microscopy now update: getting the most from your imaging · tips and tweaks for optimal...

A Sponsored Supplement to Science

Sponsored by Produced by the

Science/AAAS Custom Publishing Office

Now with additional information on superresolution microscopy

Microscopy Now Update: Getting the Most from your Imaging

www.gelifesciences.com

www.sciencemag.org

TABLE OF CONTENTS

1

Microscopy Noww Getting the Most fromYour Imaging

Introductions 2 Are your samples small, dim, or live? The cell analysis team at GE Healthcare

3 Spoiled for choice? Sean Sanders, Ph.D. Science/AAAS

White papers 4 Finding the right tool for the job: Confocal vs. widefield vs. deconvolution 7 Making your microscope work for you: Tips and tweaks for optimal performance

10 Deconvolution: Bringing clarity to widefield microscopy

13 Top Tips from the superresolution microscopy pros

Webinar: questions & answers18 Microscopy in focus: The art and science of image quality

20 Live cell imaging: The future for discoveries

22 Generating the best superresolution microscopy data: Finding the right tool for the right job

25 Science webinar series links

26 Technical notes

Imaging papers28 Light microscopy techniques for live cell imaging David J. Stephens and Victoria J. Allan

33 FAM123A binds to microtubules and inhibits the guanine nucleotide exchange factor ARHGEF2 to decrease actomyosin contractility Priscila F. Siesser, Marta Motolese, Matthew P. Walker et al.

47 Apical abscission alters cell polarity and dismantles the primary cilium during neurogenesis Raman M. Das and Kate G. Storey

52 Subdiffraction multicolor imaging of the nuclear periphery with 3D structured illumination microscopy Lothar Schermelleh, Peter M. Carlton, Sebastian Haase et al.

56 Imaging intracellular fluorescent proteins at nanometer resolution Eric Betzig, George H. Patterson, Rachid Sougrat et al.

61 Lattice light-sheet microscopy: Imaging molecules to embryos at high spatiotemporal resolution Bi-Chang Chen, Wesley R. Legant, Kai Wang et al.

62 Nanoscopy in a living mouse brain Sebastian Berning, Katrin I. Willig, Heinz Steffens et al.

63 Composition of isolated synaptic boutons reveals the amounts of vesicle trafficking proteins Benjamin G. Wilhelm, Sunit Mandad, Sven Truckenbrodt et al.

68 Extended-resolution structured illumination imaging of endocytic and cystoskeleton dynamics Dong Li, Lin Shao, Bi-Chang Chen, Xi Zhang et al.

BILL MORAN, GLOBAL DIRECTOR Custom [email protected]+1-202-326-6438

ROGER GONCALVES, SALES MANAGERCustom Publishing Europe and Middle East+41 43 [email protected]

© 2015 by The American Association for the Advancement of Science.All rights reserved. 2 October 2015

About the cover: Mitotic spindle in a PTK1 cell stained for tubulin (green) and Ncd80 (red). Image taken on DeltaVision OMX.Credit: Keith DeLuca, Colorado State University

This booklet was produced by the Science/AAAS Custom Publishing Office and supported by GE Healthcare.

Editor: Sean Sanders, Ph.D.Proofreaders/Copyeditors: Yuse Lajiminmuhip; Bob FrenchDesigner: Amy Hardcastle

http://www.sciencemag.org/search?author1=Raman+M.+Das&sortspec=date&submit=Submit

http://www.sciencemag.org/search?author1=Kate+G.+Storey&sortspec=date&submit=Submit

It was not too long ago that if your experiments called for micros-copy, your biggest decision was what magnification to use on your light microscope, or whether to splurge on an oil objective. These days, things are very different. Apart from the latest light

microscopes, complete with state-of-the-art LED illumination technol-ogy, there are a dizzying array of fluorescent microscopes with an alphabet soup of acronyms, including SPIM, STED, TIRF, and PALM, to name a few. There are systems optimized for live-cell imaging and others for looking deep into tissues. Some are automated and can count specific cell types or can image an insect larva as it develops in real time.

For the average biologist wishing to include imaging in their protocols—whether basic fixed-section imaging or advanced super-resolution with live cells—it is challenging to know which system or technology will serve them best. In particular, should they use wide-field or confocal microscopy, and what role might deconvolution play in allowing them to improve their image quality? Furthermore, once they have a system in place, what steps can they take to get the best images and data from their microscope? And what are the pitfalls they should be looking out for?

In this ebooklet we present a range of articles and information that we hope will help you, the researcher, choose a system that best suits your needs and apply it to generate the best data possible. We start with four white papers that promote a deeper understanding of the tools and techniques used in microscopy. They discuss a range of topics including the differences, similarities, and pros and cons of widefield vs. confocal vs. deconvolution microscopy, and take an in-depth look at deconvolution as an evolving tool. Also covered are tips, tweaks, and best practices for setting up your microscope, and advice from microscopy experts on optimizing your superresolution microscopy experiments.

Next, we include a selection of questions and answers from three of our past webinars, all dealing with various imaging modalities, in which our key opinion leaders respond to questions submitted by the online audience.

Finally, we have chosen a series of seminal papers from Science and Science Signaling that exemplify what can be achieved using both widefield, confocal, and superresolution microscopy. All of the work described has been made possible by recent advances in imag-ing technology and software.

We trust that you will find this ebooklet helpful and can use it as a reference guide for your forays into microscopy now and in the future.

Sean Sanders, Ph.D.Editor, Custom PublishingScience/AAAS

INTRODUCTIONSMICROSCOP Y NOW: GET TING THE MOST FROM YOUR IMAGING

Are your samples small, dim, or live?

Spoiled for choice?

L ight microscopy has been used for studying cells for many years and has advanced our understanding of key cellular processes. However, fixation involves non-physiological procedures and only provides a snapshot view of cells at a

single point in time. To truly understand cellular function, we need to extend our imaging capabilities in ways that enable us to fol-low sequential events in real time, monitor the kinetics of dynamic processes, and record sensitive or transient events. With the advent of live cell imaging and the development of high- and superresolu-tion technologies, it is now possible to acquire data on viable cells in a biologically relevant context providing us with a greater insight of cellular function than has previously been possible.

