteins of a good FRET pair separately, e.g., Cerulean (FRET donor) and Venus (FRET acceptor) as used here. In FLIM, the donor fluorescence lifetimes are measured from cells co-expressing the donor (Cerulean-bZip) and the acceptor (Venus-bZip) as well as cells that only express the donor (Cerulean-bZip). Here, the Cerulean lifetimes in live GHFT1 cells expressing only Cerulean-bZip or both Cerulean- bZip and Venus-bZip (transfected by FuGENE 6) are measured by the TPE TCSPC FLIM approach, to demonstrate the quenching of Cerulean in doubly expressed cells due to FRET between Cerulean-bZip and Venus-bZip.
1. Cells transfected with Cerulean-bZip (donor-alone control). 2. Cells transfected with both Cerulean-b Zip and Venus-bZip.
2.6 A FRET Model in Living Cells: Homo- dimerization of the CCAAT/Enhancer Binding Protein Alpha (C/EBPα) Transcription Factor
Yuansheng Sun and Ammasi Periasamy
95
3 Methods
See Note 1 for the selection of fluorescence lifetime standards for FLIM system calibration.
1. Excite the Urea Crystal sample at the 940 nm wavelength and collect its SHG signals using the 480/40 nm emission filter. The recorded SHG decay can then be imported into the BH SPCImage software as the IRF for data analysis (see more details in [46] and Note 2).
2. Acquire at least three fluorescence lifetime images from the Coumarin 6 standard solution at different fields of the solu-tion, using the 870 nm excitation wavelength and the 480/40 nm emission filter.
3. Acquire at least three fluorescence lifetime images from the Fluorescein standard solution at different fields of solution, using the 870 nm excitation wavelength and the 525/50 nm emission filter.
4. In the BH SPCImage software, fit the Coumarin 6 and Fluorescein lifetime images using the single-exponential model and the measured IRF. As shown in Fig. 3, both of their fluo-
3.1 System Calibration with Fluorescence Lifetime Standards
3.1.1 Calibration of the TPE TCSPC FLIM System with the Measured IRF
Fig. 3 Calibration of the TPE TCSPC FLIM system with Coumarin 6 in ethanol and Fluorescein in Tris buffer at pH 7.5. For both, the mono-exponential fitting of the raw decay data with the measured instrument response function (IRF) appears to be sufficient with a good χ2 of 1.1 and yields the correct lifetimes of Coumarin 6 (2.5 ns) and Fluorescein (4 ns). As expected for these homogeneous solutions, each histogram of the repre-sentative Coumarin 6 and Fluorescein fluorescence lifetime images (given in the inset) shows a very narrow Gaussian distribution—the small variation is mainly due to noise
Localizing Protein-Protein Interactions Using FLIM
96
rescence lifetimes should be accurately resolved with small variations (Fig. 3 inset), as expected from these homogeneous solutions.
In FD FLIM measurements, the system needs to be first calibrated using a standard fluorophore solution of known lifetime. In the calibration procedure, the software will adjust and set the initial DC, phase, and modulation values at each measured frequency for each detector, based on the data acquired from the standard fluo-rophore solution and its lifetime given to the software (see Note 3). Therefore, it is important to use a second fluorescence lifetime standard to verify the calibration procedure.
1. Run the system calibration with the Coumarin 6 standard solu-tion, using the 445/20 nm excitation and the 531/40 nm emission filters, given its lifetime of 2.5 ns to the ISS VistaVision software. Using the same setup, acquire at least three fluores-cence lifetime images at different fields of the Coumarin 6 solution; then switch to the Fluorescein standard solution and acquire at least three fluorescence lifetime images at different fields of the Fluorescein solution.
2. Run the system calibration with the Rose Bengal acetone standard solution, using the 560/25 nm excitation and the 630/92 nm emission filters, given its lifetime of 2.57 ns to the ISS VistaVision software. Using the same setup, acquire at least three fluorescence lifetime images at different fields of the Rose Bengal acetone solution; then switch to other Rose Bengal (water, methanol and 2-propanol) standard solutions one by one and acquire at least three fluorescence lifetime images at different fields of each solution.
3. In the ISS VistaVision software, plot the acquired fluorescence lifetime images using the phasor plot approach (see Note 4). As shown in Fig. 4, the phasor distribution of each standard solution should almost center on the semicircle, demonstrat-ing its mono-exponential decay nature, as expected from these homogeneous solutions. Placing the cursor at the center of a distribution gives both its phase and modulation lifetimes, which should be nearly identical.
1. Use untransfected cells to check if there is autofluorescence from the specimen in the donor (480/40 nm) channel for the donor excitation (820 nm) wavelength (see Note 5).
2. Use cells transfected with Venus to check if there is back bleedthrough to the donor (480/40 nm) channel for the donor excitation (820 nm) wavelength and (see Note 6).
3. Use cells expressing Cerulean alone and acquire fluorescence lifetime images from at least 10 cells using the 820 nm excitation
3.1.2 Calibration of the Confocal Digital FD FLIM System
3.2 Verification of a FLIM System for FRET Studies with FRET Standards
3.2.1 Measure FRET Standards Using the TPE TCSPC FLIM Method
Yuansheng Sun and Ammasi Periasamy
97
wavelength and the 480/40 emission filter, which were also used for the FRET-standard constructs.
4. Use cells expressing the C5V construct and acquire the donor (Cerulean) fluorescence lifetime images from at least 10 cells.
5. Use cells expressing the C32V construct and acquire the donor (Cerulean) fluorescence lifetime images from at least 10 cells.
6. Use cells expressing the CTV construct and acquire the donor (Cerulean) fluorescence lifetime images from at least 10 cells.
7. In the BH SPCImage software, fit the Cerulean-alone lifetime images using the single-exponential model and the measured IRF to estimate the unquenched donor lifetime; for the FRET- standard constructs, both single- and double-exponential fit-tings were applied with the measured IRF and the results were then compared to determine the quenched donor lifetimes—the CTV lifetime images were sufficiently fitted with the single- exponential model, while the C32V and C5V lifetime images were better fitted using the double-exponential model (see Note 7). The representative raw decays and fitting results are shown in Fig. 5, clearly demonstrating that the TPE TCSPC FLIM technique can distinguish subtle changes in FRET in the CTV, C32V, and C5V FRET-standard constructs by resolving the different lifetimes of Cerulean (FRET donor) in the absence of Venus (FRET acceptor) versus in the presence of Venus at different proximity in these FRET-standard con-structs (see Note 8).
Fig. 4 Calibration of the confocal digital FD FLIM system with Coumarin 6 in ethanol, Fluorescein in Tris buffer pH 7.5, and Rose Bengal in acetone, 2-propano, methanol, or water. (a) The FD FLIM system was calibrated with Coumarin 6, given its lifetime of 2.5 ns to the system software, and then used to measure the Fluorescein life-time, showing the correct value of 4.0 ns. (b) The FD FLIM system was calibrated with Rose Bengal in acetone and the input lifetime of 2.57 ns and then used to measure the lifetimes of Rose Bengal in other solvents—water, methanol, and 2-propanol, which are clearly resolved to be 0.14, 0.5, and 0.92 ns, respectively. For each standard solution, the raw FD FLIM data is presented by the phasor plot at the fundamental frequency of 60 MHz. Each point in the phasor plot represents a pixel in a fluorescence lifetime image, and its x, y- coordinates and intensity are determined by the phase (Φ), modulation (M), and DC values of the corresponding pixel
Localizing Protein-Protein Interactions Using FLIM
98
Fig. 5 Verification of the TPE TCSPC FLIM-FRET method using FRET-standard constructs—C5V, C32V, and CTV (adapted from [24]). The Cerulean fluorescence lifetimes in live cells expressing Cerulean alone (Cerulean) or a FRET-standard construct (CTV, C32V, or C5V) were measured using the TPE TCSCP FLIM system. (a) For each, the raw decay data points along with the corresponding fitted decay curve at a representative pixel are plotted—the comparison clearly shows a faster decay from Cerulean to CTV to C32V to C5V. Both Cerulean-alone and CTV data were analyzed by single-exponential fitting. However, C32V and C5V data were better fitted with two lifetime components, and their mean lifetimes are used for comparison. (b) The lifetime distributions and color-coded lifetime images of the representative cells are also compared, confirming that lifetimes get shorter from Cerulean to CTV to C32V to C5V
Yuansheng Sun and Ammasi Periasamy
99
1. Run the system calibration with the Coumarin 6 standard solu-tion, using the 445/20 nm excitation and the 480/40 nm emission filters, given its lifetime of 2.5 ns to the ISS VistaVision software. Using the same setup, acquire at least three fluores-cence lifetime images at different fields of the Coumarin 6 solution and then plot the acquired Coumarin 6 fluorescence lifetime images in the phasor plot to make sure to obtain a single lifetime of 2.5 ns for Coumarin 6 (Fig. 4a).
2. Use untransfected cells to check if there is autofluorescence from the specimen in the donor (480/40 nm) channel for the donor excitation (445/20 nm) wavelength (see Note 5).
3. Use cells transfected with Venus to check if there is back bleedthrough to the donor (480/40 nm) channel for the donor excitation (445/20 nm) wavelength (see Note 6).
4. Use cells expressing Cerulean alone and acquire fluorescence lifetime images from at least 10 cells using the 445/20 nm excitation and the 480/40 emission filters, which were also used for the FRET-standard constructs.
5. Use cells expressing the C5V construct and acquire Cerulean (donor) fluorescence lifetime images from at least 10 cells.
6. Use cells expressing the C32V construct and acquire Cerulean (donor) fluorescence lifetime images from at least 10 cells.
7. Use cells expressing the CTV construct and acquire Cerulean (donor) fluorescence lifetime images from at least 10 cells.
8. In the ISS VistaVision software, plot the Cerulean-alone, CTV, C32V, and C5V fluorescence lifetime images using the phasor plot approach. As shown Fig. 6, the FD FLIM technique clearly resolved the Cerulean-alone, CTV, C32V, and C5V fluorescence lifetime data distributions in the phasor plot, demonstrating that the lifetimes of Cerulean decrease from Cerulean-alone to CTV to C32V to C5V.
1. Use cells expressing Cerulean-bZip and acquire fluorescence lifetime images from at least 10 cells using the 820 nm excita-tion wavelength and the 480/40 emission filter.
2. Use cells expressing both Cerulean-bZip and Venus-bZip and acquire Cerulean (donor) fluorescence lifetime images from at least 10 cells using the 820 nm excitation wavelength and the 480/40 emission filter.
3. In the BH SPCImage software, fit the Cerulean-bZip lifetime images using the single-exponential model and the measured IRF to estimate the unquenched donor lifetime; by comparing both single- and double-exponential fittings, the “Cerulean-
3.2.2 Measure FRET Standards Using the Confocal FD FLIM Method
3.3 Localization of Homo- dimerization of C/EBPα in Living Cells Using the TPE TCSPC FLIM- FRET Method
Localizing Protein-Protein Interactions Using FLIM
100
bZip + Venus-bZip” fluorescence lifetime images were better fit-ted using the double-exponential model with the measured IRF. The representative raw decays and fitting results are shown in Fig. 7, where comparing the lifetimes of Cerulean in singly versus doubly expressed cells, measured by the TPE TCSPC FLIM method, clearly shows the quenching of Cerulean-bZip by Venus-bZip due to FRET between them, demonstrating homo-dimerization of C/EBPα-bZip in living cells (see Notes 9 and 10).
Fig. 6 Verification of the confocal digital FD FLIM-FRET method using FRET- standard constructs—C5V, C32V, and CTV. The Cerulean fluorescence lifetimes in live cells expressing Cerulean alone (Cerulean) or a FRET-standard construct (CTV, C32V or C5V) were measured using the confocal digital FD FLIM system. The representative images and the corresponding phasor plot distributions are shown. It is clearly demonstrated by the phasor plot (without any fitting) that the Cerulean lifetimes decrease from Cerulean alone to CTV to C32V to C5V. The phasor distribution of Cerulean alone is centered on the universal semicircle of the phasor plot, indicating that it has a single lifetime component. However, the C32V and C5V phasor distributions are completely located inside the universal semicircle of the phasor plot, indicating that they have more than one lifetime component. The CTV phasor distribution is close to that of Cerulean alone, as expected from the linker system
Yuansheng Sun and Ammasi Periasamy
101
4 Notes
1. It is preferable to choose standard fluorophores whose excita-tion, emission, and fluorescence lifetime properties are close to those of the fluorophores in the intended samples, so that the same imaging setup used for biological samples can be applied for the system calibration. Importantly, the standard should be carefully prepared according to the reference, such as solvent, pH, temperature, molar concentration, etc, because these fac-tors may influence its fluorescence lifetime. Typically, the con-centration of a dye would not influence its fluorescence lifetime. However, most of these standard dyes cannot be prepared in
Fig. 7 Localization of dimerized C/EBPα-bZip in living cell nuclei using the TPE TCSPC FLIM-FRET method (adapted from [46]). C/EBPα-bZip was tagged with either Cerulean (Cerulean-bZip: FRET donor) or Venus (Venus-bZip: FRET acceptor). The unquenched Cerulean lifetimes were measured from the donor-alone control cells that only express Cerulean-bZip and were determined by single-exponential fitting. The lifetimes of quenched Cerulean in cells co-expressing Cerulean-bZip and Venus-bZip were analyzed by double- exponential fitting, and the mean lifetime of the two lifetime components is used for comparison. (a) Comparing their rep-resentative raw decay data points along with the corresponding fitted curves shows a faster decay of Cerulean- bZip in the presence of Venus-bZip than in the absence of Venus-bZip, demonstrating FRET between the two and indicating the homo-dimerization of C/EBPα-bZip. (b) By applying suitable thresholds, fittings were only applied to the pixels in regions of centromeric heterochromatin of the cell nucleus. The representative intensity- overlaid lifetime images and the corresponding lifetime distributions are compared, showing the shorter Cerulean lifetimes in doubly expressed cells than singly expressed cells and demonstrating that the quenching of Cerulean-bZip by Venus-bZip in dimerized C/EBPα-bZip
Localizing Protein-Protein Interactions Using FLIM
102
100 % purity and the impurity at a high dye concentration may quench the dye, resulting in a decrease in its fluorescence life-time. Thus, one should carefully prepare the dye concentration upon the dye and the system sensitivity.
2. It is important to apply the same optics and TCPSC settings used for experimental samples, since the IRF may vary depend-ing upon both the optical path (such as the objective lens and the thickness of the emission filter) and the TCSPC configura-tion (such as the constant fraction discriminator parameters). More importantly, the IRF of a TPE TCSPC FLIM system should be periodically checked, since it can change due to reflections in the optical path, poor mode locking of the TPE laser, or instability of the TCPSC electronics.
3. To ensure the accuracy and reproducibility of the lifetime mea-surements, the FD FLIM system calibration needs to be done in every experiment and repeated periodically (every a few hours) for a long experimental course.
4. Since in most of biological applications the nanosecond-scale lifetimes are measured with high frequencies, it is important to apply a de-noising technique such as Gaussian or wavelet smoothing or median filtering to facilitate the phasor plot anal-ysis [68, 81]. The phasor plot data shown in Figs. 4 and 6 are applied with Gaussian smoothing.
5. Autofluorescence from specimens should be carefully checked. If the donor fluorescence signals are contaminated with the sample autofluorescence, the autofluorescence lifetime needs to be first determined using unlabeled specimens and then considered in the data analysis of labeled samples. In our case, no noticeable autofluorescence was observed from the unla-beled specimens in either the TPE TCSPC FLIM setup or the confocal FD FLIM setup.
6. It is important to check if there is back bleedthrough from the acceptor fluorophore to the donor channel under the donor excitation wavelength. If there is back bleedthrough, it would be better to choose another acceptor fluorophore since the back bleedthrough could significantly complicate FLIM data analysis. In our case, no noticeable back bleedthrough from the acceptor-alone specimens was observed in either the TPE TCSPC FLIM setup or the confocal FD FLIM setup.
7. The goodness-of-fit is considered an important factor for making the decision on whether or not to accept fitted FLIM results and is usually assessed by the calculated χ2 and residuals, as well as by visually comparing the fitting curve versus the measured data points. An example of evaluating the fit of TCSPC FLIM data is shown in Fig. 8. One should
Yuansheng Sun and Ammasi Periasamy
103
always be careful when accepting a more complicated model for data analysis, since it is the reproducibility of data for a particular data processing model that is critical. Most impor-tantly, more photon counts are required to obtain an accu-rate statistical fit of the lifetime data when resolving more lifetime components.
8. The FLIM E% can be calculated using Eq. 3, where the unquenched donor lifetimes (τD) are measured from the donor-alone specimens, e.g., cells expressing Cerulean-alone; the quenched donor lifetimes (τDA) are estimated from the “donor + acceptor” specimens, e.g., cells expressing a CTV, C32V, or C5V FRET-standard construct. The average of Cerulean lifetimes obtained from all donor-alone lifetime images is used as τD for the E% calculation. For doubly expressed cells, the mean Cerulean lifetime (Eq. 5) at each pixel is used as τDA for the E% calculation to produce an E% image [46]. The average FLIM E%s of CTV, C32V, and C5V were determined to be 6, 20, and 40 %, respectively [82].
9. For accurate FLIM-FRET measurements, a few cautions need to be taken into consideration to rule out the lifetime changes caused by the factors other than FRET. Photobleaching is a common problem in fluorescence microscopy imaging and can induce false information in quantitative analysis. Although
Fig. 8 Evaluation of fitting of TCSPC FLIM data. Fitting the Cerulean decay data acquired from Cerulean-alone (Cerulean) or C5V cells is used as an example—each was analyzed by both single- and double-exponential fittings. (a) For the Cerulean-alone decay, the single-exponential (Single) fit is nearly identical to the double- exponential (Double) fit, and the obtained weighted least squares (χ2) values of the two fits are almost the same (Single: 1.12 vs. Double: 1.13), indicating that the Cerulean-alone decay is mono-exponential. However, a difference is seen from comparing the single- versus double-exponential fits of the C5V decay data. (b) A zoom view into the C5V decay of 1–5 ns clearly shows a better fit of using the double-exponential model than the single-exponential model, especially at the beginning (1–2 ns) of the decay. This is also confirmed by the calculated χ2 values (Single: 2.47 vs. Double: 1.18) and residuals. Therefore, the C5V decay should be analyzed for two lifetime components
Localizing Protein-Protein Interactions Using FLIM
104
FLIM techniques are generally more tolerant to some extent of photobleaching compared to intensity-based techniques for FRET measurements, photobleaching could influence the measured lifetime [46]. Prior to FLIM measurements, the sig-nificance of photobleaching should be monitored for the time course of acquiring a FLIM data set, to establish an imaging condition where the measured donor-alone lifetimes are not affected [46]. The fluorescence lifetime of a fluorophore can be influenced by its microenvironment. Thus, it is important to make sure that the donor fluorophores reside in the same microenvironment in both donor-alone control specimens and specimens containing both donor and acceptor fluorophores. For example, suitable thresholds were applied for both donor-alone and doubly expressed cells to only analyze the donor lifetimes in the regions of centromeric heterochromatin of the cell nucleus (Fig. 7).
10. The unquenched Cerulean lifetimes have a narrow distribution with an average of about 2.75 ns (Fig. 7b) [46]. However, the quenched Cerulean lifetimes measured from cells co- expressing Cerulean-bZip and Venus-bZip vary substantially (2.0–2.6 ns) from one centromeric region to another (Fig. 7b), resulting in an E% range of 5–27 % [46]. Since the donor and acceptor fusion proteins in the C/EBPα-bZip model are expressed independently and there are also donor-donor and acceptor-acceptor dimers other than donor-acceptor dimers, the stoichi-ometry between the acceptor and the donor (the “acceptor/donor” ratio) varies from one region of interest (ROI) to another. The donor will more likely be quenched when it is surrounded by many more acceptors. Quantifying the “accep-tor/donor” ratio in addition to the quenched lifetime of Cerulean in each ROI will help to understand the changes of those quenched lifetimes in different ROIs [46].
Acknowledgments
The authors acknowledge funding from the University of Virginia, National Heart, Lung, and Blood Institute (NHLBI) PO1HL101871 and National Center for Research Resources NCRR-NIH RR027409. The authors thank Ms. Kay Christopher (Biology, University of Virginia) for preparing the samples, Dr. Steven Vogel (NIH/NIAAA) for providing the FRET-standard constructs, and Dr. Richard Day (Indiana University School of Medicine) for providing the C/EBPα-bZip constructs.
Yuansheng Sun and Ammasi Periasamy
105
References
1. Venetta BD (1959) Microscope phase fluo-rometer for determining the fluorescence life-times of fluorochromes. Rev Sci Instrum 30: 450–457
2. Biener E, Charlier M, Ramanujan VK et al (2005) Quantitative FRET imaging of leptin receptor oligomerization kinetics in single cells. Biol Cell 97:905–919
3. Gadella TWJ (2009) FRET and FLIM tech-niques. In: Pillai S, van der Vliet PC (eds) Laboratory techniques in biochemistry and molecular biology. Elsevier, Oxford, UK, p 534
4. Periasamy A, Clegg RM (2009) FLIM micros-copy in biology and medicine. CRC Press, London
5. Periasamy A, Wodnicki P, Wang XF et al (1996) Time-resolved fluorescence lifetime imaging microscopy using a picosecond pulsed tunable dye laser system. Rev Sci Instrum 67:3722–3731
6. Agronskaia AV, Tertoolen L, Gerritsen HC (2003) High frame rate fluorescence lifetime imaging. J Phys D 36:1655–1662
7. Agronskaia AV, Tertoolen L, Gerritsen HC (2004) Fast fluorescence lifetime imaging of calcium in living cells. J Biomed Opt 9:1230–1237
8. Ushakov DS, Caorsi V, Ibanez-Garcia D et al (2011) Response of rigor cross-bridges to stretch detected by fluorescence lifetime imaging micros-copy of myosin essential light chain in skeletal muscle fibers. J Biol Chem 286:842–850
9. Berezovska O, Lleo A, Herl LD et al (2005) Familial alzheimer’s disease presenilin 1 muta-tions cause alterations in the conformation of presenilin and interactions with amyloid pre-cursor protein. J Neurosci 25:3009–3017
10. Petrasek Z, Eckert HJ, Kemnitz K (2009) Wide-field photon counting fluorescence life-time imaging microscopy: application to pho-tosynthesizing systems. Photosynth Res 102:157–168
11. Lakowicz JR (2006) Principles of fluorescence spectroscopy. Springer, New York
12. Huang S, Heikal AA, Webb WW (2002) Two- photon fluorescence spectroscopy and micros-copy of NAD(P)H and flavoprotein. Biophys J 82:2811–2825
13. Bird DK, Yan L, Vrotsos KM et al (2005) Metabolic mapping of MCF10A human breast cells via multiphoton fluorescence lifetime imaging of the coenzyme NADH. Cancer Res 65:8766–8773
14. Skala MC, Riching KM, Gendron-Fitzpatrick A et al (2007) In vivo multiphoton microscopy
of NADH and FAD redox states, fluorescence lifetimes, and cellular morphology in precan-cerous epithelia. Proc Natl Acad Sci U S A 104:19494–19499
15. Galletly NP, McGinty J, Dunsby C et al (2008) Fluorescence lifetime imaging distinguishes basal cell carcinoma from surrounding unin-volved skin. Br J Dermatol 159:152–161
16. Förster T (1946) Energy transport and fluo-rescence [in German]. Naturwissenschaften 6:166–175
17. Förster T (1948) Zwischenmolekulare ener-giewanderung und fluoreszenz. Annalen Der Physik 437:55–75
18. Förster T (1965) Delocalized excitation and excitation transfer. In: Sinanoglu O (ed) Modern quantum chemistry. Academic Press Inc., New York, pp 93–137
19. Clegg RM (1996) Fluorescence resonance energy transfer. In: Wang XF, Herman B (eds) Fluorescence imaging spectroscopy and microscopy. Wiley, New York, pp 179–251
20. Jares-Erijman EA, Jovin TM (2003) FRET imaging. Nat Biotechnol 21:1387–1395
21. Sekar RB, Periasamy A (2003) Fluorescence resonance energy transfer (FRET) microscopy imaging of live cell protein localizations. J Cell Biol 160:629–633
22. Vogel SS, Thaler C, Koushik SV (2006) Fanciful FRET. Sci STKE re2
23. Piston DW, Kremers GJ (2007) Fluorescent protein FRET: the good, the bad and the ugly. Trends Biochem Sci 32:407–414
24. Sun Y, Wallrabe H, Seo SA et al (2011) FRET microscopy in 2010: the legacy of theodor for-ster on the 100th anniversary of his birth. Chem Phys Chem 12:462–474
25. Wouters FS, Bastiaens PIH (1999) Fluorescence lifetime imaging of receptor tyrosine kinase activity in cells. Curr Biol 9:1127–1130
26. Verveer PJ, Wouters FS, Reynolds AR et al (2000) Quantitative imaging of lateral ErbB1 receptor signal propagation in the plasma membrane. Science 290:1567–1570
27. Elangovan M, Day RN, Periasamy A (2002) Nanosecond fluorescence resonance energy transfer-fluorescence lifetime imaging micros-copy to localize the protein interactions in a single living cell. J Microsc 205:3–14
28. Chen Y, Mills JD, Periasamy A (2003) Protein localization in living cells and tissues using FRET and FLIM. Differentiation 71:528–541
29. Krishnan RV, Masuda A, Centonze VE et al (2003) Quantitative imaging of protein– protein interactions by multiphoton fluores-
Localizing Protein-Protein Interactions Using FLIM
106
cence lifetime imaging microscopy using a streak camera. J Biomed Opt 8:362–367
30. Chen Y, Periasamy A (2004) Characterization of two-photon excitation fluorescence lifetime imaging microscopy for protein localization. Microsc Res Tech 63:72–80
31. Biskup C, Zimmer T, Benndorf K (2004) FRET between cardiac na + channel subunits measured with a confocal microscope and a streak camera. Nat Biotechnol 22:220–224
32. Wallrabe H, Periasamy A (2005) Imaging pro-tein molecules using FRET and FLIM micros-copy. Curr Opin Biotechnol 16:19–27
33. Demarco IA, Periasamy A, Booker CF et al (2006) Monitoring dynamic protein interac-tions with photoquenching FRET. Nat Methods 3:519–524
34. Biskup C, Zimmer L, Kelbauskas T et al (2007) Multi-dimensional fluorescence life-time and FRET measurements. Microsc Res Tech 70:442–451
35. Li Q, Seeger S (2007) Label-free detection of protein interactions using deep UV fluores-cence lifetime microscopy. Anal Biochem 367:104–110
36. Li H, Li HF, Felder RA et al (2008) Rab4 and Rab11 coordinately regulate the recycling of angiotensin II type I receptor as demonstrated by fluorescence resonance energy transfer microscopy. J Biomed Opt 13:031206
37. Murakoshi H, Lee SJ, Yasuda R (2008) Highly sensitive and quantitative FRET-FLIM imag-ing in single dendritic spines using improved non-radiative YFP. Brain Cell Biol 36:31–42
38. Sun Y, Wallrabe H, Booker CF et al (2010) Three-color spectral FRET microscopy local-izes three interacting proteins in living cells. Biophys J 99:1274–1283
39. Li H, Yu P, Sun Y et al (2010) Actin cytoskeleton- dependent rab GTPase-regulated angiotensin type I receptor lysosomal degrada-tion studied by fluorescence lifetime imaging microscopy. J Biomed Opt 15:056003
40. Boutant E, Didier P, Niehl A et al (2010) Fluorescent protein recruitment assay for dem-onstration and analysis of in vivo protein inter-actions in plant cells and its application to tobacco mosaic virus movement protein. Plant J 62:171–177
41. Bu W, Lim KB, Yu YH et al (2010) Cdc42 interaction with N-WASP and toca-1 regulates membrane tubulation, vesicle formation and vesicle motility: Implications for endocytosis. PLoS One 5:e12153
42. Martin-Villar E, Fernandez-Munoz B, Parsons M et al (2010) Podoplanin associates with CD44 to promote directional cell migration. Mol Biol Cell 21:4387–4399
43. Jones PB, Adams KW, Rozkalne A et al (2011) Apolipoprotein E: isoform specific differences in tertiary structure and interaction with amyloid- beta in human alzheimer brain. PLoS One 6:e14586
44. Timpson P, McGhee EJ, Morton JP et al (2011) Spatial regulation of RhoA activity dur-ing pancreatic cancer cell invasion driven by mutant p53. Cancer Res 71:747–757
45. Kumar S, Alibhai D, Margineanu A et al (2011) FLIM FRET technology for drug discovery: automated multiwell-plate high-content analy-sis, multiplexed readouts and application in situ. Chem Phys Chem 12:609–626
46. Sun Y, Day RN, Periasamy A (2011) Investigating protein–protein interactions in living cells using fluorescence lifetime imaging microscopy. Nat Protoc 6:1324–1340
47. Becker W (2005) Advanced time-correlated single photon counting techniques. Springer, Berlin
48. Gerritsen HC, Asselbergs MA, Agronskaia AV et al (2002) Fluorescence lifetime imaging in scanning microscopes: acquisition speed, pho-ton economy and lifetime resolution. J Microsc 206:218–224
49. Becker W, Bergmann A, Hink MA et al (2004) Fluorescence lifetime imaging by time- correlated single-photon counting. Microsc Res Tech 63:58–66
50. Becker W, Bergmann A, Biskup C (2007) Multispectral fluorescence lifetime imaging by TCSPC. Microsc Res Tech 70:403–409
51. Elangovan M, Wallrabe H, Chen Y et al (2003) Characterization of one- and two-photon exci-tation fluorescence resonance energy transfer microscopy. Methods 29:58–73
52. Talbot CB, McGinty J, Grant DM et al (2008) High speed unsupervised fluorescence lifetime imaging confocal multiwell plate reader for high content analysis. J Biophotonics 1: 514–521
53. Krishnan RV, Saitoh H, Terada H et al (2003) Development of a multiphoton fluorescence lifetime imaging microscopy system using a streak camera. Rev Sci Instrum 74:2714–2721
54. Colyer RA, Lee C, Gratton E (2008) A novel fluorescence lifetime imaging system that opti-mizes photon efficiency. Microsc Res Tech 71:201–213
55. Gadella TWJ Jr, Jovin TM, Clegg RM (1993) Fluorescence lifetime imaging microscopy (FLIM): Spatial resolution of microstructures on the nanosecond time scale. Biophys Chem 48:221–239
56. Buranachai C, Kamiyama D, Chiba A et al (2008) Rapid frequency-domain FLIM spin-ning disk confocal microscope: lifetime resolu-
Yuansheng Sun and Ammasi Periasamy
107
tion, image improvement and wavelet analysis. J Fluoresc 18:929–942
57. Chen YC, Clegg RM (2011) Spectral resolu-tion in conjunction with polar plots improves the accuracy and reliability of FLIM measure-ments and estimates of FRET efficiency. J Microsc 244:21–37
58. Goedhart J, van Weeren L, Hink MA et al (2010) Bright cyan fluorescent protein vari-ants identified by fluorescence lifetime screen-ing. Nat Methods 7:137–139
59. Periasamy A, Skoglund P, Noakes C et al (1999) An evaluation of two-photon excitation versus confocal and digital deconvolution fluo-rescence microscopy imaging in xenopus mor-phogenesis. Microsc Res Tech 47:172–181
60. Verveer PJ, Squire A, Bastiaens PIH (2000) Global analysis of fluorescence lifetime imag-ing microscopy data. Biophys J 78: 2127–2137
61. Verveer PJ, Bastiaens PIH (2003) Evaluation of global analysis algorithms for single fre-quency fluorescence lifetime imaging micros-copy data. J Microsc 209:1–7
62. Pelet S, Previte MJ, Laiho LH et al (2004) A fast global fitting algorithm for fluorescence lifetime imaging microscopy based on image segmentation. Biophys J 87:2807–2817
63. Grecco HE, Roda-Navarro P, Verveer PJ (2009) Global analysis of time correlated sin-gle photon counting FRET-FLIM data. Opt Express 17:6493–6508
64. Gryczynski I, Bharill S, Luchowski R et al (2009) Nonlinear curve-fitting methods for time-resolved data analysis. In: Periasamy A, Clegg RM (eds) FLIM microscopy in biology and medicine. Chapman and Hall/CRC, Boca Raton, pp 341–369
65. Redford GI, Clegg RM (2005) Polar plot rep-resentation for frequency-domain analysis of fluorescence lifetimes. J Fluoresc 15:805–815
66. Clayton AH, Hanley QS, Verveer PJ (2004) Graphical representation and multicomponent analysis of single-frequency fluorescence lifetime imaging microscopy data. J Microsc 213:1–5
67. Hanley QS, Clayton AH (2005) AB-plot assisted determination of fluorophore mixtures in a fluorescence lifetime microscope using spectra or quenchers. J Microsc 218:62–67
68. Digman MA, Caiolfa VRM, Zamai VR et al (2008) The phasor approach to fluorescence lifetime imaging analysis. Biophys J 94: L14–L16
69. Stringari C, Cinquin A, Cinquin O et al (2011) Phasor approach to fluorescence lifetime microscopy distinguishes different metabolic states of germ cells in a live tissue. Proc Natl Acad Sci U S A 108:13582–13587
70. Sun Y, Booker CF, Kumari S et al (2009) Characterization of an orange acceptor fluo-rescent protein for sensitized spectral fluores-cence resonance energy transfer microscopy using a white-light laser. J Biomed Opt 14: 054009
71. Koushik SV, Blank PS, Vogel SS (2009) Anomalous surplus energy transfer observed with multiple FRET acceptors. PLoS One 4:e8031
72. Boens N, Qin W, Basaric N et al (2007) Fluorescence lifetime standards for time and frequency domain fluorescence spectroscopy. Anal Chem 79:2137–2149
73. Cramer LE, Spears KG (1978) Hydrogen bond strengths from solvent-dependent life-times of rose bengal dye. J Am Chem Soc 100:221–227
74. Thaler C, Koushik SV, Blank PS et al (2005) Quantitative multiphoton spectral imaging and its use for measuring resonance energy transfer. Biophys J 89:2736–2749
75. Koushik SV, Chen H, Thaler C et al (2006) Cerulean, venus, and VenusY67C FRET refer-ence standards. Biophys J 91:L99–L101
76. Rizzo MA, Springer GH, Granada B et al (2004) An improved cyan fluorescent protein variant useful for FRET. Nat Biotechnol 22:445–449
77. Nagai T, Ibata K, Park ES et al (2002) A vari-ant of yellow fluorescent protein with fast and efficient maturation for cell-biological applica-tions. Nat Biotechnol 20:87–90
78. Tang QQ, Lane MD (2000) Role of C/EBP homologous protein (CHOP-10) in the pro-grammed activation of CCAAT/enhancer- binding protein-beta during adipogenesis. Proc Natl Acad Sci U S A 97:12446–12450
79. Tang QQ, Lane MD (1999) Activation and centromeric localization of CCAAT/enhancer- binding proteins during the mitotic clonal expansion of adipocyte differentiation. Genes Dev 13:2231–2241
80. Day RN, Voss TC, Enwright JF 3rd et al (2003) Imaging the localized protein interactions between pit-1 and the CCAAT/enhancer binding protein alpha in the living pituitary cell nucleus. Mol Endocrinol 17:333–345
81. Buranachai C, Tong B, Clegg R et al (2009) General concerns of FLIM data representation and analysis. In: Periasamy A, Clegg RM (eds) FLIM microscopy in biology and medicine. Chapman and Hall/CRC, Boca Raton, pp 323–340
82. Sun Y, Periasamy A (2010) Additional correc-tion for energy transfer efficiency calculation in filter-based forster resonance energy transfer microscopy for more accurate results. J Biomed Opt 15:020513
Localizing Protein-Protein Interactions Using FLIM
109
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_7, © Springer Science+Business Media New York 2015
Chapter 7
Analysis of Biomolecular Dynamics by FRAP and Computer Simulation
Bart Geverts , Martin E. van Royen , and Adriaan B. Houtsmuller
Abstract
We present a Monte Carlo simulation environment for modelling complex biological molecular interaction networks and for the design, validation, and quantitative analysis of FRAP assays to study these. The pro-gram is straightforward in its implementation and can be instructed through an intuitive script language. The simulation tool fi ts very well in a systems biology research setting that aims to maintain an interactive cycle of experiment-driven modelling and model-driven experimentation: the model and the experiment are in the same simulation. The full program can be obtained by request to the authors.
Key words FRAP , FRET , Modelling , Simulation , Dynamics , Quantitative fl uorescence assay
1 Introduction
Cell function can be largely, if not fully, described by a set of well- orchestrated and highly compartmentalized biomolecular transformations [ 1 ]. Such transformations may be covalent modi-fi cations (e.g., enzymatic addition or removal of phosphates), intermolecular associations (e.g., heterodimerization or homodi-merization, complex formation), intracellular (re)localizations (e.g., nucleocytoplasmic shuttling, formation of DNA damage-induced foci), or synthesis from small compounds (e.g., DNA rep-lication, protein synthesis), all together acting in a highly regulated manner to perform their simple or complex tasks in the cell. Importantly, in the context of this chapter, photobleaching of a fl uorescent tag is also a molecular transformation. Although it is not relevant for cell function, it can be used to extract information on protein mobility, which is highly relevant in cell biological or biophysical research (e.g., [ 2 , 3 ]).
Mathematical or “physicochemical” modelling of cellular f unction aims to describe these transformations in terms of “ equations derived from established physical and chemical theory” [ 1 ], usually culminating in extensive sets of partial or ordinary
110
differential equations (PDEs and ODEs) [ 4 ]. These approaches have substantially contributed and are still contributing to increas-ing biological knowledge at the systems level [ 5 ]. Similarly, ODE- and/or PDE- modelling approaches have enabled quantitative analysis of fl uorescence recovery after photobleaching (FRAP) data [ 6 – 9 ]. In general, a high level of mathematical skill is needed to design, validate, calibrate, and fi nally solve these sets of interdepen-dent equations, which in highly complex scenarios are not solvable at all. Moreover, modifying or expanding models, or introducing spatial confi nements like compartmentalization in irregular struc-tures, prompts revision of the entire set, which, if still possible at all, is in general time consuming and requires expert mathemati-cians to be permanently involved in the project. Therefore, in many biology labs, where it is diffi cult to maintain a team of mathemati-cians or physicists on a stable basis, it is hard to create a sustainable modelling environment that can be used on a permanent and r outine basis.
Monte Carlo simulation provides an alternative approach to model complex systems involved in cellular function (or methods like FRAP to study them) that circumvents the intensive use of PDEs and ODEs. In the straightforward MC simulation approach we present here, it is assumed that (molecular) events in a living cell have a certain probability to occur within a specifi c short period of time. Examples of such events are binding of two molecules or molecule complexes, release of a molecule from a complex, move-ment of a molecule by random diffusion, degradation of a molecule, or bleaching of a fl uorescent dye in an intense laser beam. In MC simulation, these events occur or do not occur during small simu-lated time steps dependent on the result of generating a random number and the probability distribution related to the nature of the event. The implementation is therefore relatively simple. All the program has to do is go through cycles representing small peri-ods of time, in which for each molecule, a couple of random num-bers are generated on the basis of which is decided whether the molecule, for instance, binds, unbinds, moves, or is photobleached. Similar to the straightforward implementation of the simulation program, the required input is fully intuitive. The user defi nes mol-ecules and assigns properties like, “it moves by diffusion,” “it is able to bind to this or that other molecule,” “it is accumulating in nuclear foci,” or “it is tagged with a fl uorescent dye.” The only mathematical or rather statistical input required are the probabili-ties that molecules actually effectuate these properties, i.e., the probability that they bind, unbind, move, or bleach during the time span simulated in each cycle of the program.
In this chapter, we present a straightforward Monte Carlo (MC)-based simulation tool to model biomolecular interactions and to subject those to FRAP assays to generate simulated FRAP curves and use the output to fi t experimental data.
Bart Geverts et al.
111
The program can be instructed through a simple and intuitive script language and can therefore be used by anyone with basic knowledge of biomolecular processes. The initial motivation to develop the program was the limited availability of solutions to accurately quantitate FRAP curves using formulas or equations amenable in a biology setting. In addition, the simple modelling environment appeared to be a surprisingly strong educational tool in that it provides the possibility to understand FRAP curves and how they are infl uenced by specifi c experimental conditions like, for instance, cell size and shape. In addition, the tool is well suited to design and optimize FRAP experiments and also its combina-tion with fl uorescence resonance energy transfer (FRET) [ 10 ], in relation to the specifi c properties of the biological system under surveillance.
In the following, we will fi rst provide a brief description of the FRAP method and an overview of published quantitative FRAP analysis methods. Second, the implementation and use of the simulation tool in understanding and quantitatively analyzing FRAP and FRET experiments will be described. We then will present results of studying important properties of FRAP and combined acceptor photobleaching FRET and FRAP using the simulation tool.
2 Materials
The Java-based Monte Carlo FRAP simulation tool can be obtained by request to the authors.
3 Methods
FRAP is based on the property of fl uorescent molecules that they are photobleached, i.e., irreversibly made nonfl uorescent, when exposed to high-intensity excitation light. This often undesired property can be put to use when a small volume within a larger volume is photobleached, and the recovery of fl uorescence due to movement of unbleached mobile molecules into the bleached region is followed in time. The recovery velocity and fi nal level then represent their average displacement per unit time and the fractions of mobile and immobile molecules. FRAP is especially powerful in a situation where the system under surveillance is stud-ied in active (“wild type”) state and compared with passive or mod-ifi ed state, so that a difference in mobility can be attributed to protein function. For instance, in DNA repair, one can compare protein mobilities in unchallenged cells with those in cells exposed to a DNA damage-inducing agent. In transcriptional regulatory systems, like the nuclear hormone receptors, one can compare
2.1 Monte Carlo FRAP Simulation Tool
3.1 FRAP Analysis
Monte Carlo Based FRAP Analysis
112
mobilities of ligand-induced and inactive receptors such as either non-liganded receptors, receptors inhibited by antagonistic ligands, or receptor mutants lacking DNA-binding properties [ 11 ].
There are two straightforward ways to perform FRAP analysis, i.e., visual inspection of the curves or semiquantitative analysis, by calculating the half-life of fl uorescence recovery. In addition, a per-manently immobilized fraction can be readily estimated from the incomplete recovery of fl uorescence. Although this straightfor-ward approach is by itself useful, as can be seen from the wealth of mechanistic insight that has been gained from many studies, half- lives provide information on overall mobility but do not allow detailed quantitative analysis. For example, when relatively small fractions of proteins are engaged in immobilizing DNA-binding events, half-life will be only slightly increased compared to fully free protein. Therefore, half-life is not easily translated into quan-titative estimates of bound fractions or residence times in immobile DNA–protein complexes. To obtain more quantitative informa-tion from FRAP analysis, sophisticated analytical models have been established. The parameters that one aims to quantitatively assess are effective on- and off-rates to and from immobile DNA–protein complexes, which can be used to calculate the DNA-associated fraction and residence time in that state. When on-rates increase at constant off-rate, the immobile fraction also increases. When off- rates increase, the immobile fraction becomes smaller at constant on-rate, and, in addition, the residence time becomes shorter.
Apart from the parameters to be extracted from FRAP curves, mathematical models should include a number of fi xed parameters representing microscopic properties, like shape, size, and intensity profi le of the laser beam focused by the objective lens. In addition, FRAP curves may be infl uenced by properties like cell shape and size or heterogeneous distribution of the factors under surveil-lance. Although most of these can be obtained experimentally, incorporating them in mathematical models often complicates the situation to an extent where solving differential equations involved is hard or even impossible [ 12 ]. Therefore, a number of simplifi ca-tions have been applied.
One complicating factor in FRAP modelling is the conical shape of high aperture lenses that are in general used for single-cell imaging. Initially, the photobleached region was modelled as a cylindrical uniform profi le [ 13 ] or a cylindrical region with a radial Gaussian distribution [ 14 – 17 ]. However, although a cylin-drical profi le is justifi ed when a low numerical aperture (NA) lens is used, a double-cone profi le may be more accurate, even for low NA lenses [ 3 , 18 , 19 ]. Furthermore, unlike a cylindrical bleaching profi le, the double-cone profi le also implies a signifi cant depen-dency of the three-dimensional situation, because of the axial Gaussian distribution of the laser beam [ 19 – 21 ]. These and other experimental conditions, like photophysical properties of the fl uorophores in use (e.g., blinking) and also like the duration of
Bart Geverts et al.
113
bleaching and the consequences of diffusion during bleaching, should be carefully considered in the development of a kinetic model [ 7 , 12 , 22 , 23 ].
Another potential complication in the development of mathe-matical models is a heterogeneous distribution of target sites of the labelled proteins. Most models assume a homogeneous distribu-tion of binding sites, ignoring the typical spatial organization of nuclear processes like gene transcription or DNA replication [ 15 , 23 ]. A more accurate binding model would then be one which takes the actual inhomogeneous distribution of binding sites into account by using the acquired imaging data [ 6 ].
Because of the complications described above, which in some instances demand simplifi cation to an unwanted extent, we decided to develop a Monte Carlo-based simulation environment that avoids the necessity of designing and solving differential equations.
The simulation program we developed consists of three parts: (1) a core program which runs the simulation using basic algorithms for diffusion, binding, and bleaching (Figs. 1 and 2 ); (2) a user interface which visualizes the simulation, allows the user to modify input parameters, and provides output (Fig. 3 ); and (3) a script language which allows to design models, both of molecular inter-actions and of FRAP and FRET assays to study those.
The core program is implemented in a straightforward and easily maintainable manner. It translates the user-defi ned scripts imported through the user interface and sends output to that interface. The program goes through cycles representing small time steps, usually in the order of 10–100 ms. In each cycle, the behavior of each molecule present at that time is simulated by gen-erating random numbers on the basis of which a decision is taken whether the molecule performs specifi c actions. These decisions are based on probability distributions related to the events happen-ing to the molecule or the actions taken by it. The underlying algorithms dealing with this are described in detail below, where we present the script language, in which compartments, molecules, labels, and their properties are defi ned. In the user interface, the simulations are visualized, graph and text output relevant for the simulated system is provided, and a number of model-related parameters can be modifi ed by the user (Fig. 3 ).
In order to provide a versatile modelling environment that can be used without the need to write new or modify existing program code, an intuitive script language was developed. The script lan-guage consists of a list of defi nitions of PARTICLES (e.g., proteins, phosphates, DNA), LABELS (gfp, yfp, cfp), COMPLEXES (com-binations of particles), and COMPARTMENTS (e.g., nucleus, cytoplasm, laser cone), as well as a series of rules by which specifi c properties are assigned to the defi ned complexes. Rules like BIND
3.2 Monte Carlo Simulation Tool
3.2.1 Script Language
Monte Carlo Based FRAP Analysis
114
Fig. 1 General scheme for Monte Carlo simulation of FRAP on nuclear proteins. Overview of th e COMPARTMENTS that can be defi ned in the script language.
Bart Geverts et al.
115
COMPLEX
PARTICLE
Property List In: Nucleus Diff Coeff. D Binds p, , kon
Releases: K, koff
LABEL
A K
A
K
pNameColorSize
Type State
ParticlesLabels
PropertiesLocation
gfp yfp cfp
gfpon
Locationx, y, z
a
b
c
Fig. 2 Overview of the different components in the simulation. ( a ) PARTICLES are the basic building blocks from which complexes are created. They represent, for instance, proteins, phosphates, or DNA but only acquire properties when added to a complex. ( b ) LABELS are similar to particles in that they have to be added in a complex. Labels have a reserved name (gfp, yfp, or cfp) and have one property, representing the state of their fl uorescent behavior: on, dark, or bleached. ( c ) COMPLEXES are composed of particles and (optional) labels. When formed, they acquire a property list from a blueprint composed on the basis of the user-defi ned rules in the script language ( see text). They also have x -, y -, and z -coordinates represent-ing their location
Fig. 1 (continued) The nucleus is represented as an ellipsoid ; the laser double cone is a dedicated compartment that also contains values representing local laser intensities which are stored in a text fi le that is translated by the program into a 3-D array. ( a ) Schematic drawing of a cell nucleus ( ellipsoid ) containing randomly distrib-uted COMPLEXES ( spheres ). Random Brownian motion ( inset ) is simulated on the basis of the Einstein–Smoluchowski equation ( see text). ( b ) Simulation of binding to randomly distributed immobile target sites in the DNA ( inset ) is simulated by evaluating a chance to bind or to release based on simple binding kinetics, where the ratio between on- and off-rate constants ( k on and k off ) equals the ratio between the number of immobile and mobile molecules. ( c ) Photobleaching is simulated by evaluating a chance to get bleached based on the intensity profi le of the laser beam. (Figure adapted from [ 3 ])
Monte Carlo Based FRAP Analysis
Fig. 3 Graphical interface of the Monte Carlo simulation. The user interface contains two windows; the top window contains the visualization of the COMPLEXES, either fl uorescent ( red dot with green G ) or nonfl uorescent ( red dot only ) and bound ( red dot with white dot ) or freely mobile ( red dot ). The two graphs show the number of fl uorescent COMPLEXES in the measured strip (FRAP curve) ( top ) and the sizes of the mobile ( red curve ), shortly bound ( green curve ), and long-bound ( yellow curve ) population in time. The buttons enable export of numerical data in text fi les. The lower window is used to modify the parameters of the model using indicated variables, including diffusion coeffi cient ( Dc ) and binding properties (both short (Sh) and long (Lo) immobile fractions) of the molecules. The binding parameters are set as immobile fractions (Imf) with corresponding residence times (Rt). The software automatically calculates the corresponding probabilities of binding ( P on ) or release ( P off ). Depending on the fl uorophore used in FRAP analysis, the blinking parameters ( P blink and P fl uo ) should be adjusted. Furthermore, general parameters like nuclear position and size (Nucl), the laser profi le for FRAP analysis, and the position and size of the bleached (Bleach) and measured (Measure) areas, in addition to the bleach intensity and duration in cycles, the length of the prebleach period, and the total length of the FRAP curve can be adjusted
117
and RELEASE or ADD and REMOVE defi ne which molecules interact, including corresponding rate constants ( k on and k off ), which are translated in the program to probabilities that are evalu-ated on the basis of numbers generated by a random generator. When the script is translated, the program adds all rules assigned to a specifi c complex in a property list which serves as a blueprint that is read when the specifi c complex is actually formed at the start or during the simulation (Fig. 2 ).
Below, the script language will be explained together with the most important algorithms used in the simulation in the context of the script defi nition or rule being discussed.
COMPARTMENT(<NAME>, Compartment type, Fill-fi le.txt, Parameter list).
Compartments are predefi ned volumes, which are specifi ed by their reserved-type name followed by a number of parameters that depend on the compartment type. Simple volumes like rectangular boxes (Box) or ellipsoids (Ellipsoid) require the x -, y -, and z - coordinates of their centers and diameters. The laser cone and bleach area (Bleach area) are special volumes that in addition to center and size coordinates, require specifi cation of a formalized text fi le which contains a table with laser intensities at different positions in the laser.
EXCLUDE( compartment1 , compartment2 );
Compartments like the cytoplasm that contains other com-partments like the nucleus are defi ned by defi ning two ellipsoids and then exclude the smaller from the larger.
Examples: COMPARTMENT(nucleus, Ellipsoid, 0,0,0,10,15,5). COMPARTMENT(cytoplasm, Ellipsoid, 0,0,0,20,20,7). EXCLUDE(nucleus, cytoplasm). COMPARTMENT (stripbleach, Bleach area, Laserprofi le40X- -
1.3.txt, 0,0,0,20,1).
PARTICLE (<name>, Colour, Shape, Size).
PARTICLES (Fig. 2a ) are the basic building blocks representing units from which COMPLEXES (described below) can be constructed. Particles have a user-defi ned name, for reference in defi ning complexes, as well as drawing instructions for visu-alization in the user interface. The most basic complex con-tains one particle ( see Note 1 ).
3.2.2 Script Defi nitions
Monte Carlo Based FRAP Analysis
118
Examples: PARTICLE(transcriptionfactorX, RED, 2). PARTICLE(kinaseY, RED, 2). PARTICLE(phosphate, BLUE, 1). PARTICLE(promotersequence, WHITE, 1).
COMPLEX(particle 1 -particle 2 -…-particle N, Compartment). COMPLEXES (Fig 2c ) consist of user-defi ned combinations of
particles, which are specifi ed by the particle names (see above) separated by hyphens. All possible combinations of particles that exist at the start of the simulation or that may form during the simulation have to be defi ned. The simplest complex con-sists of a single particle. The compartment in which the complex is present is the only property defi ned here. Mobility and interaction properties are assigned by the rules explained in the next section.
Examples Complexes formed during phosphorylation: COMPLEX(factorX, nucleus). COMPLEX(kinaseY, nucleus). COMPLEX(factorX-kinaseY, nucleus). COMPLEX(phosphate-factorX-kinaseY, nucleus). COMPLEX(phosphate-factorX, nucleus).
ADDLABEL(Label, complex). One or more LABELS (Fig. 2b ) can be added to complexes.
Currently, three fl uorescent labels are supported in the pro-gram, gfp, yfp, and cfp. Labels are different from particles in that they have one property of their own (whereas particles have no properties when not in a complex), i.e., they can be in three states: on state, dark state, and bleached. Fluorescent labels are different from particles in that they do not alter the complex property list.
CREATECOMPLEXES (complex, copy number, compartment); Here, the number of copies of a specifi c complex present at the
start of the simulation is specifi ed. Note that only after this defi nition, actual complexes are formed. When complexes are created, their property list is read from a blueprint that was composed from the rules specifi ed for each complex type while the script was read.
Example:
CREATECOMPLEXES (A, 10,000, nucleus).
Bart Geverts et al.
119
Rules assign properties to COMPLEXES. When the script is read, the blueprint property list of each defi ned complexes is (which after complex defi nition contains one property, the com-partment it is in) built up from properties derived from the rules ( see also Fig. 3 ).
MOVE(complex, Diffusion Coeffi cient, Compartment). The MOVE rule adds a diffusion coeffi cient to the complex
blueprint property list. In the simulation, diffusion is simulated on the basis of the Einstein–Smoluchowski equation which defi nes the mean square displacement MSD of a pool of particles as MSD = 2nDt , where D is the diffusion coeffi cient, t is time, and n is the number of dimensions in which the particles diffuse. In the simulation core program, 3-D diffusion of individual complexes is simulated by moving each complex in x -, y -, and z -direction, over a distance obtained by generating a random number R from a uniform distribution, and subsequently computing G( R ) where G is a cumulative inverse Gaussian distribution with μ = 0 and σ 2 = 2 Dt ( see Note 2 ).