GE has long been involved in the imaging of live “samples.” The discovery of X-rays in 1895 set in motion a period in which the imag-ing of the human body became associated with GE’s brand of innova-tion. In the 1970s when the power of X-rays fully came into its own, GE became the first X-ray equipment leader to make a strong move in computed tomography. In 1983, GE Medical started investing heavily in magnetic resonance imaging technology. Today, GE Healthcare is recognized as one of the worlds’ most innovative makers of medical-imaging machines, magnetic resonance imaging, and cardiac tomog-raphy scanners for faster, clearer imaging of the human body.

Our aim is that this ebooklet will help you get the most from your cellular imaging particularly when your samples are small, dim, or live. And why have we focused on small, dim, and live? Because by enabling you to image at the limits of detection our goal is to help you generate data that is most likely to advance our knowledge of living processes.

We hope this collection of white papers, peer-reviewed papers, webinars, and Q&As will make at least a small difference to your research.

The cell analysis team at GE Healthcare

3SCIENCE sciencemag.org2 sciencemag.org SCIENCE

GE Healthcarewww.gelifesciences.com

GE and GE monogram are trademarks of General Electric Company.© 2015 General Electric Company—All rights reserved. First published Apr. 2015GE Healthcare UK Ltd, Amersham Place, Little Chalfont, Buckinghamshire, HP7 9NA, UK

K15082 04/2015

5SCIENCE sciencemag.org

SECTION ONE | WHITE PAPERS

4 sciencemag.org SCIENCE

MICROSCOP Y NOW UPDATE: GET TING THE MOST FROM YOUR IMAGING

A sk a typical cell biologist to name the best overall microscope for fluorescent microscopy applications, and chances are they’ll say a

confocal microscope. Certainly, confocal microscopy has been a boon

for cell biologists. The technology promises far sharper images of certain biological samples than routine widefield microscopy can deliver, as well as the ability to create and explore three-dimensional (3-D) renderings of those samples. For many relatively thick, relatively bright, fixed samples—in other words, samples with significant light scattering—confocal may be the ideal microscopy solution.

But in the world of imaging, there is no one-size-fits all solution, no option that is the best in all situations. “In microscopy, as in life, everything is a series of tradeoffs,” says Jason Swedlow, professor of quantitative cell biology at the University of Dundee. For instance, there are plenty of scenarios in which widefield microscopy delivers better results. It certainly delivers them faster and for less money. And despite what you may have heard, confocal does not have a monopoly on 3-D imagery, provided the microscopy system features automated focusing and the appropriate software.

Here, we compare the mechanisms, strengths and weaknesses, and applications for confocal, widefield, and deconvolution microscopy.

Confocal vs. widefield Confocal microscopy solves a very specific

problem in cell biology. In widefield fluorescence microscopy, the entire field of view is evenly bathed in excitation light. As a result, every fluorophore in the field of view fluoresces, sending its emission light towards the detector—typically an area detector of the types found in a digital camera, such as a charge-coupled device (CCD), cooled electron-multiplying

CCD, or complementary metal oxide silicon (CMOS) array.

The problem is, as the user focuses up and down through the sample (i.e., in the z direction, or along the optical axis) emission light from the focal plane competes with fluorescence originating from outside the plane, creating a confused, low-contrast image, like trying to see the stars in a city overcome by light pollution. This is especially true when imaging thicker samples (those thicker than about 15 to 30 µm) and those with a substantial fluorescence signal both within the cells and in the extracellular matrix.

Confocal microscopy addresses these problems by inserting small apertures, usually pinholes, in front of the light source and detector that are in the same conjugate plane as—that is, “confocal to”—a thin focal plane in the sample. Only a diffraction-limited volume of the sample is illuminated in this way, and only light from that focal plane reaches the detector, because the pinhole blocks most out-of-focus light. The result is a sharp, high-contrast image of a distinct optical plane. But only under certain circumstances is the confocal image better resolved than a comparable widefield one—confocal microscopes are still constrained by the same diffraction limit of light as any other conventional (i.e., non-superresolution) microscope.

By stepping the image plane up and down in the axial direction, a set of virtual optical sections (a “z-stack”) can be acquired and stored. These sections can be perused like a cellular “flipbook” to identify the most informative scans, or computationally reassembled into a 3-D representation, providing researchers the unparalleled opportunity to explore their sample’s internal structure.

To achieve that performance, confocal microscopes use a fundamentally different excitation and detection scheme than widefield systems. Rather than uniformly

illuminating the sample, a pinhole-sized point of excitation energy is projected into the sample. In the popular point-scanning confocal microscope configuration, fluorescence from only this single point reaches the detector, which in this case is usually a photomultiplier tube. Once those intensity data are collected, the point is translated to a new position and the next data point is captured, rasterizing like the electron gun in an old tube television. As a result, a confocal image cannot be viewed in real time through the microscope eyepieces; it is captured point-by-point and computationally reconstructed after the fact.

Confocal microscopes thus trade efficiency for clarity. The pinholes block most of the emission sig-nal, for instance, so relatively little reaches the detec-tor. Also, says Swedlow, confocal microscopes often have less time to dwell on any given point, in order to capture an entire frame in a reasonable time. To compensate, confocals typically use high-intensity laser excitation. But this can damage live cells and photobleach fluorescent molecules, precluding long-term imaging, especially of live cells. And, because the image is collected one point at a time, the confocal frame rate is typically lower than with widefield microscopy, complicating the study and analysis of fast biological processes.

Alternative confocal configurations can miti-gate this frame-rate problem to some extent. In a spinning-disk or multipoint-scanning confocal microscope, excitation and emission light passes through multiple holes in a spinning disk (called a “Nipkow disk”), allowing for the detection of multiple points simultaneously on a CCD camera or similar detector. A line-scanning confocal microscope im-ages through a slit aperture rather than a pinhole to achieve the same end.