Anomalous diffusion can be simulated by using the ADD and REMOVE rule described below, adding and removing a (virtual) particle, with high on- and off-rate constants. The defi nitions and rules needed are described in an example below, where the ADD and REMOVE rules are described.
BIND(complex 1 , complex 2, Compartment, rate constant, ). RELEASE(complex 1 , complex1-complex 2 , rate constant, Compartment).
The BIND rule assigns the property that complexes may bind to each other when they collide, and in which compartment this property is valid. A rate constant, k on , is specifi ed in s −1 . The chance of binding upon collision between complexes is derived by the pro-gram dependent on the duration of the simulated time step. In a relatively simple binding algorithm, the user-specifi ed rate constant and the distance between two molecules under consideration together determine the chance that the two actually bind. We applied two different strategies for calculating the binding chance. In the fi rst simple approach, the binding chance derived from k on is applied when the distance between two molecules is smaller than a user-defi ned threshold. In a second more complex approach, the chance of collision is retrieved from a 3-D look-up table that gives the chance that two molecules with diffusion coeffi cients D1 and D2 collide when they are at distance d . The chance in each table cell ( D1, D2, d ) was estimated by performing a large number of MC simulations at very small time steps of 0.1 ms, for each table cell, and counting the times the two molecules collide, i.e., come closer than 1 nm.
3.2.3 Script Rules
Monte Carlo Based FRAP Analysis
120
The RELEASE rule assigns the property to a complex that part of the complex may detach and in which compartment this prop-erty is valid. The rate constant, k off , is specifi ed here in s −1 and trans-lated by the program to the actual chance per simulated time step.
Examples: BIND (A, B, nucleus, 0.01). RELEASE (A, A-B, nucleus, 0.01).
ADD (complex 1 , complex 2 , rate constant, compartment). REMOVE (complex 1 , complex 2 , rate constant, compartment). ADD and REMOVE are similar to BIND and RELEASE, with the
very important difference that added or removed complexes do not exist independently in the simulation, but are created by ADD, and taken up in the complex, or destroyed by REMOVE. The rate constant is therefore in both cases in s −1 , that is, independent of concentration. ADD and REMOVE rules can and should be used only when variations in spatial distribution and concentration of the added or removed com-plex are negligible during the simulation, i.e., when its concen-tration is in excess compared to the complex it is added to or removed from, and no local depletion or accumulation occurs. This is, for instance, useful for abundant phosphates or other small compounds.
Examples
Addition of a phosphate by a kinase and removal by a phosphatase: ADD (phosphate, txnfactorX-KinaseY, nucleus, 0.5). REMOVE (phosphate, txnfactorX-phosphate-PhosphataseY,
nucleus, 0.5). Simulation of anomalous diffusion: ADD (anomdiffparticle, A, nucleus, 0.9). REMOVE (anomdiffparticle, A-anomdiffparticle, nucleus, 0.9). MOVE (A, 5, nucleus). MOVE (A-anomdiffparticle, 0.01, nucleus).
BLEACH (start time, stop time, label, bleach area). If a molecule contains a fl uorescent label, this label can be bleached
by a laser. We have implemented two approaches to defi ne the laser intensity profi le of the laser. In a simple approach, a theo-retical intensity distribution was used based on a Gaussian dis-tribution of intensities. In a second approach, the laser profi le was experimentally determined by photobleaching a homoge-neous fl uorescent test slide with a focused stationary laser beam and measuring the bleach intensities at different optical
Bart Geverts et al.
121
sections spaced 1 μm. These values were then translated to laser intensities. Bleaching will take place during the simula-tion between the specifi ed start and stop times in the bleach area defi ned in the defi nition section. For example:
COMPARTMENT (spotbleach, Bleach area, Laserfi le.txt, 0,0,0). BLEACH (100,101, gfp, spotbleach). bleaches a spot at (0, 0, 0), usually the center of an ellipsoid repre-
senting the nucleus, during two simulated time steps.
The simulation software was originally developed to quantitatively analyze FRAP applied to proteins in the cell nucleus. In addition, it provides a useful tool to study the infl uence of experimental con-ditions like the shape and size of the volume in which the proteins under surveillance reside as well as of the shape, size, and position of the bleach area. These two applications will be studied in the next two sections. In a third section, we will demonstrate the combination of acceptor photobleaching FRET and FRAP, which generates highly complex data.
There are basically three ways to fi t simulation-generated FRAP curves to a given experimental curve:
1. Trial-and-error interactive search (Fig. 4 ). This approach requires some experience with the specifi c FRAP method. For instance, in a strip-FRAP experiment applied to the cell nucleus, which we routinely apply, one can start comparing an inactive
3.3 Simulation of FRAP and FRET Assays
3.3.1 FRAP Simulation, Fitting Experimental Data
Fig. 4 Applications of FRAP in biological model systems. Two examples of the use of FRAP simulation of nuclear proteins ( a , b ), the androgen receptor (AR) ( a ) with its DNA-binding defi cient mutant ( b ). ( a ) Like most nuclear proteins, the transcription factor AR shows a high mobility and in addition to transient immobilizations. The wild-type AR curve was best fi tted to a curve representing a Dc of 1.9 μm 2 /s and fraction of 35 % of the ARs immobilized for 8 s. This immobilization is due to DNA binding as is shown by a FRAP analysis of ( b ), a DNA- binding defi cient AR, which fi tted best to a scenario of freely mobile molecules (Dc = 1.9 μm 2 /s). Experimental curves are red ; simulated curves are black . The fi t was performed by interactive trial-and-error modelling
Monte Carlo Based FRAP Analysis
122
situation, for instance, a non-DNA-binding mutant androgen receptor ( AR) with an active situation, the activated wild-type AR. Then assume that the inactive molecules are freely mobile and try to fi t with models of free diffusion (Fig. 4b ). Then, when fi tting the active situation, fi x the best fi tting diffusion coeffi cient and introduce an immobile fraction. This fraction can initially be estimated roughly, by checking how far recov-ery has proceeded at the time the freely mobile protein is fully recovered. The residence time can initially be estimated by checking time until full recovery (Fig. 4a ).
2. Creating a large database in which all parameters to be fi tted are systematically varied. We have extensively applied this approach in many studies investigating a variety of nuclear model systems [ 24 – 34 ]. Briefl y, the experimenter fi rst provides a number of fi xed parameters, including the experimentally derived size of the nuclei, the FRAP approach, and the lens used for photobleaching and recording. Then a database is cre-ated where diffusion and DNA-binding properties (on- and off-rate constants) are systematically varied. Note that usually the rate constants are translated to intuitively better under-standable immobile fraction and residence time (cf. Fig. 1b ). In some cases, we also fi tted a second pair of on- and off-rate constants representing short immobilization events [ 31 , 33 , 35 ]. When the database is created, a set of 20–50 simulated curves that fi t best to the experimental curve are generated. In practice, there will be two or sometimes more different sce-narios represented of which usually only one is realistic in the view of the system under surveillance. From the 5–10 best fi t-ting curves, the diffusion and binding parameters are then averaged and considered best estimates.
3. Find the best curve automatically by search algorithms. Typically, these approaches start at a number of estimated or randomized solutions; abort simulation as soon as the gener-ated curve deviates from the experimental; move to a solution nearby, in any direction of the parameter space; check if the fi t is worse; if so, move back, step in another direction, and so on, until a (local) best fi t is found. By starting at several points in parameter space, fi nding an erroneous local fi t only is avoided. Although the latter undoubtedly will in the end be the method of choice, it is the least straightforward and will not give better results than generating databases and by brute force fi nd the best fi t.
As discussed above, issues like the conical shape of the laser beam or the confi ned space in which diffusion and immobilization take place may strongly infl uence FRAP curves, complicating analytical approaches to quantify them. Below, we present results of studying
3.3.2 FRAP Simulation, Studying FRAP Principles
Bart Geverts et al.
123
a number of experimental conditions that contribute to the shape of the curve, the recovery time, and/or the degree to which recov-ery takes place these aspects. The results of all simulation below are presented after two different types of normalization, i.e., expressed relatively to the average prebleach value (further indicated as “nor-malized to prebleach”) and expressed relative to intensity immedi-ately after the bleach (0) and at full recovery (1) (further referred to as “normalized as for bleach depth and fi nal recovery” ). To get an impression of the results of a standardized bleach experiment on nuclear proteins, we fi rst generated FRAP curves of proteins in an ellipsoid volume representing the nucleus at increasing diffusion rates (Fig. 5a ) and varying immobile fractions and various resi-dence times in the immobile state (Fig. 5b ).
In Fig. 5c , the results are shown of applying FRAP to a nucleus that has two large foci (for instance, nucleoli) in which a signifi cant fraction of the protein is accumulating. The bleach area was chosen in between the accumulation, an experiment that is often applied to determine diffusion coeffi cient of the free pool. However, as can be seen, one has to be cautious when doing so, since a secondary recovery due to exchange of the free pool and the accumulation will occur. If this exchange is very slow, free diffusion can be sepa-rated from exchange by analyzing the fi rst part of the curve. However, when exchange is fast, this will not be possible, and full modelling of the obtained curve including the exchange is required to determine diffusion coeffi cient of the free pool.
In many FRAP analyses, a straightforward “inert” DNA- binding model is chosen where the distribution of residence times is exponential, representing a scenario where immobile molecules have a constant chance to get mobilized. The characteristic resi-dence time then is 1/ k , where k is the time constant of the expo-nential distribution of residence times e −kt (Fig. 5e , left panel). However, it is conceivable that residence times are variable over time, especially when initial binding factors like androgen recep-tors or DNA damage sensors are studied. Initially, these factors may have a fair chance to release after DNA binding, but after binding of additional factors, the complex may be stabilized, low-ering the off-rate constant. While the growing complex becomes more stable, it may after completion rapidly achieve its goal, i.e., launching transcription or repairing DNA damage, and the com-plex likely rapidly falls apart (Fig. 5d ). This more complex scenario has a completely different distribution of binding times but can hardly be distinguished using FRAP (Fig. 5e ) ( see Note 3 ).
The initially most popular way to perform photobleaching experiments was to bleach a spot using a stationary laser to bleach the spot [ 36 , 37 ]. Simulation of FRAP curves from differing spot locations shows that recoveries are much slower when a spot close to the edge of the nucleus is bleached compared to when the spot is approximately in the center (Fig. 6a ). Interestingly, this differ-
Monte Carlo Based FRAP Analysis
Fig. 5 Computer simulation aided analysis of diffusion and binding kinetics. Simulated spot-FRAP curves generated with ( a ) variable diffusion (2, 5, and 10 μm 2 /s) and ( b ) variable immobile fraction (0, 33, 50, and 60 %) with fi xed 15 s residence time ( D = 5 μm 2 /s). ( c ) Curves representing a scenario where accumulations of the investigated protein exist at a distance from the bleached area compared to free diffusion in a homoge-neously distributed scenario. ( d ) Complex DNA-binding scenario where the stability of the complex changes over time ( see text). ( e ) Left panel : distribution of binding times for two scenarios; red curve , simple binding model with constant k off (e −koff t ); blue curve , complex binding model with varying k off . . Right panel : FRAP curves corresponding to the simple ( red ) and complex ( blue ) binding models. Curves were normalized to prebleach only ( left column ) and normalized for bleach depth and full recovery ( right column , see text)
Fig. 6 Spot position and size variation affects FRAP curves. ( a and b ) Variable spot positions affect the FRAP curves of free diffusion ( D = 2 μm 2 /s) ( a ) and a scenario with 33 % immobile fraction with a 15 s residence time ( b ). More polar positioned spots along the longitudinal axis result in a slower recovery of fl uorescence pre-dominantly in the fi rst part of the FRAP curve suggesting a slower diffusion ( a ) rather than more or longer immobilization ( b ). ( c , d ) In contrast, variable spot sizes result predominantly in an increased permanently bleached fraction which can mistakenly be interpreted as an incomplete recovery suggesting a long immobile fraction in scenarios of free diffusion ( D = 2 μm 2 /s) with ( d , left column ) or without ( c , left column ) and a 33 % immobile fraction with 15 s residence time. In addition to the increased permanently bleached fraction, increased spot sizes introduce a delayed recovery in the fi rst part of a FRAP curve, suggesting a slower diffu-sion coeffi cient ( c , right column ). Curves were normalized both prebleach only ( left column ) and a combination of the intensity immediately after bleaching and after compete recovery ( right column )
126
ence is smaller when an immobile fraction is present (Fig. 6b ). Increasing spot size has a considerable effect on FRAP curves, obviously on the fraction bleached (Fig. 6c , left panel), but also on the shape of the recovery curve (Fig. 6c , right panel). The latter is fully abolished when a (large) transiently immobile fraction is pres-ent, showing that changing bleach area is a method to distinguish freely mobile proteins from transiently immobilized (Fig. 6d ).
The size and shape of the cell nucleus may also affect the shape of FRAP curves, dependent on which FRAP approach is used (Fig. 7 ). Nuclei with increasing length and constant width and height as well as nuclei of varying thickness showed a small but
Fig. 7 The role of nuclear size and shape in spot- and strip-FRAP analysis. ( a , b ) Variation in nuclear size on the lateral ( r = 5, 10 and 15 μm) ( a ) and longitudinal ( r = 5, 7.5 and 15 μm) ( b ) axes all result in an incomplete recovery in spot-FRAP in a scenario of free diffusion ( D = 2 μm 2 /s). Similarly to variation in spot size, variation in nuclear size introduces an additional but limited delay of recovery in the fi rst part of the curve. ( c – e ) In contrast, variation in nuclear size only on the longitudinal ( d ) and vertical axes ( e ), and, importantly, not the lateral ( c ) axis, affects recovery of fl uorescence. Furthermore, the delay in recovery in the fi rst part of the curves found in spot-FRAP ( a , b ; right column ) is absent in strip-FRAP ( c – e ; right column ). ( f ) Similar effect of incomplete recovery by nuclear size variation along the longitudinal axis is found in a scenario of 33 % immo-bile fraction, 15 s residence time ( D = 2 μm 2 /s). Curves were normalized like in Fig. 5
Bart Geverts et al.
127
signifi cant difference in a spot-FRAP experiment that could easily be confused with differences in mobility (Figs. 7a, b ; left panels). If unnoticed, this is specifi cally a threat when a protein is studied under different conditions that infl uence the size of the cell or the cell nucleus. Especially when curves are normalized for bleach depth and fi nal recovery, where the difference may be interpreted as the existence of a small transiently immobilized fraction, this leads to erroneous interpretation of the FRAP-curve. Interestingly, although bleaching a strip spanning the short axis of ellipsoid with different length results in different FRAP curves (Fig. 7d ), no difference is observed when a strip is bleached spanning the longest axis of the ellipsoid nuclei (Fig. 7c ), showing that the strip-FRAP method is less sensitive to nuclear size when applied in this way and therefore probably is a better method for these type of studies.
Fluorescence resonance energy transfer (FRET) is a method to study interactions. One of the ways to quantitatively analyze FRET is acceptor photobleaching. Briefl y, images are recorded before and after photobleaching the acceptor, leading to an increased donor
3.3.3 FRAP Simulation, Combined FRET and FRAP
Fig. 7 (continued)
Monte Carlo Based FRAP Analysis
128
signal in interacting molecules in which FRET occurred. The degree to which the donor signal is increased is a measure for the fraction of interacting molecules [ 38 ]. Interestingly, acceptor pho-tobleaching can be combined with FRAP, the acceptor signal in a bleached area within a larger area being a conventional FRAP curve. If all molecules interact, the donor signal will be the oppo-site of the acceptor signal, since all bleached acceptors are accom-panied by an unquenched donor. The curves can be compared by inverting the donor signal and normalizing both curves [ 10 ]. When not all molecules are interacting and the interaction takes place in a pool that is different from the pool of noninteracting molecules, e.g., if interaction only takes place in a DNA-bound, immobile fraction, the donor curve will be different from the acceptor curve (Fig. 8 ).
We fi rst applied this method to investigate the N/C interac-tion of androgen receptors (AR, see glossary). We studied this N/C interaction using combined FRET and FRAP applied to AR dou-ble tagged with CFP and YFP. We observed that the donor signal represented a completely mobile molecule, whereas the acceptor curve (the conventional FRAP curve) showed an additional tran-siently immobilized fraction (cf. Fig. 4b ; [ 10 ]).
The analysis of FRET–FRAP applied to double-tagged pro-teins is relatively easy. The analysis of FRET–FRAP on interaction of two single-tagged molecules is complicated by the fact that interacting molecules may be constantly releasing and rebinding again, which also contributes to the donor signal. In this case, the donor signal is constituted by three different phenomena, diffu-sion, and immobilization, like in the example with double-tagged AR, but also by exchange of interacting molecules. To interpret FRET–FRAP results, the simulation tool is very powerful (Fig. 8 ). We simulated different scenarios where a DNA-binding factor and a cofactor interact with each other, either in the nucleoplasm
Fig. 8 (continued) can be created that, e.g., represent a cofactor-binding receptor (e.g., androgen receptor (AR)) that transiently binds DNA, as is shown in the fi rst column. Three general scenarios of cofactor ( n = 10,000) binding to the receptor ( n = 10,000) are simulated; cofactor interaction with either mobile ( a , b ) or immobile ( c , d ) DNA-binding factor only and with DNA-binding factors irrespective of their mobility ( e , f ). All three scenarios are simulated both with ( b , d and f — p on = 0.4, p off = 0.2) and without ( a , c and e — p on = 1, p off = 0) cofactor turnover on the DNA-binding factors (indicated by black arrows with corresponding probabilities) resulting in shifts in the fraction sizes ( second column ). DNA-binding behavior of the DNA-binding factor ( D = 1 μm 2 /s) is simulated with fi xed probabilities ( gray arrows — p on = 0.0025, p off = 0.01), and direct cofactor binding to DNA is excluded ( D cofactor = 10 μm 2 /s). Quantitative analysis of dynamic behavior of the YFP-labelled cofactor ( third column ) and YFP labelled ( fourth column ) both in combination with their interacting subpopulation of their counterpart. To reverse the FRET–FRAP curve and enable comparison FRAP and FRET–FRAP data, curves were normalized by calculating I norm = ( I raw − I 0 )/( I fi nal − I 0 ), where I 0 and I fi nal are fl uorescence intensities after the bleach and after complete recovery, respectively. A more elaborated analysis of reciprocal labelled proteins enables distinction between models that show initially very similar curves (e.g., model in ( a ) and model in ( f ))
Bart Geverts et al.
129
Fig. 8 Computer simulations of combined FRET and FRAP analysis. FRAP simulations can be extended with a combination of FRAP with acceptor-bleaching FRET (abFRET) to quantitatively analyze the relation between protein–protein interactions and the mobility of these proteins. By assigning probabilities not only to the DNA- binding properties of molecules but also to their ability to interact either in mobility or immobile state, models
Monte Carlo Based FRAP Analysis
130
(Fig. 8a, b ), on DNA (Fig. 8e, f ), or both (Fig. 8c, d ). In order to distinguish between these scenarios, which would defi nitively con-tribute to insight in the molecular mechanisms by which these fac-tors act, it is necessary to perform FRET–FRAP experiments with cofactors bound to donor (e.g., CFP) and DNA-binding factors tagged with acceptor (e.g., YFP) and vice versa.
In their intensive correspondence almost two centuries ago, Michael Faraday concluded a letter to James Maxwell raising an important issue: “there is one thing that I would be glad to ask you. When a mathematician engaged in investigating physical actions and results, has arrived at his own conclusions, may they not be expressed in common language, as fully, clearly and defi -nitely as in mathematical formulae? (…) If so, would it not be a great boon to such as we to express them so — translating them out of their hieroglyphics that we might work upon them by exper-iment?” [ 39 ]. It may seem to many experimental biologists today dealing with analysis of complicated results from experiments like FRAP or FRET that these questions remain valid and urgent (for an in-depth discussion of the topic, see [ 40 ]). Nowadays , however, computers with their ever-increasing computational power may well be able to help solving the issue—not by translating the hiero-glyphics, but by offering the possibility to shed light on compli-cated phenomena by intuitively instructable numerical simulations rather than by complex “mathematical formulae.”
In order to contribute to opening up the analysis of complicated systems for a broad community, we aimed to develop a simulation tool that is straightforward in its implementation, maintenance, and expansion and can be used in a routine way without the need to design, validate, and solve sets of differential equations each time a new model is created or an existing one is modifi ed. In addi-tion, we aimed to provide the possibility to design models through an intuitive, easy-to-learn script language. Since our prior need was to be able to quantitate FRAP experiments, the presented model-ling tool can be used not only to model complex systems but also to simulate any FRAP (or FRET) experiment applied to the mod-elled biological system. Therefore, the simulation environment fi ts very well into a systems biology research setting where one aims to achieve an iterative cycle of experiment-driven modelling and model-driven experiments: the model and the method are united in one simulation. Apart from the advantage that models can be created in a simple intuitive script language, the complexity of the designed models with respect to molecular interactions and trans-formations is only limited by computer speed. In addition, and in sharp contrast with mathematical modelling strategies, compart-mentalization in multiple (irregular shaped) volumes of specifi c interactions poses no problem to our approach and also does not signifi cantly reduce the speed of the program.
3.4 Conclusions
Bart Geverts et al.
131
In conclusion, we have shown that Monte Carlo simulation of FRAP (1) can be used for automatically or interactively fi tting FRAP data using a large set of fi xed parameters refl ecting infl uen-tial experimental conditions that are hard to include in other approaches; (2) is well suited to understand, demonstrate, and teach FRAP principles; and (3) can be used to design, validate, and quantify more complex approaches such as the combination of FRAP and acceptor-bleaching FRET. Finally, since not only micro-scope procedures but also the investigated molecular behavior and molecular interactions are simulated, the program provides a versa-tile educational tool to model and study complex molecular inter-action schemes which contribute to understanding important molecular aspects of cell function.
4 Notes
1. Small molecules like, for instance, phosphates, which modify complex properties and which are in general highly abundant, like phosphates, are usually not defi ned as a single molecule that actually exists during the simulation when not bound, but are added to complexes, which means that they are not present before addition.
2. For simulation of collision with the boundaries of the com-partment in which the complex is diffusing, we considered the implementation of fully elastic or dampened collision. However, such algorithms require time-consuming calculation of the normal plane to an ellipsoid. Second, it is likely that the inside membrane of, for instance, the nucleus or cytoplasm is highly roughled. In addition, the nuclear membrane is covered with lamins, lamin receptors, and other macromolecules to which most likely also chromatin is attached, and the cyto-plasm membrane harbors high amounts of transmembrane proteins. In other words, it is currently not at all known what type of collision behavior we can expect at membranes. Therefore, in the current simulation, we use the simple solu-tion where each time the random-generated new position of the complex is outside the compartment, a new random step is generated until the new position is inside. This leads to a cer-tain degree of stickiness of the membrane, which does not sig-nifi cantly contribute to FRAP simulations and, moreover, is not unlikely to refl ect the actual situation.
3. Interestingly, novel methods to track single molecules may be applied to shed light on this, since the additional informa-tion obtained when tracking individual molecules is the dis-tribution of binding times, whereas bulk methods only provide averages.
Monte Carlo Based FRAP Analysis
132
References
1. Aldridge BB, Burke JM, Lauffenburger DA et al (2006) Physicochemical modelling of cell signalling pathways. Nat Cell Biol 8:1195–1203
2. van Royen ME, Dinant C, Farla P et al (2009) FRAP and FRET methods to study nuclear receptors in living cells. In: McEwan IJ (ed) Nuclear receptor superfamily, vol 505, Methods Mol Biol. Springer, Totowa, pp 69–96
3. van Royen ME, Farla P, Mattern KA et al (2009) Fluorescence recovery after photo-bleaching (FRAP) to study nuclear protein dynamics in living cells. In: Hancock R (ed) The nucleus, vol 464, Methods Mol Biol. Springer, Totowa, pp 363–385
4. Chen WW, Niepel M, Sorger PK (2010) Classic and contemporary approaches to mod-eling biochemical reactions. Genes Dev 24:1861–1875
5. Alves R, Antunes F, Salvador A (2006) Tools for kinetic modeling of biochemical networks. Nat Biotechnol 24:667–672
6. Beaudouin J, Mora-Bermudez F, Klee T et al (2006) Dissecting the contribution of diffu-sion and interactions to the mobility of nuclear proteins. Biophys J 90:1878–1894
7. Braga J, Desterro JM, Carmo-Fonseca M (2004) Intracellular macromolecular mobility measured by fl uorescence recovery after pho-tobleaching with confocal laser scanning microscopes. Mol Biol Cell 15:4749–4760
8. McNally JG (2008) Quantitative FRAP in analysis of molecular binding dynamics in vivo. Methods Cell Biol 85:329–351
9. Phair RD, Scaffi di P, Elbi C et al (2004) Global nature of dynamic protein-chromatin interac-tions in vivo: three-dimensional genome scan-ning and dynamic interaction networks of chromatin proteins. Mol Cell Biol 24:6393–6402
10. van Royen ME, Cunha SM, Brink MC et al (2007) Compartmentalization of androgen receptor protein-protein interactions in living cells. J Cell Biol 177:63–72
11. Farla P, Hersmus R, Trapman J et al (2005) Antiandrogens prevent stable DNA-binding of the androgen receptor. J Cell Sci 118:4187–4198
12. Mueller F, Mazza D, Stasevich TJ et al (2010) FRAP and kinetic modeling in the analysis of nuclear protein dynamics: what do we really know? Curr Opin Cell Biol 22:403–411
13. Carrero G, McDonald D, Crawford E et al (2003) Using FRAP and mathematical modeling
to determine the in vivo kinetics of nuclear proteins. Methods 29:14–28
14. Braeckmans K, Peeters L, Sanders NN et al (2003) Three-dimensional fl uorescence recovery after photobleaching with the confocal scanning laser microscope. Biophys J 85:2240–2252
15. Kang M, Kenworthy AK (2008) A closed-form analytic expression for FRAP formula for the binding diffusion model. Biophys J 95:L13–L15
16. Tsibidis GD (2009) Quantitative interpreta-tion of binding reactions of rapidly diffusing species using fl uorescence recovery after pho-tobleaching. J Microsc 233:384–390
17. Tsibidis GD, Ripoll J (2008) Investigation of binding mechanisms of nuclear proteins using confocal scanning laser microscopy and FRAP. J Theor Biol 253:755–768
18. Braeckmans K, Remaut K, Vandenbroucke RE et al (2007) Line FRAP with the confocal laser scanning microscope for diffusion measure-ments in small regions of 3-D samples. Biophys J 92:2172–2183
19. Hallen MA, Layton AT (2010) Expanding the scope of quantitative FRAP analysis. J Theor Biol 262:295–305
20. Mazza D, Braeckmans K, Cella F et al (2008) A new FRAP/FRAPa method for three- dimensional diffusion measurements based on multiphoton excitation microscopy. Biophys J 95:3457–3469
21. Mazza D, Cella F, Vicidomini G et al (2007) Role of three-dimensional bleach distribution in confocal and two-photon fl uorescence recovery after photobleaching experiments. Appl Opt 46:7401–7411
22. Kang M, Day CA, Drake K et al (2009) A gen-eralization of theory for two-dimensional fl uo-rescence recovery after photobleaching applicable to confocal laser scanning micro-scopes. Biophys J 97:1501–1511
23. Mueller F, Wach P, McNally JG (2008) Evidence for a common mode of transcription factor interaction with chromatin as revealed by improved quantitative fl uorescence recov-ery after photobleaching. Biophys J 94:3323–3339
24. Agarwal S, van Cappellen WA, Guenole A et al (2011) ATP-dependent and independent functions of Rad54 in genome maintenance. J Cell Biol 192:735–750
25. de Graaf P, Mousson F, Geverts B et al (2010) Chromatin interaction of TATA-binding pro-tein is dynamically regulated in human cells. J Cell Sci 123:2663–2671
Bart Geverts et al.
133
26. Farla P, Hersmus R, Geverts B et al (2004) The androgen receptor ligand-binding domain stabilizes DNA binding in living cells. J Struct Biol 147:50–61
27. Luijsterburg MS, Goedhart J, Moser J et al (2007) Dynamic in vivo interaction of DDB2 E3 ubiquitin ligase with UV-damaged DNA is independent of damage-recognition protein XPC. J Cell Sci 120:2706–2716
28. Nicassio F, Corrado N, Vissers JH et al (2007) Human USP3 is a chromatin modifi er required for S phase progression and genome stability. Curr Biol 17:1972–1977
29. Nishi R, Alekseev S, Dinant C et al (2009) UV-DDB-dependent regulation of nucleotide excision repair kinetics in living cells. DNA Repair (Amst) 8:767–776
30. Sabbioneda S, Gourdin AM, Green CM et al (2008) Effect of proliferating cell nuclear anti-gen ubiquitination and chromatin structure on the dynamic properties of the Y-family DNA polymerases. Mol Biol Cell 19:5193–5202
31. Tanner TM, Denayer S, Geverts B et al (2010) A 629RKLKK633 motif in the hinge region controls the androgen receptor at multiple lev-els. Cell Mol Life Sci 67:1919–1927
32. van den Boom V, Kooistra SM, Boesjes M et al (2007) UTF1 is a chromatin-associated pro-tein involved in ES cell differentiation. J Cell Biol 178:913–924
33. Xouri G, Squire A, Dimaki M et al (2007) Cdt1 associates dynamically with chromatin
throughout G1 and recruits Geminin onto chromatin. EMBO J 26:1303–1314
34. Zotter A, Luijsterburg MS, Warmerdam DO et al (2006) Recruitment of the nucle-otide excision repair endonuclease XPG to sites of UV-induced dna damage depends on functional TFIIH. Mol Cell Biol 26:8868–8879
35. van Royen ME, van Cappellen WA, Geverts B et al (2014) Androgen receptor complexes probe DNA for recognition sequences by short random interactions. J Cell Sci 127(Pt 7):1406–1416
36. Houtsmuller AB, Rademakers S, Nigg AL et al (1999) Action of DNA repair endonuclease ERCC1/XPF in living cells. Science 284:958–961
37. Phair RD, Misteli T (2000) High mobility of proteins in the mammalian cell nucleus. Nature 404:604–609
38. Dinant C, van Royen ME, Vermeulen W et al (2008) Fluorescence resonance energy transfer of GFP and YFP by spectral imaging and quan-titative acceptor photobleaching. J Microsc 231:97–104
39. Faraday M (2008) The correspondence of Michael Faraday, Vol 5. Institution of Engineering and Technology, Stevenage, pp 1855–1860
40. Tweney RD (2009) Mathematical representa-tions in science: a cognitive-historical case history. Top Cogn Sci 1:758–776, 2010
Monte Carlo Based FRAP Analysis
135
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_8, © Springer Science+Business Media New York 2015
Chapter 8
Fluorescence Correlation Spectroscopy
Mark A. Hink
Abstract
Fluorescence fluctuation spectroscopy techniques allow the quantification of fluorescent molecules present at the nanomolar concentration level. After a brief introduction to the technique, this chapter presents a protocol including background information in order to measure and quantify the molecular interaction of two signaling proteins inside the living cell using fluorescence cross-correlation spectroscopy.
Key words Fluorescence correlation spectroscopy, Fluorescence cross-correlation spectroscopy, Living cells, Dimerization, Molecular interactions, PI3K signaling
1 Introduction
Fluorescence correlation spectroscopy (FCS) is the most well- known member of the family of fluorescence fluctuation spectros-copy (FFS) techniques that analyzes temporal changes of the fluorescence intensity and relates these fluctuations to physical parameters of the observed molecules. FCS was developed in the early 1970s of the last century, but it took until the early 1990s, after the introduction of the confocal microscope with the use of improved lasers and detectors, to increase significantly in popularity.
The high sensitivity of the technique allows the detection of single molecules and obtaining information about the concentra-tion, diffusion rate, and interactions of the molecules [1]. FCS measurements are typically done with a standard confocal micro-scope equipped with highly sensitive detectors and an objective of high numerical aperture. The focused laser beam illuminates a sub- femtoliter volume element (ca. 1 μm3). The fluorescence photons emitted in this element pass through a pinhole and are detected by a highly sensitive detector. The signal-to-noise ratio achieved by this method is very high, since signal interference from scattered laser light, background fluorescence, and Raman emission can be largely eliminated. This allows measurements at the single- molecule level typically in the nanomolar concentration range.
136
During the measurement, molecules will, due to their Brownian motion, move in and out of the volume element and emit photons in a burst-type manner. The observed fluctuations can be used to determine the average time required for the pas-sage through the volume element of a fluorescent molecule. This is dependent on its diffusion coefficient, which, in turn, is related to the size of a molecule. Therefore, it is possible to examine if molecules are moving freely or are part of a large, slow moving complex. Meseth et al. [2] examined the resolving power of FCS to distinguish between different molecular sizes. In case of an unchanged fluorescence yield upon binding, the diffusion con-stants of the bound and unbound form have to differ at least 1.6 times, corresponding to a fourfold mass increase which is required to discriminate both species without prior knowledge of the sys-tem. To overcome these limitations of FCS, Schwille et al. [3] applied dual-color fluorescence cross- correlation spectroscopy (FCCS). In such studies, interacting molecules are tagged by spectrally different fluorescent groups, for example, a green and a red dye. The different fluorescent dyes can be either excited with different lasers or with the same laser. The emission light is split into two different detectors by which the two fluorophores can be monitored simultaneously. Now, molecular interactions can be studied by following the fluorescence fluctuations of both fluoro-phores. In this case, the discriminating factor is not the increase of the molecular mass upon complex formation but the simultaneous occurrence of fluctuations in both detection channels.
Fluorescence fluctuations can also be analyzed in different ways, for example, to study the oligomerization of molecules via photon counting histogram (PCH) [4] or number and bright-ness (N&B) analysis [5]. In order to combine temporal informa-tion with spatial information, the FFS approaches can be used to analyze images as well, which are therefore referred to as image correlation spectroscopy (ICS) techniques [6]. In addition, one could combine FCS with other spectroscopic parameters like flu-orescence lifetime [7], anisotropy [8], or Förster resonance energy transfer (FRET) [9], thereby increasing the selectivity of the measurement. In this chapter however, the focus is on single-point FCS and FCCS and its use to determine concentrations and quantify molecular interactions of fluorescently labeled molecules in living cells.
2 Materials
A standard FCS setup utilizes a confocal microscope, in order to reduce background and increase FCS sensitivity by limiting the size and therefore the number of molecules within the observa-tion volume (Fig. 1). The standard photomultiplier tube detectors
2.1 Fluorescence Microscope
Mark A. Hink
137
(PMTs) delivered with confocal systems are not sensitive enough and should be replaced by avalanche photodiodes (APDs) or hybrid photodetectors (HPDs). Back-illuminated electron multi-plying charge coupled device (EM-CCD) cameras have a higher sensitivity, but the current limitations in readout speed result in a low temporal resolution, often not sufficient to monitor the rela-tive fast molecular diffusion in the cell cytoplasm or in vitro.
Although not discussed in this chapter, alternative excitation strategies like TIRF, SPIM, STED, or spinning disk illumination can be combined with FCS, allowing the differently shaped and/or smaller observation volume. Furthermore, some of these options open the possibility to use FCS at higher dye concentra-tions [10–13].
A critical part of the equipment is the objective lens, especially when multicolor analysis is being used. It is essential to utilize an objective with a high numerical aperture (NA) in order to collect as many of the emitted photons as possible, during the limited resi-dence time of a molecule in the detection volume. Most FCS data fitting models require a 3D Gaussian or Gaussian-Lorentzian- shaped observation volume, and therefore, any mismatch between the refractive index of the objective immersion liquid and the
tautime
inte
nsity G(tau)
APD
pinhole
dichroic mirror
emission filter
objectivelens
APD
Fig. 1 FCCS setup. Light from two laser beams is focused into the sample using a highly corrected objective lens with a large numerical aperture. Fluorescence is guided via a pinhole into the detection unit where dichroic mirror and emission filters separate the light into two sensitive detectors. Diffusion of fluorescent molecules through the detection volume causes intensity fluctuations that can be analyzed by fitting the auto- and/or cross-correlation curves
Fluorescence Correlation Spectroscopy
138
s ample should be minimized. An objective with the best chromatic corrections is needed to have the best overlap between the multi-color excitation and detection volumes. Apochromat objectives fulfill these requirements and have a collar ring to manually correct for mismatches. For biological samples, a water-immersed lens is preferred since oil-immersed objectives, despite their higher NA values, lead to shape distortions of the observation volume and therefore result in poor data fits. Recently, a silicon-immersed objective with high NA was introduced by Olympus which approaches the refractive index of the cellular environment closely, resulting in high-quality intracellular FCCS curves [14].
However, even when using these highly corrected objective lenses, the Gaussian approximation of the detection volume is not perfectly met. If diffusion coefficients with extreme high accuracy should be obtained, a two-focus-FCS setup can be used where the cross-correlation signal of molecules diffusing between two closely spaced observation volumes, generated by a Wollaston prism in the excitation path, is analyzed [15].
For two-color analysis, a common problem is the presence of cross talk: the fluorescence of one type of dye detected in the “other” detector (e.g., the emission of the “green” dye in the “red” detector). The large tail in the dye emission spectrum caus-ing this cross talk could give rise to false-positive cross-correlation. One could correct the data for this artifact, as discussed later in this chapter, but this issue can be prevented by using pulsed inter-leaved excitation (PIE) [16]. Most absorbance spectra are rela-tively narrow, and therefore, it is possible to excite each dye selectively using two specific laser lines. When two pulsed laser units are emitting in an alternating mode and at a frequency much faster than the residence time of the molecule in the observation volume, one can obtain the fluorescence of both dyes in different time windows. By time-gating the detected fluorescence, one can omit the cross-talk photons in the calculation of the auto- and cross-correlation curves.
Since FCS is a method that relies on the analysis of relative fluores-cence fluctuations, the average concentration of fluorescent mole-cules should be kept low, typically between 1 nM and 1 μM. During the passage of the molecules through the detection volume, as many photons as possible should be detected. As shown by Koppel [17], the molecular brightness of the fluorophore is one of the most important parameters determining the quality of an FCS measure-ment. In order to select a high-quality FCS dye from the multitude of dyes commercially available, the product of extinction coefficient times fluorescence quantum yield gives a good indication. Therefore, fluorescent beads or quantum dots will result in FCS curves with a high signal-to-noise ratio (SNR), but due to the relative large size of these beads and limited coupling possibilities, their application to biological samples is still limited. In addition, one has to consider
2.2 Fluorescent Dyes
Mark A. Hink
139
other dye-related parameters as well like the sensitivity to photobleaching, environmental conditions that alter the dye char-acteristics (e.g., pH or hydrophobicity effects), the probability to go into a temporal dark state (triplet-state kinetics and blinking), and its stickiness to other molecules or to the sample holder.
For intracellular FCS measurements, the spectral profile of cellular autofluorescence should be taken into account. In many animal cells, the autofluorescence background is the strongest in the blue spectral region, while in plant cells, also a strong fluorescence is observed in the red region (>640 nm) when chlorophyll is present. Since (intracellular) FCS experiments are typically performed with nanomolar concentrations of the fluorescent probe, the autofluo-rescence of the cell could become a significant fraction of the total detected fluorescence. When the autofluorescence is not giving rise to an autocorrelation curve, a constant correction factor will be added to the data fitting model. It should be prevented that auto-fluorescence generates an autocorrelation curve since it is difficult to (quantitatively) correct for this contribution. It might be worth to put some effort in reducing the autofluorescence by optimizing growth conditions like varying the composition of the growth medium, changing to 37 °C and CO2 incubation of living animal cells, reducing the amount of light exposure in case of plant sam-ples, or changing the experimental setup by shifting to another spectral detection region.
Common organic dyes used for FCS experiments are members of the Alexa (Invitrogen), Atto (ATTO-TEC), and CY (Amersham) families. When fluorescent proteins (FPs) are selected as the dye of choice (Table 1), one has to keep in mind that especially the red
2.2.1 Autofluorescence
Table 1 Recommended standard monomeric fluorescent proteins for FC(C)S experiments
Fluorescent protein
λmax exc. (nm)
λmax em. (nm) Comment Reference
TagBFP2 399 456 Check maturation, UV excitation
[32]
mTurquoise2 434 477 very bright [33]
sGFP2 (or eGFP) 495 512 Photostable, bright [34]
sYFP2 (or mVenus) 515 527 Significant photophysics [35]
mCherry 587 610 Check maturation [36]
TagRFP-T 555 584 Photostable, bright, low maturation
[37]
mKate2 588 635 Check maturation [38]
Fluorescence Correlation Spectroscopy
140
variants often show reduced maturation. Due to the delay time in forming a fluorescent chromophore, the true number of proteins will be underestimated. For example, when observing fusion pro-teins involved in yeast pheromone signaling, roughly 50 % of the mCherry fusion proteins were not visible [18]. This maturation issue is highly cell-type dependent, for example, in HeLa cells, only ~5 % of the mCherry fusions are not visible. Maturation ratios can be estimated by determining the concentration of green and red proteins in an FCS experiment using a < red FP-spacer protein- eGFP > fusion construct, assuming the eGFP protein to mature instantaneously. The large spacer protein present should prevent (FRET) between the two, which would complicate the analysis [9]. In order to increase the molecular brightness, multiple copies of the FPs could be introduced in the fusion protein, although distor-tion of the biological functionality and heterogeneous maturation should be prevented [18].
For two-color FCCS studies, a critical point is the cross talk. Although one could correct for this artifact, it is desirable to dimin-ish the spectral cross talk as much as possible. The popular FP pair CFP-YFP has a CFP bleed-through factor of 30 % in a typical setup. This will lead to significant false-positive cross-correlation, and therefore, corrections are required [18–20]. If the PIE excita-tion strategy is not possible, the quality of the measurements can be improved by choosing a pair that is spectrally more separated. However, when choosing such a pair, the overlap between the two detection volumes (which depends inversely on the spectral separa-tion) will become less, and the maximum observable cross- correlation, and therefore the sensitivity, will decrease. Thus, a trade-off should be found between spectral separation and detec-tion volume overlap. A good FCCS pair is eGFP-mCherry having high molecular brightness values, low photobleaching rates, and a reasonable spectral separation. Alternatively, one could use the more blue-shifted mTurquoise2-sYFP2 pair. When the fusion pro-teins give rise to FRET, the analysis is more complicated [21]. FRET will lower the molecular brightness of the donor molecule. Since the contribution of a molecule to the autocorrelation curve scales to the square of the molecular brightness [17], non- FRETting donor molecules that are brighter will dominate the correlation curve. Therefore, analysis can be simplified when a probe pair is chosen that will have a low FRET efficiency, e.g., due to a large spectral difference or spatial distance.
In order to minimize the importance of the correction factors and the uncertainty connected to it, the experimental conditions in two-color FCCS must be optimized. To lower the effect of cross talk, one should increase the signal of the red dye in the red detector. One could couple the red dye to the most abundant protein studied, lower the brightness of the green, and increase the brightness of the
2.2.2 Two-Color FCCS
Mark A. Hink
141
red dye by optimizing both laser powers or selecting a more r ed-shifted emission filter for the red detector [20]. To minimize the non-perfect overlap between the different observation volumes, dyes like the large Stokes shift fluorescent proteins [22] can be used. Here, multiple types of fluorescent proteins are excited using only a single laser line, while each FP fluoresces in a separate spectral win-dow, due to their different Stokes shifts.
Raw or correlated data can be fitted with dedicated analysis soft-ware. Many microscope manufacturers (e.g., Zeiss, ISS, Olympus, PicoQuant, and Becker & Hickl) distribute their own analysis soft-ware. In addition, commercial software packages can be purchased (e.g., SSTC FFS Data Processor and LFD SimFCS), and some user-written scripts for programs like Matlab, Igor, ImageJ, Origin, or Mathematica can be found on the Internet.
3 Methods
In the following section, a protocol is presented that can be used to measure dual-color fluorescence cross-correlation spectroscopy in living HeLa cells. Many issues discussed below will be similar in the case of single-color and/or in vitro FCS measurements. The idea behind this protocol is to study molecular interaction in the cell cytoplasm between sGFP2 and mCherry-labeled signaling pro-teins p110 and p85 [23] using an Olympus FV1000 microscope equipped with a PicoQuant PicoHarp detection unit. Although PIE excitation is possible with this setup, many FCS system do not have this feature and therefore is not employed in the protocols below.
1. Grow HeLa cells in DMEM medium (Gibco) supplemented with GlutaMAX (Gibco) in a live cell incubator at 37 °C with 5 % CO2.
2. One day before transfection, transfer the cells from the culture flask to sterile circular (ø 24 mm, size 1) coverslips (Menzel- Gläser), stored in a six-well container (Greiner), at a conflu-ency of approximately 60 % (~5.0 × 105 cells).
3. Six hours after the cell transfer, replace the growth medium by phenol-free DMEM medium (Gibco), thereby lowering the autofluorescence.
4. Transfect the cells using Lipofectamine in Opti-MEM accord-ing to the manufacturers protocol (Invitrogen). An important point to consider, especially when using high expression pro-moters in the construct(s), is that FCS requires low expression levels (see Note 1).
2.3 Data Analysis
3.1 Transfection of HeLa Cells
Fluorescence Correlation Spectroscopy
142
1. The microscope is placed in an air-conditioned room to ensure temperature stability and the calibration samples are stored in glass-bottomed 96-well plates (Whatman). In order to reduce the amount of detected background light, a protective black-ened aluminum cover was placed on top of the sample during the measurements.
2. Turn the microscope and lasers on roughly 30 min before starting the measurements in order to stabilize.
3. Configure the FCS microscope filters: the sGFP2-labeled samples will be excited with the 488 nm laser line and the mCherry samples with the 561 nm diode laser (if available, use one or two pulsed lasers, suitable for the PIE excitation strategy, preventing cross talk). The fluorescence is separated from the excitation light by a dual-band dichroic filter, reflecting both the 488 and 561 nm excitation lines (Chroma). A sec-ondary dichroic filter, LP562 (Chroma), separates the emis-sion light into two different detection channels. Appropriate bandpass and longpass filters (e.g., BP500-550 (Semrock) for sGFP2 and LP572 (Chroma) for mCherry, respectively) are used for spectral selection and placed in front of the APD detectors (see Note 2).
4. Measure the fluorescence of a ~100 nM solution of Alexa 488 (Invitrogen) in phosphate buffered saline (PBS) at a laser power of approximately 10–50 kW · cm−2. Adjust the correction collar of the objective to the position where the highest fluorescence count rate is observed in one of the two detection channels (see Note 3). In order to prevent that one optimizes scattered exci-tation light or adsorbed probe from the bottom of the sample holder, instead of observing the wished fluorescence signal, the focus should be set at least 20 μm into the solution.
5. Optimize the detected fluorescence to its highest value by moving the X-, Y-, and, if available, Z-position of the pinhole(s) in the confocal microscope and optimize the lens position in front of the APD, if possible.
6. Perform an FCS measurement of Alexa 488 and one of Atto 565 in PBS using their corresponding laser lines at 3.2 (488 nm) and 4.0 kW·cm−2 (561 nm) (see Note 4). Adjust the measurement time such that the resulting correlation curves are smooth in the decaying part of the curve.
7. The intensity traces are imported into the FFS data processor 2.3 software, autocorrelated, and fitted to a model including terms for triplet-state kinetics and three-dimensional Brownian diffusion [24] (see Note 5):
3.2 Microscope Calibration
Mark A. Hink
143
GN
T Te
T
dif dif
xy
z
ttt
tt
ww
tt
( ) = + ×- +
-( )×
+æ
èç
ö
ø÷ + ×
æ
è
-
11
1
1
1 1
g TRIP
ççö
ø÷
æ
è
çççççç
ö
ø
÷÷÷÷÷÷
2
(1)
8. Validate if the obtained fitting parameters are within the expected range. For our system, the value for a (=ωz/ωxy) ranges from 4 to 15 and the values of τdif for Alexa 488 or Atto 565 are around 35 and 60 μs, respectively. The observed brightness val-ues vary between 8 and 12 kHz per molecule. Note that these values will be different for other microscope systems. In gen-eral, the higher the brightness per particle, the better the qual-ity of the obtained data will be. The system has to be calibrated until the highest possible brightness is reached without having photobleaching and/or saturation artifacts [25].
9. By solving Eqs. 2 and 3, using the translational diffusion coef-ficient (D) of 410 μm2·s−1 for Alexa 488 [26] and 360 μm2.s−1 for Atto 565 in buffer, the dimensions of the “green” and “red” detection volumes, V, can be calculated by approximat-ing these as cylinders (see Note 6).
t
wdif
xy
D=
2
4 (2)
V xy= × × ×2 3p wa
(3)
The effective cross-correlation observation volume is estimated from the cross-correlation curve, measuring a sample of purified sGFP2 in PBS (D = 90 μm2·s−1). Here we make use of the cross talk of sGFP2 since some of its emission will be detected in the mCherry detection channel. The emission of mCherry in the GFP channel can be omitted.
1. A coverslip with the transfected HeLa cells is sealed in an Attofluor cell chamber (Invitrogen) submerged in microscopy medium (20 mM HEPES (pH = 7.4), 137 mM NaCl, 5.4 mM KCl, 1.8 mM CaCl2, 0.8 mM MgCl2, and 20 mM glucose). Note that most growth media will acidify when incubated in the absence of 5 % CO2 during the measurement.
2. Measure a positive cross-correlation sample, e.g., a fusion protein of sGFP2 linked to mCherry separated by a large pro-tein. This control protein, like the negative control in step 3, should preferably be targeted to the same subcellular location as where the sample proteins will be localized. Here, we use
3.3 Fluorescence Fluctuation Measurements
Fluorescence Correlation Spectroscopy
144
the cytoplasmic fusion protein mCherry-p63GEFT-sGFP2 [27]. To select the cells with FCS-compatible expression lev-els, the sample is scanned by the confocal microscope using standard imaging settings for the PMT detector sensitivity but with an open pinhole. Those cells that are more fluorescent than the mock- transfected control cells, but do not give rise to detector saturation in the image, are selected for FCS mea-surements. For cross-correlation measurements, only cells are selected that express both the eGFP and mCherry fusion pro-teins. The laser power is set not higher than 2.1 kW·cm−2 for the 488 nm laser line and 1.7 kW·cm−2 for the 561 nm laser line to prevent photobleaching, cellular damage, and photo-physical effects. These excitation intensities are still sufficient to achieve reasonable SNRs [5–10] within measurement times of 60–180 seconds. Above these laser powers, photobleaching and probe saturation lead to significant distortions of the cor-relation curves (Fig. 2). Optimize the objective correction collar and the pinhole such that the amplitude of the cross-correlation curve is as high as possible. Sometimes this will lead to slightly worse structure parameter values a, but the overlap of the detection volumes is the most important param-eter in these experiments.
3. Perform a measurement of a negative cross-correlation sample. The ideal sample would consist of mutants of the two sample proteins now lacking the possibility to interact with each other due to point mutations or domain deletions. When the amino acid residues responsible for binding are unknown, HeLa cells co-transfected with free sGFP2 and free mCherry could be used as an alternative.
4. Measure the samples. All raw fluorescence intensity files are saved for processing. Data sets containing large intensity spikes
Time (s)
Inte
nsity
(kH
z)40
00 20 40
1.6
1.01.00.01 100
Tau (ms)
G (
tau)
a b
Fig. 2 Photobleaching and intensity spikes cause artifacts in the correlation curve (black) visible at the long tau values as a stepwise decay that might be erroneously assigned to slow moving molecules. By selecting only the stable part of the intensity trace (between markers) for calculation, a correct correlation curve (gray) is obtained although one has to validate that the observed decay is not biased to lower values due to photo-bleaching of mobile molecules (Note 7)
Mark A. Hink
145
(>10 times the intensity standard deviation around the mean intensity), significant signal drift, or photobleaching (>20 % intensity loss per minute) are discarded (see Note 7).
1. After loading the raw data into the FFS data processor soft-ware, a region that contains at least 50,000 photons in each channel is selected out of the two complete raw data traces. The selected intensity trace should be stable, so without an intensity drift, and the presence of (small) intensity spikes should be minimal in both channels. The selected intensity trace is auto- and cross- correlated resulting in the two autocor-relation curves GG (τ) (for the GFP channel) and GC (τ) (for the mCherry channel) and the cross-correlation curve GGC (τ). Figure 3 shows some typical experimental curves.
2. The autocorrelation curves can be fitted according to
GF
F N
T Te
Tbackground
total mm
T
ttt
( ) = + -æ
èç
ö
ø÷ ×
< >×- +
-( )
-
å1 11 1
1
2
F ××
+æ
èçç
ö
ø÷÷ + ×
æ
èç
ö
ø÷
æ
è
çççççç
ö
ø
÷÷÷÷÷÷
+1
1 12
tt
tt
wwdif m dif
xy
z,
offfset
withFm
m m
mm m
Y
Y
=( )
æ
èç
ö
ø÷å
h
h
2
2
(4)
where the first term between brackets corrects for the back-ground fluorescence in the corresponding detection channel (Fbackground). The value of this parameter is determined by
3.4 Auto- and Cross- Correlation Analysis
0.1 10 1000
Tau (ms) Tau (ms)0.1 10 1000
G (
tau)
G (
tau)
+
a b c1.3
1.0
1.3
1.0
Fig. 3 Auto- (green and red) and cross-correlation (black) curves and their fits (gray) as measured (cross) in the cytoplasm of transfected HeLa cells (a). For the negative control (b), p85-mCherry plus sGFP2, no cross- correlation amplitude is observed after cross-talk correction, while the positive cross-correlation amplitude for the measurements of the p85-mCherry plus sGFP2-p110 constructs (c) indicates a large fraction of interacting molecules
Fluorescence Correlation Spectroscopy
146
averaging the fluorescence intensity in the subcellular region of interest (e.g., cytoplasm or membrane) in mock-transfected cells. In order to correct for the noncorrelating cross-talk of sGFP2 in the mCherry channel, the background fluorescence in the mCherry channel was raised by 0.10 times the intensity observed in the GFP channel. This bleed-through factor of 10 % was determined by comparing the intensities in both detection channels using an sGFP2-only sample. Each molecu-lar species, m, contributes to the autocorrelation curve func-tion according to its fraction (Ym) and molecular brightness (ηm). The ratio ωxy/ωz is fixed to the value obtained from the calibration measurement for each channel. An offset can be included in the fit to correct for the effect of small intensity drifts in the selected data trace.
3. The amplitude of the cross-correlation curve GGC(τ) contains the information of the number of doubly labeled particles, NGC, observed in the cross-correlation observation volume, VGC. Equations 5–7 contain corrections for cross talk and assume a 1:1 stoichiometry for the complex p110-p85; for more complex situations, modified equations can be found in [19]. In the case of higher order complexes, one could deter-mine the brightness and distribution of those oligomers using PCH techniques [28] and include brightness corrections in Eqs. 5–7:
G
N N
N N N NGC
GGGC
CCCGC
GGC
CCC
G GC C G
0 1
1
( ) = +
æ
èç
ö
ø÷ + +
æ
èç
ö
ø÷
+( ) +
hh
hh
CCGGC
CCCG
GGC
CCC
N1+æ
èç
ö
ø÷ +
æ
èç
ö
ø÷
æ
èçç
ö
ø÷÷
hh
hh
(5)
GN NG
G GC
0 11( ) = ++( ) (6)
G
N N N N
C
C GC G GCGGC
CCC
0 11
2( ) = +
+ + +( )æèç
ö
ø÷
æ
èçç
ö
ø÷÷
hh
(7)
where NG, NC are the number of monomers labeled with sGFP2 or mCherry. NGC is the heterodimer and no homodi-mers are being formed. The ηprobe,excitation,emission values corre-spond to the molecular brightness values for the probes as detected for the different excitation and emission wavelengths. For example, ηGGC should be read as the molecular brightness of sGFP2 as detected in the mCherry channel using the 488 nm “GFP” laser line. These values were obtained by the calibration measurements using the same laser settings as for the samples.
Mark A. Hink
147
The relative brightness of sGFP2 to mCherry in the mCherry detection channel, ηGGC/ηCCC, is 9 % in our setup.
Equations 5–7 are linked and fitted globally including a weight factor to correct for the different detection volume sizes. An exam-ple of our experimental values is presented in Table 2. Due to the different expression levels among the cells, caused by the transient transfection procedure, the interaction values are presented as per-centage complex and as apparent dissociation constant assuming the absence of nonfluorescent and endogenous p85 and p110 proteins.
4 Notes
1. Although suitable for fluorescence imaging experiments, standard overexpression promoters like cowpea mosaic virus (CMV) will result in too high expression levels. In order to prevent this, one could change promoter sequence, reduce the amount of DNA in the transfection procedure, or shorten the time between transfection and FCS measurement. Here, 1 nanogram of DNA constructs has been used containing a CMV promoter sequence in front of p110 and p85 coding sequences. The cells were transfected the evening before the FCCS measurements. To prevent the presence of unfinished protein product, like free FP, interfering with the measure-ment the FP should be located preferably at the C-terminal of the protein of interest.
Table 2 FCCS results in HeLa cells transfected with mCherry-p110 and p85-sGFP2. The data was analyzed using Eqs. 4–7. Diffusion times are presented for the detection channel corresponding to the FP in the construct. n.a. not applicable
Construct(s) Diffusion time (ms) % Interaction (NGC/NXtotal) Apparent KD (nM) n (−)
p85-sGFP2 1.1 ± 0.2 0 ± 0 n.a. 10
mCherry-p110 1.8 ± 0.4 0 ± 0 n.a. 10
mCherry-GEFT-sGFP2 (fusion)
1.9 ± 0.31.1. ± 0.2
99 ± 298 ± 1
n.a 10
p85-mCherry +sGFP2
1.7 ± 0.31.2 ± 0.3
1 ± 10 ± 0
2,560 ± 3,984 10
p85-mCherrysGFP2-p110
2.0 ± 0.41.4 ± 0.3
85 ± 482 ± 5
172 ± 36 12
p85-sGFP2 +mCherry-p110
1.2 ± 0.31.8 ± 0.4
83 ± 580 ± 4
230 ± 41 12
Fluorescence Correlation Spectroscopy
148
2. For measurements at extreme low fluorophore concentrations, the Raman scattering signal of water molecules, present at roughly 55 M, could contaminate the fluorescence signal. The wavelength of the Raman peak can be calculated since the main peak is always red-shifted 3,382 cm−1 relative to the excitation, so proper bandpass filters can be selected blocking this signal.
3. The optimal setting of the ring, where the highest count rate is observed, corresponds very often not to the printed value on the ring that indicates the sample glass thickness.
4. Many fluorescent probes like the FPs are susceptible to photo-bleaching and show blinking behavior, due to continuous con-versions between fluorescent and dark states [29–31]. Both processes become more pronounced at higher laser powers. In addition, at higher laser intensities, probe saturation might occur in the center of the focus [25]. This will change the effective shape of the detection volume, resulting in a deviation from the 3D-Gaussian-shaped detection volume that is assumed in the fitting models. Therefore, a trade-off should be found using a laser power where the brightness of the molecule (expressed as detected counts · s−1 · molecule−1) is as high as possible but where artifacts caused by the processes described above are minimized. To find this saturation level, one should plot the fluorescence intensity of the dye versus laser power.
5. The sizes of the green, red, and cross-correlation volumes are calculated from calibration measurements in vitro since a diffu-sion coefficient has to be known in order to solve Eqs. 2 and 3. Note that the biological samples measured in living cells will introduce small optical aberrations. The obtained observation volume sizes are therefore only approximations.
6. The autocorrelation function G(τ) contains a parameter N, which corresponds to the average number of fluorescent par-ticles in the detection volume. τdif is the average diffusion time of the molecules, ωxy is the equatorial radius, and ωz is the axial radius of the observation volume. T is the fraction of molecules present in the dark state and τTRIP is the average time a molecule resides in the dark state. Take care that τTRIP is determined correctly, typically 10–200 μs dependent on dye and laser power, and is not mixing with τdif. Parameter γ describes the shape factor of the observation volume and equals 0.3535 for a 3D Gaussian or 1.0 for a cylindrical-shaped observation volume. From the autocorrelation fits the shape factor, a, describing the ratio between the axial and lateral e−2 radii of the green and red observation volumes (a = ωz/ωxy) and the τdif for Alexa 488 and Atto 565 are obtained. Although the shape of the detection volumes is close to a 3D Gaussian in “ideal conditions,” the optical aber-rations introduced by measuring inside the living cell make it possible to approximate the detection volumes by cylinders.
Mark A. Hink
149
Then the amplitude of the autocorrelation function, G(0), equals to γ/N. In all the equations used in this chapter, cylin-drical detection volumes are assumed with γ = 1.
7. Typically one observes a significant intensity drop during the first few seconds of the measurement after which the signal stabilizes. Often this is caused by photobleaching of molecules that are immobile or present in protein synthesis or degrada-tion organelles like the endoplasmic reticulum (ER) or lyso-somes. In order to distinguish this effect from a continuous photobleaching of the mobile molecules being studied, one can block the excitation light for a few seconds after which the measurement is repeated. When a signal recovery is being observed, the bleaching of the mobile molecules should be lowered using less excitation power.
Acknowledgments
The author would like to thank Kevin Crosby, Max Tollenaere, Marten Postma, and Ronald Breedijk for their assistance during the experiments. This work was supported by Middelgroot and Echo investment grants from the Netherlands Organisation for Scientific Research (NWO).
References
1. Haustein E, Schwille P (2007) Fluorescence correlation spectroscopy: novel variations of an established technique. Annu Rev Biophys Biomol Struct 36:151–169
2. Meseth U, Wohland T, Rigler R et al (1999) Resolution of fluorescence correlation mea-surements. Biophys J 76:1619–1631
3. Schwille P, Meyer-Almes FJ, Rigler R (1997) Dual-color fluorescence cross-correlation spectroscopy for multicomponent diffusional analysis in solution. Biophys J 72:1878–1886
4. Chen Y, Müller JD, So PTC et al (1999) The photon counting histogram in fluorescence fluc-tuation spectroscopy. Biophys J 77:553–567
5. Digman MA, Wiseman PW, Choi C et al (2009) Stoichiometry of molecular complexes at adhesions in living cells. Proc Natl Acad Sci U S A 106:2170–2175
6. Kolin DL, Wiseman PW (2007) Advances in image correlation spectroscopy: measuring n umber densities, aggregation states, and dynam-ics of fluorescently labeled macromolecules in cells. Cell Biochem Biophys 49:141–164
7. Kapusta P, Wahl M, Benda A et al (2007) Fluorescence lifetime correlation spectroscopy. J Fluoresc 17:43–48
8. Barcellona ML, Gammon S, Hazlett T et al (2004) Polarized fluorescence correlation spectroscopy of DNA-DAPI complexes. Microsc Res Tech 65:205–217
9. Price ES, DeVore MS, Johnson CK et al (2010) Detecting intramolecular dynamics and multiple Förster resonance energy transfer states by fluorescence correlation spectroscopy. J Phys Chem B 114:5895–5902
10. Hassler K, Leutenegger M, Rigler P et al (2005) Total internal reflection fluorescence correlation spectroscopy (TIR-FCS) with low background and high count-rate per molecule. Opt Express 13:7415–7423
11. Wohland T, Shi X, Sankaran J et al (2010) Single plane illumination fluorescence correla-tion spectroscopy (SPIM-FCS) probes inho-mogeneous three-dimensional environments. Opt Express 18:10627–10641
12. Leutenegger M, Ringemann C, Lasser T et al (2012) Fluorescence correlation spectroscopy with a total internal reflection fluorescence STED microscope (TIRF-STED-FCS). Opt Express 20:5243–5263
13. Needleman DJ, Xu Y, Mitchison TJ (2009) Pin-hole array correlation imaging: highly
Fluorescence Correlation Spectroscopy
150
parallel fluorescence correlation spectroscopy. Biophys J 96:5050–5059
14. Hink MA (2012) Single-molecule microscopy using silicone oil immersion objective lenses. Biomed Scientist 2012:83–85
15. Dertinger T, Pacheco V, von der Hocht I et al (2007) Two-focus fluorescence correlation spectroscopy: A new tool for accurate and absolute diffusion measurements. ChemPhys Chem 8:433–443
16. Müller BK, Zaychikov E, Bräuchle C et al (2005) Pulsed interleaved excitation. Biophys J 89:3508–3522
17. Koppel DE (1974) Statistical accuracy in fluo-rescence correlation spectroscopy. Phys Rev A 10:1938–1945
18. Maeder CI, Hink MA, Kinkhabwala A et al (2007) Spatial regulation of Fus3 MAP kinase activity through a reaction-diffusion mecha-nism in yeast pheromone signaling. Nat Cell Biol 9:1319–1326
19. Foo YH, Naredi-Rainer N, Lamb DC et al (2012) Factors affecting the quantification of biomolecular interactions by fluorescence cross-correlation spectroscopy. Biophys J 102:1174–1183
20. Bacia K, Petrášek Z, Schwille P (2012) Correcting for spectral cross-talk in dual-color fluorescence cross-correlation spectroscopy. ChemPhysChem 13:1221–1231
21. Eggeling C, Widengren J, Brand L et al (2006) Analysis of photobleaching in single-molecule multicolor excitation and Förster resonance energy transfer measurements. J Phys Chem A 110:2979–2995
22. Shcherbakova D, Hink MA, Joosen L et al (2012) An orange fluorescent protein with a large Stokes shift for single-excitation multi-color FCCS and FRET imaging. J Am Chem Soc 134:7913–7923
23. Yu J, Christina Wjasow C, Backer JM (1998) Regulation of the p85-p110a phosphatidylino-sitol 3′-kinase. J Biol Chem 273:30199–30203
24. Skakun VV, Hink MA, Digris AV et al (2005) Global analysis of fluorescence fluctuation data. Eur Biophys J 34:323–334
25. Gregor I, Patra D, Enderlein J (2005) Optical saturation in fluorescence correlation spectros-copy under continuous-wave and pulsed exci-tation. ChemPhysChem 6:164–170
26. Kaputsa P (2010) Absolute diffusion coeffi-cients: compilation of reference data for FCS calibration. Picoquant application note:
http://www.picoquant.com/technotes/appnote_diffusion_coefficients.pdf
27. Adjobo-Hermans MWJ, Goedhart J, van Weeren L et al (2011) Real-time visualization of heterotrimeric G protein Gq activation in living cells. BMC Biol 9:32
28. Chen Y, Johnson J, Macdonald P et al (2010) Observing protein interactions and their stoi-chiometry in living cells by brightness analysis of fluorescence fluctuation experiments. Methods Enzymol 472:345–363
29. Haupts U, Maiti S, Schwille P et al (1998) Dynamics of fluorescence fluctuations in green fluorescent protein observed by fluorescence correlation spectroscopy. Proc Natl Acad Sci U S A 95:13573–13578
30. Schwille P, Kummer S, Heikal AA et al (2000) Fluorescence correlation spectroscopy reveals fast optical excitation-driven intramolecular dynamics of yellow fluorescent proteins. Proc Natl Acad Sci U S A 97:151–156
31. Hendrix J, Flors C, Dedecker P et al (2008) Dark states in monomeric red fluorescent proteins studied by fluorescence correlation and single molecule spectroscopy. Biophys J 94:4103–4113
32. Subach OM, Cranfill PJ, Davidson MW et al (2011) An enhanced monomeric blue fluores-cent protein with the high chemical stability of the chromophore. PLoS One 6:e28674
33. Goedhart J, von Stetten D, Noirclerc-Savoye M et al (2012) Structure-guided evolution of cyan fluorescent proteins towards a quantum yield of 93 %. Nat Commun 3:751
34. Kremers GJ, Goedhart J, van den Heuvel DJ et al (2007) Improved green and blue fluores-cent proteins for expression in bacteria and mam-malian cells. Biogeosciences 46:3775–3783
35. Kremers GJ, Goedhart J, van Munster EB et al (2006) Cyan and yellow super fluorescent pro-teins with improved brightness, protein fold-ing, and FRET Förster radius. Biogeosciences 45:6570–6580
36. Shaner NC, Campbell RE, Steinbach PA et al (2004) Improved monomeric red, orange and yellow fluorescent proteins derived from Discosoma sp. red fluorescent protein. Nat Biotechnol 22:1567–1572
37. Shaner NC, Lin MZ, McKeown MR et al (2008) Improving the photostability of bright monomeric orange and red fluorescent pro-teins. Nat Methods 5:545–551
38. Shcherbo D, Murphy CS, Ermakova GV et al (2009) Far-red fluorescent tags for protein imag-ing in living tissues. Biochem J 418:567–574
Mark A. Hink
151
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_9, © Springer Science+Business Media New York 2015
Chapter 9
Homo-FRET Imaging Highlights the Nanoscale Organization of Cell Surface Molecules
Suvrajit Saha*, Riya Raghupathy*, and Satyajit Mayor
Abstract
Several models have been proposed to understand the structure and organization of the plasma membrane in living cells. Predicated on equilibrium thermodynamic principles, the fluid-mosaic model of Singer and Nicholson and the model of lipid domains (or membrane rafts) are dominant models, which account for a fluid bilayer and functional lateral heterogeneity of membrane components, respectively. However, the con-stituents of the membrane and its composition are not maintained by equilibrium mechanisms. Indeed, the living cell membrane is a steady state of a number of active processes, namely, exocytosis, lipid synthesis and transbilayer flip-flop, and endocytosis. In this active milieu, many lipid constituents of the cell membrane exhibit a nanoscale organization that is also at odds with passive models based on chemical equilibrium. Here we provide a detailed description of microscopy and cell biological methods that have served to provide valu-able information regarding the nature of nanoscale organization of lipid components in a living cell.
Key words (Homo)-FRET, Fluorescence, Anisotropy, Microscopy, Membranes
1 Introduction
The living cell surface is a complex ensemble of diverse lipid and protein molecules, where specific lipid-protein and protein-protein interactions trigger signal transduction pathways controlling the cellular physiology. The molecular organization and such interac-tions occur at a length scale of few nanometers and pose a chal-lenging question to study such processes in the living cells. Although light microscopy has allowed visualizing the cells in live, studying the nanoscale presents a physical problem for conven-tional diffraction limited optics, which cannot access the length scales below the optical resolution of ~200–300 nm. Two comple-mentary approaches have evolved to circumvent this issue. Recent developments in super-resolution imaging modalities (like STED, STORM, PALM, NSOM, SIM) have pushed the limits of the
* Authors contributed equally.
152
resolution of light microscopy to as low as 20 nm [1]. On the other hand, several parallel fluorescence-based techniques have been developed in the past couple of decades that have been used to examine protein-protein interactions in living cells [2–4]. Here we will discuss a fluorescence resonance energy transfer (FRET)-based approach, homo-FRET, and how it can be used to under-stand the nanoscale organization of cell surface molecules.
FRET (fluorescence resonance energy transfer or Förster reso-nance energy transfer) is the process of non-radiative energy trans-fer between a donor fluorophore in excited state and an acceptor fluorophore in ground state. The theoretical description of FRET [5] was developed by Theodor Förster in the 1940s. Energy trans-fer between the donor and the acceptor species is a result of dipolar interaction between them and is governed by the extent of spectral overlap between the emission spectrum of the donor species and the absorption spectrum of the acceptor, the quantum yield of the donor, and the relative orientation of the donor and acceptor tran-sition dipoles (Fig. 1a, b). Due to the dipolar nature of FRET, the efficiency of energy transfer is inversely proportional to the sixth power of distances between the fluorophores. The distance at which efficiency of energy transfer (E) is 50 % is referred to as Förster’s distance (R0), typically ranging from 1 to 10 nm depend-ing on the molecular and spectral properties of the fluorophores and the local environment. This makes FRET a suitable tool to measure distances at the biomolecular scale [6]:
E r R= + ( )é
ëùû1 1 0
6/ /
(1)
Here, r is the distance between the donor and acceptor fluoro-phores. While FRET is generally described as an energy transfer process between distinct donor and acceptor fluorophores (hetero- FRET), energy transfer can also occur between like fluorophores with low Stokes shift and hence a significant overlap between their absorption and emission spectra. Here we focus on the ways of monitoring FRET between like fluorophores, or homo-FRET, in live cells. Homo-FRET can be estimated by measuring the loss of the polarization of emitted fluorescence. Fluorescence emission polarization in turn may be monitored by determining fluores-cence emission anisotropy. This is measured by exciting fluoro-phores by plane-polarized illumination and collecting the emitted fluorescence in two orthogonal planes [7].
Fluorescence anisotropy (r) is defined by
r
I I
I I=
-
+pa pe
pa pe2,
(2)
where Ipa and Ipe are the emission intensities collected in the parallel and the perpendicular directions, respectively. Since
1.1 Homo-FRET and Fluorescence Emission Anisotropy
Suvrajit Saha et al.
153
fluorescence anisotropy (r) is normalized to the total intensity (denominator term), its readout can be considered as a sum of intensity-weighted fluorescence anisotropies of individual fluoro-phores. The polarized illumination allows selective excitation (photoselection) of a pool of fluorophores with excitation dipoles aligned to the plane of illumination. The depolarization (or loss of anisotropy) of the fluorescence emission of a population of photoselected fluorophores can be dictated by two processes, rotational diffusion and homo-FRET. In a dilute solution of fluo-rophores (say 1–10 μM), the high intermolecular distances
Fig. 1 Principle of homo-FRET process. (a) Angular orientation of donor (cyan) and acceptor (yellow-orange) transition dipoles are shown as θD and θA, respectively, where the intermolecular distance is shown as r. The extent of depolarization in the emission is dependent on the angle between the two transition dipoles, during the energy transfer process. (b) Spectral overlap integral J (λ) (shown in gray) between the donor emission (ED) and acceptor absorption (AA) is one of the factors determining the Förster distance, R0. (c) A pool of isolated fluorophores when excited by polarized light (in the absence of significant rotational motions) results in fluo-rescence emission being relatively polarized (top). When fluorophores are relatively close together (in circles below), homo- FRET can take place between two like fluorophores, if the fluorophores have a significant over-lap between their own excitation and emission spectrum, resulting in depolarized fluorescence emission and a lower value of anisotropy. Adapted with permission from Krishnan et al. [2] and Sharma et al. [18]
Homo-FRET Imaging of Nanoscale Organisation
154
(r ≫ R0) between individual fluorophores preclude homo-transfer, and hence, the anisotropy of emission is depolarized only by rota-tional diffusion. The rotational diffusion rates of molecules are governed by their size and local viscosity. If the rotational diffu-sion time is smaller than the lifetime of the probe, the probe can tumble multiple times during its lifetime and thereby depolarize the emission. The same probe when placed in a considerably more viscous medium will have much higher rotational diffusion time than its lifetime, and hence, the emission retains polariza-tion partially. The Perrin equation provides the relationship between rotational diffusion, viscosity and molecular size, and fluorescence anisotropy (r):
1 1
0 0r r
RT
r V= +
th
(3)
where T is the temperature, V is the hydrodynamic volume, η is the viscosity, τ is the rotational correlation time, and R is the uni-versal gas constant [7]. Here r0 is the fundamental anisotropy of the fluorophore in the absence of any rotational diffusion and aris-ing only from the photoselection and the angular displacement between its excitation and emission dipoles. Homo-FRET can further depolarize the fluorescence emission (and thereby reduce emission anisotropy) under conditions of low inter-fluorophore distances (r < R0) and significant spectral overlap. This is because photoselected fluorophores can transfer energy to acceptors with different relative angular orientations of their transition dipoles. Moreover, the acceptors can also undergo rotational tumbling during its fluorescence lifetime, thereby reducing the emission anisotropy of the system considerably (Fig. 1c). Since both the acceptor and donor are spectrally similar in homo-FRET, a precise measurement of the extent of loss in fluorescence emission anisot-ropy is the only way to monitor homo-FRET.
In most experiments, a comparison of the fluorescence anisot-ropy of an experimental sample is made with situations where there is no FRET. Like hetero-FRET, an expression for homo-FRET efficiency can make quantitative comparisons across different mea-surements convenient. One can use an expression based on mea-suring anisotropies rc and rm of donor fluorescence in the presence or absence of FRET conditions, respectively, i.e.:
E
r
r= -1 c
m (4)
It must be noted that the above equation is valid only under simple assumptions that homo-FRET is the only agent for the change in anisotropy and where excitation after leaving the donor never returns to the same donor species and where there is no change in the donor lifetimes [5].
Suvrajit Saha et al.
155
The spatiotemporal interaction of like fluorophores inside living cells can be imaged by implementing the measurement of fluorescence anisotropy in a microscope configuration. In subse-quent sections we describe how homo-FRET imaging is carried out in living cells to look at the nanoscale organization of cell sur-face molecules.
2 Materials
All the buffers and media are prepared in autoclaved Milli-Q- filtered water and filter sterilized prior to storage:
1. Medium 1(M1) buffer: 150 mM NaCl, 20 mM HEPES, 5 mM KCl, 1 mM CaCl2, 1 mM MgCl2, pH 7.2–7.4.
M1 can be used as the buffer system for washing, labeling, and imaging cells (both live and fixed) for fluorescence imag-ing. M1 should be supplemented with 2 mg/ml of glucose (M1-Glc) for live cell imaging.
2. Phosphate buffer saline (PBS): 137 mM NaCl, 2 mM KCl, 10 mM Na2HPO4, 2 mM KH2PO4, pH 7.4.
PBS is widely used for passaging of cells and clearing cover-slips and in immunofluorescence experiments with CHO cells.
3. Cell culture media: Chinese hamster ovary (CHO) is maintained in Ham’s F12 medium (HiMedia). The cells expressing folate receptor are specifically maintained in a folate-free variant of HF-12, which also needs to be supplemented by dialyzed serum (free of folate) when labeling the receptor with folate analogs.
1. CHO cell line stably expressing folate receptor (FR-GPI), also referred to as IA2.2F, is used to study the native nanoscale organization of FR-GPI. They are also used for labeling the cells with different lipid probes to look at their specific organization.
2. CHO cell line stably expressing EGFP-GPI, also referred to as GG8. These cells do not need any exogenous labeling as they are expressing GPI-anchored fluorescent protein. Instead, they need to be treated with cycloheximide (50–75 μg/ml) for 3 h prior to imaging to clear the pool of the protein residing in internal organelles like the Golgi and the endoplasmic reticulum.
1. Hellmanex III (Hellma Analytics): An alkaline cleaning agent for glass surfaces.
2. Piranha solution: Conc. H2SO4, 30 % H2O2 (Merck). 3. Phosphate buffer saline (PBS).
2.1 Buffers and Media
2.2 Cell Lines
2.3 Reagents for Cleaning Coverslips
Homo-FRET Imaging of Nanoscale Organisation
156
1. Ham’s F12 media without phenol red (HiMedia, India) are reconstituted in water with NaHCO3 (Merck, India) as per manufacturer’s instructions (see Note 1).
2. Cell de-adhering buffer: 5 mM EDTA (Na+ salt, Merck) in Ca2+- and Mg2+-free M1 (150 mM NaCl, 20 mM HEPES, 5 mM KCl, pH 7.4).
3. Human/bovine fibronectin (Sigma-Aldrich), used at 1 mg/ml, maintained as 10 μl aliquots at −20 °C or 4 °C as per man-ufacturer’s instructions.
HPLC grade chloroform, methanol, ethyl acetate, acetone, water (SD-Fine Chemicals, India).
Lipids are dissolved in chloroform/methanol (9:1) or ethyl acetate/acetone /methanol/water (7:1:1:1) based on their hydro-philicity, aliquoted into glass vials or Eppendorf centrifuge tubes (see Note 2), dried under nitrogen (see Note 3), and stored under nitrogen/argon at −80 °C.
1. Lipofectamine (Invitrogen). 2. Fatty acid-free BSA (Sigma-Aldrich). 3. γ-Cyclodextrin (Sigma-Aldrich).
1. Folate analogs: Folate analog (pteroyl lysine) is synthesized in the laboratory and conjugated to commercially available organic fluorophores like FITC, BODIPY-TMR, or BODIPY-FL (Molecular Probes) to generate the analogs of varying spectral ranges. These analogs can be used for the cell surface labeling of FR-GPI.
2. Lipid analogs: The fluorescent analogs (Molecular Probes, Avanti Polar Lipids) of sphingomyelin (C5-BODIPY-FL SM, chain labeled) and phosphatidylethanolamine (BODIPY-FL- DHPE, head labeled) have been used to study the nanoscale organization of incorporated probes.
1. Transferrin (Tf): Apo-transferrin (Sigma-Aldrich) is iron loaded, purified, and stored at 4 °C [8]. 1 mg of iron loaded transferrin is conjugated to Alexa 647 (Molecular Probes) such that it gives a dye to protein ratio of 1:1, followed by purifica-tion by size exclusion chromatography, and stored at 4 °C. Conjugated Tf was pulsed at 10 μg/ml concentration.
2. 10 kDa dextran (Invitrogen) conjugated to FITC as reported in [9] is stored at 4 °C and pulsed at 1 mg/ml.
2.4 Reagents for Replating Method
2.5 Solvents for Lipid Storage
2.6 Reagents for Lipid Incorporation
2.7 Reagents to Label the Cell Membrane
2.8 Reagents to Mark Intracellular Compartments
Suvrajit Saha et al.
157
3 Methods
Here we will highlight in a stepwise manner the different stages of making a homo-FRET measurement, from preparing the cells, set-ting up the microscope, followed by the analysis and interpretation of the data.
Homo-FRET is a sensitive technique, and hence, it is very impor-tant to minimize the background signals emanating due to non-specific sticking of fluorescent probes on glass coverslips. For this purpose, prior to plating cells, we resort to one of the following methods to clean the coverslips.
This is a 3:1 mixture of concentrated sulfuric acid (conc. H2SO4) and hydrogen peroxide (H2O2), which forms a strong oxidizing agent to remove organic residues from the coverslip. This solution is prepared fresh each time by adding conc. H2SO4 drop by drop (! caution: exothermic reaction!) into cold H2O2 solution in a bea-ker and with frequent mixing. To this piranha solution, coverslips are added and incubated for 30 min followed by thorough clean-ing with double-distilled water five times and then once with PBS. The treated coverslips are air-dried and UV exposed prior to plating cells for experiment.
This is a strong alkaline reagent used to clean glass surfaces of cov-erslips and cuvettes. Coverslips are incubated with 1 % Hellmanex III solution (in ion-free water, double-distilled water) at room temperature for 45 min and then washed thoroughly with copious amounts of double-distilled water five times and then one final wash with PBS. The coverslips are air-dried and UV exposed prior to plating cells for experiment (see Note 10).
In case of certain lipids especially BODIPY-tagged lipids, cleaning the coverslips by the above mentioned methods might not be enough to minimize nonspecific sticking, and in such cases, we resorted to replating method. In this method, post-lipid incorpo-ration, the cells were de-adhered and replated on to fresh cover-slips coated with fibronectin for the cells to adhere faster.
1. Fibronectin plating on coverslip dishes: Coverslips were incu-bated with fibronectin (10 μg/ml) at 37 °C for 1 h or 4 °C overnight. Excess fibronectin was gently removed and the coverslips were rinsed once with PBS before use.
3.1 Cleaning Coverslips to Minimize Background Noise Due to Sticking of Probes
3.1.1 Piranha Solution
3.1.2 Hellmanex III
3.2 Methods to Get Rid of Lipid Sticking on Coverslips
Homo-FRET Imaging of Nanoscale Organisation
158
2. Protocol: Cells grown on coverslip for 24 h and at 80 % confluence are incubated in 5 mM EDTA solution (in Ca2+- and Mg2+-free M1 or PBS, pH 7.2–7.4) for 10 min at 37 °C. During this time the cells will round up and float in the buffer, failing which it is advisable to tap gently on the sides of the coverslip (see Note 8). The cells floating in the buffer are then transferred to a 1.5 mL Eppendorf tube (see Note 9) and centrifuged at room temperature or 4 °C (as per requirement) at 82 × g for 10 min. The supernatant is carefully removed and the pellet containing cells are resus-pended in Ham’s F12 media. These cells are then trans-ferred on to fibronectin-coated dishes, followed by incubation at 37 °C for 30 min. During this time the cells will re-adhere to the coverslip completely, ready to be used for microscopic imaging.
One can use a combination of methods described in Subheadings 3.1 and 3.2 to get rid of the background sticking (Fig. 2).
CHO cells expressing folate receptor are plated in HF-12 (with dialyzed serum) at a density such that they become semi-confluent (70–80 %) in 36–42 h. The media are removed from the dishes, and the cells are washed 2–3 times gently with cold 1× M1-Glc buffer. Cells can be labeled with folate analog resuspended in M1 at saturating concentrations (~100–500 nM) on ice for 45 min [10]. Following this, the labeling mix is removed, and the cells are washed gently two times with cold M1-Glc and further imaged live in the same buffer at the desired temperatures (Figs. 2 and 3).
The power of the homo-FRET technique also lies in the fact that it can be easily extended to making measurements on fluo-rescent protein-based chimeric constructs, like EGFP-GPI, which is a lipid-tethered protein we have extensively studied. Since most of the GFP-tagged proteins in cells are fluorescent, there is considerable fluorescence from the biosynthetic pool of GFP-tagged proteins present in ER, Golgi, and secretory vesi-cles. This is undesirable when measurements are made on GFP-tagged proteins (GFP-GPI in our case) present on cell surface. Cycloheximide treatment is used to deplete GFP-tagged pro-teins from its biosynthetic pool. For GPI-APs, conditions required for depletion of proteins from biosynthetic pool was found to be highly cell type specific. In CHO cells, 50–75 μg/ml of cycloheximide reconstituted into plating medium for 3 h could remove Golgi-localized GPI-APs from >90 % of the cells [9]. Post-cycloheximide treatment, the media from the dishes are removed, and the dishes are washed 2–3 times gently with cold M1-Glc buffer. The experiments are done using M1-Glc as the imaging buffer at the desired temperature.
3.3 Preparing Cells for Microscopy
3.3.1 Exogenous Labeling of Cells with Folate Analogs
3.3.2 Preparing Cells Expressing Fluorescent Protein Constructs
Suvrajit Saha et al.
159
Lipid molecules are amphipathic in nature due to the presence of hydrophilic head group and hydrophobic lipid tails. This property of lipid molecules poses a problem for exogenous incorporation into native cell membrane. This problem can be solved by using a vehicle to transfer lipids into the cell membrane, which can be lipids that help in the formation of micelles, lipid transfer pro-teins or lipid trapping organic molecules. Based on this, lipids are incorporated into the cell membrane by three different methods. In particular, we introduce lipofectamine-based lipid incorpora-tion as a novel method: (1) BSA method, (2) γ-CD method, and
3.3.3 Exogenous Incorporation of Lipids into Cells
Fig. 2 Flowchart of cell surface labeling strategies. Protocols for exogenous labeling of FR-GPI (with fluores-cent folate analogs) and incorporation of fluorescent lipid probes via three different methods, using γ-CD, defatted BSA, and lipofectamine
Homo-FRET Imaging of Nanoscale Organisation
160
(3) lipofectamine method. In each of these methods, prior to incubation of lipid complexes with cells, it is important to get rid of the serum components which itself are capable of trapping lip-ids. For this purpose the cells are incubated in Ham’s F12 media for 15 min at 37 °C prior to incorporation of lipids.
1. BSA method: Bovine serum albumin (BSA) contains lipid pock-ets that can stably incorporate lipid molecules. The BSA-lipid complexes are incubated on cells to exchange the lipids from BSA to that of the cell membrane [11] and hence act like lipid transfer proteins. However, the BSA-lipid complex formation is dependent on the fatty acid chain length. The longer the chain length, the higher the affinity to BSA and the lower the efficiency of lipid transfer to the membrane [12, 13] (see Note 4). This method works efficiently for lipids with short fatty acid tails:(a) Preparation of BSA-lipid complexes: Fatty acid-free BSA,
dissolved in Ham’s F12 media (20 μM), is added to dried lipids such that final BSA-lipid complex attains BSA to lipid ratio 1:x (x > 1 for short lipid tails C8–C12 and x < 1 for long lipid tails). The mix was vortexed vigorously or probe sonicated (3 × 2 s) for three times (see Note 5).
(b) Protocol for lipid incorporation as BSA-lipid complexes: Firstly to get rid of serum components, the cells are incu-bated in Ham’s F12 media for 15 min at 37 °C followed by incubation with BSA-lipid complexes for 30 min either on ice for short-chain lipids or at 37 °C for long-chain lipids (see Note 11, Fig. 3b)
2. γ-Cyclodextrin (γ-CD) method: γ-Cyclodextrin is a cyclic oli-gosaccharide made up of 8 γ-d-glucopyranoside units. The glucose units are linked to each other by means of 1–4 linkages such that it becomes hydrophobic within and hydrophilic
Fig. 3 Cell surface labeling of lipid probes and lipid-tethered proteins. Representative images of CHO cells labeled exogenously with fluorescent folate analog (PL-BFL) to mark the surface expressed folate receptor (FR-GPI) (a) or incorporated with lipid probes like BODIPY-FL-DHPE (BFL-DHPE) by either γ-CD method (b) or defatted BSA (c) method and C6-NBD-SM by lipofectamine method (d) and imaged on a custom-built TIRF- anisotropy platform. Scale bar 5 μm
Suvrajit Saha et al.
161
outside the cyclic ring. The hydrophobic cavity of the ring is thus suitable to carry hydrophobic fatty acid moieties. Thus, γ-CD acts as an efficient transfer reagent for lipid molecules into the cell membrane. This method works well for lipids with long fatty acid tails [14]:(a) Preparation of γ-CD-lipid complexes: γ-CD is dissolved
in Ham’s F12 media at millimolar concentrations and mixed with lipids at micromolar concentrations such that the final γ-CD to specific lipid molar ratio (see Note 6) [14]. This mixture is further probe sonicated for 3 × 2 s for three times.
(b) Protocol for lipid incorporation as γ-CD-lipid complexes: Incubate γ-CD-lipid complexes with cells at 37 °C for 30 min (see Note 12, Fig. 3a).
3. Lipofectamine method: Lipofectamine consists of a mixture of cationic and neutral co-lipids to assist nucleic acid molecules in bypassing the cell membrane. We have developed this new method to deliver fluorescent lipid analogs using Lipofectamine:(a) Preparation of lipofectamine-lipid complexes: 2 μl of lipo-
fectamine reagent was added to 98 μl Ham’s F12 media to obtain 100 μl of labeling complex (see Note 7). 20 μl of this solution was added to dried lipid (0.2 mM, 10 μl) and incubated at room temperature for 30 min to form lipofectamine- lipid complexes (18 μg/ml of lipofectamine lipids). After the incubation, this solution is made up to 100 μl to give 20 μM lipofectamine-lipid complex solu-tion. This lipid solution is further diluted to desired lipid concentrations prior to use.
(b) Protocol for lipid incorporation as lipofectamine-lipid complexes: The cells are incubated with lipofectamine-lipid complexes at 10 °C for 10–30 min, lower incubation time required to incorporate short-chain lipids and higher incubation time for long-chain lipids (see Note 13, Fig. 3d).
The final concentration of lipofectamine used on cells is 18 μg/ml and does not change the membrane composition/fluidity, as measured from folate receptor membrane organization and traffick-ing. We find that neither the cell surface organization and diffusion nor the trafficking of GPI-AP is perturbed at this concentration of lipofectamine reagent and hence can be used as a suitable reagent for lipid incorporation into plasma membrane of live cells.
Fluorescence anisotropy measurements (homo-FRET imaging) have been implemented on a wide range of available fluorescence micros-copy platforms. To be specific, wide-field, single-photon confocal (point scanning, line scanning, and spinning disk), multiphoton, and
3.4 Microscopy Setup for Homo- FRET Imaging
Homo-FRET Imaging of Nanoscale Organisation
162
evanescent field (TIRF)-based imaging modalities have been successfully customized for high-resolution fluorescence anisotropy measure-ments [15]. Anisotropy measurements involve exciting the sample with a polarized illumination and detecting the emitted fluorescence in the planes parallel and perpendicular to the axis of polarized excita-tion. Here, we outline the basic requirements for setting up any fluo-rescence microscope (wide-field or confocal) for real-time simultaneous detection of both the polarized components of emission.
A key requirement of anisotropy measurement setups is a stable excitation source with a defined polarization output. This can be met by the following illumination options:
1. Mercury arc lamp: A very bright and broad-spectrum (UV-visible) output makes mercury arc lamps a suitable candi-date for wide-field fluorescence microscopy. However, the emis-sion from this source is not intrinsically polarized. Hence, the light produced by the lamp is collimated and selected for a spe-cific wavelength using an excitation filter before it passes through a sheet excitation polarizer (of high extinction ratio of 10−3 for visible white light) and the p-polarized component of the excitation is selected. An important issue with the arc lamps is their intrinsic temporal variability (~2–3 %) in output. This is detrimental to the anisotropy measurements by sequential detection of the polarization [16]. The real-time two-camera detection of both the orthogonal polarization components, which we discuss here, circumvents the problem, as these fluc-tuations affect both parallel and perpendicular image similarly, without significantly affecting the anisotropy measurements.
2. Lasers: Lasers are often preferred sources of illumination for confocal microscopy-based anisotropy imaging setups and in TIRF microscope. Lasers are monochromatic coherent light sources which are often intrinsically polarized [17]. The light produced by many lasers is linearly polarized, making them ideal for fluorescence polarization anisotropy measurements. The output at the laser combiner is coupled directly onto a polarization preserving optical fiber. An essential require-ment for fluorescence anisotropy measurement is to ensure polarized excitation at the sample plane with extinction coef-ficient 500–1,000:1, which is the ratio of the intensities of the two orthogonal polarized components of the excitation (as measured by a power meter on the sample plane). This is a direct measure of preservation of polarized excitation through the light path and should be carefully monitored regularly as a change in the value of this ratio can affect the anisotropy measurements. A wide range of laser sources have been used, e.g., gas lasers and solid-state diode lasers for sin-gle-photon microscopy.
3.4.1 Polarized Illumination Source
Suvrajit Saha et al.
163
In an epifluorescence setup, the excitation light (from sources described above) is delivered to the sample via the objective, and the fluorescence emission is being collected by the same objective and is selected for a specific spectral band comprising of the specific emission. This is done by a dichroic mirror (DM) followed by an appropriate emission filter (EM). To detect emission anisotropy, the fluorescence is then passed through a polarizing beam splitter (PBS) splitting it into its two orthogonal components Ipa and Ipe which may be then imaged simultaneously using suitable detectors, allowing real-time measurement of anisotropy. In this section, we present a brief description of the key optical elements specifically used for anisotropy imaging:
1. Polarizing beam splitter (PBS): A polarizing beam splitter (PBS) separates the incident emission light into two principal orthog-onal components; while the p-plane is transmitted, the s-plane is reflected (Fig. 4a). Hence, the PBS makes possible the simul-taneous detection of both the polarizations using two detec-tors, allowing real-time anisotropy imaging. The PBS should be placed after the microscope-detection port, as close to the microscope as possible to avoid light losses. Moreover, focusing light rays must be avoided since they will give rise to major aberrations in polarization at different parts of the image lead-ing to artifacts. Light should be incident on the beam splitter coating at the Brewster’s angle of 45 ± 2°. Hence, only a colli-mated beam should be incident on the PBS. There are broadly two designs of PBS used for anisotropy imaging, the broadband polarizing cube beam splitter (CVI Melles Griot, USA) and the nanowire-based sheet polarizing beam splitter (ProfluxTM Beam Splitter, MOXTEK Inc., USA).
2. Detectors for imaging: The fluorescence emission split into the p- and s-plane-polarized components at the PBS travels onward as a collimated beam, which needs to be focused at the detec-tor to generate the image. Hence, two tube lenses of similar focal lengths are fitted equidistant from the beam splitter in the path of the two orthogonally polarized components of the emission (Fig. 4b). The two detectors are placed at a distance, equivalent to the focal length of the tube lenses. Simultaneous detection of both the polarizations is done by a wide range of detectors, chiefly depending on the anisotropy measurement modality in use. On imaging setups for measuring steady-state anisotropy such as the wide-field, TIRF, spinning disk, or the line-scanning confocal system, a cooled CCD or EM-CCD is used. Two key requirements in implementing real-time dual detector imaging are the synchronization of acquisition between the two detectors to minimize time delay in simulta-neous real-time imaging and a high degree of spatial image registration between both the detectors. Cameras used for
3.4.2 Real-Time Detection of the Two Orthogonal Planes of Emission
Homo-FRET Imaging of Nanoscale Organisation
164
image acquisition play an important role in determining the signal to noise obtained in anisotropy measurements and have to be chosen for the highest possible quantum yield and lowest possible system noise. High- performance camera systems uti-lize design enhancements that dramatically reduce read noise.
Fig. 4 Basics of homo-FRET microscopy. The key component of a real-time homo-FRET microscopy platform is a polarizing beam splitter (PBS) cube (a), which divides the non-polarized collimated light into two orthogonal polarized beams at the diagonal interface (gray with stripes, a) facing the incoming fluores-cence emission. Generally the p-polarized light is transmitted, while the s-polarized light is reflected. Schematic representation (b) of wide-field fluorescence anisotropy measurement setup showing the indi-vidual components: (i) mercury arc lamp, (ii) excitation filter, (iii) sheet polarizer oriented to select p-polar-ized light, (iv) dichroic mirror with the emission filter (in detection side), (v) objective, (vi) side port prism, (vii) PBS, (viii) tube lens pair, (ix) cameras—EMCCDs. Although this is a very simple representation, the fundamental layout of components is conserved among the different microscopy modalities like confocal and TIRFM, which have been adapted for homo-FRET imaging
Suvrajit Saha et al.
165
Here we list some of the basic calibrations and measurements, which are necessary for interpretable anisotropy measurements. Their calibration also allows one to assess the suitability of the microscope for such measurements:
1. G-factor: The intrinsic differences between the dichroics, fil-ters, and other optical elements in terms of their ability to transmit s- and p-plane-polarized light result in the biased sen-sitivity of detection of one of the polarized component of the emission. Moreover, there can be differences in the relative sen-sitivities of the two detectors for real-time simultaneous imaging configurations. These can introduce potential artifacts, if not accounted for by a correction factor which estimates this bias correctly. This correction factor, referred to as the G-factor, should be introduced for quantitative analysis of such images [7]. The G-factor is defined as the bias in detection of parallel Ipa and perpendicular Ipe intensities:
G I Ifac pa pe= /
Typically, aqueous solution (1–10 μM) of fluorescein (at pH 11) can be used to estimate the G-factor. In water, the fluorescein molecules are isotropically distributed and have a rotational correlation timescale of ~120 ps and tumble rapidly many times within the fluorophore lifetime (~4 ns) to give emission anisotropy values close to zero at room temperature. During image analysis, the G-factor correction is performed on images by multiplying the perpendicular (Ipe) image (after background correction) to the G-factor image or value. The value of the G-factor varies between different setups and largely depends on the nature of optical elements and detectors used.
2. Flat-fielding of the illumination: Most of the quantitative imag-ing tools assume uniform illumination or flat-fielding across the field of view. Illumination gradients are quite common and can arise from the illumination source or the optical components in the detection and sometimes from both. This must be checked and corrected before any intensity measurements are made. This can be easily checked by imaging a solution of fluoro-phores, which is homogenous, and hence, the probe is uni-formly distributed. Generally, the same sample used for G-factor measurement meets this requirement and both flat-fielding and G-factor can be assessed from the same image.
3. Extinction ratio of the emission path: The extinction ratio of the PBS is calculated as the efficiency in transmitting p- polarized light or that of reflecting s-polarized light (Fig. 4a). To deter-mine the extinction ratio of the entire setup, a polarizer is placed on the microscope stage aligned to select p- or s-polar-ized light from the bright field source. Alternatively, the laser light can be reflected back into the microscope and sent out to the PBS. Light intensities collected by a low NA objective
3.4.3 Microscope Calibration and Characterization
Homo-FRET Imaging of Nanoscale Organisation
166
(typically less than 0.7) would be nearly perfectly polarized. The net extinction of the setup should be above 95 %.
4. Alignment of the imaging setup: System alignment is done by accurate placement of the different optical component and the detector in the light path. The PBS should be set up to face the emitted light beam at exactly 45°, the two tube lenses are placed equidistant from the PBS, and the detector is placed exactly at their focus. Alignment of the two orthogonal orien-tations is an important requirement for imaging and deter-mines the overall resolution of the imaging setup. The initial alignment is a time- consuming process, where each optical component as well as the detectors is moved manually, one at a time, and each of the alignment parameters such as the G-factor, extinction ratio, flat- fielding, and coincidence of sub-resolution beads is checked iteratively to obtain a best possible alignment (Fig. 5a). Since the real-time anisotropy setups have different light path for the two detectors, one might encounter more complicated distortions arising from chromatic or spher-ical aberrations as well as due to imprecise placement of optical components in the light path leading to a variety of defects such as skewness, shearing, tilt, rotation, translation, or a com-bination of these in different parts of the image. This is done by simultaneously imaging sub- resolution beads and assessing their spatial overlap to determine the nature of defects. Accordingly, the PBS and the detectors are moved to attain maximal alignment.
5. Steady-state anisotropy setup characterization: It is essential to finally characterize the setup for its sensitivity in detecting changes in anisotropy. A standard set of anisotropy measure-ments are made which should yield expected results based on theoretical calculations. A good test is to monitor the change in emission anisotropy of Rhodamine 6G in a glycerol-water solution as a function of increasing molar concentration of Rhodamine 6G (Fig. 5c). This will directly read out the extent of homo-FRET due to the decrease in average intermolecular distances at higher concentrations. Alternatively, one can also measure the anisotropy of a solution of fluorophore such as Rhodamine 6G dissolved in a glycerol-water mixture with vary-ing concentrations of glycerol, to estimate the effect of solvent viscosity (Fig. 5d). Soluble EGFP in phosphate-buffered saline (pH 7.4), with a mass of 27 kDa and high rotational correlation timescale (~17 ns at room temperature), has a polarized emis-sion. It can also be used as an anisotropy standard for checking the setup on a day-to-day basis after instrument calibration. These simple checks are robust indicators of the sensitivity of the real-time anisotropy setup to measure small changes in the anisotropy, which can typically be attributed to homo-FRET.
Suvrajit Saha et al.
167
A typical homo-FRET experiment can be divided into the follow-ing steps:
1. The illumination source, microscope, and the cameras are switched on, and the individual components of the light path are set up for the anisotropy measurement.
2. Sub-resolution bead sample is placed on the microscope, and images are acquired using the two detectors along the orthogonal
3.5 Image Acquisition, Processing, and Analysis
3.5.1 Image and Data Acquisition
Fig. 5 Microscope alignment and calibration. Sub-resolution bead samples are used to check the system align-ment and the image registration between the two cameras. On a well-aligned setup, it is possible to attain sub-pixel overlap precision, as shown (a). A dilute solution of fluorophores (like FITC in water, pH 11) should be used to check the flat-fielding and calculate the G-factor which should be uniform across the field of view (b); scale bar is 10 μm for both (a) and (b). The setup can be characterized by measuring anisotropy changes observed between solutions of varying concentrations of Rhodamine 6G (R6G) in 70 % glycerol-water mixtures (c). A highly concentrated solution of R6G (1 mM) undergoes random collisional homo-FRET, which leads to decrease in the ensemble anisotropy values. Similarly the effect of solvent viscosity on the emission anisotropy (d) can be monitored using solutions of same concentration of R6G in varying glycerol-water mixtures, thereby changing the viscosity (say 50 and 70 % glycerol-water as shown here). In solution of lower viscosity (50 % glycerol-water), the tumbling of the dye molecule is fast, and hence, the anisotropy is lower; increasing glyc-erol concentrations (to 70 % glycerol-water) increase the overall fluorescence anisotropy. Moreover, soluble EGFP in PBS has a higher rotational correlation timescale and hence has a polarized emission, as reported by a high value of fluorescence anisotropy (d), compared to R6G solutions (bar represents mean ± s.d.). These measurements were all made using a custom-built spinning disk anisotropy platform
Homo-FRET Imaging of Nanoscale Organisation
168
polarization axis. The quality of alignment and image registration is judged by the extent of pixel overlap, and if required the optical components are moved to attain the best possible align-ment (ideally sub-pixel precision) (see Note14).
3. Flat-fielding of the illumination is checked using a homoge-nously mixed solution of fluorophores (say fluorescein), and the G-factor is estimated. The extinction ratio is also moni-tored as described earlier (see Notes 15 and 16).
4. Fluorescence anisotropy standards like Rhodamine 6G in vary-ing glycerol concentrations or EGFP solution are imaged, and the anisotropy is measured to check the system performance (see Note 17).
5. Cells are prepared for imaging (as described in Subheading 3.1) and shifted to the microscope stage, which is pre-maintained at the desired temperature for the imaging. The imaging medium should have minimal autofluorescence to keep the background counts low (see Note 18). The cells are randomly chosen for imaging by a quick survey of the dish using a weak-illumina-tion and low-exposure setting (see Note 19). The chosen cells are placed suitably in the field and focused at the membrane plane using the fine-focus knob. The detector gain, illumina-tion power, and the camera exposure are optimized to obtain images at the highest possible signal/noise.
6. The sample background is estimated correctly at the same acquisition settings as used for imaging the cell samples. Finally, a G-factor is also re-estimated at the similar gain settings as the rest of the experiment.
The images of the two orthogonal polarizations obtained from the two detectors should be perfectly aligned for anisotropy measure-ments. However, most often they are not completely aligned for a variety of reasons as discussed earlier. The residual misalignment between the two channels can be corrected by using 2D spatial transformation algorithms. Sub-resolution beads imaged in both polarization channels are used to identify reference points (Fig. 5). A mathematically defined affine transform can be generated to correct one channel image with respect to the other [15]. Such corrections can be done using routine algorithms available in MATLAB (MathWorks, USA) or ImageJ (NIH, USA). Post align-ment an appropriate background image is subtracted from both channels and the perpendicular image is corrected with the G-factor image. To obtain reliable spatial anisotropy maps, a spatial averag-ing filter may be used on both parallel and perpendicular images to reduce the noise levels, where the average intensity of a rolling 3 × 3 pixel box is calculated and replaced at the center pixel of the box. The smoothened images are used to generate the pixel-by-pixel anisotropy map (Fig. 6a, c). These images can be further
3.5.2 Image Processing and Analysis
Suvrajit Saha et al.
169
quantified to obtain average anisotropy values of region of interest selected from the cells (Fig. 6b, d).
Homo-FRET imaging has been employed to look at the nanoscale organization of different membrane proteins like the GPI-anchored proteins and transmembrane EGF receptor [18–20]. Homo-FRET applications can be broadly divided into two modalities, the steady- state and the time-resolved modalities. Both the modalities have been used to develop a sensitive assay to study the nanoscale organization and its dynamics. Since we have discussed only the steady- state anisotropy modality here, we will only discuss applications of the same.
GPI-anchored proteins (GPI-APs) are outer-leaflet lipid- tethered membrane molecules, which have been implicated, in myriad roles like cell signaling and adhesion [21]. Based on the classical assay of lipid raft association, like the detergent resistance, these proteins were thought to be a putative raft marker, but the detection and the scale of their organization on living cells remained elusive [22]. Using homo-FRET microscopy and steady-state anisotropy as a readout, folate receptor (FR-GPI) was shown to undergo homo-FRET and form submicron domains sensitive to the cholesterol levels on the membrane [10]. The steady-state
3.6 Homo-FRET Imaging of Nanoscale Organization of Cell Surface Molecules
Fig. 6 Representative homo-FRET images and measurements on cells. CHO cells expressing folate receptor (FR-GPI) were either labeled with folate analog PL-BFL (a, b) or incorporated with BODIPY-FL-DHPE (BFL-DHPE in c, d) as shown in the concatenated images of grayscale intensities and color-coded spatial anisotropy map. The anisotropy values (mean ± s.d.) measured from such cells are plotted against the total average pixel inten-sities for both FR-GPI (b) and BFL-DHPE (d). These images were acquired on a TIRF microscope capable of homo-FRET imaging. Scale bar 5 μm
Homo-FRET Imaging of Nanoscale Organisation
170
anisotropy was measured for cells of varying surface expression or density, and it was found that the anisotropy remains fairly con-stant for a large range of surface density [10]. Cholesterol deple-tion resulted in an increase in the anisotropy values suggesting that there is a significant loss of homo-FRET. Although these measure-ments were made at lower resolution scale of whole cells, which precluded direct observation of domains, it was convincingly dem-onstrated that the GPI-APs clusters were nonrandomly distributed and exhibit a peculiar density-independent clustering over a huge range of densities [10, 22]. One way to assess whether reduced values of anisotropy or depolarized fluorescence arises from FRET is to utilize methods to reduce the density fluorophores in situ. This should predictably increase distances between fluorophores thereby reducing FRET efficiencies and lead to a systematic increase in anisotropy [18]. Fluorophore densities can be reduced by photobleaching and/or chemical quenching. It was shown that photobleaching results in predictably increasing fluorescence anisotropy of FR-GPI. This is due to the increase in distances between fluorophores and also allows the estimation of anisotropy at infinite dilution (A∞) for FR-GPI [18]. The anisotropy changes upon photobleaching can be analyzed by complementary theoreti-cal models of nanoscale clustering which can explain the experi-mental data, as detailed in Sharma et al. [18]. Based on the modeling of the homo-FRET changes upon photobleaching, Sharma et.al. showed that 20–40 % of the cell surface FR-GPI make nanoclusters of 2–4 molecules [18]. These results have now been confirmed independently by super-resolution imaging modal-ities like near-field scanning microscopy (NSOM) and photo- activated localization microscopy (PALM) [23, 24]. More recently, it has been possible to implement homo-FRET imaging on multi-ple microscopy modalities and platforms, which offer superior spa-tiotemporal resolutions, as detailed elsewhere [15]. Such modalities include the spinning disk, TIRF, line-scanning confocal, and the point-scanning multiphoton microscopes. Homo-FRET imaging at a higher spatial resolution has shown GPI-APs are patterned into a highly nonrandom and hierarchical organization in which nanoscale clusters come together to form submicron cluster-rich domains [25]. The dynamics of this clustering behavior has been studied by two homo-FRET-based photobleaching recovery assays, as elaborated in Goswami et al. [25]. These two assays are comple-mentary in the sense that they report on two different length scales of cluster remodeling. It was elegantly demonstrated that the GPI-AP clusters, though immobile, turn over more frequently at 37 °C than 20 °C, suggesting that the cluster remodeling is an activity-dependent process. It was also shown that both the steady- state organization and the dynamics of GPI-AP clusters depend on elements of actin cytoskeleton and cholesterol [25]. Recently, similar tools have also been extended to look at the nanoclustering of transmembrane molecules interacting directly with actin [26],
Suvrajit Saha et al.