DeconvolutionThere are other approaches besides confocal mi-

croscopy for cleaning up out-of-focus light. Total in-ternal reflection fluorescence (TIRF) microscopy, for instance, reduces background by confining fluores-cent excitation to a narrow sliver of tissue (about 100 nm thick) at the slide/sample interface. Thus, it typi-cally is used to image membrane-localized events. Multiphoton instruments eliminate out of focus light

by confining fluorescence to the optical sections, and light-sheet microscopes illuminate sample planes one at a time with relatively low-intensity light, imag-ing in an orthogonal direction.

But these microscopy configurations, including confocal, tend to be expensive, and often are found only in well-furnished core facilities. A less expen-sive, effective, and more readily available option for many users is widefield microscopy, especially when paired with the computational process called decon-volution.

As in every microscope, the light in a fluorescent widefield microscope passes from the excitation source, through the sample, filters, and lenses, to the detector. If every element in the light path were optically “perfect,” the projected image would match what the sample actually looks like—a spherical bead, for instance, would appear as a perfect sphere. But should there be any imperfections or light degrada-tion in the optical path, the image will be distorted. The sphere would appear, say, slightly flattened.

In practice, no microscope is perfect, and some distortion is inevitable. But it is possible to create a mathematical model of the microscope’s opti-cal characteristics using the instrument’s so-called “point spread function” (PSF), which describes the behavior of an infinitely small fluorescent point. Deconvolution uses this model to back-calculate the original appearance of the imaged field of view. Effectively, the process treats an image as an array of point spread functions and attempts to reassign every point of light back into its correct focal plane, mathematically erasing the out-of-focus noise to cre-ate a clearer image.

Deconvolution thus cleans up a preexisting image by digitally reconstructing it. By contrast, confo-cal microscopes filter that noise from their datasets at the source by using a physical barrier. Yet the net result is often the same or at least similar, and indeed, some researchers apply deconvolution to confocal sections, as well. But that also means that widefield microscopes, like confocal microscopes, are capable of generating clear 3-D datasets, at least for relatively thin samples, such as a layer of cells on a microscope coverslip. All that is required is that the microscope support sufficiently fine automated focus control to collect the necessary z-stack.

Finding the right tool for the jobConfocal vs. widefield vs. deconvolutionBy Jeffrey M. Perkel

continued>

Confocal microscopy solves a very specific problem in cell biology.

SECTION ONE | WHITE PAPERSMICROSCOP Y NOW: GET TING THE MOST FROM YOUR IMAGING

IMA

GE

: ©

DR

IMA

FIL

M/S

HU

TT

ER

ST

OC

K.C

OM





Live cell imagingThe issue of which microscopy modality to use

is particularly significant when it comes to research involving live cell imaging. For one thing, live cells, by definition, are not fixed. Thus, their internal com-

ponents—the very objects users may wish to capture—are often moving as they are being imaged. Cells and their subcellular components also are often acutely sensitive to light, and overexposure can lead to cellu-

lar damage, a phenomenon called photodamage or phototoxicity. Photodamaged cells may stop divid-ing or otherwise alter their biological behavior, nega-tively impacting experimental analysis and interpre-tation. Researchers try to minimize such damage as much as possible.

Live cell imaging therefore presents particular challenges for microscopists. Since the cells are not fixed, fluorescent staining options are limited by the ability of the fluors to get through the cell membrane in order to label internal structures. The advent of genetically encoded fluorophores, such as fluores-cent proteins has greatly enhanced the options for live cell imaging. An additional challenge is that the fluorophores must be imaged quickly, and with as little excitation energy as possible, especially if the goal is to record the cells’ behavior over extended periods.

These challenges do not preclude confocal microscopy, of course—it all depends on what the user is trying to observe. For relatively bright objects and relatively slow cellular processes (such as mitosis), confocal microscopy will work just fine, especially multipoint or line-scanning confocal microscopy. But for dimmer objects and faster processes like vesicle trafficking, widefield still may be preferred. “Anything that reduces the amount of light in live cell imaging, is something you want to do,” says Swedlow.

Indeed, widefield, or widefield plus deconvolution, offers several advantages over confocal microscopy for live cell imaging. The overall input energy is lower and distributed over the entire field of view, mean-ing the per-cell energy input may be lower. And the frame rate is faster in widefield microscopy, meaning users are more likely to capture rapid cellular pro-cesses.

Widefield microscopy also offers the advantage that it captures all of the sample’s light output for later deconvolution, whereas confocal microscopes reject any light that is out of focus, meaning there are fewer photons with which to work in the first place. Widefield also captures those images using CCD or similar detectors, which typically have higher quantum efficiencies than the point detectors used in confocal instruments. As a result, especially for weakly stained or sparse fluorescent objects, confo-cal microscopy can be akin to taking a picture in a darkened room with a cell phone camera—grainy and littered with electronic noise.

Confocal microscope users typically circumvent this problem with fixed samples by sampling each point multiple times (effectively lengthening the exposure), which boosts signal-to-noise ratios, but is also slower, more phototoxic to live samples, and more likely induce photobleaching. For live cell imag-ing, confocal users often opt to open the pinhole ap-erture to allow in more light, defeating the purpose of confocal microscopy in the first place. Instead, they are left with effectively a widefield image collected at a slower, more phototoxic pace.

The bottom line is that, as with most techniques in the lab, confocal and widefield technologies are com-plementary rather than competing and form part of a complete imaging toolbox. For many projects and applications, particularly when the sample is thick or bright, users may be well served by a powerful confo-cal instrument. Just remember that there are other options. Particularly in circumstances when you have live cells or dim samples, you may be better off with a properly outfitted widefield microscope. And there’s a good chance you already have one available.

Jeffrey M. Perkel is a freelance science writer based in Pocatello, Idaho.

R esearch-grade microscopes represent hefty investments. These are precision instruments with tightly calibrated components. Yet they

are only as good as the researchers who use them, the procedures they adhere to, and the quality of the sam-ples they study. Just as a sports car can only achieve its potential if properly maintained, a microscope will produce substandard data if it isn’t treated with care.