171
the ligand-induced oligomerization of EGF receptor [19, 20], and the organization of myosin motors on endosomes [27]. These are a few examples which clearly bring out the potential of homo- FRET microscopy and how it has helped break new grounds in exploring the nanoscale organization of cell surface molecules.
4 Notes
1. Phenol red will fluoresce and hence increases background fluorescence and interferes with the measurements.
2. Always use glass vials when chloroform is used for dissolving lipids.
3. While handling lipids it is important to wear gloves to prevent contamination of lipids from our hand to the sample.
4. The BSA to lipid ratio is 1:2 for lipids with short-chain (8–12) fatty acid and 2:1 for that with long-chain fatty acids (14–18).
5. Excess lipid or non-lipid impurity in the BSA-lipid complex is removed by dialysis.
6. The γ-CD to lipid ratio we commonly use is 1,000:1 but the reader should try a range to determine the optimum ratio for the lipid species being incorporated.
7. Always add the lipofectamine reagent to the Ham’s F12 media and not vice versa.
8. Do not tap too hard, because this will cause the cells to tear if not properly de-adhered from the coverslip. This will lead to leakage of the probe into the solution and defeats the purpose of replating.
9. Be careful not to scrape the coverslip with pipette as this will tear the adhered cells, which will lead to the same problem as described in Note 8.
10. Lipid sticking on the coverslip is specific to the fluorophore used to tag the lipids. BODIPY lipids stick more than the fluo-rescein- or NBD-tagged lipids. In the former case we resort to cell replating method, and in the latter case we resort to HellmanexTM treatment of coverslips.
11. In case of long tail lipids, 10–50 μM lipid concentration is used and for small chain lipids it is 1–10 μM.
12. In case of long-chain lipids low concentration (1–5 μM) is used, and for short-chain lipids, high concentration (10–20 μM).
13. Lipofectamine is sensitive to the polarity of the dye tagged to the lipids. This labeling protocol works well with fluorescein- and NBD-labeled lipids but not with BODIPY-TMR-labeled lipids.
14. The image on both the cameras should retain the same sample plane in focus across the entire image. This maximizes the chances of a good alignment between the two detectors.
Homo-FRET Imaging of Nanoscale Organisation
172
15. Excitation light should illuminate the entire field of view uni-formly and maintains uniform polarization over the entire field. The extinction coefficient of the laser should be moni-tored on a regular basis.
16. G-factor should be uniform across the whole field of view, and extinction coefficient should be constant and uniform across the entire field of view.
17. Anisotropy standards should be measured on a day-to-day basis to ensure the system performance.
18. The excess label should be carefully washed away after the labeling while preparing the samples for imaging. This reduces the background contributed by the label and can reduce the chances of potential artifacts. For cells expressing fluorescent protein chimeras, transient protein synthesis inhibition is a must to reduce the signals coming from the internal pools.
19. The cells should be well adhered to the glass coverslips prior to homo-FRET imaging of the cell membrane. Moreover, high-resolution FRET imaging should be done on cells, which have been freshly labeled, and imaging should be completed before significant endocytic pools develop inside the cells. This will maximize signals specific only to the membrane pool of the probes hence adding reliability to the results.
Acknowledgments
This work was supported by grants from HFSP(RGP0027/2012) and J.C. Bose Fellowship(Department of Science and Technology, India) to SM. We acknowledge support from the Wellcome Trust, the Nanoscience Mission (Department of Science and Technology, India) for the imaging stations built in the laboratory and the Central Imaging and Flow Facility (NCBS) in NCBS. S.S. would like to acknowledge fellowship support from the NCBS-TIFR Graduate programme.
References
1. Schermelleh L, Heintzmann R, Leonhardt H (2010) A guide to super-resolution fluores-cence microscopy. J Cell Biol 190:165–175
2. Krishnan RV, Varma R, Mayor S (2001) Fluorescence methods to probe nanometer- scale organization of molecules in living cell membranes. J Fluoresc 11:211–226
3. Jares-Erijman EA, Jovin TM (2003) FRET imaging. Nat Biotechnol 21:1387–1395
4. Rao M, Mayor S (2005) Use of Forster’s reso-nance energy transfer microscopy to study lipid rafts. Biochim Biophys Acta 1746:221–233
5. Agranovich V, Galanin M (1982) Electronic excitation energy transfer in condensed matter. North-Holland Publishing, Amsterdam
6. Stryer L (1978) Fluorescence energy transfer as a spectroscopic ruler. Annu Rev Biochem 47:819–846
7. Lakowicz JR (2006) Principles of fluorescence spectroscopy, 3rd edn. Springer, New York
8. Mukherjee S, Soe TT, Maxfield FR (1999) Endocytic sorting of lipid analogues differing solely in the chemistry of their hydrophobic tails. J Cell Biol 144:1271–1284
Suvrajit Saha et al.
173
9. Sabharanjak S, Sharma P, Parton RG et al (2002) GPI-anchored proteins are delivered to recycling endosomes via a distinct cdc42- regulated, clathrin-independent pinocytic pathway. Dev Cell 2:411–423
10. Varma R, Mayor S (1998) GPI-anchored pro-teins are organized in submicron domains at the cell surface. Nature 394:798–801
11. Martin OC, Pagano RE (1994) Internalization and sorting of a fluorescent analogue of glucosyl-ceramide to the Golgi apparatus of human skin fibroblasts: utilization of endocytic and nonendo-cytic transport mechanisms. J Cell Biol 125:769–781
12. Spector AA, John K, Fletcher JE (1969) Binding of long-chain fatty acids to bovine serum albumin. J Lipid Res 10:56–67
13. Eggeling C, Ringemann C, Medda R et al (2009) Direct observation of the nanoscale dynamics of membrane lipids in a living cell. Nature 457:1159–1162
14. Koivusalo M, Jansen M, Somerharju P et al (2007) Endocytic trafficking of sphingomyelin depends on its acyl chain length. Mol Biol Cell 18:5113–5123
15. Ghosh S, Saha S, Goswami D et al (2012) Dynamic imaging of homo-FRET in live cells by fluorescence anisotropy microscopy. Methods Enzymol 505:291–327
16. Varma R, Mayor S (2006) Homo-FRET mea-surements to investigate molecular-scale orga-nization of proteins in living cells. In: Stephens D (ed) Cell imaging: methods express. Scion Publishing Limited, UK, pp 247–268
17. Pawley JB (2006) Handbook of biological con-focal microscopy, 3rd edn. Springer, New York
18. Sharma P, Varma R, Sarasij RC et al (2004) Nanoscale organization of multiple GPI-
anchored proteins in living cell membranes. Cell 116:577–589
19. Bader AN, Hofman EG, Voortman J et al (2009) Homo-FRET imaging enables quanti-fication of protein cluster sizes with subcellular resolution. Biophys J 97:2613–2622
20. Hofman EG, Bader AN, Voortman J et al (2010) Ligand-induced EGF receptor oligomerization is kinase-dependent and enhances internaliza-tion. J Biol Chem 285:39481–39489
21. Fujita M, Kinoshita T (2012) GPI-anchor remodeling: potential functions of GPI- anchors in intracellular trafficking and mem-brane dynamics. Biochim Biophys Acta 1821:1050–1058
22. Mayor S, Rao M (2004) Rafts: scale- dependent, active lipid organization at the cell surface. Traffic 5:231–240
23. van Zanten TS, Cambi A, Koopman M et al (2009) Hotspots of GPI-anchored proteins and integrin nanoclusters function as nucle-ation sites for cell adhesion. Proc Natl Acad Sci U S A 106:18557–18562
24. Sengupta P, Jovanovic-Talisman T, Skoko D et al (2011) Probing protein heterogeneity in the plasma membrane using PALM and pair correlation analysis. Nat Methods 8:969–975
25. Goswami D, Gowrishankar K, Bilgrami S et al (2008) Nanoclusters of GPI-anchored proteins are formed by cortical actin-driven activity. Cell 135:1085–1097
26. Gowrishankar K, Ghosh S, Saha S et al (2012) Active remodeling of cortical actin regulates spatiotemporal organization of cell surface molecules. Cell 149:1353–1367
27. Altman D, Goswami D, Hasson T et al (2007) Precise positioning of myosin VI on endocytic vesicles in vivo. PLoS Biol 5:e210
Homo-FRET Imaging of Nanoscale Organisation
175
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_10, © Springer Science+Business Media New York 2015
Chapter 10
Practical Structured Illumination Microscopy
E. Hesper Rego and Lin Shao
Abstract
Structured illumination microscopy (SIM) is a method that can double the spatial resolution of wide-fi eld fl uorescence microscopy in three dimensions by using spatially structured illumination light. In this chap-ter, we introduce the basic principles of SIM and describe in detail several different implementations based on either a diffraction grating or liquid crystal spatial light modulators. We also describe nonlinear SIM, a method that in theory can achieve unlimited resolution. In addition, we discuss a number of key points important for high-resolution imaging.
Key words Fluorescence microscopy , Spatial resolution , Structured illumination microscopy , Diffraction grating , Reciprocal space , Image reconstruction
1 Introduction
In fl uorescence microscopy, light emitted by the sample is blurred by diffraction, often resulting in an image of insuffi cient resolution for the question being asked. However, we, as biologists or microsco-pists, are not necessarily interested in the structure of the emitted light, E ( r ). Rather, we are interested in the distribution of fl uores-cence molecules that make up the underlying sample S ( r ). S ( r ) is not identical to E ( r ), but it is related as a local product with the excita-tion light intensity distribution: E ( r ) = S ( r )· I ( r ). This relationship has led to the development of a number of techniques—so-called struc-tured illumination microscopy—that structure the incoming illumi-nation, I ( r ), in order to gain more information about the sample.
Broadly speaking, any microscope that uses a spatially varying pattern of light as the source of fl uorescence excitation can be considered a structured illumination microscope. Perhaps the most successful of these techniques is confocal microscopy, which relies on a tightly focused “point of light” for excitation and a pinhole in a conjugate image plane to discard out-of-focus light. Superb in its ability to optically section, the confocal microscope is in theory also capable of up to 2-time resolution enhancement.
176
However, in practice, confocal microscopy rarely achieves greater than diffraction- limited resolution since, to do so, would require using a very small pinhole and thus discarding valuable in-focus pho-tons. For optical sectioning, a wide-fi eld structured illumination technique has also been developed [ 1 ] and commercialized, although, like the confocal, it does little for resolution enhancement.
For the purposes of this chapter, we narrowly defi ne structured illumination microscopy (SIM) as a wide-fi eld fl uorescence micros-copy technique designed to extend the resolution of a microscope beyond the classical Abbe diffraction limit, both axially (or along the optical axis) and laterally (or perpendicular to the optical axis), by using sinusoidal illumination patterns. For references, we refer readers to [ 2 , 3 ]. We will briefl y discuss the theory of SIM in the rest of this section and then in the following sections, the practical considerations of its optical implementation in different confi gura-tions and image reconstruction algorithms. It is our hope to offer useful information for both users of the commercial SIM systems and scientists who wish to develop their own SIM setups.
The key concept behind SIM is the well-known moiré pattern. If two patterns are superposed multiplicatively, a beat pattern—moiré fringes—will appear in their product (Fig. 1 ). In the case of fl uo-rescence microscopy, one of the patterns is the unknown spatial distribution of fl uorescent dye (the sample) and the other pattern is a purposely structured excitation light intensity. Since the amount of light emitted from a point is proportional to both the dye den-sity and excitation intensity of that point, the observed image is the product of the two patterns and will thus contain moiré fringes. Such moiré fringes can be much coarser (Fig. 1c ) than either of the original patterns (Fig. 1a, b ) and, by consequence, are easily observable in the microscope even if the original patterns are too fi ne to resolve. In other words, the observed microscope images contain the normally unobservable fi ne details encoded in the form of moiré fringes. Such images can then be “decoded” to recover the information beyond the normal resolution limit, provided that
1.1 Lateral Resolution Enhancement
Fig. 1 Moiré fringes. ( a ) and ( b ) are two examples of fi ne patterns. When one is superimposed onto the other, a coarser beat pattern—moiré fringes—appears ( c )
E. Hesper Rego and Lin Shao
177
a simple illumination structure is used, such as sinusoidal stripes of known line spacing and orientations.
In optics, reciprocal space (also known as spatial-frequency space or Fourier space) is often a more informative representation of the physical reality, especially when spatial resolution is con-cerned [ 4 ]. In reciprocal space, low- and high-resolution informa-tion occupies locations close to and far away from the origin, respectively. The lateral resolution limit of a microscope can be conveniently represented by a circle whose radius is proportional to the numerical aperture and the inverse of the wavelength (Fig. 2a ). All high-resolution sample information outside of the
ky
kx
k 0+k 1
a
b
c
k 1
-k 1
k 0
d
Fig. 2 Principle of structured illumination—lateral resolution enhancement. ( a ) A microscope’s observable region of frequencies in reciprocal space; the resolu-tion limit is represented by the radius of the shaded circle , | k 0 |. ( b ) The frequen-cies of a simple sinusoidal intensity pattern, represented by the three dots in reciprocal space. When the pattern is formed by an objective lens, the highest frequencies cannot exceed the dashed circle because the pattern formation is also resolution limited. ( c ) Illuminated under the pattern shown in ( b ), the fre-quencies in the hatched areas are translated into the normal observable region and hence become effectively observable. Resolution limit is now represented by | k 0 | + | k 1 |. ( d ) Excitation patterns of the same period but of different orientations are applied to extend the lateral resolution isotropically
Practical Structured Illumination Microscopy
178
circle is lost through imaging; only the region inside the circle is observable to the microscope. A sinusoidal laterally structured excitation pattern corresponds to three frequency points (Fig. 2b ). The product of this pattern and the dye distribution amounts to the translation of sample information in reciprocal space along the arrows drawn from the origin to those points (Fig. 2c ); in particu-lar, information normally outside of the circle can be shifted into the circle and become effectively observable as moiré fringes (Fig. 2c ). However, the extra information is mixed together addi-tively with the normal resolution information and therefore needs to be separated. The only way to do so is by acquiring enough images (three in this case) to be able to solve a set of linear equa-tions. Each image is taken with a different phase of the same exci-tation pattern. As a result, lateral resolution enhancement happens only along the line perpendicular to the excitation pattern stripes (Fig. 2c ). To obtain nearly isotropic resolution, the pattern needs to be rotated to two other angles equally spaced by 60°, and addi-tional data needs to be acquired for all pattern orientations. With all the components available from all pattern orientations, repre-sented by the seven circles in Fig. 2d , they are then “stitched” back together according to their original positions in frequency space, forming a fi nal reconstructed image with extended resolu-tion (Fig. 2d ).
The axial resolution in 3D wide-fi eld microscopy is much lower than the lateral resolution, especially for low-resolution features. This is refl ected in reciprocal space by the shape of the observable region in 3D (Fig. 3a, b ): its variable axial depth is about 1/3 of its lateral diameter at most and approaches 0 near the origin. Although the same idea of lateral resolution enhancement by structured illumination can be applied for the axial case, the same excitation pattern would not work (Fig. 3c, d ) since it is struc-tured only in the lateral direction; instead, what is needed is an axially structured illumination pattern. One such illumination pattern, as seen in reciprocal space, has frequency components above and below the k x – k y plane (Fig. 3e ), translating sample information axially and making information above and below the normal observable region effectively observable (Fig. 3f–h ). Compared to lateral-only structured illumination, two more lat-eral information components are present in the raw data, and therefore, fi ve in total need to be separated. This means that fi ve raw images within each pattern orientation and at each 3D defo-cus, acquired at fi ve different lateral pattern phases, are needed in order to solve the fi ve unknowns.
1.2 Axial Resolution Enhancement
E. Hesper Rego and Lin Shao
179
2 Materials
A SIM microscope can be constructed around a commercial inverted microscope body; alternatively, it can be built on a damped optical table without requiring a microscope body. All SIM imple-mentations to date take the form of a wide-fi eld fl uorescence microscope with inserted optical components in the excitation path for generating the illumination pattern (with the exception that some setups use more than one objective, as briefl y discussed in Notes 2 and 3 ). One consideration is that of thermal drift. This is especially a problem for setups that cannot quickly switch between different phases and orientations of the pattern, as is the case in the grating-based SIM discussed below. While modest drift
2.1 SIM Hardware
2.1.1 3D or TIRF Wide-Field Microscope, Either Commercial Body or Home-Built
a b
c d
e f
g h
Fig. 3 Principle of 3D SIM. Observable regions for the conventional microscope ( a and b ), for structured illumi-nation microscopy using two illumination beams ( d ), and three illumination beams in one ( f ) or three ( g , h ) sequential orientations. ( a ) and ( g ) are the k x – k z cross section of the 3D observable regions shown in ( b ) and ( h ), respectively. The spatial-frequency components of the structured illumination intensity for the two-beam ( c ) and three-beam ( e ) case. The dotted outline in panel ( e ) indicates the set of the highest spatial frequencies that are possible to generate by illumination through the objective lens; compare with the observable region in panel ( a )
Practical Structured Illumination Microscopy
180
between different orientations of the pattern can be corrected for in the post-processing algorithm, drift within a phase series is more challenging to correct. As might be expected, thermal drift is neg-ligible for setups that are designed to capture the movement of living cells and therefore must quickly switch between illumination patterns.
In Subheading 3 , we will discuss two diffraction-based methods to generate the sinusoidally varying excitation pattern for SIM. One method uses a transmission phase grating that is mechanically manipulated to produce all the phases and orientations of the pat-tern. While simple, this method is inherently slower than an alter-native we discuss, which relies on a spatial light modulator (SLM) to electronically generate the pattern. Non-diffraction-based ways of generating an illumination pattern are also possible as discussed in Note 4 .
An important factor to consider when designing a structured illu-mination microscope is the coherency properties of the illumina-tion source (Fig. 4a ). Fully incoherent sources, like lamp illumination, can be used but will do little for resolution improve-ment. On the other hand, laser light can be tightly focused at the edge of the back pupil, generating the fi nest illumination pattern and thus the highest resolution improvement. However, the ultra- long coherence lengths of lasers can create stray interference pat-terns that are problematic for the reconstruction software. To alleviate this problem, it is possible to spatially scramble laser light ( see Note 5 ), creating a partially incoherent source. In this case, every point in a source beam is incoherent with every other point in that same source beam, but is coherent with its corresponding point in the other source beams (Fig. 4b, c ). This is especially attractive in 3D SIM imaging because it helps to confi ne the excitation light in the axial direction. However, this partial incoherence can lead to degradation in the zeros of the pattern especially at high angles of incidence that are used in TIRF or near TIRF (Fig. 4d ). For this reason and also because of the stringent requirement on the place-ment of illumination beam at the rear pupil (Fig. 4e, f ), in all the SIM–TIRF setups described [ 5 , 9 ], a fully coherent source was used. In this case, the microscope should be searched for causes of any stray interference patterns ( see Note 6 ).
It is important to note that only light of the same polarization state may interfere. Therefore, to maximize the peak-to-trough contrast in the sinusoidal excitation patterns, all interfering illumination beams must be s -polarized relative to each other; i.e., they must be linearly polarized perpendicularly to the plane of diffraction. To achieve this at all orientations of the pattern, a polarizer that can rotate with the diffraction angle is needed. Below, we will discuss two different options depending on acquisition speed requirements.
2.1.2 Diffraction Grating or Spatial Light Modulator
2.1.3 Lasers
2.1.4 Linear Polarizer or Liquid Crystal- Based Polarization- Rotating Device
E. Hesper Rego and Lin Shao
181
We recommend using a custom-made polychromatic mirror instead of the fi lter cube switching mechanism that usually comes with a modern inverted microscope. One reason for doing so is speed: with such a setup, there would be no need to switch fi lter cube for multicolor SIM, especially in live-cell imaging. Another reason is that the polychromatic mirror used in SIM should transmit shorter wavelength (i.e., excitation wavelength) and refl ect longer wavelength; this is because a refl ective optical device tends to introduce ellipticity in polarization and thus reduces the
2.1.5 Polychromatic Mirror and Emission Filters
tube lens
objective
sample
a
e
grating
b
d
c
f
2D SIM
3D SIM
Fig. 4 ( a ) A simplifi ed schematic of a 2D or 3D structured illumination excitation path. Beams diffracted from a grating (diffraction orders) are focused to the edges of the objective back focal plane (pupil). In 3D SIM, three beams are used to create a pattern with both lateral and axial components. ( b ) Spatially scram-bled laser light generates a “partially incoherent source”; that is, the points within one diffraction beam are all incoherent with each other ( blue ), but corre-sponding points ( pink ) in the other diffraction orders are coherent with each other. ( c ) In 3D structured illumination microscopy, this had the advantage that the axial excitation light is more confi ned than with a completely coherent source. ( d ) One disadvantage of this setup is that at high N.A., the beams may be clipped in the pupil, leading to imperfect zeros of the pattern. ( e ) A completely coherent illumination source can be used, which produces diffraction-limited spots in the pupil of the objective, but can create unwanted stray interference patterns. ( f ) In TIRF microscopy, small focal spots of the beams are required so that they are confi ned to the “TIRF zone” ( gray ), i.e., the region of the pupil that the produces total- internal refl ection
Practical Structured Illumination Microscopy
182
excitation pattern contrast. In addition, the mirror and the optical axis can be made closer to being perpendicular (22.5° as opposed to 45° incident angle) to minimize aberration due to the difference in optical path length because the excitation beams are generally not collimated where they pass the polychromatic mirror.
There is often a need to use a mirror or two to direct the excitation beams toward the objective lens. A mirror generally has different phase retardance along the s - and p -polarization axes and therefore would make linearly polarized excitation beams elliptically polar-ized. To avoid this, two mirrors of the same type of coating can be used to compensate each other’s retardance effect with one angled around the s -polarization and the other around the p -polarization axis. For an inverted microscope that often uses an upward turning mirror or prism directly underneath the objective, it is therefore almost necessary to have another mirror or prism before the up- turning one and after the point where the desired linear polariza-tion state has been achieved for the excitation beams. This second mirror needs to refl ect strictly within the horizontal plane.
Since SIM is a wide-fi eld technique, it is possible to image large fi elds of view without sacrifi cing acquisition speed. Of course, the read-out rate of CCD cameras does depend on pixel area. However, the latest scientifi c CMOS cameras can read out large number of pixels on millisecond timescales. Nevertheless, at extremely low light levels, a back-illuminated EMCCD may be preferred.
3 Methods
One way to generate the SIM excitation patterns is by using a diffraction grating and letting 2 (for 2D SIM) or 3 central diffrac-tion orders (for 3D SIM) interfere at the sample plane. We have outlined such a possible setup in Fig. 5 . A transmission phase grat-ing is placed at a conjugate image plane in the illumination path. Laser light comes out of the end face of an optical fi ber, is colli-mated by a lens (L1), and then diffracted by the grating. The dif-fracted beams are focused onto a secondary pupil plane by a relay lens (L2) and after passing through a dichromatic (or polychro-matic) mirror are refocused onto the objective’s back focal plane by another relay lens (L3) and the tube lens (assuming an infi nite conjugate objective lens is used). The spacing of the grating and focal lengths of the relay lenses are chosen such that the ±1 diffrac-tion orders are located on the edge of the objective’s back focal plane. If TIRF is desired, care should be taken to place the beams in the region of the back focal plane that produces total internal refl ection. The interference of 0 and ±1 diffraction orders (or only the ±1 orders in the case of 2D or TIRF SIM) at the sample space generates the sinusoidal excitation pattern used in SIM.
2.1.6 Mirrors
2.1.7 Camera
3.1 Grating-Based SIM Microscope
3.1.1 Microscope
E. Hesper Rego and Lin Shao
183
To change the phase of the excitation pattern, the grating can be mounted on a piezoelectric translator, which in turn can be mounted on a motorized rotation stage. We have successfully used a piezoelectric translator which is controlled in closed loop using a capacitive distance sensor consisting of one convex electrode on one side of the grating holder and one hollow cylindrical counter electrode that surrounds the grating holder, each fl anked by a guard electrode [ 3 ]. The counter electrode is rigidly attached to the optical table and serves as a fi xed reference, so that any drift in the rotation stage would not affect the translation precision. To maintain s -polarization at all orientations of the pattern, a linear polarizer can be corotated with the grating by mounting it on the same stage as the grating.
To achieve the highest quality in SIM images, there are a number of key implementation issues that need to be addressed. These include axial positioning of lens L3 ( see Note 7 ), lens L4, the camera ( see Note 8 ) and the grating ( see Note 9 ), locating of the secondary pupil plane ( see Note 10 ), and lateral centering and axial focusing of the fi ber end ( see Note 11 ).
2D SIM acquisition is straightforward. Nine images are taken: three phases at three orientations of the pattern. As mentioned earlier, a motorized rotation stage and a piezoelectric translator are responsible for changing the orientation and phase of the excita-tion pattern, respectively. Since all mechanical stages need time to change position and settle, it is important to ensure that camera exposure happens only when all mechanical parts involved are completely static.
3.1.2 2D SIM Acquisition
Camera
Multimodefiber
DM
Objectivelens
Sample
Primaryimageplane
Pupilplane
Conjugateimageplane
Secondarypupilplane
Conjugate image plane
Conjugate pupil plane
Diffuser
Tubelens
Laser
L1
L2L3
GratingPolarizerL4
Fig. 5 Simplifi ed diagram of 3D structured illumination apparatus. Scrambled laser light from a multimode fi ber is collimated onto a linear phase grating. Diffraction orders −1, 0, and +1 are refocused into the back focal plane of an objective lens. The beams, re-collimated by the objective lens, intersect at the focal plane in the sample, where they interfere and generate an intensity pattern with both lateral and axial structure. Emission light from the sample is observed by a camera via a dichromatic mirror (DM)
Practical Structured Illumination Microscopy
184
Image acquisition in 3D SIM is carried out in a way similar to that in regular 3D wide-fi eld microscopy, with the following differ-ences: (1) three 3D stacks are acquired in sequence, one for each lateral excitation pattern orientation; (2) at every defocus position within each orientation, fi ve images are acquired with fi ve different lateral excitation patterns of different phases evenly spaced by 2 π ; and (3) it is important that the Nyquist sampling criteria is satis-fi ed. With that in mind, each defocus step should be ½ the conven-tional defocus step.
The speed of data acquisition in a grating-based SIM system is limited by the rotational and the translational stages that switch excitation patterns. The time for the piezoelectric translator and the motorized rotator to settle is on the order of milliseconds and 100 ms, respectively, making it nearly impossible to use such a microscope for live-cell imaging. Speed can be improved by replac-ing the grating with a liquid crystal-based spatial light modulator (SLM) as the means of generating the excitation patterns (Fig. 6a ), because an SLM can be programmed to display any desired pattern with sub-millisecond switching rate. It has been shown that an SLM-based SIM microscope can be effectively used for live 3D imaging of whole cells [ 6 , 7 ].
For the highest speed achievable, a ferroelectric liquid crystal (FLC) SLM should be used, and preferably, it needs to be at least of SVGA resolution (i.e., more than 1,280 × 1,024 pixels). Each pixel of a FLC SLM behaves as a half-wave plate with its fast axis’ orientation changeable to either 0° or 45° (Fig. 6b ). Working together with a polarizing beam splitter and a half-wave plate, the SLM can function essentially as a phase grating by phase modulat-ing the parts of the incident plane wave front that is refl ected off the SLM’s ON pixels to be π phase different than the parts refl ected off the OFF pixels. To act as a phase grating for SIM, the SLM needs to be programmed into pixel patterns that resemble a regu-lar grating. Regardless of the SLM pixel pattern, however, the out-put wave fronts would share the same linear polarization. To maintain the s -polarization as discussed earlier, one can use a vari-able wave plate together with a quarter-wave plate to rapidly rotate the polarization angle. Such a variable wave plate needs to operate at about 1 ms retardance switching rate in order to not slow down the overall acquisition speed.
A thorough description of SIM reconstruction was published [ 3 ]. We briefl y review the main steps here.
1. Preprocessing: the original images are smoothed at the edges and fl at-fi eld corrected (i.e., correction of the pixel-to-pixel variation in the response of the camera to light intensity).
2. Information unmixing: for each pattern orientation, the raw data is a mixture of three or fi ve different information
3.1.3 3D SIM Acquisition
3.2 SLM-Based SIM Microscope
3.3 Image Reconstruction
E. Hesper Rego and Lin Shao
185
components that are originally centered around different fre-quency–space points, which need to be separated fi rst. This is done by a simple pixel-by- pixel matrix–vector multiplication using a decomposition matrix simulating a Fourier transform and a three- or fi ve-element vector composed of the images acquired with different pattern phases.
3. Parameter fi tting: for each pattern orientation, the spacing and the angle of the pattern is accurately estimated by maximizing the cross-correlation between each higher-resolution compo-nent and the conventional component in the frequency–space region where they overlap (Figs. 2d and 3g ); between these same overlaps, a complex numbered ratio is obtained. This ratio’s amplitude represents how close to the ideal scenario the actual pattern contrast is, and its phase indicates the phase of the fi rst of the three or fi ve patterns.
Fig. 6 Using a spatial light modulator (SLM) as a diffraction grating. The parts shown in ( a ) replace the transmis-sion grating and rotating linear polarizer shown in Fig. 5 . Each pixel of the SLM behaves as a half-wave plate with the fast axis either vertical or at 45° ( b ) depending on the state with which each pixel is programmed. As a result, a plane wave front refl ected off the SLM is phase modulated by π between the pixels in two different states ( b ). Furthermore, a liquid crystal-based variable retarder together with a quarter wave plate ( a ) act as a rapidly programmable polarization rotator to maintain s -polarization for all pattern orientations
Practical Structured Illumination Microscopy
186
4. Inverse fi ltering and reassembly: each information component is a low-pass fi ltered version of the part of the sample structure originating from various frequency–space locations; they are assigned back to where they are from and the low-pass fi ltering is reversed.
5. System calibration: to be able to perform step 4 , i.e., to reverse the low-pass fi ltering by the imaging system, the fi lter itself needs to be calibrated fi rst. This can be done by acquiring a SIM dataset of a single fl uorescent bead (usually only one pat-tern orientation is suffi cient), and the dataset is unmixed into several components as explained in step 2 . To improve the signal- to-noise ratio, it is possible to rotationally average around the optical axis (i.e., z -axis).
One example of reconstructed 3D SIM image is shown in Fig. 7 in comparison with the image acquired with conventional wide-fi eld microscopy.
So far in this chapter, we have discussed SIM under conditions of conventional fl uorescence, which can double the resolution of a fl uorescence microscope. However, it is possible to achieve even
3.4 Nonlinear SIM
Fig. 7 A comparison of images acquired with conventional wide-fi eld microscopy ( a ) and SIM ( b ). The sample is a fi xed mouse embryo fi broblast cell stained with phalloidin–TRITC. One plane out of a 3D stack is shown. The voids seen at the leading edge of the cell in panel ( b ) demonstrate the axial resolution of SIM: the lamellipodium is not fl at and there is little actin visible where it ruffl es away from the plane shown. Scale bar: 2 μm
E. Hesper Rego and Lin Shao
187
greater resolution by exploiting a nonlinear fl uorescence phenom-enon in combination with structured illumination. This idea has been realized by saturating the fl uorophore excited state [ 8 ] or the “off state” of a photoswitchable protein [ 9 ]. In principle, any non-linear fl uorescence phenomenon can be used to extend the resolu-tion by introducing higher order harmonics into the effective illumination pattern. While saturation and other nonlinear phe-nomena like stimulated emission or ground state depletion have the distinct advantage of being compatible with conventional fl uo-rescence probes, photoswitchable molecules tend to switch at lower illumination intensity, making them more suitable for bio-logical imaging. Here, we will briefl y discuss nonlinear SIM (NL-SIM) in the context of the photoswitching.
NL-SIM does not require special hardware or reconstruction software. It does, however, require a slightly more complicated acquisition sequence. First, the molecules are switched to their on state. Second, using a sinusoidal pattern of light that drives the molecules to their off state, the molecules are turned off leav-ing only the molecules in the trough of the pattern on . The width of the region containing the on-state molecules can be smaller than the diffraction limit and will depend on the line spacing of the pattern and the saturation of the off state. For a more com-plete description, we refer readers to ref. 9 . Third, the light from the on-state molecules is collected . This process is repeated for a predetermined number of phases and orientations of the pattern. Compared to linear SIM, more phases and orientations of the pattern are needed to successfully separate the information around the higher-order harmonics and obtain isotopic resolu-tion, respectively.
Since each acquisition sequence is a switching cycle, the resolu-tion of NL-SIM is inherently linked to the photophysical proper-ties of the photoswitchable molecule. For optimal imaging and maximum resolution by NL-SIM using photoswitching, probes should exhibit the following characteristics:
1. Complete turnoff . Incomplete turnoff results in molecules being on in the zeros of the pattern, which lowers the signal-to- noise of the higher order harmonics. Patterned excitation with optically clean zeros can alleviate this issue however, since that will lead to cleaner zeros of the emission distribution.
2. Complete turn-on . While it is not an absolute requirement that every molecule be switched on during an imaging cycle, incomplete turn-on does cause potential problems: (a) it will contribute to overall noise in the system in the form of shot noise since light coming from subsequent cycles could be com-ing from different molecules and (b) the fewer molecules being turned on inherently means less fl uorescent signal. However,
Practical Structured Illumination Microscopy
188
for dense labeling, it might make sense to only excite a few molecules per cycle, since this may lead to a slower photo-bleaching rate.
3. Photobleaching. During the course of an imaging experiment, the molecules should not bleach “signifi cantly.” The exact def-inition of signifi cantly is determined experimentally; for exam-ple, molecules that emit more photons may decay to a greater extent than molecules that emit less photons but still have an acceptable SNR in the images acquired last. Consequently, molecules that emit more photons are more desirable than those that are less bright.
4 Notes
1. Notation notes: f XX denotes the focal length of lens XX (e.g., f L2 stands for the focal length of lens L2 ).
2. While SIM as discussed above can double both lateral and axial resolution, the axial and lateral resolution limit ratio remains identical to normal wide-fi eld microscopy, i.e., about one third in both cases. To make the ratio closer to 1, two opposing objective lenses can be used in techniques such as I 5 S [ 10 ]. The two objectives and a beam splitter/combiner in such confi gu-ration form an interferometer such that emission light col-lected by the two objectives interferes. In addition, structured illumination beams are split into halves, incident on the sample from both objectives, and interfere to form a more compli-cated 3D excitation pattern than single-objective SIM. As a result of the interference in both emission detection and exci-tation, the axial resolution of I 5 S is extended to just below 100 nm [ 10 ]. Interferometry-based microscopy techniques such as I 5 S have stringent requirement on the optical homoge-neity of the samples, which is usually diffi cult to meet. In addi-tion, because of the higher axial resolution than regular SIM, Nyquist sampling rate means shorter focal step (~40 nm) and thus more total exposures.
3. Another confi guration of multiple-objective SIM is to use two objective lenses placed at a right angle to each other with one for excitation and the other detection [ 11 ]. Instead of scan-ning the excitation beams to form a light sheet, the beams can be spaced by a certain distance and simulate a sinusoidal excita-tion pattern. All the SIM principles discussed above apply to the so-called Bessel-beam SIM, except that lateral resolution is only extended along the beam scanning direction. Due to the geometric limit imposed by the right-angle confi guration, the numerical aperture of either objective cannot exceed 1.1
E. Hesper Rego and Lin Shao
189
usually, and therefore the achievable lateral resolution is not as high as regular SIM. The main advantage of Bessel-beam SIM is its capability to image samples thicker than a single layer of cells, due to the narrowness of the illuminated layer and there-fore reduced out-of-focus background compared to a conven-tional wide-fi eld microscope.
4. In addition to using a grating-like device to generate struc-tured illumination (diffractive beam splitting), one can also use refl ective beam splitting, in which a beam is split, one arm phase shifted, and then recombined to generate interference. Refl ective beam splitting has an advantage for multicolor applications in that the positions of the illumination beams in the pupil plane can be made independent of their wavelength, whereas with diffractive beam splitting, the distance of a beam from the pupil center is typically proportional to its wave-length and can therefore be strictly optimized only for one wavelength at a time. If diffractive beam splitting is used, simultaneous multiwavelength datasets can still be taken by optimizing for the longest wavelength, with only minor per-formance penalties at the shorter wavelength channels. On the other hand, refl ective beam splitting has a severe disadvan-tage in that it typically causes the beams to be separated spa-tially and handled by separate optical components (mainly lenses); any nanometer-scale drift of any such component translates directly into a phase error in the data. With diffrac-tive beam splitting, by contrast, all beams typically traverse the same optical components in an essentially common-path geometry, which decreases the sensitivity to component drift by several orders of magnitude. In addition, refl ective beam splitting typically requires separate mirror sets for each pattern orientation, which becomes increasingly bulky as the number of pattern orientations increases.
5. To remove the spatial coherence and speckle pattern in the light transmitted by a multimode optical fi ber, it is possible to use a fast rotating holographic diffuser before the laser beam enters the fi ber and a focusing lens to couple the diffused light into the fi ber. The diffusing angle is chosen to approximately fi ll the mode space of the fi ber. Another more light-effi cient way of laser scrambling is to mechanically shake a segment of the fi ber. To maximize the scrambling effect, at least 10 m of the fi ber should be involved in the shaking and the protective fi ber jacket should not be used. The fi ber can be wound into a fi gure-8 loop to increase the length being shaken and the cross-junction can be used as the point of attachment to a vibrator. Caution needs to be taken when working with the extremely fragile unprotected optical fi ber.
Practical Structured Illumination Microscopy
190
6. To check for stray interference patterns in the illumination light distribution, one can use as a sample a cover glass coated thinly (<300 nm) with fl uorescence dye. Allow only one of the SIM illumination beams to reach the sample at a time. In images taken under such illumination, look for any interfer-ence fringes and the causes for them because only uniform images should be observed. Usual causes for stray interference patterns include the beam intersecting the edge of the objec-tive’s rear aperture, dust on lenses and mirrors along the illu-mination path, and back refl ections.
7. To achieve optimal aberration correction, relay lens L3 (Fig. 5 ) should be placed a focal length f L3 away from the primary image plane. This can be done by directing a collimated beam from the objective lens side of the tube lens toward the tube lens and L3. When L3 is positioned precisely at f L3 away from the primary image plane, the beam should be collimated again after L3. A shear plate placed after L3 can assist in judging whether a beam is truly collimated.
8. Once L3 is correctly positioned, the same collimated beam can be further utilized to precisely position the camera axially. For this purpose, the camera should work in video mode while its axial position is being adjusted; the desired position is when the beam’s focal spot formed by lens L4 is the sharpest. One needs to make sure enough attenuation is applied to the beam to avoid camera damage. In general, L4’s position is not criti-cal as long as the distance between L3 and L4 is not so large that the diameter of L4 is not suffi ciently big to catch all the rays originating from the periphery of the fi eld of view.
9. For ideal imaging, the grating face should be perfectly conju-gate to the image plane. To optimize this parameter, one should prepare a sample composed of a single layer of fl uores-cent beads and acquire a 3D SIM dataset using only one pat-tern orientation and with focal steps much smaller than the Nyquist sampling rate (25–50 nm). The acquired dataset is fi rst unmixed into fi ve information components. The ampli-tudes of the ±1 and ±2 order components at each z section are summed, resulting in two plots of the axial variation of the illumination intensity. These two plots are referenced against a through-focus intensity plot of a selected bead. If the grating is perfectly conjugate to the image plane, the peak of the three curves should all coincide. If they do not coincide, the axial distance between the fi rst-order peak and the focus peak can be then converted into the distance by which the grating needs to be moved axially based on the overall axial magnifi cation from the sample plane to the grating.
10. Since the axial positioning of all optical components from L2 up to the optical fi ber entirely depends on the position of the
E. Hesper Rego and Lin Shao
191
secondary pupil plane (SPP), it is crucial to locate it precisely. The SPP is usually not focal length f L3 from relay lens L3 because the distance between the objective and the tube lens is usually not strictly f obj + f tube and unknown to the users (unless a cus-tom-designed microscope is used). It is thus often best to locate the SPP experimentally. This can be achieved by shining a point-like light source, such as a laser diode or laser light com-ing out of a single-mode fi ber, from near the SPP toward the objective with the tube lens and L3 in place. At far fi eld above the objective, an Airy pattern should be visible if the point source is well centered and near the secondary pupil plane. One can then adjust the light source’s axial position until the Airy pattern looks the sharpest and note down that position as the SPP. With the SPP precisely located, lens L2 should be posi-tioned f L2 away from the SPP, grating or SLM f L2 away from L2, and fi nally the fi ber end f L1 away from the grating.
11. The fi ber end surface ideally needs to be positioned laterally on the optical axis and axially conjugate to the SPP. Severe lateral misalignment leads to either of the ±1 diffraction order beams being partially or completely stopped by the back pupil when it is severe, resulting in ineffi cient interference and thus low contrast in the excitation patterns. Minor lateral misalign-ment causes the confi guration of 3D SIM information com-ponents in reciprocal space (the ideal case of which is shown in Fig. 3 ) asymmetric with respect to the k x – k y plane, which can lead to reconstruction artifacts because such asymmetry cannot be accounted for by the ideal imaging model assumed in the reconstruction algorithm. One can precisely center and focus the fi ber end by checking the illumination spot formed by the objective lens at the ceiling (or any far-fi eld location): the spot needs to be round with distinct edge and centered around the point where the extension of the optical axis inter-sects the ceiling.
References
1. Neil MA, Juskaitis R, Wilson T (1997) Method of obtaining optical sectioning by using struc-tured light in a conventional microscope. Opt Lett 22:1905–1907
2. Gustafsson MG (2000) Surpassing the lateral resolution limit by a factor of two using struc-tured illumination microscopy. J Microsc 198:82–87
3. Gustafsson MG, Shao L, Carlton PM et al (2008) Three-dimensional resolution doubling in wide-fi eld fl uorescence microscopy by struc-tured illumination. Biophys J 94:4957–4970
4. Goodman JW (2005) Introduction to Fourier optics. Roberts & Co., Englewood
5. Kner P, Chhun BB, Griffi s ER et al (2009) Super-resolution video microscopy of live cells by structured illumination. Nat Methods 6:339–342
6. Fiolka R, Shao L, Rego EH et al (2012) Time- lapse two-color 3D imaging of live cells with doubled resolution using structured illumination. Proc Natl Acad Sci U S A 109:5311–5315
7. Shao L, Kner P, Rego EH et al (2011) Super- resolution 3D microscopy of live whole cells using structured illumination. Nat Methods 8:1044–1046
8. Gustafsson MG (2005) Nonlinear structured- illumination microscopy: wide-fi eld fl uorescence
Practical Structured Illumination Microscopy
192
imaging with theoretically unlimited resolu-tion. Proc Natl Acad Sci U S A 102:13081–13086
9. Rego EH, Shao L, Macklin JJ et al (2012) Nonlinear structured-illumination microscopy with a photoswitchable protein reveals cellular structures at 50-nm resolution. Proc Natl Acad Sci U S A 109:E135–E143
10. Shao L, Isaac B, Uzawa S et al (2008) I 5 S: wide-fi eld light microscopy with 100-nm-scale resolution in three dimensions. Biophys J 94:4971–4983
11. Planchon TA, Gao L, Milkie DE et al (2011) Rapid three-dimensional isotropic imaging of living cells using Bessel beam plane illumina-tion. Nat Methods 8:417–423
E. Hesper Rego and Lin Shao
193
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_11, © Springer Science+Business Media New York 2015
Chapter 11
4Pi Microscopy of the Nuclear Pore Complex
Martin Kahms , Jana Hüve , and Reiner Peters
Abstract
4Pi microscopy is a far-fi eld fl uorescence microscopy technique, in which the wave fronts of two opposing illuminating beams are adjusted to constructively interfere in a common focus. This yields a diffraction pattern in the direction of the optical axis, which essentially consists of a main focal spot accompanied by two smaller side lobes. At optimal conditions, the main peak of this so-called point spread function has a full width at half maximum: fi xed phrase of 100 nm in the direction of the optical axis, and thus is 6–7-fold smaller than that of a confocal microscope. In this chapter, we describe the basic features of 4Pi microscopy and its application to cell biology using the example of the nuclear pore complex, a large protein assembly spanning the nuclear envelope.
Key words 4Pi microscopy , Nuclear pore complex , Topographic analysis , Nuclear transport recep-tors , Binding site distribution
1 Introduction
Among the currently available methods for the analysis of protein complexes in living cells, fl uorescence microscopy is a promising candidate as it allows for live cell imaging, molecular specifi city, high time resolution and spectral multiplexing, and at the same time can be regarded as minimally invasive [ 1 ]. However, an obvi-ous shortcoming of conventional fl uorescence microscopy is spa-tial resolution. But recently, several concepts emerged to overcome the classical resolution limit in light microscopy opening a highly attractive perspective to analyze single protein complexes in living cells and tissues. Among these concepts are, e.g., stochastic detec-tion of individual switchable fl uorophores (photoactivated local-ization microscopy, PALM [ 2 , 3 ]; stochastic optical reconstruction microscopy, STORM [ 4 ]); nonlinear de-excitation of fl uorescent dyes (stimulated emission depletion, STED [ 5 ]); and superposi-tion of coherent wave fronts, constituting the basic principle of 4Pi microscopy [ 6 ].
194
In 4Pi microscopy, a classical point-scanning confocal microscope is combined with an interferometer, i.e., a laser beam is split into two beams that are focused by two opposing objectives onto a common focal point. This arrangement leads to a doubled total aperture angle, and though the ideal case of covering the complete solid angle of 4Pi is not possible, the acronym 4Pi is used as an approximation. When the phases of the two counter-propa-gating beams are carefully adjusted, this optical confi guration results in a characteristic and peculiar 4Pi point spread function (PSF), consisting of a sharp main peak accompanied by two weaker side lobes in the direction of the optical axis [ 6 ], while in the focal plane, the 4Pi PSF consists of a single peak that has virtually the same dimensions as the confocal PSF. At optimal conditions, the main peak of the PSF has a full width at half maximum (FWHM) of ~100 nm in axial direction and thus is 6–7-fold smaller com-pared to a confocal PSF. When employing one-photon excitation, the intensity of the side lobes is frequently larger than half of the intensity of the main peak. However, the relative size of the side lobes can be decreased by two-photon excitation [ 7 ], and with a relative intensity <50 %, ghost images present in 4Pi raw data can be removed by mathematical procedures [ 8 ].
Different optical confi gurations for 4Pi microscopy have been described in which either interference of the excitation light (type A), interference of the emitted fl uorescence light (type B), or inter-ference of both, excitation and fl uorescence light (type C), is used for resolution improvement. While type A 4Pi microscopy usually requires two-photon excitation for achieving side lobe reduction [ 7 ], type C 4Pi microscopy can be performed with one-photon excitation [ 9 ]. 4Pi microscopy of type B, in contrast, is only of conceptual interest [ 10 ].
One of the largest protein complexes in the cell is the nuclear pore complex (NPC) (for review see [ 11 , 12 ]). In vertebrates, the NPC is composed of several hundreds of proteins with a total mass of 125 MDa. The NPC spans the nuclear envelope (NE), and one of its major function is to mediate transport of matter, energy, and information between the cell nucleus and cytoplasm. Being the only gate between the cytoplasm and nucleus, the NPC is involved in almost all basic cellular functions like chromatin organization, gene activation, and cell cycle regulation (for review see [ 13 , 14 ]).
The structure of the NPC has emerged from electron micros-copy studies, providing currently three-dimensional models at a resolution of 6–10 nm [ 15 – 17 ]. According to these models, the vertebrate NPC consists of an approximately cylindrical scaffold of ~125 nm diameter and ~70 nm height containing a large channel of ~50 nm in diameter. The central scaffold is composed of three connected rings, the cytoplasmic ring, the spoke ring, and the nuclear ring, exhibiting on a gross level eightfold symmetry with
Martin Kahms et al.
195
regard to the central axis of the NPC and twofold symmetry with regard to the plane of the NE. Eight fi laments of ~50 nm in length are attached to the cytoplasmic ring, while the nuclear ring is deco-rated with eight fi laments which distally conjoin to form a basket-like structure.
NPCs are composed of about 30 different protein species, known as nucleoporins (Nups). Each Nup occurs in eight copies or multiples of eight per NPC [ 18 , 19 ]. Several Nups have been crys-tallized and their structures solved at atomic resolution. These studies indicate that Nups are composed of only a few different folds, which somewhat reduces the complexity of Nup structures and NPC architecture [ 20 , 21 ].
Although intensively studied, the detailed mechanism of cargo transport mediated by the NPC is still vividly debated. Several transport models have been proposed to explain the remarkable features of cargo transport through the NPC in terms of speed and selectivity (for review see [ 22 , 23 ]). In this context, structure and arrangement of Nups in the most inner layer of the NPC, forming the walls of the central channel, are of particular interest [ 24 ]. These Nups, comprising about one third of all proteins in the NPC, contain domains with multiple repeats of phenylalanine and glycine and are therefore termed FG-Nups. FG domains serve as interaction sites for nuclear transport receptors (NTRs) and their cargo complexes, and therefore it is essential to analyze the arrange-ment of these FG domains and the interaction with NTRs in great detail to further elucidate the mechanism of cargo translocation.
Here, we describe the application of 4Pi microscopy for the analysis of topographic features of single NPCs as well as determi-nation of binding site distributions of transport factors along the central axis of the NPC. This work demonstrates that super- resolution fl uorescence microscopy is a promising method for ana-lyzing single protein complexes and the cellular nanomachinery in general.
2 Materials
The basic parts of the commercially available 4Pi setup used in our studies (TCS 4Pi microscope, type A, Leica Microsystems, Mannheim, Germany) are depicted in Fig. 1a and comprise a confocal laser scanning microscope of type TCS SP2, laser systems for one- and two-photon excitation, and photon counting by avalanche photodiodes. The standard objective turret of the SP2 microscope is replaced by a 4Pi attachment which operates comparable to a classical Michelson interferometer. In the attachment, the excitation laser light is split into two illumination beams. The two beams are defl ected by mirrors into the entrance pupil of two opposing objective lenses which focus
2.1 4Pi Microscope
4Pi Microscopy of the NPC
196
the counter-propagating wavefronts in a common focal point. Lateral scanning is performed by the scan unit of the SP2 system, while axial scanning is accomplished by a piezo-scanning stage. The position of the lower objective relative to the upper one can be adjusted in all three dimensions by means of piezo-elements.
Fig. 1 4Pi setup. ( a ) Scheme of the commercially available 4Pi setup (TCS 4Pi microscope, type A, Leica Microsystems, Mannheim, Germany). O1 and O2: objective lenses; S: sample; BS: beamsplitter; PC: phase control; PZ: piezo- element; SU: scan unit; PH: pinhole; APD: avalanche photodiode; TiSa: Ti:sapphire laser. ( b ) 4Pi PSF profi les along the optical axis for glycerol immersion lenses (100×, NA = 1.35, circles ) and water immersion lenses (63×, NA = 1 . 2, triangles ) in a two-photon 4Pi microscope type A determined at immobilized beads of sub-resolution size
Martin Kahms et al.
197
The correct adjustment of the whole system, i.e., the achievement of a common focus, is visualized and checked by imaging the entrance pupils of both objective lenses by a CCD camera located at the beam splitter. Additionally, the relative phase of the two beams can be changed by moving the lower defl ecting mirror to achieve constructive interference in the focal point. Especially for analyzing live specimens, in which the refractive index step between the cover glass and immersion medium induces a phase change which increases linearly with penetration depth, an active external compensation can be additionally implemented to ensure the same relative phase of the counter- propagating beams while scanning in z -direction.
The two interferometer arms can be independently blocked by shutters. Thus, the microscope can be used in the confocal mode with the upright or inverted beam path and single- or two-photon excitation or as a two-photon excitation 4Pi microscope. Unlike in a classical type A setup, where detection of the emission light is blocked in one interferometer arm, emission light is collected through both objectives and detected in a noncoherent fashion. For two-photon excitation, a mode-locked Ti:sapphire laser (MaiTai, Spectra Physics, Mountain View, USA) with a pulse length stretched to >1 ps is employed. Its wavelength can be tuned in the range of 710–990 nm. Fluorescence originating from the sample is imaged onto the confocal pinhole of the SP2 system, then passed a fi lter cube (SP700, BS560, BP500–550, and BP607–683; Chroma Technology Corporation, Bellow Falls, USA), and fl uorescence intensity was measured by photon-counting avalanche photodiodes (PerkinElmer, Foster City, USA).
As mentioned above, in 4Pi microscopy constructive interference of two opposing wave fronts yields a main diffraction maximum that is 3–7 times narrower than that of a single lens. But the drawback of this optical confi guration is the appearance of elevated side lobes along the optical axis. To obtain the highest possible image resolution, the height of the side lobes has to be <50 % of that of the main maximum [ 25 ]. Otherwise spatial information is partially lost, as indicated by a frequency gap in the corresponding optical transfer function (OTF). The relative height of the side lobes sharply decreases with increasing semi-aperture angle α of the objectives, and hence, the best performance is reached with the highest possible α . Oil immersion lenses such as the HCX PL APO CS 100/1.46 Oil (Leica Microsystems, Mannheim) provide a semi-aperture angle of α = 74° and, in combination with a high refractive index embedding medium (97 % thiodiethanol, TDE [ 26 ]), yield a resolution along the optical axis of ~100 nm [ 27 ]. In addition, it has been reported that these lenses enable 4Pi confocal fl uorescence microscopy of type C and even type A employing
2.2 Objectives
4Pi Microscopy of the NPC
198
one-photon excitation [ 9 , 28 ]. However, the application of oil immersion lenses requires embedding media with a refractive index close to 1.52.
Glycerol immersion lenses such as the HCX PL APO 100/1.35 Gly Corr (Leica Microsystems, Mannheim) provide a semi- aperture angle of 68.5° [ 29 ]. Calculations suggested that at α > 68° 4Pi microscopy of type A should be feasible employing one-photon excitation, but experiments showed insuffi cient side lobe suppres-sion [ 28 ]. Hence, two-photon excitation has to be employed. Typically, a FWHM of ~110 nm for the main peak can be achieved accompanied by side lobes of ~30 % relative height. Although this resolution is slightly smaller than that achieved with oil immersion lenses and that two-photon excitation is required, glycerol immer-sion objectives appear advantageous for imaging fi xed specimens. The correction collar of these objectives tolerates refractive indices ranging from ~1.44 to ~1.46, corresponding to glycerol concen-trations between 72 and 88 %. In combination with quartz cover-slips, an isotropic refractive index of 1.46 throughout the optical path is generated, and proper interference of the opposing beams can be optimized. In our hands glycerol immersion yielded the most reliable 4Pi measurements and was least sensitive to sample preparation.
For live cell imaging or analysis of cellular preparations under physiological conditions, water immersion objectives are required (e.g., HCX PL APO 63/1.2W Corr, Leica Microsystems, Mannheim). The reduced NA of 1.2 results in a decreased FWHM of 150–180 nm for the main maximum [ 30 , 31 ], and due to the discontinuity of the refractive index at the glass/water interface, this confi guration is highly sensitive to any inaccuracy along the optical path like poorly adjusted correction collars with respect to coverslip thickness or even slight tilting of the probe. 4Pi PSF pro-fi les along the optical axis for glycerol immersion lenses (100×, NA = 1.35) and water immersion lenses (63×, NA = 1.2) are shown in Fig. 1b (for a specialized confi guration see Note 1 ).
3 Methods
For 4Pi microscopy, the sample is sandwiched between two cover-slips. This confi guration permits focusing the coaxial but antiparal-lel laser beams into a common point within the sample. Because the working distance of high NA objectives is very small, the layer thickness between the two coverslips should not exceed 30 μm. Furthermore, great care has to be taken to avoid a deformation of the 4Pi point spread function (PSF), which can be induced by any refractive index mismatch. It is therefore essential to use the highest- quality coverslips of well-defi ned thickness ( see Note 2 ) and to adjust the correction collar of the objectives properly ( see
3.1 General Aspects
3.1.1 Sample Preparation
Martin Kahms et al.
199
Note 3 ). Additionally, the sample has to contain a reference which can be used to align the focus and phase of the counter- propagating waves. This reference can consist of either fl uorescently labeled endogenous point- or rodlike structures of sub-diffraction dimen-sions or of artifi cial immobilized fl uorescent beads ( see Note 4 ). To prepare such samples use the following protocol:
1. Chose coverslips (30 mm in diameter) depending on the objec-tive confi guration of the 4Pi microscope (quartz coverslips for glycerol immersion lenses, BK7 coverslips for water immersion lenses, see Note 2 ).
2. Coat a coverslip with poly-(L)-lysine (0.01 %) by distributing a drop of solution on the coverslip surface. Wait for 10 min, wash with ddH 2 O, and let dry.
3. Add 10 μl of bead solution ( see Note 4 ). After evaporation of ddH 2 O most of the beads are immobilized on the glass surface.
4. Glue the coverslip (intended to be at the bottom of the cham-ber) carefully to a sample holder (available from Leica Microsystems, Mannheim). Try to avoid any tilting.
5. Put a thin layer of mounting medium (15–25 μl) on top of the coverslip. The choice of mounting medium depends on the objectives used in the 4Pi experiment (see below).
6. Mount the second coverslip with the specimen upside down in the sample holder.
7. Finally seal the samples with a special two-component knead-ing silicone, and check the thickness of the complete sample holder. The thickness of the layer between the two coverslips should not exceed 30 μm.
For oil immersion lenses in combination with 97 %, TDE as embedding medium ( see Note 5 ), the best results have been reported by a stepwise increase of TDE concentration during sev-eral washing steps of the fi xed specimen [ 26 ].
With glycerol immersion lenses, all components between the two opposing objective lenses have to be adjusted to a refractive index close to 1.46. This can be achieved by using quartz cover-slips and immersing cells in 87 % PBS-buffered glycerol ( see Note 6 ). Of course, quartz coverslips are much more costly than stan-dard BK7 glass coverslips.
With water immersion lenses, samples can be mounted in aqueous buffer ( n = 1.33–1.34), but special care has to be taken to avoid even the slight tilting of the coverslips.
To obtain full image resolution, 4Pi data have to be post-processed to remove ghost images. Various methods have been described for this purpose ranging from three- or fi ve-point linear fi lters [ 32 ] to
3.1.2 Deconvolution of Raw Data
4Pi Microscopy of the NPC
200
nonlinear mathematical image restoration algorithms [ 8 ]. Like in any imaging system, the image can be described by a convolution of the unknown object function with the PSF of the microscope. In a fi rst approximation, the 4Pi PSF can be separated into inde-pendent radial and axial contributions allowing for fast deconvolu-tion algorithms. Here, the axial profi le is approximated by a convolution of a peak intensity function describing a single maxi-mum and a lobe function representing relative position and height of the three maxima. Side lobes can be then removed by convolu-tion of the image function with the inverse of the lobe function. This so-called three- or fi ve-point linear deconvolution is fast and robust and does not require highly accurate PSF measurements. But two requirements have to be fulfi lled to make this approach feasible. First, the relative height of the side lobes must be constant with respect to a lateral offset from the optical axis. Second, the PSF has to be spatially invariant in the image. When these require-ments are met, linear point deconvolution enables easy side lobe removal without highly accurate PSF measurements, but at the expense of a decreased effective signal-to-noise ratio and no fur-ther enhancement of resolution.
In contrast, resolution can be further enhanced by direct linear deconvolution, i.e., inverse fi ltering of the complete image with the 4Pi PSF. This works well for noise-free images and a spatially invariant and perfectly known PSF. In reality, however, an addi-tional fi lter (like, e.g., Wiener fi lter) has to be implemented to account for noise. Though an additional resolution enhancement up to 50 % can be achieved using linear deconvolution, a limited signal-to-noise ratio and improper description of the 4Pi PSF make this approach unfeasible in many cases.
Nonlinear image restoration methods, e.g., based on Richardson-Lucy deconvolution, use a maximum-likelihood algo-rithm to iteratively minimize the deviation between the measured image and estimated object. These algorithms offer the advantage of noise reduction and resolution enhancement but are slow and susceptible to poor PSF measurements.
One prerequisite for all deconvolution methods described here is a spatial invariant PSF in the image. In practice, this is not always the case, i.e., the PSF depends on the position in the sample most often manifested in a phase shift of the interference pattern along the optical axis. Methods for deconvolution with a variable PSF have been described to account for this problem [ 33 ].
The position of single point-like objects, although imaged as dif-fraction limited spots, can be determined with a precision that is one order of magnitude higher than the optical resolution by fi t-ting the measured photon distribution to an ideal Gaussian [ 34 ]. In 4Pi microscopy, the main peak of the PSF has an FWHM of 100 nm in the direction of the optical axis at optimal conditions
3.1.3 Localization of Single Epitopes
Martin Kahms et al.
201
and thus is 6–7-fold smaller than that of a confocal PSF. The local-ization precision of diffraction-limited signals scales with ~FWHM/√ n , with n being the number of photons detected, meaning that improved optical resolution, at least in principle, pro-vides higher accuracy. Additionally, two opposing lenses are capa-ble of detecting more photons than a single lens which leads to a further increase of localization accuracy. As the 4Pi PSF is intrinsi-cally elliptical, a true 2D fi tting routine is essential for proper cen-troid localization ( see Note 7 ). Taking these parameters into account, a localization accuracy of only a few nm has been demon-strated for 4Pi microscopy [ 35 , 36 ].
Labeling of mammalian cell nuclei with fl uorescent NPC-specifi c antibodies typically results in a punctate staining of the nuclear periphery. The dots of the punctate pattern usually have a density of ~5/μm 2 and a corresponding nearest neighbor distance of ≤0.4 μm. By confocal scanning microscopy it has been shown that the dots predominantly represent single NPCs, while NPCs which occur as part of a cluster cannot be resolved [ 37 , 38 ]. The central axis of the NPC is perpendicular to the plane of the nuclear enve-lope, and therefore the central axis of NPCs in the upper and lower nuclear membrane co-aligns with the optical axis of the micro-scope. All these parameters—size, density, and orientation—make the NPC an ideal object for 4Pi studies. We started 4Pi analysis of the NPC by using a primary antibody against Nup358, the major component of the cytoplasmic fi laments of the NPC.
1. Grow HeLa cells on quartz coverslips (220 μm thickness) to 70 % confl uency.
2. Fix cells in 4 % paraformaldehyde for 20 min at room tempera-ture, wash three times with phosphate-buffered saline (PBS, 138 mM NaCl, 2.7 mM KCl, 10 mM Na 2 HPO 4 , 2 mM KH 2 PO 4 , pH 7.4), and permeabilize in 0.25 % Triton X-100.
3. Wash three times with PBS, block unspecifi c binding with 0.5 % fi sh skin gelatin for 30 min at room temperature, and incubate with a primary antibody directed against Nup358 (1:2,000, kind gift from Dr. E. Coutavas) for 1 h at room temperature. Wash for 20 min in PBS, and incubate with sec-ondary antibody (goat-α-rabbit, Alexa488 conjugated, 1:1,000, Invitrogen) for 1 h at room temperature. Wash cells for 20 min in PBS.
4. Prepare the coverslip intended to be at the bottom of the sam-ple, and glue it in a sample holder as described above. Apply 15 μl of PBS-buffered glycerol (87 % w/w) onto the coverslip.
5. Mount the second coverslip with attached cells upside down. Seal the sample with two-component silicone.
3.2 4Pi Microscopy of the Nuclear Pore Complex
3.2.1 Imaging of Single Nuclear Pores in Fixed Specimens
4Pi Microscopy of the NPC
202
6. Transfer the sample to the 4Pi microscope equipped with glycerol immersion lenses using glycerol (87 % w/w) as immer-sion medium. Set the excitation wavelength of the Ti:sapphire laser to 820 nm, the beam expander to 6, and the confocal pinhole size to 0.8 Airy units to further reduce relative side lobe height [ 35 ].
7. Align focus and phase of the counter-propagating beams using immobilized beads as a reference. When focus and phase are properly adjusted, the FWHM of the main peak of the PSF should be 110–115 nm with relative side lobe height of 30 %.
8. Switch the focal plane to the upper coverslip with attached cells and record xz scans of cell nuclei.
A comparison of confocal and 4Pi axial sections is shown in Fig. 2 . In confocal images NPCs appear as diffraction limited spots which are blurred in xz -direction (Fig. 2a, d ). In contrast, 4Pi
Fig. 2 Imaging of single NPCs by 4Pi microscopy. HeLa cells were labeled by secondary immune fl uorescence techniques using a primary antibody against the NPC protein Nup358 and an Alexa488-labeled secondary antibody. Images show an axial section of a HeLa cell nucleus imaged in confocal ( a ) and 4Pi mode, the latter as raw data in ( b ) and after removal of side lobes by deconvolution in ( c ). A detailed view from the same nucleus acquired at a higher zoom factor is shown for the confocal mode in ( d ), corresponding 4Pi raw and deconvolved images are shown in ( e ) and ( f ). Scale bars: 2 μm and 0.5 μm, respectively (adapted by permis-sion from Elsevier Ltd: Biophysical Journal, copyright 2008 [ 35 ])
Martin Kahms et al.
203
images of NPCs consist of a narrow main maximum (FWHM ~ 110 nm) accompanied by two ghost images of ~30 % relative intensity (Fig. 2c, e ). These ghost images could be removed by using simple 3-point linear deconvolution ( see Note 8 ) resulting in a true resolution of 110 nm along the optical axis (Fig. 2f ). Since the density of NPCs is relatively large in HeLa cells, the increase in resolution simplifi es discrimination of single NPCs in 4Pi images. However, a close inspection of the representative optical section through a complete HeLa cell nucleus reveals that even using an almost isotropic refractive index throughout the optical path, the 4Pi PSF still depends on optical path length in the specimen (Fig. 2e ). While in the lower nuclear membrane, the 4Pi raw images of the NPC consist of strong main peaks and weak side lobes, one of the side lobes is almost as strong as the main peak at the upper nuclear membrane, making the PSF asymmetric. Therefore, the applied deconvolution scheme is quite successful on the lower, but not on the upper nuclear envelope. An improvement in such cases could be achieved by active external phase compensation or more sophisticated deconvolution algorithms.
In the past, structural properties of the NPC have been mainly studied by electron microscopy of fi xed or frozen specimen and by X-ray diffraction of isolated crystallized proteins. However, protein complexes are highly dynamic entities, and to account for this dynamics, functional and structural properties have to be studied at physiological conditions and eventually even in living cells. Therefore we asked if it is possible to resolve topographic features of single NPCs by 4Pi microscopy. As a fi rst step towards live cell imaging, we used fi xed specimens and aimed to determine the dis-tance along the central axis between different epitopes in single NPCs using a two-color localization method [ 35 ].
Two different epitopes in the NPC of HeLa cells were labeled by means of immune fl uorescence. We chose the NPC proteins Nup358 and Tpr, which are both present in eight copies per sin-gle NPC, but should be arranged around the central axis of the NPC in a relatively fi xed distance to the midplane of the NPC. While Nup358 is the major component of the cytoplasmic fi laments, Tpr is a major component of the nuclear basket. The anti-Tpr antibody was directed against an epitope in the C-terminus of Tpr, which according to immuno-electron microscopic studies is localized at the distal part of the nuclear fi laments. The second-ary antibodies against Nup358 and Tpr were labeled with Alexa488 (green fl uorescence) or Alexa594 (red fl uorescence). These fl uorophores can be spectrally separated by means of their different emission spectra. They can be excited, however, at the same wavelength (800 nm) using two-photon excitation thus avoiding chromatic aberrations due to different excitation vol-umes. A small axial offset, introduced by the pinhole, remains
3.2.2 Distance Determination Between Two Epitopes Using a Two-Color Localization Method
4Pi Microscopy of the NPC
204
when two different detection colors are used. But this effect was estimated to be <3 nm. After immuno- labeling cells were embed-ded in 87 % glycerol and imaged by two- photon 4Pi microscopy with glycerol immersion lenses. A representative example is shown in Fig. 3a . In confocal sections, the green (Nup358) and red (Tpr) fl uorescent epitopes overlap to a large extent, while in 4Pi sections
Fig. 3 Distance determination of epitopes in single NPCs. ( a ) In HeLa cells, epitopes of the NPC proteins Nup358 and Tpr were labeled by secondary immune fl uorescence methods and imaged by confocal and two-color two-photon 4Pi microscopy. Representative images of a single NPC, derived from the green Nup358-associated fl uorescence and the red Tpr-associated fl uorescence comparing confocal with 4Pi images, are shown. In the confocal mode, epitopes overlap largely in both detection channels. However, epitopes can be clearly separated in 4Pi images. Scale bar: 0.2 μm. ( b ) Distance between molecular sites in single NPCs as determined by two- color localization 4Pi microscopy. A sketch of the NPC indicates the positions of the sites that were fl uorescently labeled, negative values indicate the cytoplasmic face of the NPC, and posi-tive values indicate nuclear contents. Distances were analyzed in single NPCs of PtK 2 cells between the Tpr epitope and POM121-eGFP 3 , and in NRK cells between the Tpr epitope and POM121-eGFP 3 , as well as between POM121-eGFP 3 and the Nup358 epitope. Obtained distances are represented as the mean ± SD of n measure-ments (adapted by permission from Elsevier Ltd: Biophysical Journal, copyright 2008 [ 35 ])
Martin Kahms et al.
205
the epitopes are clearly separated. These results demonstrate unequivocally that it is possible to resolve the nanoscopic sub-structure of single protein complexes by light microscopy. However, it should be recalled that eight copies of Nup358 and Tpr, respectively, are arranged around on a radius of 80–120 nm the central axis of the NPC. This octagonal arrangement around the central axis could not be resolved so far.
For quantitative distance analysis, only those NPCs were con-sidered, which were axially co-aligned with the optical axis of the microscope, i.e., the direction of the highest resolution. The epit-opes were individually localized using a routine that provides a true two-dimensional elliptical Gaussian fi t, and their distance along the central axis of the NPC was determined. These distance measure-ments were based on raw, i.e., non-deconvolved images. The dis-placement between main image and ghost images is large when using two-photon excitation, and the localization of the main max-imum can be performed even more accurately when deconvolution is omitted. Measurements for a total of n = 218 NPCs yielded a distribution with a mean value of 152 ± 30 nm, a value in good accordance with predictions based on electron microscopy data.
In a further series of experiments dual-color measurements were extended to cells constitutively expressing a fl uorescent Nup. We used HeLa and Ptk 2 cells, in which the Nup POM121 was fused to three molecules of enhanced GFP (eGFP). POM121 anchors the NPC in the nuclear envelope and is localized in or very close to the center plane of the NPC. Cells expressing Pom121- eGFP 3 were fi xed and labeled with the anti-Tpr or Nup358 anti-body. After reaction with an Alexa594-labeled secondary antibody, NPCs were imaged by 4Pi microscopy, and topographic analysis was performed. The results are summarized in Fig. 3b . All values obtained were in good agreement with recent electron microscopy estimates.
The mechanism of cargo translocation through the NPC is still debated. Crucial but yet incompletely understood parameters are the molecular arrangement of FG domains and the distribution of NTR binding sites within the central channel of the NPC. While the topographic analysis described above was performed on fi xed specimens using glycerol immersion lenses, in the following we applied water immersion lenses, permitting to image functional NPCs of permeabilized cells at physiological concentrations of transport factors. We started our analysis focusing on Ran, a GTP- binding protein mediating directionality of nucleocytoplasmic transport. Ran does not directly interact with FG repeats. However, the NPC contains a number of Ran-binding sites located in non- FG sites of certain Nups. In addition, shuttling of Ran through the NPC is facilitated by a special transport factor, NTF2, which binds both Ran and FG repeats:
3.2.3 Determination of Binding Site Distributions Under Physiological Conditions at Single NPCs
4Pi Microscopy of the NPC
206
1. Grow NRK cells, stably expressing a Pom121-GFP 3 fusion protein, on BK7 coverslips to 70 % confl uency.
2. Permeabilize the plasma membrane with 50 μg/ml digitonin in transport buffer (50 mM Hepes/KOH, pH 7.3, 110 mM potassium acetate, 5 mM sodium acetate, 2 mM magnesium acetate, 1 mM EGTA, and 2 mM DTT) for 3 min on ice ( see Note 9 ).
3. Wash cells with transport buffer and incubate with 5 μM Ran- Alexa633 in transport buffer in the absence or presence of 10 μM NTF2.
4. Prepare the coverslip intended to be at the bottom of the sam-ple and glue it in a sample holder as described above.
5. Add 15 μl of transport buffer containing 5 μM Ran-Alexa633 (±NTF2) onto the prepared coverslip with immobilized beads, and mount the second coverslip with attached cells upside down.
6. Choose the excitation wavelength of the Ti:sapphire laser such that eGFP and Alexa633 fl uorescence show approximately equal brightness in their respective detection channel (870–890 nm). Set the beam expander to 3 and the pinhole to 1 Airy unit.
7. Mount the sample onto the 4Pi microscope equipped with water immersion lenses and ddH 2 O as immersion medium. Align focus and phase of the counter-propagating beams using immobilized beads as a reference. When focus and phase are properly adjusted, the FWHM of the main peak of the PSF should be 150–160 nm with relative side lobe height of <50 %.
8. Image single NPCs, which happened to be oriented with their main axis in direction of the z -axis of the optical system and record xz stacks with a pixel size of 14.5 × 14.5 nm in xz - direction and a step size of 480 nm in y -direction to avoid pre- bleaching of fl uorescence in the following image section.
9. Fit the intensity profi les of the two PSFs main maxima inde-pendently to an elliptical 2D Gaussian, and calculate the dis-tance between the two centers ( see Note 7 ).
10. Correct distance values for offset introduced by the two detec-tion colors ( see Note 10 ).
Examples for the imaging of single NPCs by water immersion two-photon 4Pi microscopy are shown in Fig. 4a . The recorded signals showed the typical 4Pi intensity distribution, even though the background in the nucleus was high because of imported Ran molecules. The use of water immersion objectives broadens the point spread function and reduces the axial resolution to 150–170 nm as compared with 110–130 nm with glycerol immersion. However, the localization accuracy, which directly scales with resolution but
Martin Kahms et al.
207
also depends on the signal-to-noise ratio, was virtually not affected, and the distance between single epitopes within the NPC could be determined with a precision similar to the one obtained with glyc-erol immersion lenses. Again, removal of the side lobes by decon-volution was dispensable in our analysis, and therefore the confocal pinhole could even be set to >1 Airy unit to increase the number of detected photons. The distribution of the obtained distance
Fig. 4 Binding of Ran-Alexa633 to single NPCs analyzed by 4Pi microscopy. 4Pi images using water immersion lenses of single NPCs in permeabilized NRKPom121-eGFP 3 cells are shown in ( a – c ). Left : eGFP fl uorescence detected in the green emission channel. Middle : Alexa633 fl uorescence detected in the red emission channel. Analysis was performed for 5 µM Ran-Alexa633 (a), 5 µM Ran Alexa633 with an excess of 10 µM NTF2 (b), and 5 µM Ran-Alexa633 with an excess of 10 µM NTF2 E42K (c), a mutant Scale bar in both dimensions: 0.6 μm. The corresponding distance distributions of Ran- Alexa633 relative to Pom121-eGFP 3 are shown at the right, directed from the cytoplasm (negative values) through the NE center (origin) into the nuclear contents (positive values) (adapted by permission from John Wiley and Sons: Traffi c, copyright 2009 [ 31 ])
4Pi Microscopy of the NPC
208
values is shown in Fig. 4b . In the absence of NTF2, the distribution could be well fi tted by a Gaussian with a maximum at x c = −30 nm and an FWHM of 78 nm. In the presence of NTF2, the distribu-tion of Ran was also almost symmetric and had a maximum at x c = −9 nm and an FWHM of 76 nm. We defi ned the localization such that negative distance values report binding sites shifted in the direction of cytoplasm relative to Pom121. The difference between the distributions is highly signifi cant (signifi cance level <10 −6 in a one-way ANOVA). Thus, the NTF2-induced redistribu-tion of RanGDP in the NPC was clearly resolved by 4Pi micros-copy. As a control, we repeated the analysis in presence of the NTF2 mutant E42K. This mutant binds to FG repeats with nor-mal affi nity but has lost the ability to interact with Ran. 4Pi analysis of the Ran distribution in the presence of 10 μM NTF2 (E42K) showed a localization maximum at x c = −29 nm, not signifi cantly different from the value obtained in the absence of NTF2.
In the same manner we analyzed the distribution of the NTR Kapβ1- and a Kapβ1-based transport complex and found them to have also binding maxima at approximately −10 nm. These obser-vations support transport models in which NTR binding sites are distributed all along the transport channel and argue against mod-els in which the cytoplasmic entrance of the channel is surrounded by a cloud of binding sites.
In essence, this chapter shows that structural aspects of single cellular protein complexes and the distribution of binding sites on single complexes can be studied by super-resolution fl uorescence microscopy. Of course, the studies presented are just a humble starting point. The resolution obtained currently by 4Pi micros-copy is still insuffi cient for drawing conclusions on the detailed arrangement of FG repeats and NTR binding site within the cen-tral channel of the NPC. However, fl uorescence techniques such as STED or STORM have emerged recently further enhancing reso-lution. Thus, live cell studies at nanometer resolution eventually are becoming more than just a dream.
4 Notes
1. In an attempt to increase the speed of data acquisition, lens arrays have been employed. By these means up to 64 4Pi foci could be produced and simultaneously scanned across the specimen, reducing acquisition times by one order of magni-tude [ 39 ].
2. The glycerol immersion lenses (HCX PL APO 100/1.35 GLYC CORR) require quartz coverslips of 220 μm thickness, while the water immersion lenses (HCX PL APO 63/1.2W CORR) are designed for BK7 glass of 170 μm thickness.
Martin Kahms et al.
209
Custom-made coverslips of well-defi ned thickness (e.g., 170 ± 2 μm) have to be used, and new batches should be tested with a sliding caliper.
3. To optimize the adjustment of the correction collars, cover-slips with a refl ective coating at the periphery might be useful. Maximizing the detected signal in refl ection mode can help to fi nd the best position of the correction collars for an individual sample.
4. It is essential to have a suffi ciently bright and stable reference probe of sub-diffraction size for the alignment of the 4Pi PSF. We recommend a 1:2,500 dilution of TransFluoSpheres NeutroAvidin, 0.1 μm microspheres in ddH 2 O (Invitrogen, NY, USA). Due to the large Stokes shift ( λ Ex : 488 nm, λ Em : 605 nm), they show good performance in two-photon micros-copy with a broad excitation spectrum ranging from 760 to 960 nm.
5. It has to be mentioned that some fl uorophores such as eGFP are quenched at high TDE concentrations [ 26 ].
6. The refractive indices of glycerol-based immersion and embed-ding media should be checked using a refractometer.
7. To determine the position of single epitopes, we used a routine that provides for a true two-dimensional (2D) fi t (IL-Tracker, Ingo Lepper Software/Consulting, Münster, Germany). Some fi tting algorithms separate the measured intensity distribution into orthogonal intensity profi les which are independently fi t-ted to a 1D Gaussian. Although this procedure is sometimes referred to as a 2D fi t, it is basically a 1D fi t, and diffi culties arise when combining the two 1D Gaussians. Thus, this method usually leads to less accurate results than fi tting a proper 2D Gaussian function to the whole 2D intensity profi le. IL-Tracker is a commercial plug-in for ImageJ, and 2D Gaussian fi tting is based on nonlinear least-squares routine according to the Levenberg-Marquardt method [ 35 ].
8. For simple 3- or 5-point linear deconvolution, we used a rou-tine provided by the Leica software. For more sophisticated linear as well as nonlinear deconvolution, we recommend Imspector Image Analysis and Acquisition Software ( http://www.imspector.de ).
9. Treatment of mammalian cells with a defi ned concentration of digitonin results in selective perforation of the plasma mem-brane, leaving the nuclear envelope intact [ 40 ]. Integrity of the nuclear envelope can be checked by adding a fl uorescently labeled substrate >40 kDa (e.g., BSA-Alexa488) which should be still excluded from the nucleus after treatment.
4Pi Microscopy of the NPC
210
10. In contrast to glycerol immersion lenses, where an offset intro-duced by the two detection colors was negligible, we found a substantial chromatic error when using water immersion lenses. We quantifi ed this error by imaging a thin layer of Atto532, which showed similar brightness in both detection channels. The position of the main maxima of the 4Pi profi les in the two detection channels was determined by indepen-dently fi tting the intensity profi les in z -direction row by row to a 1D Gaussian. From the resulting histograms of maxima posi-tions, the chromatic error was determined to be ~14 nm in z -direction.
Acknowledgments
We thank Dr. R. Wesselmann and P. Lehrich for their experimental contributions and Dr. E. Coutavas for providing anti-Nup358 and anti-Tpr antibodies. This work was supported by the National Institutes of Health, Bethesda, MD, grant No. 1 R01 GM071329- 01, and the Deutsche Forschungsgemeinschaft, grant PE-138/19-1.
References
1. Peters R (2006) Checking and fi xing the cel-lular nanomachinery: towards medical nanos-copy. Trends Mol Med 12:83–89
2. Betzig E, Patterson GH, Sougrat R et al (2006) Imaging intracellular fl uorescent proteins at nanometer resolution. Science 313:1642–1645
3. Hess ST, Girirajan TP, Mason MD (2006) Ultra-high resolution imaging by fl uorescence photoactivation localization microscopy. Biophys J 91:4258–4272
4. Rust MJ, Bates M, Zhuang X (2006) Sub-diffraction- limit imaging by stochastic optical reconstruction microscopy (STORM). Nat Methods 3:793–795
5. Hell SW, Wichmann J (1994) Breaking the diffraction resolution limit by stimulated emis-sion: stimulated-emission-depletion fl uores-cence microscopy. Opt Lett 19:780–782
6. Hell SW, Stelzer EH (1992) Properties of a 4Pi confocal fl uorescence microscope. J Opt Soc Am A 9:2159–2166
7. Hell SW, Stelzer EH (1992) Fundamental improvement of resolution with a 4Pi-confocal fl uorescence microscope using two-photon excitation. Opt Commun 93:277–288
8. Schrader MS, Hell SW, van der Voort HTM (1998) Three-dimensional super-resolution
with a 4Pi-confocal microscope using image restoration. J Appl Phys 84:4033–4042
9. Lang MC, Engelhardt J, Hell SW (2007) 4Pi microscopy with linear fl uorescence excitation. Opt Lett 32:259–261
10. Hell SW, Stelzer EH, Lindek S et al (1994) Confocal microscopy with an increased detec-tion aperture: type-B 4Pi confocal microscopy. Opt Lett 19:222
11. Peters R (2006) Introduction to nucleocyto-plasmic transport: molecules and mechanisms. Methods Mol Biol 322:235–258
12. Wente SR, Rout MP (2010) The nuclear pore complex and nuclear transport. Cold Spring Harb Perspect Biol 2:a000562
13. Strambio-De-Castillia C, Niepel M, Rout MP (2010) The nuclear pore complex: bridging nuclear transport and gene regulation. Nat Rev Mol Cell Biol 11:490–501
14. Wozniak R, Burke B, Doye V (2010) Nuclear transport and the mitotic apparatus: an evolv-ing relationship. Cell Mol Life Sci 67:2215–2230
15. Beck M, Forster F, Ecke M et al (2004) Nuclear pore complex structure and dynamics revealed by cryoelectron tomography. Science 306:1387–1390
Martin Kahms et al.
211
16. Beck M, Lucic V, Forster F et al (2007) Snapshots of nuclear pore complexes in action captured by cryo-electron tomography. Nature 449:611–615
17. Stoffl er D, Fahrenkrog B, Aebi U (1999) The nuclear pore complex: from molecular archi-tecture to functional dynamics. Curr Opin Cell Biol 11:391–401
18. Cronshaw JM, Krutchinsky AN, Zhang W et al (2002) Proteomic analysis of the mammalian nuclear pore complex. J Cell Biol 158:915–927
19. Rout MP, Aitchison JD, Suprapto A et al (2000) The yeast nuclear pore complex: com-position, architecture, and transport mecha-nism. J Cell Biol 148:635–651
20. Alber F, Dokudovskaya S, Veenhoff LM et al (2007) The molecular architecture of the nuclear pore complex. Nature 450:695–701
21. Devos D, Dokudovskaya S, Williams R et al (2006) Simple fold composition and modular architecture of the nuclear pore complex. Proc Natl Acad Sci U S A 103:2172–2177
22. Kahms M, Hüve J, Wesselmann R et al (2011) Lighting up the nuclear pore complex. Eur J Cell Biol 90:751–758
23. Peters R (2009) Translocation through the nuclear pore: Kaps pave the way. Bioessays 31:466–477
24. Solmaz SR, Chauhan R, Blobel G et al (2011) Molecular architecture of the transport channel of the nuclear pore complex. Cell 147:590–602
25. Nagorni M, Hell SW (2001) Coherent use of opposing lenses for axial resolution increase in fl uorescence microscopy. I. Comparative study of concepts. Vis J Opt Soc Am A Opt Image Sci 18:36–48
26. Staudt T, Lang MC, Medda R et al (2007) 2,2′-Thiodiethanol: a new water soluble mounting medium for high resolution optical microscopy. Microsc Res Tech 70:1–9
27. Lang MC, Staudt T, Engelhardt J et al (2008) 4Pi microscopy with negligible sidelobes. New J Phys 10:1–13
28. Lang M, Muller T, Engelhardt J et al (2007) 4Pi microscopy of type A with 1-photon exci-tation in biological fl uorescence imaging. Opt Express 15:2459–2467
29. Martini N, Bewersdorf J, Hell SW (2002) A new high-aperture glycerol immersion objec-tive lens and its application to 3D-fl uorescence microscopy. J Microsc 206:146–151
30. Bahlmann K, Jakobs S, Hell SW (2001) 4Pi-confocal microscopy of live cells. Ultramicroscopy 87:155–164
31. Kahms M, Lehrich P, Hüve J et al (2009) Binding site distribution of nuclear transport receptors and transport complexes in single nuclear pore complexes. Traffi c 10:1228–1242
32. Hänninen PE, Hell SW, Salo J et al (1995) Two-photon excitation 4Pi confocal micro-scope: enhanced axial resolution microscope for biological research. Appl Phys Lett 66:1698–1700
33. Baddeley D, Carl C, Cremer C (2006) 4Pi microscopy deconvolution with a variable point-spread function. Appl Opt 45:7056–7064
34. Thompson RE, Larson DR, Webb WW (2002) Precise nanometer localization analysis for individual fl uorescent probes. Biophys J 82:2775–2783
35. Hüve J, Wesselmann R, Kahms M et al (2008) 4Pi microscopy of the nuclear pore complex. Biophys J 95:877–885
36. Schmidt M, Nagorni M, Hell SW (2000) Subresolution axial distance measurements in far-fi eld fl uorescence microscopy with preci-sion of 1 nm. Rev Sci Instrum 71:2742–2745
37. Kubitscheck U, Wedekind P, Zeidler O et al (1996) Single nuclear pores visualized by con-focal microscopy and image processing. Biophys J 70:2067–2077
38. Kubitscheck U, Grunwald D, Hoekstra A et al (2005) Nuclear transport of single molecules: dwell times at the nuclear pore complex. J Cell Biol 168:233–243
39. Egner A, Verrier S, Goroshkov A et al (2004) 4Pi-microscopy of the Golgi apparatus in live mammalian cells. J Struct Biol 147:70–76
40. Adam SA, Marr RS, Gerace L (1990) Nuclear protein import in permeabilized mammalian cells requires soluble cytoplasmic factors. J Cell Biol 111:807–816
4Pi Microscopy of the NPC
213
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_12, © Springer Science+Business Media New York 2015
Chapter 12
Application of STED Microscopy to Cell Biology Questions
Natalia H. Revelo and Silvio O. Rizzoli
Abstract
The increasing interest in “seeing” the molecular environment in biological systems has led to the recent quest for breaking the diffraction barrier in far-field fluorescence microscopy. The first nanoscopy method successfully applied to conventional biological probes was stimulated emission depletion microscopy (STED). It is based on a physical principle that instantly delivers diffraction-unlimited images, with no need for further computational processing: the excitation laser beam is overlaid with a doughnut-shaped depleting beam that switches off previously excited fluorophores, thereby resulting in what is effectively a smaller imaging volume. In this chapter we give an overview of several applications of STED microscopy to biological questions. We explain technical aspects of sample preparation and image acquisition that will help in obtaining good diffraction-unlimited pictures. We also present embedding techniques adapted for ultrathin sectioning, which allow optimal 3D resolutions in virtually all biological preparations.
Key words Super-resolution microscopy, STED, Diffraction barrier, Cell imaging, Live imaging
1 Introduction
In recent decades far-field fluorescence microscopy has become an important tool in the investigation of cellular components, both in functional and structural terms. The easy sample preparation (com-pared with other techniques such as electron microscopy) and the possibility of imaging live specimens account for its popularity. Nevertheless, the acuity of lens-based instruments is limited by the diffraction of the light passing through the optical components of the microscope. In 1873 Ernst Abbe postulated that a point source emit-ting light of a wavelength λ, which travels through a medium with refractive index n and converges to a lens with an angle θ, will produce a spot with radius d = λ/2(nSinθ), where nSinθ is equivalent to what came to be known as the numerical aperture (NA) of the objective [1]. In practical terms, this means that conventional fluorescence microscopes using light sources in the range of the visual spectrum and objectives with NA close to 1 cannot resolve two structures that are closer than approximately 200–300 nm. This is an obvious diffi-
214
culty when dealing with protein complexes or small organelles (typically one order of magnitude smaller than the diffraction of light).
The first technique that overcame the Abbe’s diffraction bar-rier was stimulated emission depletion (STED) microscopy. Here, the sample is scanned with an excitation beam spatially overlapped with a red-shifted depletion beam (see Fig. 1a). The wavelength of the depletion beam is carefully selected to fall in the red side tail of the fluorophore emission spectrum. The depletion beam is physi-cally modified to produce a central area of zero intensity surrounded by an outer ring or “doughnut” of non-zero intensity. In the doughnut center, where only the excitation beam is present,
Fig. 1 STED microscopy breaks Abbe’s diffraction barrier. (a) Working principle of a STED microscope. The sample is illuminated with an excitation beam (blue ) and a doughnut-shaped depletion beam (orange ) that are spatially aligned. In the center of the depletion beam, spontaneous fluorescence is allowed to occur, whereas fluorophores in the bright surrounding area are depleted. The result is a lateral reduction of the effective exci-tation volume with concomitant improvement of resolution. (b) STED microscopy applied on biological sam-ples. Neuronal synaptic terminals revealed by immunostaining of the synaptic vesicle protein synaptotagmin were imaged with confocal and STED microscopy for comparison. Individual terminals can be differentiated only under the increased resolution of STED microscopy, due to their subdiffraction size. Reproduced with permission from Ref. [7]
Natalia H. Revelo and Silvio O. Rizzoli
215
fluorescence excitation and emission will occur just as in conventional fluorescence imaging techniques. In the outer area, however, where both excitation and depletion light coincide, the depletion beam will stimulate the excited fluorophores to emit photons at a longer wavelength than usual, far from the fluorescence detection window [2, 3]. As a result, fluorescence photons are only collected from a narrower central excitation volume that is smaller than the diffraction limit, its size Δr given by Dr NA I IS» +( )l / / .2 1 Here, I is the maximum intensity of the depletion beam, and the saturation intensity IS is the intensity required to reduce the fluo-rescence probability by half [4]. Thus, larger depletion laser intensities will not only increase the likelihood of stimulated emission to occur but will also increase the resolution of the microscope by broadening the effective area of the depletion doughnut. Thus, STED microscopy allows the immediate acqui-sition of diffraction-unlimited images without requiring further computational post processing, as it is the case for other fluores-cence nano-resolution techniques.
The principle of STED microscopy was first published by Stefan Hell and Jan Wichmann in 1994 [5]. In 2000 the technique was applied to living Saccharomyces cerevisiae and Escherichia coli cells labeled with organic dyes [3]. Since then, STED has been applied to diverse biological preparations, especially in the neuroscience field (see Fig. 1b). Once STED had successfully been applied on immu-nostained microtubules [6], later studies using immunofluorescence became relatively trivial. For example, the vesicle membrane protein synaptotagmin I was shown to remain clustered in the plasma mem-brane after vesicle fusion in cultured mammalian neurons [7]. Also, the doughnut-shaped structure of Bruchpilot, a protein important for accurate synapse assembly in the neuromuscular junction of Drosophila, was described with STED by using immunostaining [8]. Live imaging of antibody- labeled synaptic vesicles (using synapto-tagmin I antibodies) was possible in 2008, featuring imaging speeds of ~28 frames per second [9]. The adaptation of STED microscopy for use with the green fluorescent protein (GFP) [10] opened fur-ther pathways for studying living preparations. Thus, STED imaging of fluorescent proteins has been used to reveal the axon arbors of serotonergic neurons in intact living organisms like Caenorhabditis elegans [11] and to study movements of dendritic spines in both brain slices [4, 12, 13] and, finally, the living mouse brain [14].
This general summary evidences how STED microscopy has been improved to fulfill the requirements of nowadays cell biology research. However, it is important to highlight that subdiffraction resolution will not only depend on the microscope setup or the strength of the depletion laser but also on the thickness of the sam-ple, the fluorophore selection, and the sample preparation. In this chapter we will focus on the technical aspects needed for obtaining high-quality images and points relevant to the data interpretation.
STED Microscopy
216
2 Materials
In the last decade STED has undergone major technical improve-ments and enhancements. The achievable resolution in the lateral and axial dimensions can be changed by using different phase masks, which are used to modify the wavefront of the depletion beam to generate the necessary depletion patterns. The most commonly used is the vortex phase plate, which creates the well-known dough-nut pattern for increasing the lateral resolution [15, 16]. Two-color STED has been performed by combining two pairs of excitation and depletion beams (one pair for each fluorophore) [17–19] or by selecting the fluorophores thus that the same depletion beam could be used for both fluorophores [20]. Three-color STED imaging was achieved (using only two different laser pairs) by separating the used dyes both spectrally and by their excited state lifetimes. This way, two spectrally similar dyes (KK114 and Atto 647 N) could be distinguished, even though they were excited and depleted with the same pair of lasers [21]. Such multicolor STED imaging enables nanoscale colocalization studies of macromolecules in biological systems. Moreover, an isotropic spatial resolution of ~30 nm has been demonstrated by applying STED through two opposing high numerical aperture objectives thereby generating an almost spheri-cal focal spot. This technique was coined isoSTED [20, 22].
STED has been used in conjunction with other microscopy tech-niques, thus proving to be a tool with versatile implementations in the biological field. Two-photon excitation has been combined with STED microscopy [23], becoming useful in the study of neuronal architecture in deep planes of brain slices [24]. STED and electron microscopy were used in a correlative microscopy study, where fluo-rescently tagged proteins first localized by STED microscopy were spatially overlaid with subcellular structures later revealed by electron microscopy, all in the same samples [25]. Furthermore, STED microscopy and fluorescence correlation spectroscopy (FCS) were combined to characterize the diffusion dynamics of single protein and lipid molecules on the plasma membrane of living cells [26].
STED microscopy has remained technically challenging, because most of the improvements and fine-tuning to the particu-lar fluorophores and sample requirements required the construc-tion of customized STED setups. In the meantime, however, Leica Microsystems (Wetzlar, Germany. www.leica-microsystems.com/products/super-resolution/) has been constantly developing and refining commercial STED microscopes, which are also offered as upgrade to their confocal counterparts.
The first setup developed for STED microscopy uses synchronized trains of light pulses for excitation and depletion [3]. The optimal durations of such excitation and depletion pulses range in the order of tens to hundreds of picoseconds; see, for example, [27].
2.1 STED Setups
2.1.1 STED with Pulsed Lasers
Natalia H. Revelo and Silvio O. Rizzoli
217
Typically, an excitation pulse is immediately followed by a depletion pulse in order to prevent early spontaneous fluorescence coming from fluorophores located outside of the doughnut center. Keeping the duration of the excitation-depletion cycle shorter than the excited state lifetime of the fluorophore reduces the probability of detecting those early spontaneous events, which is important for setups where photons are continuously gathered throughout the experiment. However, in cases where this situation cannot be prevented, photons can be selectively gathered right after the depletion pulse by using a time-gated detection system [28].
The temporal separation of excitation and depletion in a pulsed STED configuration implies that relatively low average power is required for depletion. As only a given pool of fluorophores is excited at very specific time points (by the pulsed excitation laser), the depletion beam is only needed at specific points in time, at a moderately high intensity. Although the STED beam intensity required for pulsed systems is two orders of magnitude larger than the one for single-photon excitation (i.e., conventional laser scan-ning confocal imaging), this is still three orders of magnitude lower than the one for multiphoton excitation [29], thus representing a moderate damage to biological probes. Another benefit of pulsed STED is that excited fluorophores can relax from triplet states between pulses before being re-excited. This diminishes fluoro-phore photobleaching (by reducing further excitation and damag-ing of dyes found in the triplet state) and thus allows repeated imaging rounds, important for time-lapse experiments.
Pulsed STED setups are fairly expensive due to the high cost of most pulsed laser sources. As a further drawback, many pulsed depletion lasers are typically tunable in the far-red spectral range (~700–1000 nm), limiting the readily usable fluorescent dyes to the red and infrared spectral range. Imaging more commonly used dyes in the bluer spectral range (e.g., GFP and yellow fluorescent protein YFP) requires additional efforts, such as using an optic parametric oscillator (OPO, for nonlinear frequency conversion) to attain shorter wavelength light (by doubling the frequency of the wavelength to make it match with the red-sided tail of emission of the fluorophore) [7, 10]. Alternative light sources that are easily tunable include stimulated Raman scattering fiber sources [30] or supercontinuum lasers [28, 31].
In most figures presented below, we used a Leica pulsed STED microscopy setup (Leica Microsystems GmbH, Mannheim, Germany), based on a TCS SP5 confocal microscope equipped with a 100× 1.4 NA HCX PL APO oil objective. This microscope uses a pulsed diode laser (18 mW, 80 MHz, 640 nm emission, PicoQuant, Germany) for excitation and a pulsed infrared titanium/sapphire (Ti:Sa) tunable laser (1 W, 80 MHz, 720–1000 nm) for depletion (Mai Tai Broadband, Spectra-Physics, Santa Clara, CA, USA). Detection devices include two ultrasensitive avalanche photodiodes and high-sensitivity, low-noise PMTs (see Subheading 2.2).
STED Microscopy
218
With the aim of making STED more easily applicable, the technique has also been implemented with continuous wave lasers [32]. In this configuration, the spatially overlapped excitation and deple-tion beams are continuously illuminating the sample during image acquisition, making the setup much simpler to implement as no synchronization or modification of laser pulses is necessary. CW STED is more affordable and flexible, since continuous wave lasers are less expensive and are commonly available throughout the visual spectral range, broadening the palette of usable dyes.
The main disadvantage of CW STED is that since the excita-tion beam functions continuously, the population of excited fluo-rophores is continuously renewed and thus needs to be constantly “surrounded” by sufficient depletion photons—resulting in a depletion beam power 3–5 times larger than the time-averaged power of the pulsed STED laser [32]. This difficulty in depleting all excited fluorophores leads to a less-defined fluorescence con-trast at the doughnut inner border, resulting in less sharp images [27, 33]. CW STED was first introduced using for depletion a Ti:sapphire laser operating in the CW mode at 750 nm [32], but later visible fiber lasers emitting at 592 nm were used instead [34]. Another difficulty of CW STED is that the constant illumination of the sample promotes bleaching of the dyes through dark state exci-tation, requiring dyes less prone to reach triplet dark states.
Sensitive detectors are necessary for STED microscopy, as a smaller focal volume results in a smaller amount of photons emitted from the sampled spot. To compensate for the reduced signal, pixel dwell times can be increased and more sensitive detectors can be implemented, so as to record more photons from the same spot.
Most commonly, avalanche photodiodes (APDs) are used as detection devices in STED setups. However, hybrid detectors (e.g., GaAsP-based detectors) have been used more recently, since they combine the wide dynamic range of photomultipliers (PMTs) and the high sensitivity of APDs, producing more contrasted images [35].
Particularly bright samples can be imaged with PMTs, although resolution might be suboptimal due to lower signal to noise ratio. As mentioned above, gating systems can be included in the detec-tion process, to avoid collection of early produced photons that could contaminate the image [33].
As mentioned in the Introduction, STED microscopy could pro-duce in theory an infinitely small excitation volume from which fluorophores could be detected, by arbitrarily increasing the deple-tion laser intensity. Nevertheless, the main limitation for the appli-cation of STED microscopy to biological preparations is the availability of dyes that are stable enough under increasing powers of the depletion laser. Moreover, suitable dyes should have a high
2.1.2 STED with Continuous Wave (CW) Lasers
2.2 Detectors
2.3 Fluorophores and Sample Labeling
Natalia H. Revelo and Silvio O. Rizzoli
219
quantum yield for good detection and should have appropriate fluorescence lifetimes, in case of pulsed STED. At the beginning of STED microscopy, an additional limitation was the matching of emission spectrum of the fluorophores with the available pulsing STED lasers. Dyes in the red side of the spectrum like RH-414 and Pyridine 4 were used for labeling the vacuolar membrane of yeast cells and E. coli membranes, respectively [3].
Atto dyes have been successfully used in the commercial Leica pulsed STED setup, with Atto 647 N and Atto 655 being the most widely used. In our hands, Atto 647 N has worked well in combi-nation with Chromeo 494 for two-color Leica pulsed STED. Technical improvements to double the wavelength fre-quency of a Ti:Sa pulsing laser allowed depletion of Alexa Fluor 594 [24]. With the advent of STED with CW lasers, blue-shifted fluorophores like Alexa Fluor 488 could be also used [34]. Dyes recommended for the Leica CW setup include BD Horizon V500, Oregon Green 488, Chromeo 488, Chromeo 505, and Atto 488, among others (see Note 1).
Most of the aforementioned dyes have been used conjugated to antibodies for immunofluorescence assays of fixed or live sam-ples, easing the study of intermolecular interactions and structural details at the subcellular level. A big step in the study of live cells not only in culture but also in tissues was the application of STED on fluorescent proteins. The first report used a GFP-tagged pro-tein imaged in rotavirus-like particles and GFP fused to an ER-targeting sequence in mammalian cells [10]. Morphological plasticity of neuronal dendrites was assessed in YFP-transgenic mice [4, 13]. A more stable and brighter version of YFP, citrine, was also targeted to the ER and imaged in living cells to detect morphological changes of this organelle in time [36]. A new monomeric far-red fluorescent protein, TagRFP657, was engi-neered by site-specific and random mutagenesis to fit into the excitation and depletion wavelengths of the commercial pulsing STED microscope [37]. The reversible switchable fluorescent proteins (RSFPs), also called photochromic fluorescent proteins, Dronpa and Padron were combined for double-label STED microscopy. Since both proteins are spectrally similar, only one excitation (488 nm) and one depletion beam (595 nm) were needed in combination with a beam at 405 nm that switches Padron off and Dronpa on [38].
Other mechanisms for protein labeling include genetically encoded tags. The SNAP-tag technology is based on the reaction of the protein O6-alkylguanine-DNA alkyltransferase for DNA repairing. The reaction can be used to transfer fluorophores from benzylguanines (BG) to the SNAP-tag, which can be fused with a protein of interest. Vimentin, MAP2, caveolin, and connexin-43 were SNAP-tagged and labeled with tetramethylrhodamine for STED imaging [39]. Another type of tag is the fluorogen- activating
STED Microscopy
220
proteins (FAPs), which render nonfluorescent organic dyes fluorescent upon specific binding. Thus, actin molecules fused with FAPs were evidenced by activation of malachite green added to the media of living HeLa cells [40].
3 Methods
Different types of biological samples, fixed or living, can be studied under the STED microscope (see Subheading 1). Thin samples prepared on glass coverslips, including cell monolayer cultures, organelles, or membrane sheets, are very easy to image. The close proximity of the objects to the glass excludes major aberrations due to changes in the refractive index along the sample.
For live cell imaging the sensitivity of the detection device and the speed of movement can be limiting. Despite this, time-lapse record-ings have been performed to follow morphological changes of the endoplasmic reticulum in living cells [36] or to track single vesicle fast movements in synaptic boutons and axons at video-rate imaging [9].
In thick preparations (e.g., tissue sections or in vivo imaging) the differences in refractive index between the oil-immersion medium and the biological specimen reduce the STED effect when focusing at deeper planes. Glycerol-immersion objectives with a correction collar have been successfully used to overcome this type of aberration, allowing subdiffraction resolution as deep as 120 μm into living brain slices [12] or up to 15 μm in the living mouse brain [14]. Nonetheless, these studies required abundant cytosolic expression of EYFP in neurons to get a sufficiently bright signal that could be imaged at such depths.
Immunostaining of thick fixed samples can be performed as for confocal microscopy, but special attention should be paid to the selection of the embedding medium (see below).
Thin fixed samples (e.g., cell cultures, membrane sheets, isolated organelles) can be easily embedded in Mowiol, a polyvinyl alcohol- based, water-soluble polymer. For thicker samples the fluorescence intensity and STED resolution will drop at deeper planes due to spherical aberrations caused by a mismatch between the refractive index (n) of the embedding medium and the ones of the glass cov-erslip (n = 1.515) and the immersion oil (n = 1.518). Commonly used embedding media like Mowiol (n = 1.49) and glycerol (n = 1.45) are therefore not suitable [41]. 2,2′-Thiodiethanol (TDE) has been characterized as the best embedding medium to solve this problem, since its n can be easily adjusted by dilution with water (n = 1.515 for a 97 % TDE solution). TDE also com-pensates for changes of n generated by large cellular components and preserves the quantum yield of many fluorophores [41]. Drawbacks of TDE embedding are short life of samples (1–2
3.1 Samples
3.2 Embedding Procedures
Natalia H. Revelo and Silvio O. Rizzoli
221
weeks) and higher mobility compared to Mowiol embedding (see Note 2). Figure 2 shows differences in STED resolution at different depths between a sample embedded in Mowiol and one embedded in TDE.
For even bigger fixed samples like whole brain preparations or small organisms, tissues can be embedded in a more stable matrix to allow subsequent slicing in ultrathin sections. This is an extremely important procedure, as (1) such samples are rarely easy to image in high-resolution microscopy, and correct embedding and sec-tioning procedures remove this difficulty; (2) ultrathin sectioning can be performed down to at least ~40 nm, thus offering a trivial (and cheap) method of obtaining super-resolution in the Z-axis.
Several plastic resins are available in the market with different polymerization temperatures, hardness, and pH values. Preliminary tests with different resins are advisable according to the type of sample and fluorophores used. For example, glycol methacrylate (GMA) was selected for imaging fluorescent proteins expressed in C. elegans, due to its good penetration and polymerization at pH 8, important for stability of the fluorophore [25]. In our laboratory we have successfully used melamine, a highly water- compatible resin that does not require previous dehydration of the tissue,
Fig. 2 TDE embedding preserves STED resolution along the axial plane in thick samples. The cranial muscle leva-tor auris longus, responsible for ear movement, was dissected from a mouse. After immunostaining against the protein complexin (involved in the regulation of synaptic vesicle fusion), sections of the muscle were embedded in Mowiol (see Subheading 3.2.1) or TDE (see Subheading 3.2.2) for comparison. Complexin was imaged at the neuromuscular junctions (NMJs) by STED microscopy at four different planes: at the surface of the muscle and around 10, 20, and 30 μm of depth into the tissue. Note that STED resolution decreases at deeper planes of the Mowiol-embedded sample, due to spherical aberrations produced when the light beam reaches a medium of a refractive index different to those of the oil and the glass coverslip. In contrast, the perfect match in refractive index offered by TDE allows imaging at deep planes with preserved STED resolution
STED Microscopy
222
ensuring its better preservation. Melamine embedding followed by thin sectioning has also been used with cell cultures in order to improve axial resolution for 3D reconstruction [42]. Figure 3 shows STED imaging of melamine sections from immunostained neurons. We have also used the acrylic resin LR White for samples labeled with chemical dyes, although dehydration steps are required, preserving membranes more poorly.