Given a top-notch sample, even a mediocre micro-scope can produce quality data. But place a poorly prepared sample on the stage, and the best micro-scope in the world will be of little help. As the saying goes, “garbage in, garbage out.”

But a successful microscopy experiment is about more than just sample prep. Will you image live cells or fixed cells? How will you stain them, and what will you grow them on? What objectives will you use to image your cells, and what tradeoffs does that deci-sion represent?

Here we review a few of the many variables researchers need to consider when planning their microscopy work in order to get the most from their imaging.

Coverslip choiceLike your objectives, filters, and light source, the

coverslip is part of the optical path that produces your image. As such, its properties—and in particular, its thickness—can have a tremendous influence on image quality.

The coverslip most typically used in biological microscopy is the #1.5 coverslip, which is around 170 μm thick. Coverslip type #1 are a tad thinner at about 150 μm, while #2 coverslips are thicker (220 μm).

That thickness matters because as light transitions from one medium to another—from air, through the coverslip, through the sample, then back into the air and into the objective—it diffracts. Modern

microscopes typically are designed to account for a certain thickness of glass between the objective and the sample, and can correct for this diffraction.

Typically, the expected amount of glass is 170 μm—that is, the objective “expects” the sample to be mounted using a #1.5 coverslip. If a #1 coverslip is used instead, image quality could suffer, especially if the sample’s refractive index (RI) does not match that of the glass itself (RI ~1.52), or when a live, thick sample is imaged in culture medium (RI ~1.33). If a sample is fixed and embedded in plastic or resin, which has approximately the same RI as glass, this is less of an issue. In that case, the sample is effectively an extension of the glass itself.

Though the nominal thickness of a #1.5 coverslip is 170 μm, coverslips actually have a range of thicknesses, anywhere from 160 to 190 μm. In order to obtain sharper images and more reliable, comparable data, users can purchase coverslips with tighter tolerances (that is, closer to 170 μm). Or, they can use an objective with an “adjustment collar,” which allows the image quality to be tweaked by correcting for coverslip thickness.

Choose your immersion mediaAs light passes from the sample towards the

objective, the media it travels through dictate how much of it can be captured and how much will be lost. For instance, light passing from a glass coverslip into air diffracts away from the optical axis, effectively distorting the image and reducing the amount of light reaching the detectors. Microscope manufacturers have developed immersion media and immersion objectives intended to mitigate these effects.

Five types of objectives are commonly used in bio-logical research: dry, water-immersion, oil-immersion, glycerin-immersion, and water-dipping objectives. Dry objectives are used in air. Immersion lenses

Making your microscope work for you Tips and tweaks for optimal performanceBy Jeffrey M. Perkel

Confocal and widefield technologies are complementary rather than competing and form part of a complete imaging toolbox.

Total internal reflection

fluorescence (TIRF)

microscopy reduces

background by confining

fluorescent excitation to

a narrow sliver of tissue.


continued>IMA

GE

: ©

VSH

IVK

OV

A/S

HU

TT

ER

ST

OC

K.C

OM





are intended to have a droplet of liquid between the sample and the front lens of the objective. Water-dip-ping objectives are designed to allow the objective to be dipped into culture media or buffer without being damaged by the salts in the media.

Immersion media effectively reduce diffraction by erasing RI differences as light travels from the sample into the objective. Oil, for instance, has an RI (~1.52) very similar to glass. Thus, an oil immersion lens ef-fectively couples the glass coverslip and glass objec-tive in a block of uniform material. Glycerin has an RI of ~1.47, making it compatible with some common sample mounting media, while water’s RI is ~1.33.

Matching the immersion media to the sample is critical for microscopy. For instance, an oil-immersion objective imaging into an aqueous media (e.g., cul-ture media) will create spherical aberrations due to a mismatch in RIs. Spherical aberration is an optical distortion that causes objects to appear stretched or compressed, and which also results in an apparent loss of photons.

According to a microscopy primer published in 2006 in the Journal of Cell Biology, “Spherical aber-ration describes the phenomenon whereby light rays passing through the lens at different distances from its center are focused to different positions in the z axis. It is the major cause of the loss in signal intensity and resolution with increasing focus depth through thick specimens” (1).

Among other advice for reducing spherical aber-ration, the primer suggests mounting the specimen “on or as close to the coverslip as possible,” as well as using the correct coverslip and avoiding air bubbles in the immersion medium.

Microscopy tools vendors offer multiple immer-sion media to satisfy a range of experimental condi-tions. To help identify the best medium for a given microscopy experiment, GE Healthcare Life Sciences has developed an Immersion Oil Calculator web and smartphone application (2). Simply supply the objec-tive working distance, coverslip thickness, specimen RI, and distance from coverslip to specimen, as well as the working temperature and wavelength used, and the calculator will suggest the optimal RI of the im-mersion oil that will yield the best performance.

For instance, using a 60x/1.42 NA Oil PlanApoN objective lens with a 0.15 mm working distance, a #1.5

coverslip, a specimen in glycerol, a sample directly on top of the coverslip, and imaging at 37°C with 550 nm excitation light, the calculator suggests an immersion medium RI of 1.521. (NA, or numerical aperture, describes an objective’s “light-gathering ability,” and is also crucial in minimizing spherical aberration.)

Once you’ve selected the proper medium and ob-jective, be sure to protect your hardware investment. Beginning microscopists tend to overdo it on the oil or glycerin, which can gum up the objective’s delicate parts. For instance, modern lenses are spring-loaded to prevent the lens from punching through a sample. Excess oil can collect in that mechanism, potentially ruining the objective. Similarly, water-immersion ob-jectives, if not properly cared for, can be fouled by dried salt crystals from culture media. Media can also contaminate the body of the microscope itself if there are open slots in the objective turret, an even more expensive repair.

One quick and simple method for keeping your microscope safe is to wrap the objective barrel with an inexpensive cloth hair “scrunchie,” which can catch overflowing oil or medium. It’s far easier to replace that, than the objective itself.