1. Bring 24 g glycerol, 9.6 g Mowiol 4-88 reagent (Merck Millipore, Merck KGaA, Darmstadt, Germany), 62.4 mL dis-tilled water, and 9.6 mL 1 M Tris buffer into a conical cylinder. Mix with a magnetic stirrer for 5–7 days. Heat the mixture at 40–50 °C to help Mowiol dissolving. Let the mixture settle and aliquot only the supernatant in 1.5 mL tubes. Freeze the aliquots by dipping them into liquid nitrogen. Store at −20 °C.
2. Sample embedding: thaw an aliquot of the Mowiol solution. Place a 10–12 μL drop on a glass microscope slide (for 18 mm
3.2.1 Mowiol Embedding
Fig. 3 Improving resolution in the Z-axis. Imaging large specimens (several millimeters thick) or generating high-resolution 3D reconstructions of cultured cells is possible using melamine embedding followed by ultra-thin sectioning. This is an easy method to obtain subdiffraction axial resolution that complements the improved lateral resolution achieved with STED. As example, hippocampal neurons immunostained for the proteins synaptosomal-associated protein 25 (SNAP-25), bassoon, or syntaxin 16 (Syx16), embedded in melamine (see Subheading 3.2.3.) and sliced to 100–150 μm thick sections, were imaged in confocal and STED microscopy. The result in the STED image is clearly defined organelles, well differentiated from their surrounding environ-ment, which is hardly achieved in the confocal image
Natalia H. Revelo and Silvio O. Rizzoli
223
diameter coverslips). With the help of curved forceps, hold the coverslip containing the cell culture vertically and dry it with a soft tissue from the lower border. Place the coverslip on top of the Mowiol drop. Mowiol hardens overnight at room tem-perature or in 20 min at 37 °C (see Notes 3 and 4).
1. Prepare solutions of 30, 50, 70, and 90 % TDE (Sigma Chemical Company, St. Louis, MO, USA) in distilled and fil-tered water.
2. After the last wash step of the labeling or immunostaining pro-cedure, incubate the sample in 30 % TDE solution for 10 min. Repeat this step with 50, 70, and 90 % TDE solution sequen-tially. Incubate in 100 % TDE three times more for 10 min.
3. Sample embedding: place the sample on the center of a glass coverslip of 30 mm diameter. Add a drop of 100 % TDE on top of the sample and place a coverslip of 18 mm diameter on top. Apply nail polish on the rim of the 18 mm coverslip and let the sample dry for 1 h before imaging. This combination of cover-slips allows imaging on either side of the sample—important for thick samples (the only ones for which TDE embedding is supe-rior to Mowiol, as indicated above). Store the samples at 4 °C and image within 5 days after sample preparation, since the mor-phology of the samples is not well preserved beyond a few days.
1. Weight 48 mg p-Toluenesulfonic acid monohydrate (Cat. No. 402885, Sigma-Aldrich Inc., St. Louis, MO, USA) and bring it to a 15 mL conical tube. Add 0.576 mL distilled water. Vortex until complete homogenization of the solution. Add 1.344 g 2,4,6-Tris[bis(methoxymethyl)amino]-1,3,5-triazine (melamine, Cat. No. T2059, TCI Europe, Zwijndrecht, Belgium) and vortex for several minutes. Flip the tube several times until all the melamine is in contact with the solution. Fix the tube with adhesive tape on a horizontal shaker and agitate it at 250 rpm for 2 h or until the melamine is completely dis-solved. Vortex shortly every 30 min to speed up this process.
2. Label the sample fluorescently as for any other imaging proce-dure (e.g., by immunostaining).
3. Sample embedding: place the labeled sample on an 18 mm cov-erslip. Take into account the orientation of the tissue for later processing. With the help of a micropipette, remove as much buffer as possible surrounding the sample. Cut small pieces of Whatman filter paper and with their tips dry remaining buffer around the sample. Cut the bottom and the lid of a BEEM cap-sule (Beem Inc., West Chester, PA, USA) and place the capsule with the upper rim down surrounding the sample. Carefully add 200 μL of freshly prepared melamine solution (from step 1) drop by drop into the BEEM capsule while covering the sample.
3.2.2 TDE Embedding
3.2.3 Melamine Embedding
STED Microscopy
224
Put the coverslip into the lid of a small plastic dish and this one into a box containing silica gel for desiccation (Sigma Chemical Company, St. Louis, MO, USA). Store in a dry and dark place 24 h at room temperature to ensure penetration of the embed-ding medium into the specimen.
4. With the sample still on the silica gel, incubate for 24 h at 40 °C.
5. Prepare Epon resin (Epofix kit, Cat. No. 40200029, Struers, Ballerup, Denmark) and pour over the melamine layer, filling the BEEM capsule to the top. Incubate for 72 h at 60 °C or until it is completely hardened (see Note 6).
6. The coverslip can be detached from the block by pressing it between the thumb and the index finger. Samples can be also dipped into liquid nitrogen. Check if the melamine is com-pletely hard. If not incubate on silica gel for 24 h more at 60 °C.
7. Cut the BEEM capsule to obtain the polymerized block.
8. Trim around the sample where necessary. Use a microtome to cut sections of 50–200 nm (this procedure is identical to those performed in electron microscopy ultrathin processing). These sections can be dried on a glass coverslip and mounted with Mowiol on a slide for STED imaging (see Notes 5 and 6).
For this embedding procedure one should use LR White medium grade resin and LR White accelerator (London Resin Company Ltd., Reading, Berkshire, England).
1. Prepare solutions of 30, 50, 70, 90, and 95 % EM grade etha-nol in distilled and filtered water.
2. After the last wash step of the labeling or immunostaining pro-cedure, a dehydration process with increasing concentrations of ethanol is required. Incubate the sample in a 30 % ethanol solution for 10 min. Repeat this step with 50, 70, 90, 95, and 100 % ethanol solutions, each 10 min. Incubate once more in 100 % ethanol for 10 min.
3. Sample embedding: place the sample on an 18 mm coverslip. Take into account the orientation of the tissue for later pro-cessing. With the help of a micropipette, remove as much etha-nol as possible. Cut small pieces of Whatman filter paper and with their tips dry remaining ethanol around the sample. Cut the bottom and the lid of a BEEM capsule (Beem Inc., West Chester, PA, USA) and place the capsule with the upper rim down surrounding the sample. Carefully add a small volume of the LR White resin, enough to cover the sample. Incubate overnight at room temperature to allow penetration of the embedding medium into the specimen.
4. Prepare a mixture of 10 mL LR White resin and 1 drop of LR White accelerator. Fill the BEEM capsule with the mixture.
3.2.4 LR White Embedding for Large Specimens
Natalia H. Revelo and Silvio O. Rizzoli
225
The resin will harden within 10–20 min. Cooling the preparation to 4 °C, or even lower, should be used at this step if better morphology preservation is required.
5. The coverslip can be detached from the block by pressing it between the thumb and the index finger. Remaining glass pieces can be removed under a microscope by manipulating them with a forceps or with razor blades.
6. Cut the BEEM capsule to obtain the polymerized block. 7. Trim the sample where necessary. Use a microtome to cut sec-
tions of about 50–200 nm and mount and embed them as above.
We have successfully used LR White embedding in cell cultures for cases when not only lateral, but also high axial resolution is required. One difficulty is that some cell types (e.g., COS7 cells) do not easily detach from the coating of the glass coverslip, making difficult the transfer into the embedding medium. The following protocol is a modification to the protocol from Subheading 3.2.4, introduced to overcome this difficulty.
1. After the last wash step of the labeling or immunostaining pro-cedure, incubate the cultured coverslip in a solution of 30 % EM grade ethanol, three times each 5 min.
2. Incubate the coverslip in a 1:1 mixture of LR White resin and 30 % ethanol, for 30 min.
3. Incubate in 100 % LR White resin for 60 min. 4. Prepare a mixture of 10 mL LR White resin and one drop of
LR White accelerator and vortex. 5. Dry out the excess of LR White from the coverslip by cleaning
the border with a soft tissue. Place the coverslip with the cells facing up on a metal block covered with parafilm and previ-ously cooled at 4 °C. Cut the bottom and the lid of a BEEM capsule (Beem Inc., West Chester, PA, USA) and place the capsule with the upper rim facing the cell culture.
6. Add the mixture of resin: accelerator (from step 4) into the BEEM capsule up to the upper border. Let it harden for 10–20 min.
7. Continue in step 5 from the previous protocol (Subheading 3.2.4).
8. See also Note 7.
Special care must be taken when analyzing STED images from immunostained samples. The use of primary antibodies fol-lowed by fluorescently labeled secondary antibodies increases the distance between the recognized epitope and the fluoro-phore, affecting interpretation of the spatial distribution of the target protein. Moreover, the volume of an antibody recogniz-
3.2.5 LR White Embedding for Cell Cultures
3.3 Data Interpretation
STED Microscopy
226
ing one epitope may hinder the recognition of adjacent epit-opes, reducing in this way the accuracy and density [43] of the staining and offering and incomplete picture in a diffraction-unlimited image (see Fig. 4). In many cases the zoom in a STED picture will show a collection of spots that would not reflect the real molecular structure of the target. Although fluorescently labeled primary antibodies can be used, signal intensity might be reduced. It is important to optimize the immunostaining protocols according to the antibodies used and the protein of
Fig. 4 Deficient immunostaining protocols show artifactual protein distributions. The large size of primary and secondary antibodies used in immunostaining methods imposes two difficulties: poor recognition of epitopes adjacent to an already recognized epitope and location of fluorophores decorating the secondary antibody far from the actual epitope position. To demonstrate this drawback, hippocampal neurons were incubated with primary antibodies against one cytoskeletal protein (actin, neurofilaments, or tubulin) or the cytoplasmic pro-tein N-ethylmaleimide-sensitive fusion protein (NSF). Atto 647-labeled secondary antibodies were later used. Low-zoom images of neuronal processes show apparent filaments for actin, neurofilaments, and tubulin, while NSF is distributed more homogeneously. A larger zoom (insets) reveals that the filaments are not continuous and are actually featured by fairly separated spots, now difficult to differentiate from the enlarged NFS picture. Correlation and distribution analyses can be negatively affected by misleading information about the true loca-tion of the protein of interest
Natalia H. Revelo and Silvio O. Rizzoli
227
interest. Different concentrations of fixatives, blocking solutions, and antibodies should be tested. With appropriate epitope access good quality images can be acquired, even suit-able for correlation analyses (Fig. 5).
In a recent study, conventional antibodies (~150 KDa, ~10 nm in size) were replaced by smaller anti-GFP nanobodies (~13 KDa, 1.5 × 2.5 nm) coupled to Alexa Fluor 647, to label GFP-fused proteins in yeasts. High-resolution images showed increased label density and reduction of the full width at half maximum (FWHM) values almost by half, when compared to a conventional immunostaining protocol with primary and secondary antibodies [44]. In our laboratory we have also per-formed, for the first time, high- resolution STED microscopy of nucleotide-based aptamers. These are single-stranded oligonucleotides,
Fig. 5 Good immunostaining procedures allow reliable protein colocalization analyses. As an example, the synaptic vesicle proteins synaptophysin and synaptotagmin were shown to remain clustered on the plasma membrane after synaptic vesicle fusion. (a) Experimental paradigm: following stimulated exocytosis the lumi-nal domain of synaptotagmin is labeled by Atto 647-coupled mouse anti-synaptotagmin antibodies (red), and synaptophysin is bound by rabbit primary anti-synaptophysin antibodies (black). After fixation, Atto 590-cou-pled anti-rabbit secondary antibodies (green) are added. (b) STED images of synaptotagmin (upper panel) and synaptophysin (middle panel) show perfect colocalization at several vesicle release sites in a neuronal process (lower panel). Reproduced with permission from Ref. [46]
STED Microscopy
228
RNA or DNA, that can specifically bind to other molecules and therefore can be used in a similar fashion to antibodies. Aptamers present several advantages over antibodies. They are syntheti-cally generated, can be directly coupled to fluorophores, are small (up to a few tens of kDa), and can be easily engineered according to the experimenter needs [45].
Images acquired with STED microscopy will need the standard noise reduction by filtering (using, e.g., median filtering). When imaging two molecular species by two-color STED, simple signal overlap will not be sufficient to estimate colocalization. Spatial analysis methods like measurement of distance between the spots can be a good option [46]. The Pearson’s correlation analysis has also been applied in this kind of experiments [47]. For accurate results, it is important to verify that satisfactory resolution values were attained in both channels.
4 Notes
1. For a more detailed list of dyes used in STED microscopy, see [35] or visit nanobiophotonics.mpibpc.mpg.de/old/dyes.
2. We recommend to image TDE embedded samples within 5 days after preparation.
3. Application of nail polish on the rim of the coverslip adds to sample preservation and stability.
4. Samples embedded in Mowiol can be stored at 4 °C for several months depending on the quality of the fluorescent labeling.
5. Although the Z-resolution is obviously higher in thinner sam-ples, they will be substantially dimmer than the thicker ones.
6. Melamine embedding can be also applied to cells cultured on coverslips (Fig. 3). When embedding cells, we recommend to trim the melamine only after it is completely hardened, to avoid modifying the flat surface of the cell layer, through the pressure applied during trimming. In contrast, when embed-ding large tissue preparations, we recommend to trim the sam-ples when the melamine is still soft (~48 h after incubation at 60 °C, step 5) to prevent fracture of the sample. Different batches of reagents can have different hardening speeds, and therefore it is necessary to identify the appropriate length of the hardening period at 60 °C.
7. LR White embedding is preferred over melamine embedding when the samples need to be also analyzed by electron microscopy.
3.3.1 Data Analysis
Natalia H. Revelo and Silvio O. Rizzoli
229
Acknowledgments
We thank Nicolai T. Urban for advice and for reading the manu-script, Felipe Opazo for providing images used in Fig. 4, and Christina Schäfer and Katharina Kröhnert for technical assistance. S.O.R. acknowledges the support of a Starting Grant from the European Research Council, Program FP7 (NANOMAP). N.H.R. acknowledges the support of the Deutsche Forschungsgemeinschaft (SFB 889).
References
1. Abbe E (1873) Beiträge zur theorie des mikroskops und der mikroskopischen wahrne-hmung. Arch für Mikroskopische Anat 9:413–418
2. Hell SW, Dyba M, Jakobs S (2004) Concepts for nanoscale resolution in fluorescence micros-copy. Curr Opin Neurobiol 14:599–609
3. Klar TA, Jakobs S, Dyba M et al (2000) Fluorescence microscopy with diffraction reso-lution barrier broken by stimulated emission. Proc Natl Acad Sci U S A 97:8206–8210
4. Nägerl UV, Willig KI, Hein B et al (2008) Live-cell imaging of dendritic spines by STED microscopy. Proc Natl Acad Sci U S A 105:18982–18987
5. Hell SW, Wichmann J (1994) Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy. Opt Lett 19:780–782
6. Dyba M, Jakobs S, Hell SW (2003) Immunofluorescence stimulated emission depletion microscopy. Nat Biotechnol 21:1303–1304
7. Willig KI, Rizzoli SO, Westphal V et al (2006) STED microscopy reveals that synaptotagmin remains clustered after synaptic vesicle exocy-tosis. Nature 440:935–939
8. Kittel RJ, Wichmann C, Rasse TM et al (2006) Bruchpilot promotes active zone assembly, Ca2+ channel clustering, and vesicle release. Science 312:1051–1054
9. Westphal V, Rizzoli SO, Lauterbach MA et al (2008) Video-rate far-field optical nanoscopy dissects synaptic vesicle movement. Science 320:246–249
10. Willig KI, Kellner RR, Medda R et al (2006) Nanoscale resolution in GFP-based micros-copy. Nat Methods 3:721–723
11. Rankin BR, Moneron G, Wurm CA et al (2011) Nanoscopy in a living multicellular organism expressing GFP. Biophys J 100:L63–L65
12. Urban NT, Willig KI, Hell SW et al (2011) STED nanoscopy of actin dynamics in synapses deep inside living brain slices. Biophys J 101:1277–1284
13. Tønnesen J, Nadrigny F, Willig KI et al (2011) Two-color STED microscopy of living synapses using a single laser-beam pair. Biophys J 101:2545–2552
14. Berning S, Willig KI, Steffens H et al (2012) Nanoscopy in a living mouse brain. Science 335:551
15. Keller J, Schönle A, Hell SW (2007) Efficient fluorescence inhibition patterns for RESOLFT microscopy. Opt Express 15:3361–3371
16. Wildanger D, Medda R, Kastrup L et al (2009) A compact STED microscope providing 3D nanoscale resolution. J Microsc 236:35–43
17. Donnert G, Keller J, Wurm CA et al (2007) Two-color far-field fluorescence nanoscopy. Biophys J 92:L67–L69
18. Blom H, Rönnlund D, Scott L et al (2012) Nearest neighbor analysis of dopamine D1 receptors and Na(+)-K(+)-ATPases in dendritic spines dissected by STED microscopy. Microsc Res Tech 75:220–228
19. Meyer L, Wildanger D, Medda R et al (2008) Dual-color STED microscopy at 30-nm focal- plane resolution. Small 4:1095–1100
20. Schmidt R, Wurm CA, Jakobs S et al (2008) Spherical nanosized focal spot unravels the interior of cells. Nat Methods 5:539–544
21. Bückers J, Wildanger D, Vicidomini G et al (2011) Simultaneous multi-lifetime multi- color STED imaging for colocalization analy-ses. Opt Express 19:3130–3143
STED Microscopy
230
22. Schmidt R, Wurm CA, Punge A et al (2009) Mitochondrial cristae revealed with focused light. Nano Lett 9:2508–2510
23. Moneron G, Hell SW (2009) Two-photon excitation STED microscopy. Opt Express 17:14567–14573
24. Ding JB, Takasaki KT, Sabatini BL (2009) Supraresolution imaging in brain slices using stimulated-emission depletion two-photon laser scanning microscopy. Neuron 63:429–437
25. Watanabe S, Punge A, Hollopeter G et al (2011) Protein localization in electron micro-graphs using fluorescence nanoscopy. Nat Methods 8:80–84
26. Eggeling C, Ringemann C, Medda R et al (2009) Direct observation of the nanoscale dynamics of membrane lipids in a living cell. Nature 457:1159–1162
27. Leutenegger M, Eggeling C, Hell SW (2010) Analytical description of STED microscopy performance. Opt Express 18:26417–26429
28. Galiani S, Harke B, Vicidomini G et al (2012) Strategies to maximize the performance of a STED microscope. Opt Express 20:7362–7374
29. Donnert G, Keller J, Medda R et al (2006) Macromolecular-scale resolution in biological fluorescence microscopy. Proc Natl Acad Sci U S A 103:11440–11445
30. Rankin BR, Hell SW (2009) STED microscopy with a MHz pulsed stimulated-Raman- scattering source. Opt Express 17: 15679–15684
31. Wildanger D, Rittweger E, Kastrup L et al (2008) STED microscopy with a supercontin-uum laser source. Opt Express 16:9614–9621
32. Willig KI, Harke B, Medda R et al (2007) STED microscopy with continuous wave beams. Nat Methods 4:915–918
33. Vicidomini G, Moneron G, Han KY et al (2011) Sharper low-power STED nanoscopy by time gating. Nat Methods 8:571–573
34. Moneron G, Medda R, Hein B et al (2010) Fast STED microscopy with continuous wave fiber lasers. Opt Express 18:1302–1309
35. Müller T, Schumann C, Kraegeloh A (2012) STED microscopy and its applications: new insights into cellular processes on the nanoscale. Chemphyschem 13(8):1986–2000. doi:10.1002/cphc.201100986
36. Hein B, Willig KI, Hell SW (2008) Stimulated emission depletion (STED) nanoscopy of a fluorescent protein-labeled organelle inside a living cell. Proc Natl Acad Sci U S A 105:14271–14276
37. Morozova KS, Piatkevich KD, Gould TJ et al (2010) Far-red fluorescent protein excitable with red lasers for flow cytometry and super-resolution STED nanoscopy. Biophys J 99:L13–L15
38. Willig KI, Stiel AC, Brakemann T et al (2011) Dual-label STED nanoscopy of living cells using photochromism. Nano Lett 11:3970–3973
39. Hein B, Willig KI, Wurm CA et al (2010) Stimulated emission depletion nanoscopy of living cells using SNAP-tag fusion proteins. Biophys J 98:158–163
40. Fitzpatrick JA, Yan Q, Sieber JJ et al (2009) STED nanoscopy in living cells using Fluorogen Activating Proteins. Bioconjug Chem 20:1843–1847
41. Staudt T, Lang MC, Medda R et al (2007) 2,2′-thiodiethanol: a new water soluble mount-ing medium for high resolution optical micros-copy. Microsc Res Tech 70:1–9
42. Punge A, Rizzoli SO, Jahn R et al (2008) 3D reconstruction of high-resolution STED microscope images. Microsc Res Tech 71:644–650
43. McKinney SA, Murphy CS, Hazelwood KL et al (2009) A bright and photostable photo-convertible fluorescent protein. Nat Methods 6:131–133
44. Ries J, Kaplan C, Platonova E et al (2012) A simple, versatile method for GFP-based super- resolution microscopy via nanobodies. Nat Methods 9(6):582–584
45. Opazo F, Levy M, Byrom M et al (2012) Aptamers as potential tools for super- resolution microscopy. Nat Methods 9:938–939
46. Hoopmann P, Punge A, Barysch SV et al (2010) Endosomal sorting of readily releasable synaptic vesicles. Proc Natl Acad Sci U S A 107:19055–19060
47. Neumann D, Bückers J, Kastrup L et al (2010) Two-color STED microscopy reveals different degrees of colocalization between hexokinase-I and the three human VDAC isoforms. PMC Biophys 3:4
Natalia H. Revelo and Silvio O. Rizzoli
231
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_13, © Springer Science+Business Media New York 2015
Chapter 13
Three-Dimensional Photoactivated Localization Microscopy with Genetically Expressed Probes
Kelsey Temprine , Andrew G. York , and Hari Shroff
Abstract
Photoactivated localization microscopy (PALM) and related single-molecule imaging techniques enable biological image acquisition at ~20 nm lateral and ~50–100 nm axial resolution. Although such techniques were originally demonstrated on single imaging planes close to the coverslip surface, recent technical developments now enable the 3D imaging of whole fi xed cells. We describe methods for converting a 2D PALM into a system capable of acquiring such 3D images, with a particular emphasis on instrumentation that is compatible with choosing relatively dim, genetically expressed photoactivatable fl uorescent proteins (PA-FPs) as PALM probes. After reviewing the basics of 2D PALM, we detail astigmatic and multiphoton imaging approaches well suited to working with PA-FPs. We also discuss the use of open-source localiza-tion software appropriate for 3D PALM.
Key words Super resolution , Single-molecule imaging , 3D microscopy , Cell biology
1 Introduction
Modern fl uorescence microscopy is an invaluable tool for the biologist, combining contrast, molecular specifi city, and biocom-patibility to enable the visualization of cellular constituents. Unfortunately, diffraction limits the spatial resolution of a wide- fi eld fl uorescence microscope to ~250 nm laterally and 500–750 nm axially, and achieving even this “diffraction-limited” performance is diffi cult in practice. Methods for imaging beyond the diffraction limit are thus of great practical interest in cell biol-ogy, and the last decade has seen the emergence of such “super- resolution” optical imaging techniques [ 1 ].
In this chapter, we focus on photoactivated localization micros-copy (PALM) [ 2 ] and related single-molecule imaging techniques (e.g., STORM [ 3 ], fPALM [ 4 ]). In these techniques, the specimen is labeled with a photoswitchable fl uorescent probe. During data acquisition, the fl uorescence emitted from a sparse subset (“sparse” means that the image of each molecule is resolvable) of probe
232
molecules is isolated by switching on only a few molecules at a time. The fl uorescence is recorded using a camera, the probe molecules are switched off, and the process repeated until most or all mole-cules have been photoswitched. The raw data thus collected consist of thousands of individual frames, each containing diffraction- limited fl uorescence images of single photoswitchable molecules. The image of each molecule (point-spread function or PSF) is analyzed to localize its center with high precision, and the resulting localizations are combined to generate a super-resolution (in opti-mal conditions, resolutions of ~20 nm or better can be achieved) image of the positions of the molecules (Fig. 1 ). Resolution is ultimately dependent on (a) localization precision (how precisely each probe molecule may be localized) and (b) density of localiza-tions assembled in the fi nal image ( see Note 1 ). Localization preci-sion and density are in turn dependent critically on the choice of switchable molecule, and it is this choice that largely determines the success or failure of a single-molecule super-resolution imaging experiment.
Broadly speaking, photoswitchable probes may be divided into two broad categories: those that may be genetically expressed (e.g., photoactivatable fl uorescent proteins, PA-FPs [ 5 ]) and those synthetic dyes that may be exogenously introduced via anti-bodies or small-molecule labeling strategies [ 6 ]. Each class has its
Fig. 1 Concept behind PALM. When labeled with a fl uorescent probe and viewed through a conventional micro-scope, high spatial frequency detail in an object ( leftmost column ) is obscured, producing a diffraction-limited image ( middle left column ). PALM may be used to obtain a much higher-resolution image, if two criteria are met. First, the sample must be labeled with a photoswitchable probe and only a sparse subset of probes induced to activate and fl uoresce in any given frame ( middle right column ). This “isolation” procedure alone is insuffi cient, as combining the recorded fl uorescence from all N frames recovers the diffraction-limited image. However, if the optically resolved images of the individual molecules are also localized ( rightmost column ) with high precision, the aggregate image resulting from all such localizations may be combined to obtain a super-resolution image
Kelsey Temprine et al.
233
own strengths and weaknesses: PA-FPs are small (a few nm in size), may be targeted with molecular precision and are thus highly spe-cifi c, and may possess a high (>1,000) contrast ratio between bright and dark states, helpful in isolating a single fl uorescent mol-ecule from its neighbors and acquiring high-density (and thus high resolution) images. Unfortunately, PA-FPs are in general dimmer than most synthetic dyes, limiting localization precision and chal-lenging the detection of activated single-molecule signal above cellular autofl uorescence and residual faint background emission from the potentially large number of inactive molecules. They are also available only in limited colors, and the brightness and con-trast ratio of the green-emitting PA-FPs is generally lower than the red-emitting PA-FPs (although the palette of PA-FPs continues to expand [ 7 ] and the performance of green PA-FPs continues to improve [ 8 ]). Switchable synthetic dyes can be very bright and are available in many colors [ 9 ], signifi cant advantages over PA-FPs. However, the coupling method (e.g., to antibodies) can introduce signifi cant uncertainty in their position, thus defeating the bright-ness advantage and effectively limiting resolution. Exogenously introduced dyes often label their target less specifi cally than PA-FPs, leading to an increased level of background in the reconstructed super- resolution image and sometimes introducing uncertainty as to whether localizations are real or spurious. Finally, many exoge-nous dyes require an additional chemical cocktail (typically reduc-ing agents or oxygen scavengers) and/or high excitation intensities, limiting application in live cells. In this chapter, we focus on instru-mentation particularly suited to the somewhat dimmer PA-FPs.
Initial PALM experiments were conducted in 2D [ 2 , 10 ] near the coverslip surface, where optical aberrations and scattering are minimized, and the full numerical aperture (NA) of oil objectives may be used to maximize signal collection and aid in the localiza-tion precision of dim PA-FPs. An additional advantage of 2D imag-ing near the coverslip is that total internal refl ection (TIR [ 11 ]) may be used to limit excitation to the focal plane and further aid in single-molecule detection. As it forms the basis for 3D experi-ments, we cover 2D PALM instrumentation in Subheading 3.1 .
While suited to membrane [ 12 ] and focal adhesion [ 13 ] stud-ies in live cells, the obvious drawback of TIRF is that the vast majority of the cell is inaccessible. The simplest way around this problem is to move the excitation out of TIRF, thus employing epi- or wide-fi eld imaging to illuminate the entire cell. This imag-ing modality also creates several problems for 3D super-resolution imaging: (a) in addition to x and y , a molecule’s z coordinate must be localized with subdiffractive precision, and the resulting infor-mation used in assembling a 3D PALM dataset; (b) axial drift must be measured and/or controlled; and (c) background photoactiva-tion and subsequent fl uorescence of probe molecules at axial locations other than the focal (imaging) plane must be minimized,
Three-Dimensional PALM
234
lest their localizations be wasted and the extraneous background signal so created impede localization in the focal plane. We discuss technical solutions to (a–b) in Subheading 3.2 and (c) in Subheading 3.3 . We also refer to the reader to [ 14 , 15 ], as this chapter draws heavily from these sources.
2 Materials
1. IX-81 microscope frame equipped with TIRF illuminator, left- side port, DIC optics, fi lter cube turret, and 1.6× internal beam expander (Olympus America, IX2-RFACB2-R).
2. Automated XY stage (Applied Scientifi c Instrumentation, MS-2000).
3. High numerical aperture (1.49 NA) 60× objective lens, suit-able for TIRF (Olympus America, APON 60XOTIRF).
4. Filters in microscope frame, used for PA-FP imaging: (a) mEos2/pa-mCherry1 dichroic mirror (Semrock, FF562-
Di02- 25x36); bandpass. (b) mEos2/pa-mCherry1 emission fi lter (Semrock, FF01-617/
73-25). (c) Dronpa/PS-CFP2 dichroic mirror (Chroma, T495lp). (d) Dronpa/PS-CFP2 bandpass emission fi lter (Chroma,
ET525/50). 5. Optical table (Technical Manufacturing Corporation). 6. Overhead shelf system (Newport, ATS-10). 7. Sticky pads to remove dust (Cole Static Control). 8. Temperature/humidity USB monitor (Practical Design Group
LLC, THUM). 9. Excitation optics sled, optomechanics, and spacers for raising
sled to TIRF port height (Thorlabs, PBH11106, BLP01, PF175, PS1, PS2, PS3, PS4).
10. Lasers for common PA-FPS: (a) Excitation for mEos2, pa-mCherry1, 200 mW, 561 nm
(Crystalaser, CL561-200). (b) Excitation for Dronpa/PS-CFP2, 150 mW, 488 nm
(Newport, Cyan scientifi c laser). (c) Photoactivation of PA-FPs, 100 mW, 405 nm (Coherent,
405-100 CIRCULAR). 11. Sled optics:
(a) 561 nm beam expansion, 2.7×, Thorlabs f = 15 mm LA1540-A and f = 40 mm LA1304-A.
(b) 488 nm beam expansion, 4×, Thorlabs f = 25 mm AC127-025- A-ML and f = 100 mm AC254-100-A-ML.
2.1 2D PALM Instrumentation
Kelsey Temprine et al.
235
(c) 405 nm beam expansion, 2×, Thorlabs f = 15 mm LA1540-A and f = 30 mm LA1289-A.
(d) Neutral density fi lters (Edmund Scientifi c, NT54-460). (e) Cage optomechanics for mounting beam expanders and
neutral density fi lters (Thorlabs, CP02T, ER4, ER6). (f) Shutter, controller, and power supply (Thorlabs, SH05,
TSC001, TCH002). (g) Visible mirrors and holders (Thorlabs, BB1-E02 and KS1). (h) External dichroic mirror for integrating 488 nm laser
(Semrock, FF506-Di02-25x36). (i) External dichroic mirror for integrating 405 nm laser
(Semrock, FF458-Di02-25x36). (j) Dichroic mirror holders (Thorlabs, KM100C). (k) Beam steering assembly (Thorlabs, GN05 and CVI/
Melles Griot 12.5 mm diameter/5 mm thick fused silica windows).
(l) Focusing lens (Edmund Optics, f = 50 mm achromatic lens, NT49-356-INK) and holder (Thorlabs, ST1XY-S and SM1ZM).
(m) Collimating lens (Edmund Optics, f = 100 mm achromatic lens, NT49-360-INK) and holder (Thorlabs, ST1XY-S and SM1ZM).
(n) Translation stage for imaging focused beam at back focal plane of microscope (Newport, 461-X-M and SM-06).
12. Sample holder and stage insert (Applied Scientifi c Instrumentation, I-3033 and I-3033-25D).
13. 25 mm, #1.5 coverslips (Warner Instruments, CS-25R15). 14. Back-thinned, electron-multiplying CCD (Andor Technology,
DU-888E-C00-#BV). 15. 1.2× emission-side beam expander (Diagnostic Instruments,
DD12BXC). 16. Acquisition Computer (Advantech, SYS-4U-BTO with 2 TB
hard drive, 2.5 GB RAM, Dual 2.13 GHz processors).
1. Automated XY translation stage with z piezo (Applied Scientifi c Instrumentation, PZ-2000).
2. High numerical aperture (1.2 NA) 60× water-immersion objective lens (Olympus America, UPLSAPO60XW).
3. Focus-lock system to minimize z drift (Mad City Labs, C-Focus).
4. Gold fi ducial particles, 100 nm (Microspheres-Nanospheres, 790122-010).
2.2 Additional Components for 3D PALM
Three-Dimensional PALM
236
5. Data analysis computer (Xi Computer Corp, 2P64 workstation with 500 GB hard drive, 8 GB RAM, 2× AMD Opteron pro-cessors 2 GHz, 8 cores).
6. Network-attached storage device (Amazon.com, Netgear ReadyNAS + 2 TB internal hard drives).
7. Cylindrical lenses for astigmatic localization (Thorlabs, f = 100 mm, LJ1567RM-A; f = 150 mm, LJ1629RM-A; f = 300 mm, LJ1558RM-A).
8. Optomechanics for inserting cylindrical lenses between EM-CCD and microscope left-side port: (a) Filter wheel (Thorlabs, CFW6). (b) Adapter with external SM1 threads and internal C-mount
threads (Thorlabs, SM1A10). (c) Adapters with (a) external C-mount threads and internal
SM1 threads and (b) external SM1 threads (Thorlabs, SM1A9 and SM1T2).
(d) 1″ travel translation stage (Thorlabs, DT25). (e) Support for 1.2× expander (Thorlabs, RS2P, extra spacers
as needed).
1. Laser for two-photon activation (2PA), capable of producing pulses of 140 fs duration and tuned to 800 nm (Coherent, Chameleon Ultra II).
2. Power control: (a) Glan-Laser calcite polarizer (Newport, 10GL08AR.16). (b) Half-wave plate for infrared wavelengths (Newport,
10RP52-2). (c) Motorized rotation mount (Thorlabs, PRM1Z8E). (d) Beam dump (Newport, PL15).
3. Temporal focusing optics: (a) Periscope system for directing beam onto scanning mirror
from above (Thorlabs, KCB1, CP02T, ER4, ER6, PH3, TR2).
(b) Galvonometric scanner (Cambridge Technology, 6215HB). (c) Function generator for driving galvo (Stanford Research
Instruments, DS340). (d) Focusing cylindrical lens (Thorlabs, LJ1653L2-B). (e) Cylindrical beam expander (Thorlabs, f = 30 mm,
LJ1212L2-B, and f = 200 mm, LJ1653L1-B). (f) Low-dispersion silver mirrors, 1″ and 2″ diameter (Femto-
lasers, OA022 and OA248).
2.3 Components for Two-Photon Activation of PA-FPs
Kelsey Temprine et al.
237
(g) 830-groove-per-millimeter gold-coated refl ective diffraction grating with 21.4° blaze angle (Newport, 53107BK02-035R).
(h) Magnetic mount for diffraction grating (Thorlabs, KB3X3). (i) Lens pairs for demagnifying grating onto sample
(Thorlabs, f = 300 mm AC508-300-B-ML and f = 100 mm AC254-100- B-ML) and (Thorlabs, f = 500 mm AC508-500-B-ML and Olympus, UPLSAPO60XW f = 3 mm objective).
4. Coupling optics for combining 800 nm light with visible wave-lengths used in single-photon excitation: (a) Removable mirror mounted on a magnetic base (Thorlabs,
BB1-E02 and KB1X1). (b) Visible achromatic lens, f = 400 mm (Edmund Optics,
NT49-369-INK). (c) External dichroic mirror for combining visible and 800 nm
illumination prior to microscope body (Semrock, Di01-R561-25x36).
(d) Periscope unit for raising 800 nm illumination and direct-ing it into the microscope frame (Thorlabs, KS2, RS4P, RS6, RS1, RM1C).
(e) Right-side port attachment on microscope frame for illu-mination (Olympus, IX2-RFACB2-R).
(f) Filter cube suitable for right-side port illumination (Olympus, IX2-MFB-SP-R).
(g) Custom dichroic mirror to be placed inside fi lter cube (Chroma, ZT405/488/561/IR-RPC, refl ects 405 nm, 488 nm, 561 nm, 700–1,100 nm).
(h) Shortpass emission fi lter, for blocking 800 nm pump light, to be placed inside fi lter cube (Semrock, FF01-680/SP-25).
5. Diagnostics for measuring quality of temporal focus: (a) Quantum dot solution (Q dots; Ocean NanoTech,
QSO-520-0010). (b) Ethylene vinyl acetate copolymer resin (DuPont, Elvax
410). (c) Toluene (Sigma, 244511). (d) Glass vial (Fisher Scientifi c, 03-339-22C). (e) Hot plate (Corning, PC-220). (f) Spin coater (Laurell Technologies Corporation, WS-650S-
8NPP-LITE).
Three-Dimensional PALM
238
3 Methods
As instrumentation for 2D PALM has been covered extensively elsewhere [ 15 ], here we review major concepts only. Much of the following description also applies to 3D PALM; exceptions for TIRF are noted when appropriate. Data acquisition and processing are covered in Subheading 3.2 .
PALM instrumentation should be housed in a clean, dust-free environment to reduce sample contamination and avoid the stray light and scattering caused by dust in the optical path. We place sticky pads outside the room housing the PALM to help reduce dirt brought in by users. While the undesirable effects of thermal drift may be compensated for by post-processing (Subheading 3.2.4 ), it is advantageous to control the temperature as much as possible to mitigate this problem. We periodically monitor the temperature using a USB device in order to ensure temperature is within speci-fi cation. To reduce mechanical vibrations, we bolt all optics on a vibrationally isolated air table ( see Note 2 ).
Although it is possible to build a PALM system entirely from “the ground up” [ 2 ], we fi nd it helpful to use a commercial microscope frame (Olympus, IX-81) as the base because the eyepieces, bright- fi eld optics, objective holder, and fi lter cube turret make it easy to both screen samples and examine them with additional, non-PALM imaging modes (e.g., differential interference contrast (DIC)). Another advantage of buying a commercial frame is that the TIRF illuminator add-on facilitates TIRF by enabling the user to free- space couple high-intensity lasers of the appropriate wavelength into the system, thus minimizing PALM acquisition time. Also, the microscope manufacturer can provide an automated XY stage (e.g., ASI, MS-2000), helpful in scouting for appropriate samples and in marking their positions. With the frame bolted to the optical table, the main additional components are the excitation and activation optics, the optics housed within the microscope frame itself, and the detection optics (Fig. 2a ).
As almost all PA-FPs are photoactivated with near-ultraviolet light, we use the commonly available 405 nm laser for PALM (Coherent, 405-100 CIRCULAR). Red PA-FPs (pa-mCherry1 or mEos2) are effi ciently excited with a 561 nm laser (Crystalaser, CL561-200); for green PA-FPs (Dronpa or PS-CFP2), we use a 488 nm laser (Newport, Cyan scientifi c laser). We bolt all lasers to a small optical sled (Thorlabs, PBH11106), beam expand each to a common diameter, and combine their output with dichroic mirrors (Fig. 2b ). We introduce beam-steering assemblies in front of each laser to correct for small differences in laser height or lateral position. Cage-mount optomechanics (Thorlabs, CP02T, ER4, ER6) are
3.1 2D PALM Instrumentation
3.1.1 Basic Considerations
3.1.2 System Overview
3.1.3 Excitation and Photoactivation
Kelsey Temprine et al.
239
particularly useful in mounting beam expanders or neutral density fi lters and assure colinearity of optics along the beam path. Finally, we place a computer-controlled, mechanical shutter (Thorlabs, SH05, TSC001, and TCH002) after the coaligned beams, allow-ing the user to easily turn off the illumination to the sample.
As discussed elsewhere [ 15 ], TIRF is achieved by focusing the coaligned beams at the edge of the back focal plane (BFP) of the objective lens. We focus the laser illumination on a sled with lens FL and recollimate the beam with lens CL. By positioning the entire laser system such that the back focal plane of lens CL is coin-cident with the front focal plane of the 200 mm relay lens internal to the TIRF illuminator ( see Note 3 ), the focus created by FL is imaged at the BFP. In this geometry, the image of the focus can be moved at the BFP by translating a mirror placed on translation stage TS. The “U-DP” attachment on the TIRF illuminator is use-ful for switching between arc lamp and TIRF illumination, but only the 200 mm relay lens within the illuminator is necessary; the entire fi ber-coupling assembly can be removed.
3.1.4 Achieving TIRF
Fig. 2 Basic (2D) PALM system. ( a ) Front ( top ) and side ( bottom ) views, showing major items referred to in text including microscope frame, vibration isolation table, overhead shelf, and excitation sled. ( b ) Expanded sche-matic ( top ) and photograph ( bottom ) of excitation/activation lasers, associated optics, and optomechanics. Symbol key: DC dichroic mirror, BS beam-steering assembly, ND neutral density fi lter, BE beam expander, SH shutter, TS translation stage, FL focusing lens, CL collimating lens. Lasers are labeled according to wavelength, as referred to in the text. Lenses are drawn as ellipses in the schematic, refl ective mirrors as fi lled - in rectan-gles . Note that DC1 and the 532 nm laser are not used for exciting PA-FPs and are thus not referenced in the text. Panel ( b ) is modifi ed from York 2011, supplementary Fig. 1 with permission from Nature Methods
Three-Dimensional PALM
240
Dichroics and emission fi lters suitable for green or red PA-FPs can be placed directly into commercially available fi lter cubes and placed into the fi lter turret assembly ( see Note 4 ). Most PALM samples are cells that may be cultured directly on glass coverslips. We use 25 mm round coverslips (Warner Scientifi c) and place them in a round coverslip holder (ASI, I-3033-25D). The holder and coverslip may be placed inside an insert (ASI, I-3033) that can be secured on the automated stage.
For optimal single-molecule localization, the pixel size of the detector must properly sample the point-spread function [ 16 ]. Given the pixel sizes of most cameras, this usually implies addi-tional magnifi cation besides the magnifi cation provided by the objective lens. With a 60× objective, we use the 1.6× internal mag-nifi er housed within the microscope frame and a 1.2× external lens, thus providing a total magnifi cation of 115.2 between sample and camera.
We currently employ a back-thinned EM-CCD (Andor, DU-888E-C00-#BV, see Note 5 ) with 13 μm pixels, for a fi nal imaging pixel size of 113 nm.
With relatively few additional components, the microscope described above can be converted to a 3D PALM system capable of sub-100 nm resolution in all three dimensions. Major additional requirements include a water-immersion objective better suited for 3D samples, minimization and compensation of drift during the acquisition period, cylindrical optics for converting a molecule’s axial position into a PSF shape change, and data processing soft-ware capable of processing the raw data into a list of 3D coordi-nates and rendering the coordinates into a PALM image.
For 3D samples that are immersed in aqueous buffer and that are more than a few 100 nm thick, water-immersion objectives are bet-ter suited for PALM than the oil objectives used in TIRF. This is for several reasons: (a) the refractive index of most cellular samples is between 1.33 and 1.38; thus, an objective with greater NA does not realize its full resolution potential or collection effi ciency. (b) Spherical aberration (due to refractive index mismatch between the oil immersion media and the aqueous sample) causes an oil objec-tive’s performance to degrade at increasing depths from the cover-slip. The resulting broadening of the PSF (especially axially) spreads the available signal out, decreasing the signal-to-noise ratio (SNR) per pixel in the raw images. This is especially problematic when localizing dim PA-FPs, where signal is precious and a signifi cant decrease lowers the localization precision or renders the molecule undetectable. At the macroscopic level, the “focal shift” intro-duced by the refractive index mismatch introduces undesirable axial image distortion. Such deleterious effects may be partially
3.1.5 Microscope Body and Sample Holder
3.1.6 Detection Optics and Camera
3.2 3D PALM
3.2.1 Water-Immersion Objective
Kelsey Temprine et al.
241
compensated for by adding index-matching reagents to the sample (e.g., glycerol) or by scaling the axial localizations by the ratio of refractive indices [ 17 ]. We fi nd it simpler to use a water-immersion objective (Olympus America, UPLSAPO60XW), matching the refractive index of our fi xed cellular samples ( see Note 6 ) and avoiding these effects while simultaneously maintaining a relatively high NA (1.2) ( see Note 7 ). This objective also has good infrared transmission, important for activating PA-FPs with 800 nm light (Subheading 3.3 ).
To assemble a “z stack” or collection of imaging planes that span a 3D sample, it is necessary to move the sample relative to the imag-ing plane. We use an automated XY stage with an additional z piezo upgrade that permits 100 μm axial displacement of the sam-ple (Applied Scientifi c Instrumentation, PZ-2000). This replaces the 2D stage in Subheading 3.1 .
Drift during single-molecule imaging is a serious problem and must be reduced and/or removed before the fi nal images are ren-dered. Although it is possible to track fi ducial markers (Subheading 3.2.6 ) to remove the drift after acquisition, the drift during the experiment (especially in z) is sometimes so bad that real-time drift cancellation is needed. “Focus-lock” systems [ 17 ] that use a refl ection (e.g., from an infrared laser beam) from the coverslip/buffer interface to feedback directly to the piezoelectric stage or objective are likely the best approach, as drift is measured at or near the sample plane. Most major microscope manufacturers and several smaller companies provide these drift-canceling sys-tems as “add-ons” to commercially available microscope frames. Unfortunately, the magnitude of the refl ection feedback signal diminishes greatly when refractive index mismatches are minimized (as happens when using a water-immersion objective), so we opt instead for a system that corrects objective drift only (Mad City Labs, C-Focus). This feedback system consists of a piezoelectric collar that houses the objective and contains a glass scale grating and a sensing arm containing a laser and detector. If the reading from the grating changes in response to axial drift, the piezoelectric collar shifts the objective to compensate.
One drawback of the C-Focus is that it requires both the objective and the sample stage to be raised to accommodate the piezoelectric collar. Although the manufacturer includes spacers for this purpose, we fi nd that the additional circumference of the piezoelectric collar precludes easy use of other objectives in the Olympus fi lter turret. Rotating the focus-lock housing away from the imaging position in the turret causes it to smash into the turret base. Other objectives may be screwed into the piezoelectric collar to inspect the sample at other magnifi cations, but this process is somewhat tedious. Nevertheless, we fi nd that we can reduce axial drift to <200 nm over a 15-min period (Fig. 3 ) using this system, long enough for PALM imaging a single plane in a z stack.
3.2.2 Piezoelectric z Stage and Axial Drift Reduction
Three-Dimensional PALM
242
A variety of methods can be used to reshape the PSF of a subdiffractive particle in order to estimate both axial and lateral position. Such techniques include biplane imaging (introducing defocus into the emission path and collecting two images located at offset z planes, thus providing extra shape information that aids in axial localiza-tion) [ 18 ], generation of a helical PSF (converting axial position into angular position of a “dumbbell” PSF, usually by placing a spatial light modulator or phase mask in the emission path) [ 19 ], and astigmatic imaging (using a cylindrical lens to force rays in vertical and horizontal directions to focus at different axial planes, thus generating a PSF whose ellipticity varies with axial position) [ 20 ].
Regardless of the specifi c implementation, a feature common to all of these methods is that the detected signal in each PSF image is spread out over more camera pixels than in the ideal 2D Airy function. This fact has two consequences: (a) the signal-to-noise per pixel is reduced, reducing localization precision relative to 2D PALM, particularly for dim PA-FPs; and (b) extra care must be taken with the photoactivation laser, to ensure that too many sin-gle molecules are not simultaneously activated as their images are larger and thus more likely to overlap.
For dim PA-FPs, the fi rst consequence suggests that maxi-mizing detection effi ciency is critical in achieving acceptable SNR and 3D localization precision for PA-FPs. As cylindrical or astig-matic imaging introduces only one additional optical element, minimizing light loss in the emission path, and because it is easy to implement, we chose to adopt this method for 3D localization.
3.2.3 Astigmatic Imaging for 3D Subdiffractive Localization
Fig. 3 Performance of commercial focus-lock system in reducing axial drift. A 100 nm gold fi ducial marker stuck on a glass coverslip was localized repeatedly over a 15-min period. Blue curve indicates native axial drift; red curve demonstrates signifi cant reduction in axial drift when the Mad City Labs C-Focus module is used
Kelsey Temprine et al.
243
When deciding on the placement of the cylindrical lens in the emission path, we considered fi rst placing it beneath the fi lter cube, in the accessible space where the DIC analyzer usually sits. Unfortunately, we found cylindrical lenses with focal length <1 m (conveniently available from Thorlabs) placed in this location introduced far too much astigmatism in the PSF, thus spreading out the available signal too much and rendering such PSFs unus-able for single- molecule localization ( see Note 8 ). Instead, we place the cylindrical lens directly after the 1.2× expander external to the microscope and immediately before the EM-CCD (Fig. 4 ). We fi nd it useful to install the cylindrical lens in a fi lter holder, permitting us to switch between cylindrical lenses of varying focal length and allowing us the option of operating the microscope with minimal or no astigmatic distortion (to maintain 2D PALM/TIRF operation). Smaller focal length cylindrical lenses introduce greater axial variation of the PSF, but more image distortion (Fig. 5 ). The correct pixel sizes to restore the image to the cor-rect aspect ratio can be calculated by measuring the apparent shift of gold fi ducial markers ( see Note 9 ).
In order to maintain the proper distance between sample and camera, given the extra path introduced by the fi lter wheel and the adapters, we found it necessary to shave the fi lter wheel down to the proper thickness. This thickness is best determined empirically, by adjusting the camera position until the image of a thin fl uores-cent layer is in focus ( see Note 10 ). In our system, the total dis-tance between the external surface of the camera and the left edge of the expander is 32 mm, but this distance may vary depending on the exact optics used. Fine adjustments of the axial distance between the fi lter wheel assembly and left microscope side port may be made by mounting the entire assembly on a translation stage (Fig. 4 ). In this case, we place a mechanical post beneath the fi nal resting position of the 1.2× expander unit, for mechanical stability ( see Note 11 ).
When using the water-immersion objective, we found it very sen-sitive to tilt of the coverslip (Fig. 6 ). This feature has been previ-ously documented [ 21 ] and usually introduces aberrations into the PSF that make it hard to model theoretically. Naively fi tting a 2D elliptical Gaussian function to the molecular images in Fig. 6 can result in a localization error of hundreds of nanometers. We fi nd it diffi cult to completely remove this aberration by adjusting the tilt of our coverslip holder, and every time we perform an experiment with a different coverslip, the observed PSFs differ slightly (presumably in response to different degrees of sample tilt). Furthermore, most localization code that relies on experi-mentally fi tting a model function to a PSF also attempts to account for background levels and noise. Experimental noise is diffi cult to model accurately and background subtraction is complicated in
3.2.4 Cross-Correlation Code for 3D Subdiffractive Localization
Three-Dimensional PALM
244
thick, 3D samples where out-of-focus fl uorescence varies spatially and temporally during acquisition.
For these reasons, rather than using a model PSF that imper-fectly refl ects reality, we prefer to directly measure the PSF with a 100 nm gold fi ducial and perform a cross-correlation between the
Fig. 4 Optomechanics for coupling cylindrical lenses to emission path of micro-scope. ( a ) Photograph and ( b ) accompanying schematic. Adapter 1 has external SM1 threads and internal C-mount threads and connects the 1.2× expander to the fi lter wheel that holds the cylindrical lenses. Adapter 2 is made up of two parts that mate together, one that has external C-mount threads and internal SM1 threads (for connecting the camera) and a second with external SM1 threads (for connecting the fi lter wheel). We fi nd it useful to position a small sup-port beneath the 1.2× expander to provide mechanical stability. A translation stage is used to position the imaging plane in an optimal position with respect to the excitation (see discussion in Subheadings 3.2.3 and 3.3.5 )
Kelsey Temprine et al.
245
experimentally measured PSF and the raw single-molecule images to localize each molecule’s xyz position. This functionality is accomplished using our open-source software program, “palm3d.py,” which: (a) crops and smoothes the raw measured PSF data into a calibration stack, (b) identifi es candidate molecules from the raw PALM data, (c) localizes the candidates via cross-correlation with the calibration stack, (d) links localizations into molecules and then re-localizes linked localizations (this step is optional), and (e) bins localizations into a 3D histogram. More information about “palm3d.py” and a step-by-step user tutorial can be found at: http://code.google.com/p/palm3d/ . Other useful programs ref-erenced in the text and notes may be obtained at http://code.google.com/p/palm3d/source/browse/#hg%2Ftest .
The majority of our hardware is controlled manually during data acquisition. However, our drift correction method requires fre-quent adjustment of the piezoelectric z stage, so we automate this step. We initially used MicroManager to simultaneously control our camera and sample positioning stage but found that MicroManager
3.2.5 Control of Data Acquisition
Fig. 5 Effect of cylindrical lenses with different focal length on imaging fi eld and PSF. The same fi eld ( z = 1 μm above the coverslip surface) was imaged conventionally and with 300, 150, and 100 mm focal length cylindri-cal lenses. Selected slices (−2, −1, 0, 1, and 2 μm from the coverslip surface) from PSFs measured on the same gold bead are also displayed next to each imaging condition. Shorter focal lengths introduce greater distortion of the imaging fi eld (particularly in horizontal direction) but introduce more astigmatic character in the PSF. Scale bar in top left image : 5 μm
Three-Dimensional PALM
246
could not operate the Andor EM-CCD at top speed without crashing occasionally. Since whole-cell PALM can take hours, crashes are unacceptable, and we now use the Andor software (Andor SOLIS for Imaging) that comes with the EM-CCD. Andor’s software does not crash during acquisition but presents several serious annoy-ances. First, the Andor software is “deaf” and cannot be controlled by an external program. Next, it saves in a proprietary, undocu-mented “.sif” format, which ineffi ciently uses 32-bit fl oating-point numbers, even though the camera outputs unsigned 16-bit data. Andor’s software can convert the “.sif” format to other formats for processing and viewing, but the batch conversion is buggy and fre-quently crashes. The Andor software has extremely limited scripting capability and cannot write to a USB serial port, so it cannot directly control our axial positioner. However, the Andor scripting language can issue DOS commands, which is its saving grace. We use this functionality to launch scripts written in Python, which lets us over-come the Andor software’s annoyances. During acquisition, one such Python script writes data to the USB serial port to move our piezoelectric z stage (“piezo.py”). After each acquisition, Andor’s
Fig. 6 Images of 100 nm gold fi ducial markers at indicated z position, taken on different coverslips with different tilt and thickness. Arrows indicate features that cannot be easily accounted for with a theoretical PSF, such as “wings” in horizontal or vertical directions or an asymmetric intensity distribution in the center of the PSF. Reproduced from York 2011, supplementary material with permission from Nature Methods
Kelsey Temprine et al.
247
software launches another Python script to convert Andor’s “.sif” fi les to raw binary ( see Note 12 , “process_palm.py”). Finally, after acquiring calibration images, the Andor software launches a Python program to process the calibration images into a smoothed, cropped estimate of the microscope’s point-spread function. We save image data directly to our network-attached storage device (Amazon.com, Netgear ReadyNAS), so the operator can process each dataset as soon as it is acquired without interrupting data acquisition.
1. Alignment : It is a good idea to check the alignment of activa-tion and excitation lasers before data acquisition, especially if the temperature fl uctuates. We use a quantum dot sample ( see Note 10 ) for alignment, as it fl uoresces over a broad excitation wavelength range. After alignment, we set a region of interest (ROI) using camera software. We take care to cover only the illuminated area covered by coaligned activa-tion and excitation beams (we typically use a ~40 μm × 40 μm area, 350 by 350 pixels).
2. Fiducials : We minimize axial drift during our potentially long acquisition periods by using a commercial focus-lock system (Subheading 3.2.2 , Mad City Labs, C-Focus). To correct for residual drift, we add 100 nm gold particles (Microspheres- Nanospheres, 790122-010) as fi ducials (bright particles that do not bleach) to the sample and track their xyz displacement over data acquisition.
We typically add ~0.2–1 mL of 1× fi ducials to our samples and then wait 15–20 min for them to settle. This process should be monitored on the microscope as we have found vari-ability in the “stickiness” of different batches of fi ducials and in the same batch over time, which could affect the necessary time or volume to be added. Once a typical fi eld of view has about 1–3 fi ducials, the sample is washed with 1× PBS to remove fi ducials that have not settled.
3. Calibration : Calibration is the fi nal step that must be performed before data acquisition commences. During calibration, a z stack (PSF) is taken of a single fi ducial, which is later used dur-ing data processing to localize each PA-FP. Calibration should be performed for each sample, as the PSF changes slightly each time a new sample is loaded onto the microscope. We take the z stack from −2 μm to 2 μm (0 μm is the “best focus” position) in 50 nm steps with ten images taken at each z position (later averaging these frames to increase SNR), and repeat this pro-cess once to check if thermal drift occurred (a sample calibra-tion program, “calibration.pgm,” is available online).
After the z stack is taken, the user selects and crops the fi du-cial out from the entire fi eld of view. This is a crucial step, and the cropped area should be big enough to include the entire
3.2.6 Sample Data Acquisition Procedure
Three-Dimensional PALM
248
fi ducial with the fi ducial centered in the cropped area. We fi nd it useful to compute the axial cross-correlation function of the fi ducial and display it for the user (“process_palm.py”). Large positive areas in the cross-correlation plot indicate the shape change of the fi ducial does not depend strongly on z. This is a problem because the calibration data is used during data pro-cessing to determine the z coordinate of each molecule, and if many different z positions look similar, it is diffi cult to perform a precise z localization (Fig. 7 ).
4. Data acquisition : Once an acceptable calibration has been taken, the user may start acquiring data. An ideal imaging fi eld includes (a) 1–3 fi ducials ( see Note 13 ), (b) photoactivatable molecules (this can be quickly checked by turning the activa-tion laser on and off and observing the effect on the level of blinking molecules), and (c) low background (i.e., there should only be blinking where your sample is and not on the coverslip itself). After fi nding a good fi eld of view, the user should focus on the fi ducial (if there are multiple fi ducials in the fi eld of view, select the best one as the primary fi ducial and focus on it each time) and turn on the focus-lock system. The “track_fi d.pgm” program (available online) can then be used for data acquisition.
With thick samples, often the fi ducial and the imaging plane are at different z positions, resulting in a dim and out-of-focus fi ducial at the desired z position. The “track_fi d.pgm” program addresses this potential issue via jumping between the imaging plane and plane of best fi ducial focus. Two hundred images are taken at the desired plane (slice images), then 20 images are taken at the z position of the fi ducial (tracking images), and this process is repeated a total of 50 times. The program treats the fi ducial’s z position as “ z = 0” and allows the user to determine the z position to image at by modifying the “slicePosition = 0 // ×100 nm” line in the code. During each acquisition round, a total of 11,000 frames are taken, divided amongst slice and tracking images. After completing one acquisition round, the entire process can be repeated at either the same plane or a different plane until all the photoac-tivable molecules have been bleached. Imaging the same plane multiple times can help add additional detail to that plane, while imaging many different planes (especially when using the cylindrical lens) is advantageous for 3D imaging.
Close to the coverslip surface (within a few μm), the methods in Subheading 3.2 enable 3D super-resolution imaging with <50 nm lateral and <100 nm axial resolution, if bright and high-contrast photoactivable proteins are used and a suffi ciently high density of localizations are extracted during PALM. Further from the coverslip and for thicker samples, astigmatic imaging becomes increasingly
3.3 Two-Photon Photoactivation of PA-FPs for 3D PALM
Kelsey Temprine et al.
249
Fig. 7 Example cross-correlation plots. Comparing the maximum cross- correlation between each slice of the calibration stack (Subheading 3.2.6 ) is use-ful in judging the suitability of the imaging conditions for 3D PALM. With no cylindrical lens ( top ), most axial slices of the fi ducial marker resemble each other, resulting in large areas of high cross-correlation. This results in a PSF that will confuse the cross-correlation localization algorithm and cause poor axial local-ization. With an f = 150 mm cylindrical lens ( bottom ), the usable axial localization range is extended as there are fewer regions of high cross-correlation
Three-Dimensional PALM
250
diffi cult for PA-FPs. The main reason for this is that the wide-fi eld illumination employed in Subheading 3.2 activates and excites out-of-focus molecules, increasing background and overwhelming the fl uorescence from dim PA-FPs. If an out-of-focus molecule is not detected and localized, it is wasted, decreasing the effective label density and reducing image resolution.
Methods that confi ne photoactivation to the focal plane, such as two-photon microscopy [ 22 ], alleviate such out-of-focus back-ground. As PALM images the entire fi eld at once, the ideal imple-mentation of two-photon activation (2PA) would create a plane of activation at the focal plane. Temporal focusing microscopy [ 23 ] achieves such a planar illumination geometry by scanning an intense line focus across the sample on a picosecond timescale. We originally implemented 2PA with temporal focusing in conjunc-tion with the green PA-FP Dronpa, confi ning activation to ~±1 μm of focus and enabling PALM imaging with lateral localization pre-cisions of ~50 nm in single planes throughout whole cells. More recently, we implemented an improved, line-scanning version of temporal focusing [ 24 ] that further confi nes activation to ~±0.5 μm of focus [ 14 ]. When combined with the better PA-FP pa-mCherry1, this implementation enables 3D imaging of whole fi xed cells at depths up to 8 μm (Fig. 8 and [ 14 ]). We describe how to add such functionality to the microscope described in Subheading 3.2 here.
A multiphoton laser system suitable for line-scanning temporal focus, 2PA of PA-FPs, must satisfy the following requirements: (a) suffi cient average power to drive photoactivation, ideally several watts at 800 nm for maximal photoactivation of pa-mCherry1; and (b) ability to generate pulses of 100–200 fs duration, providing enough dispersion for temporal focusing but not so much dispersion that the pulse broadens unacceptably before reaching the sample. We use the Chameleon Ultra II (Coherent) laser as it satisfi es both criteria and is additionally a turnkey system that is easy to use. We note that this laser system is quite expensive (~$150,000). Cheaper alternatives are available that provide simi-lar pulses and repetition rates but are not tunable. Tunability is not essential, as our experience with Dronpa and pa-mCherry1 sug-gests that wavelengths near 800 nm are suitable for 2PA of these PA-FPs. The optical setup (Fig. 9a ) covered in the following description is optimized for the Ultra II; general considerations that may help in the selection of other laser sources are covered in [ 14 ] Supplementary Note 1.
As the average power delivered by the laser system may be as high as 4 W, we use polarization optics to control power instead of neutral density fi lters. As the light from the laser is polarized, we place a half-wave plate (Newport, 10RP52-2) in a motorized rota-tion mount (Thorlabs, PRM1Z8E) and follow it with a polarizer
3.3.1 Femtosecond Laser Choice
3.3.2 Power Control
Kelsey Temprine et al.
251
(Newport, 10GL08AR.16) and beam dump (PL15) (Fig. 9b ). During the experiment, we control the degree of photoactivation by controlling the rotation of the half-wave plate from our acqui-sition computer.
Line-scanning temporal focusing offers the sectioning quality of point-scanning two-photon microscopy but the increased scan speed of line-scanning two-photon microscopy by combining spa-tial focusing and temporal focusing in orthogonal directions. The temporal focus scans an intense point from left to right on a pico-second timescale (by illuminating the diffraction grating with a line focus), and a galvanometric mirror scans this illuminated line up- down on a millisecond timescale to cover the full imaging fi eld. The task for the user thus involves focusing the 2P excitation into
3.3.3 Optical System for Achieving Line- Scanning Temporal Focus
Fig. 8 Example of 3D PALM. Selected slices from a volumetric 3D PALM dataset of pa-mCherry1 lamin fusion proteins at indicated position above the coverslip surface highlighting two nuclei in fi xed Cos7 cells. The sample was photoactivated with 2P line-scanning temporal focus as described in Subheading 3.3 , resulting in a high volumetric localization density. XYZ pixel sizes are 60 × 60 × 60 nm; Gaussian smoothing of 1 pixel was applied to the data. Sections of the nuclear membrane are resolved to better than 100 nm
Three-Dimensional PALM
252
a line focus onto the grating, and imaging the surface of the grat-ing onto the sample plane. After directing the beam onto a galva-nometric scanner (Cambridge Technology, 6215HB), we place a 200 mm focal length cylindrical lens (Thorlabs, LJ1653L2-B) 200 mm from the galvo and focus the excitation into a line shape on a diffraction grating (Newport, 53107BK02-035R) placed 200 mm from the cylindrical lens. The galvo is oriented so that the scan direction is parallel to the grooves on the grating; this neces-sitates a mounting scheme that directs the beam onto the galvo from above. We achieve this by periscoping the beam up and directing it downwards onto the galvo (Fig. 9c ). We direct the beam onto the grating face at 41.6° incidence angle (Fig. 9d ), cho-sen so that the fi rst diffracted order at 800 nm emerges normal to the grating face. In the intermediate space between galvo, cylindri-cal lens, and grating, we place a cylindrical lens beam expander, thus expanding the beam 6.7× in the direction perpendicular to the grating grooves (this cylindrical lens expander is thus oriented so that it expands the beam in a direction perpendicular to the focusing cylindrical lens).
Fig. 9 Optical setup for 2PA of PA-FPs. ( a ) Optical schematic, including 1P optical sled (Fig. 1b ), 2P optics, and coupling optics. Symbol key is as in Fig. 1b , with the following additions: HWP half-wave plate, POL polarizer, PER periscope assembly, GAL galvanometric scan mirror, CFL cylindrical focusing lens, CBE cylindrical beam expander, GR grating, RM removable mirror. Dashed boxes indicate setup components further detailed in pho-tographs ( b )–( e ). ( b ) 2P power control. ( c ) Periscope assembly that directs beam onto scanning mirror. ( d ) Cylindrical optics and diffraction grating. ( e ) Periscope and coupling optics that direct 2P and 1P illumination into laser port B. In ( b )–( e ), arrows indicate direction of illumination through the setup. Panel ( a ) is modifi ed from York 2011, supplementary Fig. 1 with permission from Nature Methods
Kelsey Temprine et al.
253
To achieve near-diffraction-limited sectioning, it is important to demagnify the image of the grating enough so that the back focal plane is overfi lled [ 14 ]. We achieve this condition by using a demagnifi cation of 500× between grating and sample. Given the 3 mm focal length of our objective lens, a single f = 1,500 mm lens after the grating would accomplish this job but would require a total of ~2 × (1,500 + 3) ≈ 3 m path length on our optical table. A more space-effi cient solution is to break up the demagnifi cation with two telescopes: a 3:1 demagnifi cation followed by a 167× demagnifi cation (Fig. 9a ).
We fi nd it easiest to couple 2P illumination into the optical system by using the Olympus laser port B addition to the fi lter cube tur-ret. Using this accessory requires: (a) periscoping the 2P illumina-tion beam to the height of laser port B, (b) combination with 1P excitation light from the excitation sled (Fig. 9e ), and (c) appropri-ate optics housed within the microscope body.
We periscope the scanned temporal focus activation beam with the aid of 2″ mirrors (Femtolasers, OA022 and OA248) attached to a pillar post (Thorlabs, RS4P, RS6, RS1) with appropriate opto-mechanics (Thorlabs, KS2, RM1C). To couple the visible laser output from the excitation sled into laser port B, while maintaining the possibility of TIRF excitation, we place a removable mirror in front of the TIRF port. With the removable mirror in place, visible light is refl ected into laser port B with a dichroic mirror (Semrock, Di01-R561-25x36) designed to refl ect wavelengths <561 nm but transmit the 800 nm illumination emerging from the periscope. We place a 400 mm lens (Edmund Optics, NT49-369-INK) in the visible path before the dichroic, thus focusing the visible light to the objective back focal plane and enabling wide-fi eld excitation. Finally, we place a custom dichroic mirror (Chroma, ZT405/488/561/IR-RPC) that refl ects 405, 488, 561, and 700–1,100 nm illumination while transmitting the intermediate wavelengths into a side-mounted fi lter cube (Olympus, IX2-MFB-SP-R) placed in the fi lter turret. We also place a shortpass emission fi lter (Semrock, FF01-680/SP-25) into this cube, thus blocking transmitted infra-red light from reaching the camera. As this fi lter does not leave room for the emission fi lter typically used for PA-mCherry1 (Semrock, FF01-617/73-25), we instead place it in an empty DIC fi lter holder and place it beneath the fi lter turret.
If the temporal focus optics are correctly aligned, it should be pos-sible to obtain near-diffraction-limited two-photon illumination, or a thin sheet of illumination ~1 μm in thickness. We measure the performance of the temporal focusing by scanning a thin quantum dot layer ( see Note 10 ) axially through the fi xed illumination path. An example of optimal performance is shown in Fig. 10 . An appar-ent thickness >1.5 μm (using the 60×, 1.2 NA water-immersion
3.3.4 Injection of 2P Illumination into Microscope and Combination with Visible Illumination
3.3.5 Alignment of 2P Excitation Optics and Matching Plane of Temporal Focus to the Imaging Plane
Three-Dimensional PALM
254
objective) indicates either a problem with optical alignment or a thicker than ideal test sample.
Achieving optimal temporal focus sectioning is contingent on proper alignment, yet how to achieve optimal alignment is not obvious. We fi nd it useful to work backwards from the microscope, by following these steps:
1. Starting with the objective and dichroic in the Olympus micro-scope, we set a reference beam. This beam must be well colli-mated, since it will be used to set the axial position of several lenses in the system. We collimate our beam by adjusting a beam expansion telescope and measure its collimation by defl ecting it off the optical table with a mirror. If the beam is made to travel several meters before hitting a wall, we can tell accurately if it’s expanding or contracting. The beam diameter should be reasonably large (at least a centimeter), ideally with an adjustable iris to allow easy changing of the beam diameter. The back refl ection of the reference beam from the objective should be very well aligned with the reference beam, indicat-ing normal incidence on the back focal plane of the objective. Note that any back refl ection which makes its way back into the femtosecond laser source will cause instability and prevent mode- locking, so we align just poorly enough to prevent this.
Fig. 10 Temporal focus optical sectioning. An 800 nm thick quantum dot layer was scanned in 100 nm axial steps through a line-scanning temporal focus illumination and the fl uorescence recorded at each step. The normalized, integrated fl uorescence at each frame is shown. Note the greatly reduced axial FWHM of 2P, line- scanning temporal focus illumination (~1.3 μm) relative to wide-fi eld, 488 nm illumination
Kelsey Temprine et al.
255
With the objective removed, the reference beam should propagate in the same direction the objective points. We fi nd it helpful to use a plumb line to mark a spot on the ceiling directly above the objective (assuming the objective is mounted vertically) and ensure that the center of the beam falls on this spot with/without the objective lens.
2. With a reference beam set, the next lens (working backwards from the objective, this is the f = 500 mm lens in Fig. 9a ) can now be aligned. The axial position of the lens is correct when the combination of the lens and the objective produces a nearly collimated beam, i.e., the diameter of the beam on the ceiling mark is minimized (position the objective’s z position near its operating position while making this alignment). The trans-verse position of the lens is correct when it does not defl ect the reference beam; the reference beam should hit the same spot on the ceiling that it does with this lens and the objective removed. The tilt of this lens is correct when the back refl ec-tion from this lens points back along the reference beam. The curved surface of the lens is normally placed towards a colli-mated beam. Note that with temporally focused beams, it can be confusing whether or not a beam is “collimated” or con-verging/diverging. In this case, note that the beam in the sample is “temporally focused,” which at any given instant is equivalent to a point focus. Therefore, we can regard the beam between the f = 500 mm lens and the objective as collimated and place the curved side of the f = 500 mm lens facing this direction. With this lens in place, we can remove the objective and check if the reference beam still hits the reference mark on the ceiling. If it doesn’t, this means the beam was not well centered on the objective. Reset the reference beam, and repeat the alignment of this fi rst lens.
3. Two more lenses ( f = 100 mm and f = 300 mm in Fig. 9a ) are now added by the same procedure: back refl ection, non- defl ection, and preserving collimation ( see Note 14 ).
4. Next we align our diffraction grating. Since some alignment processes are easier without the grating, we use a magnetic mount (Thorlabs, KB3X3) that allows us to switch the grating for a fl at mirror. Since different wavelengths of light diffract off the grating at different angles, we deliver light to the grating with a mirror on a linear translation stage (Fig. 9a, d ) very close to the grating to allow adjustment when we tune our femtosecond laser to different wavelengths. In practice, how-ever, we almost always use this laser at 800 nm. To position the grating axially, we fi nd it is suffi cient to measure the distance from the grating to the next lens ( f = 300 mm in Fig. 9a ) with a ruler, placing it one focal length away. To position the grat-ing transversely, we center it on the reference beam before
Three-Dimensional PALM
256
redirecting the reference beam to the mirror on the linear translation stage. We then set a new reference beam, roughly parallel to the old one but displaced to the side, bouncing off the mirror on the linear translation stage towards the grating. We tune the input angle of the grating by moving the linear translation stage and rotating the mirror mounted on it. We choose the input angle so that the fi rst-order diffraction of the grating hits the next lens ( f = 300 mm in Fig. 9a ) directly on center. We choose the tilt angle of the grating such that the output beam is normal to the grating face ( see Note 15 ).
5. With the grating positioned, we next place a cylindrical lens one focal length away from the grating, centered on the refer-ence beam (CFL in Fig. 9a ). This lens produces a horizontal line focus on the grating. Since the beam does not impact the grating at normal incidence, the line is only truly in focus at one point on the grating. However, the axial demagnifi cation of the lens system between the grating and the sample is quite large (250,000×), so the effects of this defocus are negligible in the sample. As with previous lens alignment, the vertical position can be set by non-defl ection, and the tilt can be set by back refl ection. The horizontal position of this lens is unim-portant, as long as it doesn’t clip the beam.
6. Next we place a cylindrical beam expander in the reference beam (6.7× CBE in Fig. 9a ). The axial position of this expander is not critical, but the relative axial positions of the two cylin-drical lenses in the expander are quite important. As before, non- defl ection, back refl ection, and collimation tests are used to insert these lenses in the reference beam. We make sure to position this expander close enough to the diffraction grating that enough room is left for the galvanometric mirror.
7. Positioning the galvanometric mirror may be the most diffi cult step in the process. We measure one focal length from the cylindrical focusing lens (in the direction away from the micro-scope, towards the reference beam) and place a reference object at this point, centered on the reference beam. We typi-cally use a small- diameter Allen key attached to the end of an optical post. The location of this reference object is used to determine where the galvanometric mirror should be posi-tioned. Once the reference object is set, we place the galvano-metric mirror with its periscope assembly in its fi nal position and send a new reference beam through the optical system. Even with a good reference object, fi nely aligning the galvano-metric mirror is challenging, since the entire assembly must be moved to rotate the galvanometric mirror about a vertical axis, but this motion disrupts the reference beam passing through the galvo. We iterate, adjusting the rotation and resetting the reference beam several times.
Kelsey Temprine et al.
257
8. With two-photon illumination optics aligned, we check the depth at which the two-photon beam reaches maximum inten-sity. We use a thin layer of quantum dots for this measurement ( see Note 10 ) and illuminate the quantum dot sample with 800 nm light while viewing its fl uorescence on the EM-CCD. The layer has bumps and features that inform us when the camera emission path is focused on the quantum dots, and the fl uorescent brightness of this informs us when the illumination path is focused on the quantum dots. Hopefully, these two depths match closely. If not, we move the camera axially using the attached linear translation stage (Thorlabs, DT25) to align the camera’s focus with the illumi-nation’s focus.
For 3D PALM, the best localization precision is often obtained for molecules which are moderately out of focus. To bias activation towards slightly out-of-focus molecules, we translate the camera axially a small degree. The optimal level of defocus can be esti-mated by imaging a sample of gold fi ducials, with a cylindrical lens in the imaging path. We start with the camera initially axially coaligned with the illumination then switch to a gold fi ducial sam-ple. We focus on the gold fi ducials such that they appear maximally round and then axially translate the camera until they appear slightly elliptical.
Taking data follows the same scheme as Subheading 3.2.6 , but with a few differences. During acquisition, we photoactivate the sample using 2PA illumination by scanning the galvanometric mir-ror at 500 Hz. As this speed is much higher than our acquisition rate (usually 10–13 Hz), we drive the galvo with the aid of a function generator (Stanford Research Instruments, DS340), fi nd-ing it unnecessary to synchronize scanning with camera exposure. During acquisition we modify the 2P illumination power by rotat-ing the half-wave plate placed in the automated rotation mount, as appropriate for the desired degree of photoactivation.
4 Notes
1. Localization precision (how well the center of each PSF can be determined) and the density of available photoswitchable mol-ecules are two factors that determine the resolution of PALM. In the case of negligible background, the error in the PSF center determination is inversely proportional to the square root of the number of collected photons [ 16 ]. Thus, a photoactivatable tag that emits 100-fold more photons than another would give a 10-fold increase in localization precision. Although a high number of collected photons are desirable
3.3.6 Data Acquisition
Three-Dimensional PALM
258
from a localization standpoint, in practice other factors (errors in the localization estimate, drift correction, background due other fl uorescent tags present at high density, or autofl uores-cence of the sample) limit the localization precision to ~10 nm for the majority of photoactivatable molecules.
The other key determinant of resolution is the density of molecules. According to the Nyquist-Shannon sampling theo-rem, the sampling interval (mean distance between neighbor-ing localized molecules) must be at least twice as fi ne as the desired resolution; otherwise, the feature of interest will be undersampled and unresolved. As an example, to achieve 10 nm resolution in two dimensions, molecules must be spaced 5 nm apart in each dimension, for a minimum density of 4 × 10 4 molecules/μm 2 or ~2,000 molecules in a diffraction- limited region 250 nm in diameter. Achieving a high density of localizations in practice is dependent not only on the labeling strategy (PA-FPs vs. exogenous photoswitchable dyes) but also on the contrast ratio (ratio in fl uorescence intensity between the inactive and active state of the photoswitchable molecule). If the contrast ratio is less than the density of the probe in a given diffraction-limited region, the weak intensity from the inactive molecules will overwhelm the intensity from a single activated molecule, confounding both localization and the interrogation of densely labeled samples.
2. We fi nd an overhead shelf system (e.g., Newport, ATS-10) useful for mounting laser (and other) power supplies in a space-effi cient manner near the optical table.
3. The laser sled’s axial position may be adjusted with appropriate “feet” and spacers from Thorlabs (BLP01, PF175, PS1, PS2, PS3, PS4).
4. We sometimes fi nd it useful to place an additional bandpass fi lter (or laser-line blocking fi lter) into the emission path in order to provide more rejection of the excitation light. A par-ticularly useful component for this is an extra analyzer holder (used in DIC) without the polarizer. This part is available from the manufacturer and provides a convenient space for addi-tional 25 mm diameter optics.
5. An electron-multiplying charge-coupled device (EM-CCD) camera is still the detector of choice for most single-molecule imaging studies, due to its high quantum effi ciency, moderate frame rate, and negligible read noise. We note that EM-CCDs may soon be surpassed by cheaper and faster scientifi c complementary- symmetry metal-oxide-semiconductor detec-tors, especially as their quantum effi ciency improves.
6. We typically fi x samples in a 2–4 % paraformaldehyde/1× PBS (or similar buffer) solution for 15 min, rinsing 3× in 1× PBS before imaging in 1× PBS. A complete protocol is available in ref. 15 .
Kelsey Temprine et al.
259
We note that fi xing in this way and imaging with a water-immer-sion objective obviates the need for Mowiol or other oil-based mounting schemes and minimizes perturbations to the sample.
7. Silicone oil objectives may provide a higher NA (1.3) while still minimizing refractive index mismatch, but we have not tested their suitability for PALM.
8. We note that Rocky Mountain Instruments manufactures cylindrical lenses with up to 10 m focal length.
9. In order to construct a histogram of the processed PALM data, the user must know the nanometers per x and y pixel (“nm_per_x/y_pixel” in the palm3d processing code, code.google.com/p/palm3d). These values are determined by the magnifi -cation of the microscope, the camera’s pixel size, and whether and what focal length cylindrical lens is used. An easy way to determine these quantities is by using the “xy_grid.pgm” and “xy_stage.py” programs ( http://code.google.com/p/palm3d/source/browse/#hg%2Ftest ). These programs record images of a single fi ducial marker at its initial position and after defi ned stage movements in x and y . Creating a maximum pro-jection of all the generated images in ImageJ results in a 5-by-5 grid with 3 μm spacing between the fi ducial location at each node in the grid. The number of pixels between each location in each dimension can then be determined, and dividing the 3,000 nm step size by this value will give the “nm_per_x/y_pixel” to be used for that lens. Before using “xy_grid.pgm,” we set the “kinetic series length” in our camera software to 1, so that only one image is taken at each xy position. Using a fi eld of view with only one fi ducial is best, since the presence of additional fi ducials will result in a confusing maximum projec-tion with multiple 5-by-5 grids overlapping each other.
10. A good alignment sample is essential for checking the overlap of excitation and activation lasers, as well as measuring the thickness of the 2P temporal focus activation plane. We fi nd a particularly good choice is a thin (<1 μm) layer of quantum dot nanoparticles, spin coated on a glass coverslip. The quantum dots are bright and photostable, fl uoresce over a broad spectral region, and last months if prepared properly.
The quantum dots we start with are bare, i.e., do not have a polymer shell encapsulating them (Q dots; Ocean NanoTech, QSO-520-0010). For this reason, they are hydro-phobic (shipped in toluene) and must be embedded in a poly-mer resin before attachment to a glass coverslip. We prepare a dilution of 4 % (weight by volume) solution of Elvax 410 eth-ylene-vinyl acetate copolymer resin (DuPont) in toluene by overnight heating on a hotplate. Care must be taken to ensure that the polymer does not bubble, by keeping the temperature stable at or below 40 °C.
Three-Dimensional PALM
260
We next dilute this solution to 1 % by diluting in a 1:1.5:1.5 4 % Elvax 410/quantum dots/toluene ratio. This solution is stable for ~1 month, if stored in a capped glass vial with a good seal (capping is necessary as the toluene is volatile; glass is important as it is inert and will not be dissolved by toluene). The 1 % Elvax/Q dot solution is spin coated onto a cleaned 25 mm diameter glass coverslip. We use model # WS-650S- 8NPP-LITE (Laurell Technologies Corporation) and ramp the speed from 0 to 500 r.p.m. over 7 s then maintain 500 r.p.m. for 40 s. Other spin coaters work too, and the proper settings may require repeated trials to get a thin layer repeatedly. We fi nd it useful to verify the thickness of the layer with atomic force microscopy, as <1 μm thickness is preferable for measure-ment of the thickness of the temporal focus layer. Once made, we store the samples in the dark at room temperature.
11. When taking data, we drape black fabric around the entire assem-bly, thus minimizing external light leakage into the camera.
12. The main functions of “process_palm.py” are: (a) Conversion of the “.sif” images generated by Andor into “.dat” images (a much easier to use format). Once we realized the “.sif” format appears to consist of a header, followed by numerical data stored as fl oating-point 32-bit numbers, reverse engineering consisted of learning how to read the header to determine where the numerical data would start and stop within the fi le. Different versions of the Andor software use slightly different headers, however, so use caution if attempting to reuse our conversion code. (b) Taking the user through cropping and then smoothing the calibration fi ducial before generating the cross-correlation plot ( see Subheading 3.2.6 ).
13. Sometimes trash on the coverslip can be mistaken as fi ducials; it is useful to inspect the putative fi ducial for a few minutes to verify it does not bleach before using it.
14. When there is an even number of lenses (including the objec-tive), this process works very well. When there is an odd num-ber of lenses, the output of the objective diverges rapidly and defl ection is very hard to judge. In this case, unscrew the objective during alignment, and restore it when it’s time to add the next lens.
15. Pay attention to the blaze of the grating; there are two orienta-tions of the grating which will work, but one will be substan-tially more effi cient at refl ecting the fi rst-order diffracted beam.
Acknowledgments
This research was supported by the Intramural Research Program of the NIH, including the National Institute of Biomedical Imaging and Bioengineering.
Kelsey Temprine et al.
261
References
1. Hell SW (2007) Far-fi eld optical nanoscopy. Science 316:1153–1158
2. Betzig E, Patterson GH, Sougrat R et al (2006) Imaging intracellular fl uorescent proteins at nanometer resolution. Science 313:1642–1645
3. Rust MJ, Bates M, Zhuang X (2006) Sub-diffraction- limit imaging by stochastic optical reconstruction microscopy (STORM). Nat Methods 3:793–796
4. Hess ST, Girirajan TPK, Mason MD (2006) Ultra-high resolution imaging by fl uorescence photoactivation localization microscopy. Biophys J 91:4258–4272
5. Lukyanov KA, Chudakov DM, Lukyanov S et al (2005) Photoactivatable fl uorescent pro-teins. Nat Rev Mol Cell Biol 6:885–890
6. Chen I, Ting AY (2005) Site-specifi c labeling of proteins with small molecules in live cells. Curr Opin Biotechnol 16:35–40
7. Patterson GH (2011) Highlights of the optical highlighter fl uorescent proteins. J Microsc 243:1–7
8. Chang H, Zhang M, Ji W et al (2012) A unique series of reversibly switchable fl uorescent pro-teins with benefi cial properties for various applications. Proc Natl Acad Sci U S A 109:4455–4460
9. Dempsey GT, Vaughan JC, Chen KH et al (2011) Evaluation of fl uorophores for optimal performance in localization-based super- resolution imaging. Nat Methods 8:1027–1036
10. Shroff H, Galbraith CG, Galbraith JA et al (2007) Dual-color superresolution imaging of genetically expressed probes within individual adhesion complexes. Proc Natl Acad Sci U S A 104:20308–20313
11. Axelrod D (2001) Total internal refl ection fl u-orescence microscopy in cell biology. Traffi c 2:764–774
12. Manley S, Gillette JM, Patterson GH et al (2008) High-density mapping of single- molecule trajectories with photoactivated local-ization microscopy. Nat Methods 5:155–157
13. Shroff H, Galbraith CG, Galbraith JA et al (2008) Live-cell photoactivated localization
microscopy of nanoscale adhesion dynamics. Nat Methods 5:417–423
14. York AG, Ghitani A, Vaziri A et al (2011) Confi ned activation and subdiffractive localiza-tion enables whole-cell PALM with genetically expressed probes. Nat Methods 8:327–333
15. Shroff H., White H., Betzig E. (2008) Photoactivated localization microscopy (PALM) of adhesion complexes. Curr Protoc Cell Biol 41, 4.21.21–24.21.27
16. Thompson RE, Larson DR, Webb WW (2002) Precise nanometer localization analysis for individual fl uorescent probes. Biophys J 82:2775–2783
17. Huang B, Jones SA, Brandenburg B et al (2008) Whole-cell 3D STORM reveals inter-actions between cellular structures with nanometer- scale resolution. Nat Methods 5:1047–1052
18. Juette MF, Gould TJ, Lessard MD et al (2008) Three-dimensional sub-100 nm resolution fl u-orescence microscopy of thick samples. Nat Methods 5:527–529
19. Pavani SRP, Thompson MA, Biteen JS et al (2009) Three-dimensional, single-molecule fl u-orescence imaging beyond the diffraction limit by using a double-helix point spread function. Proc Natl Acad Sci U S A 106:2995–2999
20. Huang B, Wang W, Bates M et al (2008) Three-dimensional super-resolution imaging by stochastic optical reconstruction micros-copy. Science 319:810–813
21. Arimoto R, Murray JM (2004) A common aberration with water-immersion objective lenses. J Microsc 216:49–51
22. Denk W, Strickler JH, Webb WW (1990) Two- photon laser scanning fl uorescence microscopy. Science 248:73–76
23. Oron D, Tal E, Silberberg Y (2005) Scanningless depth-resolved microscopy. Opt Express 13:1468–1476
24. Tal E, Oron D, Silberberg Y (2005) Improved depth resolution in video-rate line-scanning multiphoton microscopy using temporal focus-ing. Opt Lett 30:1686–1688
Three-Dimensional PALM
263
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_14, © Springer Science+Business Media New York 2015
Chapter 14
Direct Stochastic Optical Reconstruction Microscopy (dSTORM)
Ulrike Endesfelder and Mike Heilemann
Abstract
Single-molecule localization-based super-resolution microscopy can be performed with regular, bright, and photostable organic fluorophores. We review a concept termed direct stochastic optical reconstruction microscopy (dSTORM), which operates conventional fluorophores as photoswitches and provides an opti-cal resolution of ~20 nm. We introduce the principle of dSTORM, illustrate experimental schemes, and discuss approaches for data analysis.
Key words Super-resolution fluorescence microscopy, dSTORM, Single-molecule fluorescence imaging, Single-molecule localization, Photoswitching, Organic fluorophores, Cluster analysis
1 Introduction
Single-molecule based super-resolution techniques employ photo-switchable fluorophores, single-molecule localization, temporal separation, and image reconstruction and achieve an optical reso-lution of down to ~20 nm in the imaging plane [1–3]. Among the various techniques introduced, direct stochastic optical reconstruc-tion microscopy (dSTORM) [2, 4] operates conventional organic fluorophores as photoswitches by making use of their photophysi-cal and photochemical transitions. dSTORM imaging relies on three key points (Fig. 1): (1) the use of photoswitchable fluoro-phores, (2) stochastic activation and temporal separation of the fluorescence signal to separate single molecules in space and time, and (3) single-molecule localization with nanometer precision [2, 3]. dSTORM imaging was demonstrated with multiple colors [5–7], in combination with photoactivatable fluorescent proteins [8] or chemical tags [9], for dynamics in vitro [10] as well as in live cells [11, 12], also extended to three-dimensional by optical [13] or physical [14] sectioning of thick tissue.
In a dSTORM experiment, single fluorophores are stochastically activated and read out over time, and an image stack is recorded.
264
In each single image of the stack, the fluorescence emission pattern (or point-spread function) of single molecules is approximated with a Gaussian function and provides information on the location of a fluorophore. From a large number of single- molecule coordi-nates, an artificial image can be reconstructed (exemplary dSTORM images are shown in Fig. 2). As a result, dSTORM and similar methods provide a list of single-molecule coordinates, which is in contrast to conventional microscopy techniques which provide intensity information directly. In addition to generating super-res-olution images and resolving cellular structures, this technique allows developing analytical localization- based algorithms to extract single-molecule information.
2 Materials
dSTORM utilizes conventional organic fluorophores for super- resolution fluorescence microscopy, which are “programmed” to become photoswitches in the presence of reducing agents. The main classes of common organic fluorophores that can be photo-switched are carbocyanines [2], rhodamines, and oxazines [15].
Fig. 1 Principle of single-molecule super-resolution microscopy. (a) Photoswitchable fluorophores are transferred from a fluorescent bright state into a nonfluorescent dark state by the irradiation of light. (b) By activating only a sparse subset of fluorophores at a time, the position of single fluorophores can be determined by approximating their point-spread function. (c) A super-resolved image is reconstructed from single-molecule coordinates
Ulrike Endesfelder and Mike Heilemann
265
3 Methods
Here we describe the basic approach for measuring dSTORM images in two dimensions in a static sample. It is possible to extend these methods for imaging dynamic samples, multiple colors, and three-dimensional structures (see Notes 1–3).
Organic fluorophores are available as active esters and can be chemically conjugated to various functional groups found in bio-molecules (e.g., –NH2 or –SH). This strategy enables labeling of small receptor-binding peptides, antibodies and proteins, oligonucleotides (e.g., molecular beacons), or small specific drug molecules (e.g., the actin-binding peptide phalloidin) as well as other substrates.
3.1 Labeling Strategy
Fig. 2 dSTORM images of cellular structures. (a) Exemplary dSTORM images of microtubulin immunostained with different fluorophores. (b) Dual-color measurement of a cell stained for actin (EosFP) and mitochondria (Alexa Fluor 647). Scale bars represent 1 μm
Direct Stochastic Optical Reconstruction Microscopy
266
The most common labeling strategies for cellular structures that are compatible with dSTORM are:
1. Immunocytochemistry: two complementary antibodies are applied as sandwich system, with a primary antibody targeting a specific antigen and a secondary antibody (or F(ab)2 frag-ment) carrying a fluorophore and targeting the first antibody.
2. Chemical tags: The introduction of a chemical tag like the TMP tag [16], the SNAP tag [17], the CLIP tag [18], and the Halo tag [19] represents a simple, live cell compatible method for target-specific labeling with organic fluorophores. The pro-tein of interest is genetically fused to a marker protein, which then covalently or non-covalently binds to a fluorescent tag.
3. Click chemistry: The basic chemical reaction that underlies “click chemistry” is a Cu(I)-catalyzed cycloaddition of an azide and an alkyne to a triazole. This approach is very specific, as neither of the reactive groups is found in biological samples. Cells are treated with an azide or alkyne analogon (e.g., 5-ethynyl-2′-deoxyuridine (EdU) instead of thymidine for DNA labeling, or l-homopropargylglycine (HPG) or l- azidohomoalanine (AHA) for amino acid labeling), which is incorporated into biomolecules and subsequently labeled with a fluorophore carrying an alkyne or azide group [20, 21].
Redox-induced photoswitching is best visualized and explained with a Jablonski diagram (Fig. 3). A fluorophore is first excited from the ground state (S0) into a higher electronic state (S1, S2, …, SN). After internal conversion and vibrational relaxation to the lowest excited state (S1), spontaneous emission of a photon (fluorescence) brings the molecule back into the ground state S0. Alternatively, a non-radiative transition into S0 can occur, as well as intersystem crossing into the triplet state (T1).
In the presence of reducing agents which match the redox potential of the fluorophores, an electron transfer can occur. In the presence of thiol group containing reducing agents, many fluorophores are reduced out of the triplet state and form stable radical anions [22]. This radical state represents the nonfluores-cent OFF- state of the fluorophore, which was demonstrated to exhibit a thermal stability of many minutes to hours in aqueous solution [22]. Upon reaction with molecular oxygen or irradia-tion with UV light, the molecule returns to the ground state, and fluorescence is recovered. The radical anion of some fluorophores (e.g., ATTO 655) exhibits a high electron affinity and can be fur-ther reduced to the leuco-form of a fluorophore [23].
The rate for the transition from the fluorescent ON-state into the nonfluorescent OFF-state (koff) is controlled by the concentra-tion of the reducing thiol reagent and the irradiation intensity. The lifetime of the OFF-state and thus the rate of the transition into the
3.2 Photoswitching of Organic Fluorophores
Ulrike Endesfelder and Mike Heilemann
267
ON-state (kon) depends on the thermal stability and the concentration of molecular oxygen. In addition, the OFF-state can be depopu-lated by irradiation with low-intensity near-UV light, which is in resonance with the absorption band of the radical anion of many fluorophores.
An important experimental parameter for high-quality dSTORM super-resolution imaging is an appropriate ratio of the ON- and OFF-switching rates, r = koff/kon. The higher the fluoro-phore density, the higher must be the ratio r, to ensure that each fluorophore is recorded as a single molecule at a given time. This implies rates koff ≫ kon which guarantee sufficiently low spot densi-ties per imaging frame as the active fraction of fluorophores is low compared to the total number [24].
Localization-based super-resolution microscopy is a technique operated on microscopy setups with wide-field illumination (Fig. 4).
1. Excitation pathway: a multiline laser (e.g., argon-krypton laser) or single laser sources (e.g., diodes or Ti:sapphire lasers) are filtered by an acousto-optic tunable filter to select for the appropriate excitation wavelength. The laser beam is then focused on the back focal aperture of the objective. A movable mirror enables to switch between wide-field illumination, total internal reflection (TIR) illumination, and highly inclined and laminated optical sheet (HILO) microscopy [25].
3.3 Microscope Setup
Fig. 3 Reversible photoswitching of organic fluorophores in the presence of reducing agents. Upon irradiation, the fluorophore is excited from its singlet ground state S0 into higher electronic states. From the first excited state S1, either fluorescence emission or intersystem crossing into the triplet state T1 occurs. The long-lived triplet state can further react with molecular oxygen to recover the singlet ground state or react with reducing agents (such as thiols) to form a radical anion (F•−). The singlet ground state can be recovered by oxidiza-tion with oxygen or excitation of the radical with near-UV light. For some fluoro-phores (e.g., the oxazine fluorophore ATTO 655) were found to become fully reduced to the leuco-form (FH), which can also recover into ground state by reaction with oxygen
Direct Stochastic Optical Reconstruction Microscopy
268
2. Microscope configuration: an inverted microscope is equipped with an oil-immersion objective suitable for total internal reflection fluorescence microscopy (TIRFM). Excitation and fluorescence light are separated using a dichroic mirror.
3. Detection pathway: the fluorescence light is directed onto a single-molecule sensitive electron-multiplying charge-coupled device (EMCCD) camera. Additional lenses may be intro-duced in the detection path to adjust for an optimal pixel size (70–160 nm) on the camera chip.
Examples for some commonly used fluorophores and appro-priate filter sets are highlighted in Tables 1 and 2.
Typically, a stack of 4,000–20,000 images with a frame rate between 10 Hz and 2 kHz is recorded. The EMCCD camera should be operated in the most sensitive mode, applying sufficient cooling for
3.4 dSTORM Imaging
Fig. 4 Experimental setup for single-molecule super-resolution microscopy. An acousto-optical filter (AOTF) is used to select an excitation wavelength. The laser light is focused (e.g., by a telescope consisting of the lenses L1 and L2) onto the back focal (BF) plane of a high numerical aperture oil-immersion objective (OBJ). Excitation and emission light are separated by a dichroic mirror (DM). Fluorescence light is spectrally filtered by long- pass and/or bandpass filters (F) and projected (L3) on a camera. Additional lenses (not sketched) can be arranged in the detection path to adjust for a different pixel size
Ulrike Endesfelder and Mike Heilemann
269
reduced thermal background noise, the highest dynamical range settings, and the fastest readout option by, e.g., a frame transfer mode and optimal gain settings. To achieve an optimal signal-to-noise ratio, the frame rate should be adjusted with respect to the average time a fluorophore resides in the fluorescent ON-state, which is controlled by the applied irradiation intensity. The appro-priate recording time is dependent on the target (e.g., abundance and dimensionality), the labeling density, and the switching prop-erties of the dye. The spot density, controlled by the transition rates between fluorescent “ON”- and nonfluorescent “OFF”-states, has to be sufficiently low for robust single-molecule detection. Furthermore, all fluorophores should be detected at least once, in order to allow for a proper visualization of a structure.
Table 1 Exemplary fluorophores suitable for dSTORM
Fluorophore Absorption/emission (λabs/λem) Switching buffer
Alexa Fluor 488 491/517 100 mM MEA, pH 8
ATTO 488 501/523 100 mM MEA, pH 9
Alexa Fluor 532 532/552 100 mM MEA, pH 7.4–8
Alexa Fluor 568 572/600 100 mM MEA, pH 8–8.5
Alexa Fluor 647/Cy5 649/670 100 mM MEA, pH 7.4 (+oxygen scavenger system)
ATTO 655 663/684 10 mM MEA, pH 7.4–8
Table 2 Exemplary filter sets for single-color dSTORM measurements
Dyes Cleanup Dichroic mirror Detection filter set
ATTO 488 Z488/10, Chroma HC-Quad 410/504/582/669, Semrock
LP 488 RU RazorEdge, Semrock
Alexa Fluor 488 HC 550/88 Brightline, Semrock
ATTO 568 Z568/10, Chroma HC-Dual 560/659, Semrock LP 568 RU RazorEdge, Semrock
Alexa Fluor 568 HQ 610/75, Chroma
Cy5 Z488/568/647 RPC, Chroma
HC-Dual 560/659, Semrock LP 647 RU RazorEdge, Semrock
Alexa Fluor 647 ET 700/75, ChromaATTO 655
Direct Stochastic Optical Reconstruction Microscopy
270
Due to the separation of single fluorophores in time, it is necessary to acquire enough images in order to reconstruct a representative super-resolution image, such that a dSTORM experiment typically lasts several minutes. As a consequence, localization-based super- resolution microscopy is very sensitive to drift occurring during the acquisition time. Small movements in the nanometer range can already introduce image artifacts and impact image quality and interpretation. Drift can be corrected for by adding photostable fluorescent beads as fiducial markers to the sample. Due to the stable and long-lasting fluorescence emission of the beads, the drift can be traced, and the individual frames of image stack can be realigned. Alternatively, image correlation can be used for drift correction, but requires a minimum of structural information.
Single-molecule super-resolution methods such as dSTORM pro-vide coordinates of individual point-spread functions recorded from single fluorophores. For each frame, the subset of active fluo-rophores needs to be identified and fitted precisely and rapidly using the post-processing software.
Commonly, the point spread function (PSF) of a single fluoro-phore is approximated by a Gaussian function:
I x y A BA x x y y
Bc cx y
c
x
c
y
, , , exp( ) = --( )
--( )æ
èçç
ö
ø÷÷+
2 2 2
2
2
2
2ps s s s
with σx and σy as standard deviations, (xc, yc) the coordinates of the center, and the amplitude A and the background signal B.
Single-molecule localization software, like the freely available rapidSTORM [24, 26] and QuickPALM [27]), are based on two consecutive main tasks performed for each frame of the imaging stack:
1. Spot identification: the PSF of single fluorophores is identified, e.g., by blurring and non-maximum suppression routines.
2. Spot fitting: for each identified single fluorophore, the optimal fit is computed and validated, e.g., by least-squares fitting. Subsequently, the parameters of the fit are evaluated, and those which are, e.g., below an intensity threshold or exhibit asym-metry are discarded.
The accuracy of single-molecule localization primarily increases with the number of detected photons. It is also dependent on the pixel size, the shot noise of the camera, and the background signal originating from out-of-focus fluorescence and detector noise. Several formulas have been introduced to estimate the uncertainty of single-molecule localization [28–30]. For high signal-to-noise
3.4.1 Drift Control
3.5 Image Reconstruction
3.5.1 Localization Accuracy
Ulrike Endesfelder and Mike Heilemann
271
ratios, the contribution of background signal and pixilation is negligible, and the localization error ϵ can be estimated from
e ~
sPSF
N
with σPSF being the standard deviation of the PSF and N the num-ber of photons detected within the PSF. The localization accuracy can also be determined experimentally by measuring the standard deviation of multiple localizations of the same single fluorophore. Knowledge of the localization accuracy is necessary to understand the resolution that can be achieved (see Notes 4 and 5).
Different to other microscopic techniques, single-molecule localization- based super-resolution techniques provide no direct intensity information, but single-molecule coordinates. Various ways of reconstructing intensity-like images from single-molecule localization data have been introduced. The simplest approach is to map the single-molecule localizations in a scatter plot. This does not reflect intensities, and information is lost at higher localization densities. A more image-like representation is achieved by blurring each single-molecule localization with a Gaussian function with its width reflecting the localization accuracy. The blurred spots are summed up to an artificial image in which the intensity is propor-tional to the local density of single-molecule localizations. These images can be used for, e.g., image-based colocalization analysis. Another possibility is the binning of all single-molecule localiza-tions into a 2D histogram. This generates a pixelated intensity image in which the intensity information per pixel reflects the number of events detected in this area. Furthermore, also more complex representation schemes were introduced [31].
In contrast to microscopy techniques which provide intensity images, single-molecule localization methods provide a list of single- molecule coordinates. On the basis of reconstructed images, all kinds of data analysis can be performed as if for conventional fluorescence microscopy. In addition, coordinate-based analysis for, e.g., clustering or colocalization, can be performed.
The choice of an appropriate cluster algorithm depends on the individual data set and the specific question to be addressed. Different cluster algorithms are optimized for different tasks, such as the efficient identification of cluster numbers, shape, or separa-tion. Single-molecule localization microscopy data provides single- molecule coordinates, such that cluster algorithms that operate with the Euclidean distance as basic parameter can be used. However, in single-molecule super-resolution imaging, a single fluorophore is often detected more than once, either in multiple
3.6 Image Representation
3.7 Data Interpretation
3.7.1 Cluster Analysis
Direct Stochastic Optical Reconstruction Microscopy
272
adjacent imaging frames or with interruptions induced by blinking or reversible photoswitching. Single-molecule localizations at the same position in adjacent frames most likely do not correspond to two molecules, but to the same single fluorophore which was detected multiple times. These multiple localizations can be cor-rected for, which was demonstrated by either using correlation algorithms [32] or tracking algorithms [33, 34].
Cluster analysis can be performed by, e.g., calculating nearest neighbor distances of single-molecule localizations. As these distri-butions are measured by a single distance parameter, characteristic distances can be detected and compared to a random distribution of single-molecule localizations. Other algorithms consider all single- molecule positions, such as Ripley’s K function [35].
The colocalization of two different fluorophores can be investigated using image-based spatial colocalization algorithms [36, 37]. In a similar way, these algorithms can be applied to reconstructed super-resolution images. However, the results depend on post-processing parameters chosen for the reconstruction of the artificial images, e.g., blurring or pixelation. On the other hand, colocalization anal-ysis can be performed on the basis of single- molecule coordinates. This way, it is possible to provide a measure on how each localiza-tion of one species is surrounded by a second species, and heteroge-neities and subpopulations can be revealed [38].
4 Notes
1. For following dynamic processes, the acquisition rate has to be sufficiently faster than the dynamics of the observed biological process (Fig. 5a). Dynamic processes can be visualized using a sliding window algorithm [10], or heterogeneous dynamics of small unresolved assemblies can be observed [11]. Most cells already contain glutathione (GSH), a tripeptide containing a thiol group, at millimolar concentrations, depending on the cellular environment [39]. This has been proven to be suited for direct photoswitching of some organic fluorophores like rhodamine or oxazine derivatives in living cells [11, 12].
2. Addressing biological questions often requires observing multiple targets at once, such that multicolor imaging is needed. In dSTORM imaging, this translates to the need of (1) at least a dye pair with spectrally separated fluorescence emission, (2) similar switching properties, and (3) similar brightness. Various solutions for multicolor imaging were introduced [6–8]. Chromatic aberrations need to be corrected carefully, e.g., by using multicolor fiducial markers or by recording a reference sample (Fig. 5b). Live cell imaging is only practicable with simultaneous or alternating imaging schemes.
3.7.2 Colocalization Analysis
Ulrike Endesfelder and Mike Heilemann
273
3. In its simplest experimental configuration, information on the axial position of a single fluorophore is not accessible, as a two- dimensional detector is used. However, the axial position can be determined by introducing a distortion into the PSF of single fluorophores (Fig. 5c), e.g., through astigmatism introduced by
Fig. 5 Advanced imaging schemes. (a) Video-like dSTORM data of dynamic samples can be obtained by apply-ing a sliding window algorithm to an image stack. Here, the example of in vitro actin filaments gliding on a myosin surface is shown, each dSTORM image reconstructed from 100 individual TIRF images (integration time 10 ms). A reconstructed image including all localizations color-coded for time projects the movement of the filaments (right). Gray arrows indicate the disruption of a filament, typically occurring for filaments sub-jected to strong curvature (scale bar 1 μm). (b) A possible detection pathway for two-color imaging with paral-lel data acquisition and one EMCCD camera. The two detection channels are separated using two identical dichroic mirrors. The two images are superimposed and corrected for aberration by a nonlinear local weighted mean matrix. Exemplary aligned dual-color images of cells stained for actin (green) and β-tubulin (red) or cytochrome c oxidase (red) were superimposed with an accuracy of below 10 nm (scale bar 2 μm). (c) Schematic view of two approaches for single-molecule localization in 3D by distortion of the single-molecule PSF. Astigmatic distortion by a cylindrical lens (CL) causes lengthening of the PSF in either the x- or y-plane (left). The biplane approach records images in two focal planes (right)
Direct Stochastic Optical Reconstruction Microscopy
274
a cylindrical lens [40], biplane imaging with two cameras at different focal planes [41], or the generation of a double-helical PSFs [42] using a spatial light modulator. Thick tissue samples can be imaged in 3D by applying physical or optical sectioning [14, 43].
4. The spatial resolution is the distance at which two objects can still be discerned as individual objects [44]. In single-molecule local-ization microscopy, the spatial resolution can be related to the localization accuracy and can be estimated by calculating the full width at half maximum FWHM = »( )2 2 2 2 35ln .´e ´e [45].