Go liveLive cell imaging is perhaps one of the most re-

warding and challenging of microscopy applications. Rather than viewing a frozen moment in cellular time and trying to work out the events that preceded and followed that moment, live cell imaging provides re-searchers with a real time, video record. Yet collecting reliable data requires some careful attention to detail.

First is the question of how you will visualize the action. Intracellular staining options are limited in live cell work, as few dyes can cross the cellular mem-brane. One option is to image live cells without stain-ing them, for instance in transmitted-light differential interference contrast (DIC) mode. Or, to focus on spe-cific molecules, researchers can genetically tag them with a fluorescent protein. (Often, researchers image in both fluorescence and DIC modes, and overlay the two images.)

Genetic tagging requires molecular engineering to make the cells express the protein, and possibly some tinkering, too, as it is necessary to strike a balance: Too little protein, and you may not see

what you’re looking for; too much, and you may end up with a noisy image and possibly poison the cell, or divert it too far from its normal tasks, potentially skewing your data.

Live cell imaging also requires balancing imaging with cell behavior. You need to use sufficient excita-tion energy to efficiently stimulate fluorescence, for instance, but not so much as to induce phototoxicity or photobleaching. And you need to ensure a suf-ficiently fast frame rate to capture the behavior you hope to document.

If you hope to image the cells for an extended period of time, you have another concern: Keeping the cells alive and healthy. Eukaryotic cells are usu-ally grown in an incubator at 37°C and fixed carbon dioxide concentration. If you plan to image them for, say, 30 minutes, it’s probably enough to maintain their temperature on the microscope stage; otherwise, the cells could experience a temperature shock that alters their behavior.

For longer imaging experiments, other conditions come into play, such as humidity, CO2 concentration, and gas exchange—all of which can usually be con-trolled using an environmental control chamber. This adds additional cost and complication to your experi-mental setup, but will be worthwhile for maintaining the optimal environment for normal cell behavior and for obtaining accurate and meaningful results.

Correct your colorsWhether performing live- or fixed-cell microscopy,

researchers increasingly are imaging multiple fluorescent colors. But as anyone who has used a prism knows, light of different colors bends to different degrees in glass. As a result, green light and red light will focus to different locations, making it difficult to determine, for instance, whether two proteins are coincident, or lie on opposite sides of a membrane. This is especially true in the z (optical) axis, causing two coincident points to appear in separate optical sections.

Apochromatic lenses are color-corrected to account for these differences. Of course, researchers imaging just a single color (for instance, green fluorescent protein) can utilize uncorrected objectives, as they don’t need to worry about registering different images. But as most researchers use at least two

colors, often overlaying, say, GFP expression with DAPI nuclear staining, apochromatic objectives are typically key components of the microscopy toolkit.

It is important to note that apochromatic lenses tend to be less efficient light collectors than their uncorrected alternatives. Another key consideration is that color-corrected lenses tend only to be corrected for specific wavelengths. If, say, your objective is optimized for red light at 620 nm, but your fluorescent dye produces its strongest emission at 650 nm, then for that particular experiment the correction is only partial.

Another tool for ensuring color alignment is multi-colored microspheres. These beads contain a shallow layer of fluorescent dye that, when optically sectioned using confocal microscopy, produces a thin ring of fluorescence. In a non-color-corrected microscope, these beads appear as a set of non-overlapping col-ored rings when imaged at different wavelengths. But by adjusting the microscope alignment, researchers can correct for that issue, thereby ensuring their data will be in proper register.

Color dispersion and spherical aberrations can both influence microscopy operation, and both can be influenced by immersion media. Generally speak-ing, researchers can optimize one or the other, but not both. If absolute color alignment is key—for instance, for colocalization studies—pick an oil that works best for that purpose. On the other hand, if every photon counts, for instance under low-light, live cell imaging conditions, an oil optimized to correct spherical aber-ration may be a better choice.

Bottom line: Ask your microscope manufacturer for the precise specifications of your hardware and re-agents so that you can get as close as possible to the optimal conditions. And be prepared for some optimi-zation. It might take time to get everything operating just right. But it will all be worth it, once you see the beautiful images that emerge.

1. A. J. North, “Seeing is believing? A beginners’ guide to practical pitfalls in image acquisition,” J. Cell Biol. 172, 9–18 (2006).2. www.gelifesciences.com/oilcalculator


Live cell imaging is perhaps one of the most rewarding and challenging of microscopy applications.


IMA

GE

: ©

VSH

IVK

OV

A/S

HU

TT

ER

ST

OC

K.C

OM

MICROSCOP Y NOW: GET TING THE MOST FROM YOUR IMAGING

http://www.gelifesciences.com/oilcalculator

It is true that confocal microscopy can easily collect high-contrast data, but it’s also true that another, equal-ly accessible approach can be used to reduce blur and restore contrast: deconvolution.

Deconvolution defined Deconvolution is so-called because it is the reverse

of convolution, the optical process by which light is blurred by microscope optics when producing an image. More specifically, deconvolution is a form of image processing in which the blurred signal in a 3-D image stack is erased or reassigned to correct blurring introduced by the optics.

Imagine a tiny (i.e., subresolution, ~100 nm) fluo-rescent bead on a microscope slide. Ideally, that bead would appear as a discrete dot with no signal above or below the focal plane. In reality, it will appear as a se-ries of concentric rings in the x-y plane and as an hour-glass shape in the axial direction (x-z or y-z) (see images below). The latter is called the “point spread function” (PSF) and is a representation of how light blurs in a mi-croscope system.

Such blurring may not pose much of a problem when imaging a single bead or an object with a distinctive structure. But in a complex biological sample, every single fluorescent molecule acts as a point source of light with its own PSF, all of which will overlap and interact with each other, resulting in a confused, or blurred, image.


The images captured by even top-of-the-line microscopes do not reflect biological samples as they truly are, but rather are slightly distorted

representations of those samples. That’s because light doesn’t simply pass from the sample directly to the eyepieces or camera—it interacts with biological mate-rials in the sample itself as well as the lenses, mirrors, filters, and other optical elements of the microscope, all of which can distort the image of the object.