5. In fluorescence microscopy, the target is not directly observed itself, but the fluorescent label which is attached to a target. In order to resolve a structure, not only the spatial or optical reso-lution is important but also the labeling density which deter-mines the structural resolution. This fact is addressed in the Shannon-Nyquist sampling theorem, which states that the sampling interval must be at least twice as small as the desired resolution [46]. Applied to single-molecule localization microscopy, a structural resolution of 20 nm thus requires that every 10 nm a fluorophore is localized.
Acknowledgments
The authors thank Marina Dietz, Franziska Fricke, and Bianca Nouvertné for providing super-resolution images. This work was supported by the Bundesministerium für Bildung und Forschung (FORSYS program, grant number 0315262).
References
1. Betzig E, Patterson GH, Sougrat R et al (2006) Imaging intracellular fluorescent proteins at nanometer resolution. Science 313:1642–1645
2. Heilemann M, van de Linde S, Schuttpelz M et al (2008) Subdiffraction-resolution fluo-rescence imaging with conventional fluores-cent probes. Angew Chem Int Ed Engl 47:6172–6176
3. Rust MJ, Bates M, Zhuang X (2006) Sub-diffraction- limit imaging by stochastic optical reconstruction microscopy (STORM). Nat Methods 3:793–795
4. van de Linde S, Loschberger A, Klein T et al (2011) Direct stochastic optical reconstruction microscopy with standard fluorescent probes. Nat Protoc 6:991–1009
5. Bates M, Dempsey GT, Chen KH et al (2012) Multicolor super-resolution fluorescence imag-ing via multi-parameter fluorophore detection. ChemPhysChem 13:99–107
6. Lampe A, Haucke V, Sigrist SJ et al (2012) Multi-colour direct STORM with red emitting carbocyanines. Biol Cell 104:229–237
7. van de Linde S, Endesfelder U, Mukherjee A et al (2009) Multicolor photoswitching micros-copy for subdiffraction-resolution fluorescence imaging. Photochem Photobiol Sci 8:465–469
8. Endesfelder U, Malkusch S, Flottmann B et al (2011) Chemically induced photoswitching of fluorescent probes – a general concept for super-resolution microscopy. Molecules 16: 3106–3118
9. Eckhardt M, Anders M, Muranyi W et al (2011) A SNAP-tagged derivative of HIV-1 – a versatile tool to study virus-cell interactions. PLoS One 6:e22007
10. Endesfelder U, van de Linde S, Wolter S et al (2010) Subdiffraction-resolution fluorescence microscopy of myosin-actin motility. ChemPhysChem 11:836–840
Ulrike Endesfelder and Mike Heilemann
275
11. Wombacher R, Heidbreder M, van de Linde S et al (2010) Live-cell super-resolution imaging with trimethoprim conjugates. Nat Methods 7:717–719
12. van de Linde S, Heilemann M, Sauer M (2012) Live-cell super-resolution imaging with syn-thetic fluorophores. Annu Rev Phys Chem 63:519–540
13. Jones SA, Shim SH, He J et al (2011) Fast, three-dimensional super-resolution imaging of live cells. Nat Methods 8:499–508
14. Nanguneri S, Flottmann B, Horstmann H et al (2012) Three-dimensional, tomographic super-resolution fluorescence imaging of serially sec-tioned thick samples. PLoS One 7:e38098
15. Heilemann M, van de Linde S, Mukherjee A et al (2009) Super-resolution imaging with small organic fluorophores. Angew Chem Int Ed Engl 48:6903–6908
16. Miller LW, Cai Y, Sheetz MP et al (2005) In vivo protein labeling with trimethoprim conju-gates: a flexible chemical tag. Nat Methods 2:255–257
17. Keppler A, Gendreizig S, Gronemeyer T et al (2003) A general method for the covalent labeling of fusion proteins with small molecules in vivo. Nat Biotechnol 21:86–89
18. Gautier A, Juillerat A, Heinis C et al (2008) An engineered protein tag for multiprotein label-ing in living cells. Chem Biol 15:128–136
19. Los GV, Encell LP, McDougall MG et al (2008) HaloTag: a novel protein labeling tech-nology for cell imaging and protein analysis. ACS Chem Biol 3:373–382
20. Salic A, Mitchison TJ (2008) A chemical method for fast and sensitive detection of DNA synthesis in vivo. Proc Natl Acad Sci U S A 105:2415–2420
21. Besanceney-Webler C, Jiang H, Zheng T et al (2011) Increasing the efficacy of bioorthogo-nal click reactions for bioconjugation: a com-parative study. Angew Chem Int Ed Engl 50:8051–8056
22. van de Linde S, Krstic I, Prisner T et al (2011) Photoinduced formation of reversible dye radicals and their impact on super-resolution imaging. Photochem Photobiol Sci 10: 499–506
23. Kottke T, van de Linde S, Sauer M et al (2010) Identification of the product of photoswitching of an oxazine fluorophore using fourier trans-form infrared difference spectroscopy. J Phys Chem Lett 1:3156–3159
24. Wolter S, Endesfelder U, van de Linde S et al (2011) Measuring localization performance of super-resolution algorithms on very active sam-ples. Opt Express 19:7020–7033
25. Tokunaga M, Imamoto N, Sakata-Sogawa K (2008) Highly inclined thin illumination enables clear single-molecule imaging in cells. Nat Methods 5:159–161
26. Wolter S, Sauer M (2012) Follow-up to paper by S. Wolter, M. Schuttpelz, M. Tscherepanow, S. van de Linde, M. Heilemann and M. Sauer, entitled Real-time computation of subdiffraction- resolution fluorescence images. J Microsc 245:109
27. Henriques R, Lelek M, Fornasiero EF et al (2010) QuickPALM: 3D real-time photoacti-vation nanoscopy image processing in ImageJ. Nat Methods 7:339–340
28. Thompson RE, Larson DR, Webb WW (2002) Precise nanometer localization analysis for individual fluorescent probes. Biophys J 82: 2775–2783
29. Mortensen KI, Churchman LS, Spudich JA et al (2010) Optimized localization analysis for single-molecule tracking and super-resolution microscopy. Nat Methods 7:377–381
30. Smith CS, Joseph N, Rieger B et al (2010) Fast, single-molecule localization that achieves theoretically minimum uncertainty. Nat Methods 7:373–375
31. Baddeley D, Cannell MB, Soeller C (2010) Visualization of localization microscopy data. Microsc Microanal 16:64–72
32. Veatch SL, Machta BB, Shelby SA et al (2012) Correlation functions quantify super- resolution images and estimate apparent clustering due to over-counting. PLoS One 7:e31457
33. Annibale P, Vanni S, Scarselli M et al (2011) Quantitative photo activated localization microscopy: unraveling the effects of photob-linking. PLoS One 6:e22678
34. Lando D, Endesfelder U, Berger H et al (2012) Quantitative single-molecule microscopy reveals that CENP-ACnp1 deposition occurs during G2 in fission yeast. Open Biol 2:120078
35. Ripley BD (1977) Modelling spatial patterns. J R Stat Soc B 39:172–212
36. Bolte S, Cordelieres FP (2006) A guided tour into subcellular colocalization analysis in light microscopy. J Microsc 224:213–232
37. Zinchuk V, Zinchuk O (2008) Quantitative colocalization analysis of confocal fluorescence microscopy images. Curr Protoc Cell Biol, Unit 4.19
38. Malkusch S, Endesfelder U, Mondry J et al (2012) Coordinate-based colocalization analy-sis of single-molecule localization microscopy data. Histochem Cell Biol 137:1–10
39. Jones DP (2010) Redox sensing: orthogonal control in cell cycle and apoptosis signalling. J Intern Med 268:432–448
Direct Stochastic Optical Reconstruction Microscopy
276
40. Huang B, Jones SA, Brandenburg B et al (2008) Whole-cell 3D STORM reveals interactions between cellular structures with nanometer- scale resolution. Nat Methods 5:1047–1052
41. Juette MF, Gould TJ, Lessard MD et al (2008) Three-dimensional sub-100 nm resolution flu-orescence microscopy of thick samples. Nat Methods 5:527–529
42. Pavani SR, Thompson MA, Biteen JS et al (2009) Three-dimensional, single-molecule fluorescence imaging beyond the diffraction limit by using a double-helix point spread function. Proc Natl Acad Sci U S A 106: 2995–2999
43. Cella ZF, Lavagnino Z, Perrone DM et al (2011) Live-cell 3D super-resolution imaging in thick biological samples. Nat Methods 8:1047–1049
44. Rayleigh L (1896) On the theory of optical images, with special reference to the micro-scope. Philos Mag 42:167–195
45. Huang B, Wang W, Bates M et al (2008) Three-dimensional super-resolution imaging by stochastic optical reconstruction micros-copy. Science 319:810–813
46. Shannon CE (1949) Communication in the presence of noise. Proc Inst Radio Eng 37: 10–21
Ulrike Endesfelder and Mike Heilemann
277
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8_15, © Springer Science+Business Media New York 2015
Chapter 15
Optogenetics: Optical Control of a Photoactivatable Rac in Living Cells
Taofei Yin and Yi I. Wu
Abstract
Recent developments in optogenetics have extended optical control of signaling to intracellular proteins, including Rac, a small G protein in the Rho family. A blue light-sensing LOV (light, oxygen, or voltage) domain derived from Avena sativa (oat) phototropin was fused to the N-terminus of a constitutively active mutant of Rac, via an α-helix (Jα) that is conserved among plant phototropins. The fused LOV domain occluded binding of downstream effectors to Rac in the dark. Exposure to blue light caused a conforma-tional change of the LOV domain and unwinding of the Jα helix, relieving steric inhibition. The LOV domain incorporates a fl avin as the photon-absorbing cofactor and can be activated by light in a reversible and repeatable fashion. In cultured cells, global illumination with blue light rapidly activated Rac and led to cell spreading and membrane ruffl ing. Localized and pulsed illumination generated a gradient of Rac activity and induced directional migration. In this chapter, we will describe the techniques in detail and present some examples of applications of using photoactivatable Rac (PA-Rac) in living cells.
Key words Optogenetics , Rho GTPases , Rac or Rac1 , Photoactivatable , PA-Rac , LOV domain , Cell migration
1 Introduction
Signaling events inside a living cell are often subject to stringent spatial and temporal control. Such regulatory mechanisms are being increasingly appreciated because they are essential to achieve specifi city in signaling transduction, preserve fi delity of cellular pro-cesses, and enable coordination between complex cellular behaviors. A thorough understanding of how cells accomplish this requires complementary methods both in measurement and perturbation. While increasing numbers of fl uorescence imaging technologies are developed to resolve subcellular structures [ 1 , 2 ] and to capture fast dynamics [ 3 ], the majority of methods of perturbation rely on global perfusion of pharmacological agents and slow genetic modifi cations such as RNAi and overexpression. A method of perturbation of signaling that achieves comparable spatiotemporal
278
resolution is needed. With the emergence of a research fi eld called optogenetics [ 4 ], it became feasible to switch on or off specifi c signaling proteins with light at precise times and subcellular loca-tions. For example, microbial channelrhodopsins are broadly applied in neuroscience to trace a transient fi ring of a single neu-ron to a map of functionally connected neuronal networks [ 5 ]. Similarly, engineering of light-induced gene expression [ 6 – 8 ] or protein-protein interactions [ 9 ] enables rewiring of signaling net-works for synthetic biology applications. Despite these rapid developments, direct control of the activities of signaling proteins in optogenetics has been limited to homologous proteins such as ion channels [ 10 ] and GPCRs [ 11 ] due to the reliance on the existence of native photoreceptor homologues. The vast majority of intracellular signaling proteins remains out of reach.
The LOV (light, oxygen, or voltage) domain denotes a sub-group of the Per-ARNT-Sim (PAS) domains, named for homolo-gous domains that occur within signaling proteins initially identifi ed in Drosophila period circadian rhythm (Per), single-minded (Sim), and the vertebrate aryl hydrocarbon receptor nuclear transporter (ARNT) [ 12 ]. Photochemistry studies of the LOV domain revealed that excitation of the fl avin molecule leads to the formation of a covalent linkage between the C4(a) atom in the fl avin and a thiol from a conserved Cys residue in the LOV domain. This reaction is reversible and undergoes recovery in the dark over seconds [ 13 ]. C-terminal to the LOV domain in plant phototropin is an α-helix (named Jα) that is docked on the β-sheet of the LOV domain. Upon light illumination, the LOV domain undergoes conforma-tional changes including dissociation and unwinding of a C-terminal helical extension (Jα) [ 14 ]. We tethered these sequences to the N-terminus of a constitutively active mutant of Rac, a member of the Rho family of small G proteins, so that the LOV domain steri-cally blocked its interactions with downstream effectors until illu-mination with blue light ( see Fig. 1 ) This produced a genetically encoded photoactivatable analogue of Rac (PA-Rac) that enables precise modulation of Rac activity at regions that are submicrons in size and capable of controlling activation with microsecond preci-sion in living cells and animals [ 15 , 16 ]. Here we describe in detail how to apply the PA-Rac reagents in live cells and how conven-tional imaging systems can be customized to be compatible with such applications.
2 Materials
The majority of the DNA constructs used in the current protocol can be obtained from the addgene website ( www.addgene.org , search for reagents from Dr. Klaus Hahn’s laboratory) or from our
2.1 DNA Plasmids for PA-Rac and Its Controls
Taofei Yin and Yi I. Wu
279
lab upon request (Table 1 ). PA-Rac contains the following protein sequences: LOV(404-546)-Rac(4-192, Q61L/E91H/N92H). Q61L is a constitutively active mutation in Rac that renders it GTP hydrolysis defi cient. The mutations E91H and N92H were intro-duced empirically to disrupt Rac-GAP interactions. Overexpression of the LOV-Rac fusion that harbors only the Q61L mutation led to apparent membrane ruffl es in cells in the dark, consistent with an elevated background of Rac activation. We speculated that this is due to a dominant-negative effect of this earlier construct on endogenous Rac because it may sequester GAP proteins. The GAP-binding mutations in PA-Rac resulted in improved binding of effector PAK in the light yet substantially reduced the back-ground Rac activity in the dark.
We generated PA-Rac constructs with various fl uorescent protein tags, including mCerulean, mVenus, and mCherry, to facilitate the identifi cation of cells expressing PA-Rac in transient experiments. Based on the excitation wavelengths of these fl uo-rescent proteins, mCherry (~590 nm) can be monitored with-out concern for excitation of the LOV domain. The standard commercial excitation fi lters for mVenus pass light below 500 nm which can activate the LOV domain. As a result, we recommend using different, somewhat suboptimal, excitation fi lters that
Fig. 1 Cartoon diagram of the photoactivatable Rac (PA-Rac). A LOV domain and its C-terminal Jα helix are fused to the N-terminus of a constitutively active mutant of Rac1 and block its interaction with downstream effectors (e.g., PAK) in the dark. Upon illumination, the Jα helix dissociates from the LOV domain and unwinds, restoring PAK binding and activation
Table 1 Available constructs of PA-Rac from addgene website or upon request
Vector: pTriEx-4 (transient transfection) or pBabe-TetCMV-puro (stable cell line and inducible expression)
Fluorescent protein tags: mCerulean, mVenus, or mCherry
Modifi cations of the LOV domain:
WT (photoactivatable), C450A and C450M (dark), or I539E (lit)
Modifi cations of Rac1: Q61L/E91H/N92H (active) or T17N (dominant negative)
Optogenetics
280
blocks transmission below 510 nm, together with imaging rou-tines that use attenuated excitation intensity and brief exposure time, to avoid unintended photoconversion of the LOV domain. Because the excitation wavelength of mCerulean (433 nm max) falls in the photoactivation spectrum, fl uorescence imaging of mCerulean is best used for confi rmation of expression at the end of experiments.
For transient transfections, plasmids in pTriEx-4 vector (Novagen) should be used. Alternatively plasmids in tetracycline- inducible expression (Tet-Off) retroviral vector pBabe-TetCMV- puro can be used to generate cell lines that have stable expression. With retroviral gene delivery, lower but more uniform expression of PA-Rac can be achieved, which is benefi cial in obtaining consistent photoactivation responses among different cells. The inducible expression system is of favor because of the lack of expression of PA-Rac in cells in the presence of low concentration of doxycycline in the medium. The cells can be maintained and passaged in the light without worrying about unintended photoactivation. We rec-ommend using the included fl uorescent protein tag to gauge expression level, determining optimum expression for light- induced Rac activation with minimal background activity in the dark. The expression level, if needed, can be further controlled through titra-tion with different concentrations of doxycycline. Alternatively, populations of stable cells with optimal expression level can be iso-lated using fl uorescence activated cell sorting (FACS).
1. HeLa, mouse embryonic fi broblast (MEF), and LinXE cells are maintained in high-glucose DMEM supplemented with 10 % fetal bovine serum (FBS), GlutaMAX, and Pen/Strep (Invitrogen). For convenience, we use a prior established Tet- Off MEF cell line (Clontech) that stably expresses tetracycline transactivator (tTA). When the Tet-Off cell line is not avail-able, coinfecting other murine cells with retroviral particles produced from the plasmid pRev-Tet-Off (Clontech) may be necessary. The Tet-Off MEFs are maintained in above culture medium with 100 μg/ml of G418 for drug-resistance selection of the tTA gene. LinXE is a HEK293-derived retroviral pro-ducer cell line. It stably expresses packaging and envelope pro-teins for the production of ecotropic, MoMuLV-based retroviruses that infect dividing murine (mouse and rat) cells. LinXE cells are maintained in above medium with 100 μg/ml of hygromycin.
2. X-tremeGENE (Roche) and Opti-MEM I reduced serum medium (Invitrogen) are needed for transient transfection.
3. Medium 199 without phenol red (Invitrogen) or F-12K nutri-ent mixture without phenol red (custom made from Invitrogen) is used in epifl uorescence imaging.
2.2 Cell Culture, Transfection, and Retroviral System
Taofei Yin and Yi I. Wu
281
4. 25 mm round coverslips that are cleaned with detergent and by sonication. After extensive rinsing, the coverslips can be stored in 70 % ethanol in a sealed jar.
5. A stock solution (10,000×) of doxycycline (Clontech) at 100 μg/ml is prepared for switching off protein expression in Tet-Off MEFs and a stock solution (1,000×) of puromycin (Gold Biotechnology) at 10 μg/μl for selection of stable cells after retroviral infection.
6. 2× HBS (HEPES-buffered saline): 50 mM HEPES, 280 mM NaCl, and 1.5 mM Na 2 HPO 4 , adjust pH exactly to 6.95 with HCl. Sterile fi lter the solution through 0.2 μm fi lter and store at 4 °C.
7. 2 M CaCl 2 . Filter sterilize through 0.2 μm fi lter and store at 4 °C. 8. 0.45 μm PES fi lter and 5 ml syringes with luer lock. 9. Polybrene (Sigma) stock solution (1,000×) at 6 mg/ml.
Other common reagents for cell culture can be obtained from Invitrogen. Additional reagents for specifi c imaging applications will be described in Subheading 3 .
Localized illumination with precise computer control can be accomplished using most laser scanning confocal microscopes, and applications using PA-Rac on these systems should be relatively straightforward ( see Note 1 ). Alternatively, localized illumination can be achieved using most conventional wide-fi eld microscopes that have a fi eld stop/diaphragm in place or can be modifi ed to incorporate a pinhole in a conjugate image plane. The broad- spectrum light source used for epifl uorescence excitation can be used to illuminate a small region of the cell. Commercial vendors have devised a variety of add-on solutions for laser irradiation of small portions of the fi eld of view, usually for FRAP studies. A laser beam can be coupled into the light path and focused on the focal plane to a diffraction-limited spot or dilated to bigger areas through z offset. These spots can also be mobilized to scan across different shapes either manually or using galvanometer-driven mirrors. Complex patterns can be achieved using digital micromirror-based devices such as the commercially available Mosaic active illumina-tion system (Andor) or more advanced methods based on liquid- crystal spatial light modulators, realized in applications involving laser tweezers [ 17 ] and adaptive optics [ 18 ]. We opted for the Mosaic active illumination system because of its ability to simulta-neously illuminate multiple regions of interest and fast switching time. One of the concerns with this method is insuffi cient intensity output. Because the LOV domain has a reasonable extinction coef-fi cient ( ε 449 = 12, 200 mol −1 cm −1 ) but high quantum yield (0.44) [ 19 ], we found that even collimated consumer LED light source (<1 W output) can induce membrane protrusion in live cells upon
2.3 Illumination Control and Microscope Customizations
Optogenetics
282
brief (seconds), localized illumination. We found that recent versions (TE2000 and Ti) of Nikon inverted epifl uorescence microscopes are convenient for such applications because they have a “stage-up” design ( see Note 2 ), proven advantageous over sys-tems from other microscope manufacturers. The stage-up design allows an ideal incorporation of the Mosaic system so that epifl uo-rescence excitation and Mosaic active illumination can be incorpo-rated independently in the infi nity space between objective and camera. Two motorized turrets for fi lter blocks also permit rapid and independent control of fi lter settings for imaging and photo-activation ( see Fig. 2 ). Additional components needed for the cus-tomization of such an imaging system are:
1. LED light source to be mounted on the Mosaic active illumi-nation system: a high-power LED (Thorlabs) at 455 nm wave-length, with a power output rated at 900 mW, and with a beam collimator and an adapter to the Mosaic system.
Fig. 2 Schematic diagram of customization of an epifl uorescence microscope for application of PA-Rac. A digital micromirror device (DMD) is used to generate active spatial patterns of illumination. The DMD is coupled to the image conju-gate plane of the microscope and refl ects a collimated light beam from a high- power LED source (455 nm) which activates the LOV domain. An additional dichroic mirror (495LP) is inserted in the infi nity space of the conventional epi-fl uorescence imaging path to incorporate the activation light beam. The trans-mitted light source, typically halogen, is replaced with a red light-emitting (625 nm) high-power LED to avoid activation of the LOV domain during acquisi-tion of DIC images
Taofei Yin and Yi I. Wu
283
2. Modifi cation of transmitted light: replace the halogen light source with a high-power LED (Thorlabs) at 625 nm wave-length, with a power output rated at 440 mW, and with a beam collimator and an adapter to the transmitted lamp house.
3. LED DC current drivers with TTL inputs (Thorlabs) and a DAQ board for shutter-free switching of LEDs using TTL outputs (NI, USB-6259).
4. A 495LP dichroic mirror is installed in the top fi lter block to refl ect the photoactivation beam “shaped” by the Mosaic sys-tem to the samples while passing excitation and emission at longer wavelengths during image acquisition. Because the LEDs have well-confi ned spectrum range (~20 nm full width at half maximum (FWHM)), it was unnecessary to use cleanup fi lters in front of the LEDs in our hands.
We also installed a small CO 2 incubator near the microscope in the darkroom and yellow or red consumer LEDs (PAR38) for darkroom lightings for convenience in sample handling.
3 Methods
1. Day 1: trypsinize a well-maintained stock of HeLa cells, and seed the cells at 30–50 % confl uency in a 35 mm culture dish in the morning. The cells should adhere and spread within a few hours and will be ready for transfection at the end of the day.
2. At the end of day 1, prepare transfection of PA-Rac constructs using X-tremeGENE DNA transfection reagent according to manufacturer’s instructions. Briefl y, 0.25 μg of PA-Rac DNA plasmid and 3 μl of X-tremeGENE reagent are added sequen-tially to 100 μl of Opti-MEM in a 0.5 ml centrifuge tube. Gently mix the solution by tapping the tube and briefl y centri-fuge the tube at low speed, followed by incubation at room temperature for 30 min.
3. Add the DNA mix dropwise onto the cells covering different areas of the 35 mm culture dish, and gently rock the dish hori-zontally to help spread the DNA mix onto cells. Replace the culture dish back in the incubator until day 2.
4. Pick a 25 mm round coverslip that was previously cleaned. Rinse the coverslip with PBS for three times in a new 35 mm culture dish and fl ip the coverslip with a pair of forceps between washes to remove residue amount of ethanol trapped under-neath the coverslip. Add 2 ml of 10 μg/ml fi bronectin in PBS and move the culture dish on a rocker and incubate with gentle shaking at 4 °C overnight.
5. Day 2: fi rst thing in the morning, check transfection effi ciency and expression level of PA-Rac by monitoring the fl uorescent
3.1 Transient Expression of PA-Rac in HeLa Cells
Optogenetics
284
protein tag using a tissue culture microscope equipped with epifl uorescence. Take out the culture dish containing the fi bronectin- coated coverslip and rinse briefl y with PBS at room temperature. Trypsinize the transfected cells and seed them at 10–20 % confl uency on the coated coverslip. Move the culture dish to the incubator located in the dark room where the imag-ing system is located ( see Note 3 ).
6. Warm up imaging medium and equilibrate it with CO 2 in the same incubator until imaging.
1. In the morning of day 1, passage well-maintained LinXE cells at 70–90 % confl uency into a 60 mm culture dish with 4 ml medium.
2. At the end of day 1, set up calcium-phosphate transfection of viral constructs. Equilibrate plasmid, buffers, and sterile water to room temperature before mixing the solutions. Add 300 μl of 2× HBS in a polystyrene tube.
3. Prepare DNA mix (fi nal volume 300 μl) in a 0.5 ml centrifuge tube as follows: 10 μg pBabe-TetCMV-puro plasmid, dilute DNA with sterile water to 262.5 μl, add 37.5 μl 2 M CaCl 2 , and mix gently.
4. Add DNA mix dropwise to the polystyrene tube containing 2× HBS while vortexing the polystyrene tube at low speed. Incubate the mixture at room temperature for 30 min. The solution should become cloudy due to DNA-calcium phos-phate complex formation.
5. Briefl y vortex the DNA-calcium phosphate complex and add dropwise onto LinXE cells.
6. In the morning of day 2, the LinXE cells should become highly fl uorescent. Replace the medium with 2 ml fresh culture medium. Note that the supernatant already contains infectious virus particles and safety procedures for retroviral production should be in place.
7. At the end of day 2, also seed MEF-Tet-Off cells at 50 % con-fl uency in a 35 mm culture dish.
8. Day 3: collect the fi rst round of viral supernatant and add 2 ml fresh medium to the cells for the second round collection. Centrifuge the supernatant at 1,500 × g for 5 min to remove cell debris and pass the supernatant through a 0.45 μm PES fi lter using a 5 ml syringe.
9. Add polybrene to the viral supernatant at a fi nal concentration of 6 μg/ml. After briefl y mixing, replace the culture medium of MEF cells with the viral supernatant and incubate overnight.
10. Day 4: check whether the MEF cells start expressing fl uores-cent protein markers. Depending on viral titer, one round of
3.2 Viral Production and Infection, and Inducible Expression of PA-Rac
Taofei Yin and Yi I. Wu
285
infection may be suffi cient. The maximal expression level may be achieved between 24 and 48 h after infection. If necessary, a second or up to third round of infection can be performed by repeating steps 8 and 9 with the transfected LinXE cells.
11. Replace with fresh medium and proceed with steps 5 and 6 in Subheading 3.1 for imaging experiments. Alternatively, the expression can be switched off for long-term culture ( see Note 4 ) by adding doxycycline at a fi nal concentration of 10 ng/ml. Expression can be restored by passaging cells into doxycycline- free medium and incubate for 24–48 h.
Global illumination of cells expressing PA-Rac or its variants can be conveniently achieved by exposing the culture dish to ambient light or designated light source with ample irradiance in the blue light range to induce (PA-Rac) or inhibit (T17N) Rac activity. The C450A or C450M mutants can be used in these experiments as light-insensitive controls. The cells can be lysed for biochemical analysis or can be fi xed after a brief dose of irradiation, followed by standard immunocytochemistry protocols that stain F-actin (fl uo-rescently labeled phalloidin) or detect translocation and phosphory-lation of proteins downstream of Rac. Alternatively, live-cell imaging experiments can be performed with global illumination using a motorized epifl uorescence microscope. We found that traditional mercury Arc lamp used in epifl uorescence has suffi cient irradiance in the blue light range. Photo cytotoxicity can be minimized by using a standard fi lter for cyan fl uorescent protein (CFP) or green fl uorescent protein (GFP) excitation and neutral density fi lters. Depending on the power of the light source, the expression level of PA-Rac, and how sensitive the cells used are to Rac activation, the illumination can be applied as milliseconds or seconds of pulses or constant during image acquisition. Because the LED transmitted light source has a well-defi ned emission range (625/17 nm FWHM) that is distant from the absorption spectrum of the LOV domain, we typically acquire DIC images at high frame rate (2–10 frames/s) to capture the dynamics of cell shape changes without interfering activation of the LOV domain. We found that 10–20 min post illu-mination is suffi cient for capturing cellular response to Rac activa-tion with an initial cellular response becoming apparent within 30 s, although we typically acquire equal duration/numbers of images prior to illumination to set the background cellular activity. The DIC images can be assembled to time-lapse movies for viewing or analyzed in kymograph for activities of membrane protrusion and retraction using line scans at regions of interest.
As described before, limited spatial control can be implemented using the fi eld diaphragm or pinholes in an image conjugate plane. By modulating illumination intensity, low-intensity excitation light at a “safe” wavelength can be used to target and shape the illumination,
3.3 Photo Manipulation of PA-Rac in Living Cells
3.3.1 Global Illumination to Induce or Inhibit Rac Activity
3.3.2 Precise Spatial and Temporal Control of Photoactivation
Optogenetics
286
followed by high-intensity illumination below 500 nm to activate protein. Acquisition and analysis of DIC images can be conducted at cellular regions near the site of illumination, using other parts of the cell body as control regions. Because of the impact of active Rac on actin dynamics, mCherry-actin or Lifeact- mCherry [ 20 ] can be used to aid detection of Rac activation. To analyze Rac initiation of down-stream signaling, we also tracked translocation of downstream effec-tors. For wide-fi eld microscopy, artifacts of changing cell volume were avoided using ratiometric imaging. The fl uorescence intensity of the tagged target protein was divided by the fl uorescence intensity of a separately expressed volume indicator, i.e., yellow fl uorescent protein (YFP) or mCherry. As demonstrated previously [ 15 ], PA-Rac redistributed slowly out of sites of illumination with a diffusion coef-fi cient of 0.55 μm 2 /s. This led to detectable accumulation of down-stream effectors where Rac was activated. Phosphorylation of endogenous PAK can also traced using immunofl uorescence of fi xed cells after local activation of PA-Rac ( see Fig. 3 ). MEF cells stably expressing PA-Rac were plated onto special coverslips with etched grids and numbers (Bellco). The grids and numbers are visible through a 20× phase-contrast objective and can be used to locate the cells that had been irradiated. Cells can be locally illuminated. Immediately after protrusions were induced, the cells were fi xed in 3.7 % formalin (Sigma), permeabilized in 0.2 % Triton X-100, incu-bated with anti-phospho-PAK antibody (Cell Signaling), and fi nally incubated with Alexa Fluor 594-conjugated secondary antibody (Molecular Probes).
4 Notes
1. The action spectrum of plant phototropin is in the UV-A and blue light range (360–500 nm). We tested several common laser lines for their ability to induce membrane ruffl es in MEF cells expressing PA-Rac. The wavelengths 405, 458, 473, and 488 nm all proved to be effective. The power dosage of the PA-Rac to 458 nm line was measured in stable MEF cell lines, where expression levels could be well controlled and the areas of induced protrusions readily measured. A light dose of 6.2 μJ over a 10 μm spot at 458 nm induced a cellular response with a single exposure. This was the lowest power setting (0.1 % of total power on the mW scale) of our Fluoview 1000 confocal microscope at very fast scan rate (10 μs/pixel). Following this exact photoactivation regime but with increasing laser power or scan duration, we determined that the cellular response (protrusion area) stopped increasing when we reached 1,000-fold higher dose.
Taofei Yin and Yi I. Wu
287
2. The “stage-up” confi guration of Nikon microscope refers to an elevated nosepiece (objectives) and sample stage that permit installation of an additional fi lter block turret in the infi nity space of a microscope, benefi ting from the use of infi nity- corrected objectives.
3. Because of the light sensitivity of the LOV domain, cells expressing PA-Rac should not be exposed to light within the LOV action spectrum (<500 nm) immediately before imaging experiments. It is safe to prepare live samples under yellow or red light, which can be implemented in dark rooms. Even light emitted from bright computer monitors can be of concern, but this is less consequential at a distance. We transferred the cells immediately after transfection into incubators located in
Fig. 3 Local photoactivation of PA-Rac followed by immunocytochemistry. MEF cells expressing mVenus-PA- Rac were plated onto special coverslips that have photoetched grids of numerically labeled squares and were locally irradiated at 473 nm ( white circle ) to generate protrusion (phase-contrast images shown before and after irradiation). The cells were immediately fi xed and stained for phospho-PAK. The numbers on the etched grids were used to locate irradiated cells after fi xation (DIC). The acute staining of phospho-PAK ( arrowhead ) at PA-Rac-induced protrusions indicates local PAK activation (pPAK). Scale bars, 10 μm
Optogenetics
288
isolated dark rooms to avoid unintended light exposure. In some experiments, extensive manipulations were required prior to imaging. When this was diffi cult in a dark room, cells could be incubated in the dark for an hour to reduce the effect of prior light exposure and restore responsiveness to photoactivation.
4. For reasons that remain unclear to us, MEF cells selected for stable expression of PA-Rac appeared to downregulate the activity of endogenous Rac, evidenced by the reduction of lamellipodia of these cells prior to induction of expression. In contrast we noticed an upregulation of Rac activity in PA-Rac- T17N overexpressing cells prior to induction of expression. This could be due to mechanisms of cellular compensation. Nevertheless, such phenomena seemed to enhance the effect of illumination to switch on or off Rac in living cells.
Acknowledgments
This work was supported by NIH grant NS071216 and UCHC start-up funds.
References
1. Hell SW (2009) Microscopy and its focal switch. Nat Methods 6:24–32
2. Betzig E, Patterson GH, Sougrat R et al (2006) Imaging intracellular fl uorescent pro-teins at nanometer resolution. Science 313:1642–1645
3. Salomé R, Kremer Y, Dieudonné S et al (2006) Ultrafast random-access scanning in two- photon microscopy using acousto-optic defl ec-tors. J Neurosci Methods 154:161–174
4. Deisseroth K, Feng G, Majewska AK et al (2006) Next-generation optical technologies for illuminating genetically targeted brain cir-cuits. J Neurosci 26:10380–10386
5. Deisseroth K (2011) Optogenetics. Nat Methods 8:26–29
6. Shimizu-Sato S, Huq E, Tepperman JM et al (2002) A light-switchable gene promoter sys-tem. Nat Biotechnol 20:1041–1044
7. Kennedy MJ, Hughes RM, Peteya LA et al (2010) Rapid blue-light-mediated induction of protein interactions in living cells. Nat Methods 7:973–975
8. Wang X, Chen X, Yang Y (2012) Spatio-temporal control of gene expression by a light- switchable transgene system. Nat Methods 9:266–269
9. Levskaya A, Weiner OD, Lim WA et al (2009) Spatiotemporal control of cell signalling using a light-switchable protein interaction. Nature 461:997–1001
10. Nagel G, Szellas T, Huhn W et al (2003) Channelrhodopsin-2, a directly light-gated cation-selective membrane channel. Proc Natl Acad Sci U S A 100:13940–13945
11. Airan RD, Thompson KR, Fenno LE et al (2009) Temporally precise in vivo control of intracellular signalling. Nature 458:1025–1029
12. Möglich A, Ayers RA, Moffat K (2009) Structure and signaling mechanism of Per-ARNT- Sim domains. Structure 17:1282–1294
13. Swartz TE, Corchnoy SB, Christie JM et al (2001) The photocycle of a fl avin-binding domain of the blue light photoreceptor pho-totropin. J Biol Chem 276:36493–36500
14. Harper SM, Neil LC, Gardner KH (2003) Structural basis of a phototropin light switch. Science 301:1541–1544
15. Wu YI, Frey D, Lungu OI et al (2009) A genetically encoded photoactivatable Rac con-trols the motility of living cells. Nature 461:104–108
16. Wang X, He L, Wu YI et al (2010) Light- mediated activation reveals a key role for Rac in
Taofei Yin and Yi I. Wu
289
collective guidance of cell movement in vivo. Nat Cell Biol 12:591–597
17. Curtis JE, Koss BA, Grier DG (2002) Dynamic holographic optical tweezers. Opt Commun 207:169–175
18. Girkin JM, Poland S, Wright AJ (2009) Adaptive optics for deeper imaging of biological samples. Curr Opin Biotechnol 20:106–110
19. Salomon M, Christie JM, Knieb E (2000) Photochemical and mutational analysis of the FMN-binding domains of the plant blue light receptor, phototropin. Biochemistry 39:9401–9410
20. Riedl J, Crevenna AH, Kessenbrock K et al (2008) Lifeact: a versatile marker to visualize F-actin. Nat Methods 5:605–607
Optogenetics
291
Peter J. Verveer (ed.), Advanced Fluorescence Microscopy: Methods and Protocols, Methods in Molecular Biology,vol. 1251, DOI 10.1007/978-1-4939-2080-8, © Springer Science+Business Media New York 2015
A
Abbe diffraction limit ...............................................1 76, 214 AB plot ...............................................................................9 0 Acceptor ....................... 3 8, 67–75, 77–79, 84–86, 90, 91, 93,
94, 97, 101, 102, 104, 111, 121, 127–129, 152–154 Acceptor fluorophore .................3 8, 68, 72, 74, 102, 104, 152 Adaptive optics ...........................................................3 6, 281 Agarose beaker ................................................. 4 5, 47–54, 56 Amplitude modulation .......................................................8 9 Angle of incidence ................................ 5 –8, 10–13, 180, 182 Anisotropy
steady-state ......................................... 1 63, 166, 169, 170 time-resolved .................................................. 8 8, 89, 169
Astigmatic imaging .................................. 2 42–243, 245, 248 Autocorrelation ..........................1 39, 140, 145, 146, 148, 149 Autofluorescence ......................................7 8, 84, 96, 99, 102,
139–141, 168, 233, 258 Automated microscopy .........................1 8, 20, 21, 55, 59, 60,
64, 116, 122, 131, 234, 235, 238, 240, 241, 245, 257 Automated sample preparation ...........................................5 9 Axonal bouton ..............................................................3 1, 33
B
Binding site distribution ...................................1 95, 205–208 Biplane imaging ................................................ 2 42, 273, 274 Brownian motion ......................................................1 15, 136
C
Calcium imaging .............................................. 2 6, 30, 34–37 genetically encoded calcium indicators ...................3 4, 37
cAMP ............................................................... 6 7, 72, 77, 78 CCDs. See Charge-coupled devices (CCDs) CCPs. See Clathrin coated pits (CCPs) Cell adhesion ................................................................2 , 169 Cell migration ........................................................1 6–18, 44 Channelrhodopsin ......................................................3 5, 278 Charge-coupled devices (CCDs) ......................2, 4, 9, 11–13,
80, 88, 163, 182, 197, 235 Cholesterol depletion .......................................................1 70 Clathrin coated pits (CCPs) .........................................13, 14 Clathrin-mediated endocytosis (CME) ................. 13, 14, 17 Cluster analysis .........................................................2 71–272 CMOS camera .................................................................1 82
Confocal microscopy ........................2 , 10–12, 26–28, 30, 43, 48, 62–64, 71–73, 77, 79, 80, 87, 88, 91–93, 96, 97, 99, 100, 102, 135–137, 142, 144, 161–164, 170, 175, 176, 194, 195, 197, 201, 202, 204, 207, 214, 216, 217, 220, 222, 281, 286
Conformational changes ................................. 3 8, 67, 84, 278 Cranial window ............................................................3 1–32 Cross-correlation ............................... 8 8, 136–138, 140–148,
185, 243–245, 248, 249, 260 Cylindrical lens ........................................ 2 36, 242–245, 248,
249, 252, 256, 257, 273, 274
D
Data mining ..................................................... 5 9, 60, 63, 64 Dc. See Diffusion coefficient (Dc) Deconvolution ............................................. 4 3, 89, 199–200,
202, 203, 205, 207, 209 Dendritic spine ......................................... 3 1, 33, 34, 36, 215 Depletion beam ........................................................2 14–219 Diffraction barrier ............................................................2 14 Diffraction grating ..................................... 8 0, 181–183, 185,
189, 237, 251–253, 255, 256, 260 Diffraction limit .......................................1 51, 176, 181, 187,
200–202, 215, 231, 232, 253, 258, 281 Diffusion coefficient (Dc) .........................116, 119, 121–123,
125, 136, 138, 143, 148, 286 Digital micromirror-based devices ...................................2 81 Digital scanned light sheet microscope (DSLM) .........43, 45 Dimerization ................................................ 9 4, 99–101, 109 Direct stochastic optical reconstruction microscopy
(dSTORM) ..................................................263–274 DNA
damage ....................................................... 1 09, 111, 123 DNA–protein complexes ............................................1 12 repair ...................................................................1 11, 219
Donor ................................. 3 8, 67–79, 84–86, 90, 91, 93, 94, 96, 97, 99, 101–104, 127, 128, 130, 140, 152–154
Donor fluorophore ....................................... 3 8, 90, 104, 152 Drift ..................................2 0, 74, 77, 81, 145, 146, 179, 180,
183, 189, 233, 235, 238, 240–242, 245, 247, 258, 270 DSLM. See Digital scanned light sheet microscope
(DSLM) dSTORM. See Direct stochastic optical reconstruction
microscopy (dSTORM)
INDEX
292 ADVANCED FLUORESCENCE MICROSCOPY
Index
Dynamics ..................................... 2 , 4, 16–18, 20, 25, 30, 33, 44, 64, 73, 77, 78, 80, 83, 84, 109–131, 169, 170, 203, 216, 218, 263, 265, 269, 272, 273, 285, 286
E
Electron-multiplying CCD (EM-CCD) .................. 2, 4, 19, 137, 163, 164, 182, 235, 236, 240, 243, 246, 257, 258, 268, 273
Electroporation ...................................................................3 1 Emission beam splitter .....................................................4 , 9 Emission spectrum ......................6 8, 138, 152, 153, 214, 219 Endocytosis
clathrin-coated pits .................................................1 3, 14 CME ................................................................ 1 3, 14, 17
EPAC ...........................................................................7 2, 77 Evanescent field .............................................. 1 , 5–10, 12, 13 Evanescent field microscopy ....................1 , 5–10, 12, 13, 162 Evanescent wave ................................................... 1 , 9, 10, 15 Evanescent wave microscopy ................................ 1 , 9, 10, 15 Excitation beam ............................................. 7 –9, 11, 12, 37,
182, 188, 214, 218, 247 Excitation spectrum ..........................................................2 09 Exocytosis ......................................................... 2 , 15–16, 227 Exponential ..............................................6 , 7, 13, 67, 87, 89,
90, 95–101, 103, 123 Extinction coefficient ..........................8 5, 138, 162, 172, 281
F
FAs. See Focal adhesions (FAs) FCCS. See Fluorescence cross-correlation spectroscopy
(FCCS) FCS. See Fluorescence correlation spectroscopy (FCS) FD-FLIM. See Frequency-domain FLIM (FD-FLIM) Fiducial markers ................2 41–243, 245, 246, 249, 270, 272 Fiducials ............................ 2 35, 244, 247, 248, 257, 259, 260 FilterFRET ...................................................... 6 8, 69, 73–76 FLIM. See Fluorescence lifetime imaging microscopy
(FLIM) FLIM-FRET ............ 2 6, 85, 86, 90, 91, 93–94, 98–101, 103 Fluctuations ...........................7 2, 81, 135–138, 143–145, 162 Fluorescence
anisotropy ............ 1 52–155, 161, 162, 164, 166–168, 170 autofluorescence ..................................7 8, 84, 96, 99, 102,
139–141, 168, 233, 258 fluctuations ................................. 1 35, 136, 138, 143–145 lifetime ................. 8 3, 86–89, 91–102, 104, 136, 154, 219
Fluorescence anisotropy ................................... 1 52–155, 161, 162, 164, 166–168, 170
Fluorescence correlation spectroscopy (FCS) .................................................... 135–149, 216
Fluorescence cross-correlation spectroscopy (FCCS) ................................. 136–138, 140–141, 147
Fluorescence fluctuations .................. 1 35, 136, 138, 143–145
Fluorescence lifetime ............................... 8 3, 86–89, 91–102, 104, 136, 154, 219
Fluorescence lifetime imaging microscopy (FLIM) ..................................................... 26, 83–104
Fluorescence recovery after photobleaching (FRAP) ........................................... 26, 109–131, 281
Fluorescence resonance energy transfer/Förster resonance energy transfer (FRET) ............................. 26, 35, 37, 38, 67–81, 84–86, 90, 91, 93–94, 96–101, 104, 111, 113, 121–131, 136, 140, 151–172
efficiency ................................6 8, 71, 78, 90, 94, 140, 154 microscopy .................................6 7–81, 94, 164, 169, 171
Fluorescent beads ..............................1 38, 186, 190, 197, 270 Focal adhesions (FAs) ........................................... 16–20, 233 Folate receptor ...................................1 55, 158, 160, 161, 169 Fourier space .....................................................................1 77 FRAP. See Fluorescence recovery after photobleaching
(FRAP) Frequency-domain FLIM (FD-FLIM) ......................86–102 FRET. See Fluorescence resonance energy transfer/förster
resonance energy transfer (FRET) FRET-FRAP ...........................................................1 28, 130 Full-width at half maximum (FWHM) .................. 194, 198,
200–203, 206, 208, 227, 254, 274, 283, 285 Functional imaging .......................................................3 2, 67
G
Galvanometric mirror ............................... 2 51, 252, 256, 257 Gene knock-down ..............................................................5 9 Gene silencing ....................................................................5 9 G-factor ............................................................1 65–168, 172 GFP. See Green fluorescent protein (GFP) Gradient refractive index lenses (GRIN) ......................32, 38 Grating .............. 1 79–185, 189–191, 237, 241, 251–256, 260 Green fluorescent protein (GFP) ................ 15, 16, 143–146,
158, 215, 217, 219, 227, 285 eGFP ...................................1 39–141, 144, 204–207, 209
GRIN. See Gradient refractive index lenses (GRIN)
H
Heating system .............................................................5 2, 53 HepaRG spheroid ............................................ 4 7, 48, 53–55 Hetero-FRET ..................................................................1 54 High content screening ......................................................8 8 High throughput screening ................................................6 3 Homo-FRET ...........................................................1 51–172
I
Image reconstruction ...........7 6, 184–186, 263, 270–271, 273 Immobile fraction ......................1 12, 116, 122, 124–126, 128 Instrument response function (IRF) ................ 89, 92, 95–97,
99, 100, 102 In vivo imaging ............................................... 2 6, 30, 32, 220
ADVANCED FLUORESCENCE MICROSCOPY
293
Index
L
Light, oxygen, or voltage (LOV) domain ................2 78–282, 285, 287
Light sheet-based fluorescence microscopy (LSFM) ............................................................43–56
holder ...............................................................5 1–52, 56 Living cells ..................... 6 7, 68, 83–104, 136, 148, 151, 152,
155, 169, 180, 193, 203, 216, 219, 220, 272, 277–288 Localization accuracy ....................... 2 01, 206, 270–271, 274 Localization-based super-resolution
microscopy ............................................ 2 67, 270, 271 Localization precision ...................................... 2 01, 232, 233,
240, 242, 250, 257, 258 Localized illumination ..............................................2 81, 282 LSFM. See Light sheet-based fluorescence microscopy
(LSFM)
M
Modelling ................................................. 1 09–113, 123, 130 Modulation ..........................................8 7–90, 92, 96, 97, 278 Moiré pattern ...................................................................1 76 Molecular brightness ........................................ 1 38, 140, 146 Molecular interactions ....................................... 3 6, 110, 113,
130, 131, 136, 141, 219 Molecular ruler ...................................................................6 8 Monte Carlo simulations .......................... 1 10, 113–120, 131
N
Neurolucida ........................................................................3 0 Nexin ..................................................................................7 1 Nonlinear SI .....................................................................1 87 Nonlinear structured-illumination microscopy
(NL-SIM) ....................................................186–188 Nuclear envelope .......................1 94, 195, 201, 203, 205, 209 Nuclear pore complex (NPC) ...................................193–210 Nuclear transport receptors (NTRs) ................. 195, 205, 208
O
Objective-based TIRF ..................................................2 , 7–9 Off-rates ............................................1 12, 115, 119, 122, 123 On-rates ...........................................................................1 12 Optogenetics ........................................................ 3 5, 36, 278 Organic fluorophores .................1 56, 263, 264, 266, 267, 272 Oxygen scavengers ....................................................2 33, 269
P
PA-FP. See Photoactivatable fluorescent protein (PA-FP) PALM. See Photo activated localization microscopy
(PALM) PCH. See Photon counting histogram (PCH) Perfusion chamber ..............................................................5 0 Perrin equation .................................................................1 54
Phase ....................................................4 4, 55, 61, 86, 87, 89, 90, 92, 96, 97, 178–180, 182–185, 187, 189, 194, 196, 197, 199, 200, 202, 203, 206, 216, 242, 286, 287
delay ........................................................................8 9, 90 Phasor plot ........................................9 0, 96, 97, 99, 100, 102 Photoactivatable .......................................................2 77–288 Photoactivatable fluorescent protein
(PA-FP) ......... 232–234, 236, 238–242, 247–258, 263 Photoactivatable Rac (PA-Rac) ................................278–288 Photoactivated localization microscopy
(PALM) .................................151, 170, 193, 231–260 Photobleaching ................................. 3 , 13, 14, 26, 43, 44, 80,
103, 104, 109–111, 115, 120–123, 127, 128, 139, 140, 143–145, 148, 149, 170, 188, 217
Photoconversion ...............................................................2 80 Photomanipulation ...................................................2 85–286 Photomultiplier tube (PMT) ........................... 27–29, 72, 80,
88, 91, 136–137, 144, 217, 218 Photon counting histogram (PCH) ..........................136, 146 Photoselection ..........................................................1 53, 154 Photoswitchable fluorescent probe ...................................2 31 Photoswitchable fluorophores ..................................2 63, 264 Photoswitchable molecules ....................... 1 87, 232, 257, 258 Photoswitches ...........................................1 87, 231, 232, 257,
258, 263, 264, 266–267, 272 PIE. See Pulsed interleaved excitation (PIE) 4Pi microscopy .........................................................1 93–210
type A .................................................................1 94–198 type B .........................................................................1 94 type C .................................................................1 94, 197
PMT. See Photomultiplier tube (PMT) Point spread function (PSF) .............................. 11, 194, 196,
198, 200–203, 206, 209, 232, 240, 242–247, 249, 257, 264, 270, 271, 273, 274
Polarization .................................................. 6 , 152–154, 162, 163, 168, 172, 180–185, 250
Polar plot ............................................................................9 0 Prism-based TIRF ............................................................7 –9 Protein-protein interactions ..........6 7, 83–104, 151, 152, 278 PSF. See Point spread function (PSF) Pulsed interleaved excitation (PIE) .................. 138, 140–142
Q
Quantum dots ........................................... 2 5, 138, 237, 247, 253, 254, 257, 259, 260
Quantum yield ..................... 8 5, 138, 152, 164, 219, 220, 281
R
Rac ...........................................................................2 77–288 Rac1 .................................................................................279 Ratio imaging ...............................................................6 8, 72 Reciprocal space ............................................... 1 77, 178, 191 Redox-induced photoswitching ........................................2 66
294 ADVANCED FLUORESCENCE MICROSCOPY
Index
Reducing agents ....................................... 2 33, 264, 266, 267 Residence time ..........................1 12, 116, 122–126, 137, 138 Resolution limit ...................................2 6, 176, 177, 188, 193 RNA interference (RNAi) .................................... 59–65, 277
screening .................................................................5 9–65 Rotational diffusion ..................................................1 53, 154
S
Sensitized emission ..........................6 7, 68, 70, 71, 73–77, 86 SIM. See Structured-illumination microscopy (SIM) Simulation ................................................................1 09–131 Single molecule imaging .................................. 2 31, 232, 241,
245, 258, 271 Single-molecule localization ..................... 2 40, 263, 270–274 Single plane illumination microscope
(SPIM) ..................................................... 44, 45, 137 Sinusoidal illumination patterns .......................................1 76 siRNA ..........................................................................6 0–64 Spatial light modulator (SLM) .......................... 37, 180, 184,
185, 242, 274, 281 Spatial resolution ........................................2 6, 31, 35, 68, 77,
83, 170, 177, 193, 216, 231, 274 Spectrum
emission .................................6 8, 138, 152, 153, 214, 219 excitation ....................................................................2 09
Spheroid formation ......................................................4 3–56 SPIM. See Single plane illumination microscope (SPIM) Steady-state anisotropy ..................................... 1 63, 166, 169 Stimulated emission depletion (STED) .................... 26, 137,
151, 193, 208, 213–228 Stochastic activation .........................................................2 63 Stochastic optical reconstruction microscopy
(STORM) ............................................ 193, 263–274 Structured-illumination .............1 76–179, 181, 183, 187–189 Structured-illumination microscopy (SIM) ...... 151, 175–191 Super-resolution .......................................1 51, 170, 195, 208,
216, 221, 231–233, 248, 263, 264, 267, 270–272
Super-resolution microscopy .............................. 2 6, 264, 268 Synthetic dyes ..................................................... 3 0, 232, 233
T
TCSPC FLIM. See Time-correlated single photon counting (TCSPC) FLIM
TD FLIM. See Time-domain (TD) FLIM Temporal focusing ................................... 2 36, 237, 250–257,
259, 260 Three-dimensional cell cultures ..........................................4 4 3D microscopy ..........................................................2 31–260 Time-correlated single photon counting (TCSPC)
FLIM .................................................. 8 8–92, 94–103 Time-domain (TD) FLIM ..........................................8 6–89 Time-resolved anisotropy ..................................... 8 8, 89, 169 Topographic analysis ........................................................2 05 Total internal reflection fluorescence (TIRF)
objective-based .........................................................8 –10 prism-based ..............................................................7 –10
Tumor spheroids .................................................................4 4 Turnover ......................................1 7, 18, 20, 26, 33, 128, 170 Two-photon (2-photon) .........................25–38, 84, 194–198,
203–206, 209, 216, 236–237, 248–257 Two-photon excitation (2PE) ......................... 25–38, 84, 91,
92, 94–102, 194, 195, 197, 198, 203, 205, 216 Two-photon illumination .........................................2 53, 257
U
Uncaging ................................................................3 0, 35–37
V
Variable angle TIRF microscopy (VA-TIRFM) .......4, 10–11 Voltage sensing proteins .....................................................3 5<