One of these distortions is blur. As detailed in a 2001 review, blur is “a nonrandom spreading of light caused by its passage through the imaging system and lenses” (1).

Blur is most acutely visible in images from widefield microscopes, in which samples are uniformly bathed in light through a three-dimensional (3-D) volume, and fluorescent molecules throughout the sample emit light simultaneously. An area detector, such as a charge-coupled device (CCD), electron multiplying CCD, or scientific complementary metal-oxide semi-conductor (sCMOS) camera, captures light from the focal plane. It also detects blurred light from planes above and below, which degrade the contrast of the raw image.

Some researchers choose laser-scanning confo-cal microscopy (LSCM) as an alternative. In LSCM, samples are illuminated point by point with a diffrac-tion-limited patch of light and read out as a series of intensities using a point detector, such as a photomul-tiplier tube (PMT). Out-of-focus light is rejected using pinhole apertures that, in ideal circumstances, allow only in-focus light to reach the detector. Spinning disk confocal systems take a hybrid approach using multiple pinholes in parallel with fast, sensitive camera detection. Neither confocal method fully eliminates blur, but contrast and optical sectioning are improved. Yet these improvements come at the cost of speed (over a large field of view) for point scanners and high noise in the signal for both confocal methods.

Because the PSF can be measured by imaging sub-resolution beads, this measurement can be used to reverse or remove the blur from experimental images during deconvolution, creating an image that more accurately represents the actual sample.

However, the accuracy of the deconvolved data is only as good as the deconvolution algorithm itself. Several image-processing methods have been de-veloped over the years, taking fundamentally differ-ent approaches to image correction (1). Deblurring algorithms treat an image stack as a series of two-di-mensional planes, processing each plane one by one. These deblurring algorithms can increase contrast but are highly subtractive, significantly decreasing the image signal-to-noise ratio (1). Image restoration algorithms, on the other hand, treat the image stack as a single 3-D entity in which the signal is preserved and simply reassigned from one axial (z) plane to another.

GE’s softWoRx software implements a constrained iterative image restoration deconvolution algorithm. First described in 1964 and refined two decades later, this algorithm uses the captured image to estimate, given the PSF, what the actual image looks like. It then compares that estimate to the captured image, adjusts the estimate, and repeats. Thus, it is iterative. That the algorithm is restorative refers to the fact that intensity values are neither added nor subtracted (as in deblur-ring algorithms), only reassigned. The algorithm is constrained because it disallows negative intensity values that arise during deconvolution processing (because it is not possible to have a negative fluores-cent intensity).

When applied to a fluorescence image, deconvolu-tion can reduce the blurring often seen in widefield raw data, thus dramatically increasing contrast. In practice, deconvolution can improve contrast up to an order of magnitude, especially in the axial (z) direc-tion, where blur is most severe. But importantly, the overall intensity of the image—that is, the sum of inten-sity values throughout the 3-D stack—remains essen-tially unchanged. In addition to increasing contrast, deconvolution can also slightly improve resolution, though deconvolved images, like all standard micro-scopic data, are diffraction-limited.

When to deconvolve As a general rule, most fluorescence images can

be improved with deconvolution, even those col-

lected with a confocal microscope. But the approach can be particularly beneficial when a signal is limiting. For instance, in live-cell imaging, researchers typically want to capture data using endogenous fluorescent proteins and minimal excitation energy to extend cell viability and minimize phototoxicity. Deconvolution al-lows users to collect all the photons and computation-ally sort them out later, whereas confocal approaches reject photons in an attempt to increase contrast, dis-carding potentially useful information and necessitat-ing the use of more potentially damaging light.

For instance, Jason Swedlow and his colleagues at the University of Dundee, Scotland, investigated weakly fluorescent cytoskeletal elements in Toxoplasma gondii parasites growing within fibroblast hosts (2). The team first tried to image these structures in parasites expressing yellow fluorescent protein (YFP)-tubulin using LSCM, yet were unable to resolve the structures due to high background levels. Using widefield microscopy, the team was able to detect the structures of interest, even without deconvolution. But adding deconvolution increased image contrast as much as 10-fold, producing data the researchers could use to tease apart fine structural details.

As part of this study, Swedlow and his team di-rectly compared the performance of LSCM, widefield microscopy, and widefield with deconvolution on mixtures of beads containing varying (but known) levels of fluorescence. They found that widefield microscopy alone outperformed LSCM in imaging the dimmest fluorescent beads (whose fluorescence intensity roughly matched that of the structures they hoped to image). Adding deconvolution increased the signal-to-noise ratio in this bead population, re-ducing the coefficient of variation by about one-third. All three methods performed comparably on higher-intensity beads.

They concluded that “combining [widefield microscopy] with a CCD detector and image deconvolution generated the most accurate measurement of sample fluorescence for weakly fluorescent objects and therefore might be used for quantitative analysis of images of living cells with limited signal levels.” (2).

For samples thicker than about 60 µm, or those plagued by high background or nonspecific autofluorescence, researchers may obtain superior results using confocal microscopy or one

DeconvolutionBringing clarity to widefield microscopyBy Jeffrey M. Perkel

Most fluorescence images can be improved with deconvolution.


continued>

IMA

GE

: (T

OP

) ©

VSH

IVK

OV

A/S

HU

TT

ER

ST

OC

K.C

OM


Make sure that the raw data matches that model as accurately as possible.


of the other available contrast-enhancing techniques, such as total internal reflection fluorescence (TIRF) or light-sheet microscopy.

A 2007 study provides a useful rule of thumb for helping researchers decide among widefield, point-scanning confocal microscopy, and multipoint confocal microscopy for different samples (3). Using a value comparable to a signal-to-background ratio, the study defines H, the “haziness index,” as “the ratio of background intensity to the intensity of the smallest objects of interest in the focal plane.” For H of less than about 20, the authors reported, widefield plus deconvolution represents an ideal approach. Multipoint (e.g., spinning disk) confocal microscopes excel over the range of 20–200, while point-scanning confocals are preferred for H values of 200–1,000. “Beyond H = 1,000, none of these methods is likely to be satisfactory,” the authors wrote. [See (3) for more information on calculating H for biological samples.]

A question of trustFor all its potential benefits, deconvolution is

relatively underutilized. Researchers may be reluctant to embrace deconvolution due to its reliance on mathematical models of microscope performance, which can be difficult to match to real-world conditions as well as concerns that image processing might actually introduce artifacts rather than remove them.

Yet deconvolution has a long history. In use for at least 30 years, the technique is featured in some 21,000 publications, 1,400 using the softWoRx algorithm alone.

To ensure that deconvolution is effective and alleviate any concerns about the process, it is important to do two things well: 1) select a robust model of microscope performance (the PSF), and 2) make sure that the raw data matches that model as accurately as possible. Here are some tips on how to do this:

• Check your deconvolution algorithm. Within the deconvolution computation, the PSF isn’t used directly; instead, the Fourier transform of the PSF, called the “optical transfer function” (OTF), is utilized. OTFs can be generated based on the theoretical performance of a microscope or determined from an experimentally measured PSF. It could be assumed that the theoretical OTF is the best choice for decon-volution because it represents the best performance theoretically possible, but surprisingly, when using

high numerical aperture (NA) objectives, some mi-croscopes actually perform slightly better than the theory suggests (4). Therefore, when using a theoreti-cal OTF, the algorithm may be too aggressive in its correction, reassigning the signal from real structures rather than just blurred light. This can result in de-convolved images where fine or dim structures have been processed away. For this reason, when decon-volving data acquired with high NA objectives, it is important to use empirical, or experimentally mea-sured OTFs. These tend to be more conservative than theoretical OTFs but can ensure that real data is not lost in processing.

• Choose the right oil. Once an appropriate model has been identified, the next step is to ensure that the raw data collected accurately matches that model. Selection of immersion oil whose refractive index (RI) matches the sample, for instance, can minimize spherical aberration (5), a significant source of PSF error (1) (see images below).

• Use the best quality hardware. In addition to oil selection, researchers can equip their microscopes with higher-end hardware such as high-quality objectives, uniform light sources, and ultrastable stages to improve raw data quality.

Of course, as with most techniques in the life sciences, widefield deconvolution and confocal microscopy both have strengths and weaknesses that should be understood and considered before imaging begins. As you plan your experiments, it’s a good idea to keep these variables in mind:


A lthough German microscopist Ernst Abbe postulated in 1873 that light microscopy could not surpass the diffraction-limited resolution

of 200 nanometers, superresolution microscopes are able to zoom in on objects just tens of nanometers apart. That’s enough resolution to easily see viruses and even individual proteins. This technology was so revolutionary that the 2014 Nobel Prize in Chemistry was awarded to Eric Betzig (Howard Hughes Medical Institute’s Janelia Research Campus, Ashburn, Virgina), together with Stefan W. Hell (Max Planck Institute for Biophysical Chemistry, Göttingen, Germany) and Wil-liam E. Moerner (Stanford University, Stanford, Califor-nia) for “the development of super-resolved fluores-cence microscopy.” A panel comprising Betzig, Raman Das, and Justin Taraska (see “Webinar participants”) provided advice for beginners and tips for experts on superresolution microscopy during a Science webinar, “Generating the Best Superresolution Microscopy Data: Finding the Right Tool for the Right Job,” broad-cast in July 2015.

Today’s superresolution microscopy includes sev-eral general methods, each of which has been imple-mented in specific forms (see “Superresolution sur-vey.”). The dominant ones include structured illumina-tion (SIM), single-molecule localization methods such as PALM, and point-scanning methods such as STED.

The wide variety of choices creates a complex decision for a biologist considering superresolution microscopy: What is the right tool for the scientific question at hand?

Top Tips from the Superresolution Microscopy ProsThree experts, including Nobel laureate Eric Betzig, explore the best applications and optimal conditions when using this diffraction limit–breaking technology.By Mike May

continued>IMA

GE

: (T

OP

) ©

VSH

IVK

OV

A/S

HU

TT

ER

ST

OC

K.C

OM

How thick is the sample you plan to image? Samples thinner than ~60 μm can be imaged using either technique, but thicker samples are best imaged using confocal microscopy.

How intense is your signal? In samples where the signal is high, both widefield with deconvolution and confocal microscopy will do a good job. In dim or live samples, however, widefield microscopy is probably a better choice. If you’re unsure, consider calculating the haziness index (H) for your sample.

How high is your background? Deconvolution algorithms perform best with relatively low H values (that is, high signal-to-noise), whereas confocal systems can better tolerate high autofluorescence and elevated background signals as found in thicker samples.

Are you imaging live or fixed samples? Live-cell samples typically require lower-energy excitation over extended periods, for which widefield strategies (plus deconvolution) are ideal.

If you do opt for a widefield/deconvolution strat-egy, be sure to ask your supplier about the quality of their lenses: Do they utilize theoretical or empirical OTFs during deconvolution, and how well does the algorithm OTF match the characteristics of your partic-ular system? Also, look into the details of your decon-volution algorithm to ensure that it meets your needs.

With the right combination of hardware and software, there’s no reason deconvolution shouldn’t be a key and frequently used piece of your microscopy toolbox.

1. W. Wallace, L. H. Schaefer, J. R. Swedlow, BioTechniques 31, 1076 (2001). 2. J. R. Swedlow, K. Hu, P. D. Andrews, D. S. Roos, J. M. Murray, Proc. Natl. Acad. Sci. U.S.A. 99, 2014 (2002). 3. J. M. Murray, P. L. Appleton, J. R. Swedlow, J. C. Waters, J. Microsc. 228, 390 (2007).4. Y. Hiraoka, J. W. Sedat, D. A. Agard, Biophys J. 57, 325 (1990).5. D. A. Agard, Y. Hiraoka, P. Shaw, J. W. Sedat, in Methods in Cell Biology, D. L. Taylor, Y. L Wang, Eds. (Academic Press, San Diego, 1989), vol. 30, pp. 353–377.



SIM is one of the more accessible technologies.


Success from the systemsAlthough many scientists think of superresolution

as a microscope, it actually arises from several forms of interacting technology. In fact, the most interest-ing development in the world of superresolution microscopy, according to the experts on this panel, is the creation of systems of tools that address specific biological questions. The Nobel Prize even celebrated the development of a system: fluorescence for super-resolution microscopy. As Betzig puts it, “It’s been as much a story about the chemistry, the labels, delivery, attachment strategies, and basically the interdisciplin-ary synergy of all the physical sciences to try to make this stuff more and more real.”

The years of research on methods of superreso-lution microscopy resulted in advances from labs around the world being developed into commercial products. Utilizing this form of microscopy once re-quired experience in a wide range of physics and instrumentation, and scientists could only use the technology by building their own devices in the lab. Today, commercially available systems make super-resolution microscopy accessible to a much broader array of biologists.

The commercial devices also let biologist s work with existing techniques, such as formalin-fixed, paraffin-embedded (FFPE) sections. Scientists in Das’s lab use a similar method—using paraformaldehyde instead—and they image the sections with superreso-lution microscopy. Moreover, Das and his colleagues prepare these sections as they would for traditional imaging—using the same staining procedure, section-ing, and labels—and then image them with SIM.

Despite a long tradition of using fixed cells, biolo-gists also work increasingly on imaging live cells. In that situation, anyone moving to superresolution imaging will worry about cell viability and optimizing sample preparation for microscopy. Instead of worry-ing about that just yet, biologists should first decide if they need superresolution at all.

Getting startedIn discussing superresolution imaging, the panel

recommended first assessing the needs at hand. In fact, they all encouraged exploring the limits of wide-field microscopy and standard deconvolution before even considering superresolution methods. The stan-dard tools often resolve many of the questions that biologists might think require superresolution.

Every form of imaging includes compromise. As Betzig said, “There’s always a trade-off, and the more resolution you ask for the more pain you can expect.” To keep the method as painless as possible, the panel encouraged researchers to start with the lowest-resolution method that can answer the biological question under consideration. If a standard widefield or confocal microscope supplies that an-swer, use that technology instead of superresolution.

As an example of imaging trade-offs, Betzig de-scribed a tetrahedron (three-sided pyramid) with the vertices labeled as “spatial resolution,” “speed of im-aging,” “toxicity to the cell,” and “maximum imaging depth.” Imagine a point inside that tetrahedron as a representation of an imaging method’s combination of features. If it improves one feature, say spatial reso-lution, it moves closer to that vertex and away from the others, thereby reducing its capabilities in the remaining three characteristics. To attain higher spatial resolution, for example, an imaging method requires more pixels, and that requires more measurements, which takes more time and might damage the cells by exposing them to more light.

As a biologist reviews imaging options, a range of elements must be considered. Are the samples fixed or live cells? The experiments might also include an-tibodies and several fluorophores. If traditional meth-ods do not provide the desired answer in an imaging experiment, then a biologist should start with the easi-est forms of superresolution. “SIM is one of the more accessible technologies,” Das said.

Living cellsAlthough the experts on this panel agreed that all

forms of imaging come with pros and cons, Betzig pointed out that the “big divide” comes from imaging fixed or live cells, in some cases with traditional imag-ing and with superresolution microscopy in others. In Das’s lab, the scientists use very standard deconvolu-tion-widefield microscopes for some live-cell imaging.

For the best overall superresolution imaging, Das said, “SIM seems to be the best compromise.” Das and Betzig both mentioned the benefit of SIM using tradi-tional labels, while other forms of superresolution mi-croscopy require modified biological markers. Betzig even noted his disappointment that the Nobel Prize did not recognize SIM. Unfortunately, SIM’s key devel-oper, Mats Gustafsson, passed away in 2011, and the Nobel only goes to living candidates.

When combining superresolution with live-cells, however, a biologist must carefully consider the amount of light required by the method. “The trade-off you have when you want to get super high resolution is that you’re taking thousands, sometimes millions of images,” Taraska said, “and so, that generally requires the cells [to be] dead” and the sample fixed. For live cells, a scientist should expose the cell to as little light as possible.

To give listeners an idea of the light involved, Betzig pointed out that cells evolved under light inten-sities of less than 1/10th of a watt per square centime-ter. PALM uses about four orders of magnitude more light than cells usually experience, reversible saturable optical fluorescence transition (RESOLFT) takes four to five orders of magnitude more light that PALM, and STED needs seven to eight orders of magnitude more light than RESOLFT. Of all the methods for superreso-lution microscopy, SIM requires the least light.

With any method of superresolution microscopy, biologists should use the least light the experiment requires and do their best to analyze its impact on the cells. That information should be included in articles published about the results.

For live-cell imaging, biologists also need tech-niques that provide adequate temporal resolution. Overall, widefield imaging allows faster acquisition. For a 10-micron field of view with RESOLFT or STED, imaging requires seconds to minutes. PALM takes a similar amount of time, and SIM can be much faster. Betzig said, “If it’s a brightly enough labeled specimen [as with] the endoplasmic reticulum, we can image that [with SIM] at 50 hertz at a little under 100-nano-meter resolution.” This statement reveals how the speed of acquisition must also be considered with live-cell imaging.

Depth and resolutionFor live-cell imaging, the structure of the cells

themselves makes a difference in deciding on the

Webinar participants

Eric Betzig (group leader, Howard Hughes Medical Institute’s Janelia Research Campus, Ashburn, Virgina) is corecipient of the 2014 Nobel Prize in Chemistry for his work on the superresolution technique PALM, developed during his time at Bell Laboratories.

Raman Das (research associate, University of Dundee, Dundee, United Kingdom) is focused on the cell biological mechanisms driving neurogenesis, for which he has played an instrumental role in pioneering new imaging technologies. This work, pub-lished recently in Science, led to the discovery of a new form of cell subdivision.

Justin Taraska (investigator, Laboratory of Molecular and Cel-lular Imaging, National Heart, Lung, and Blood Institute, National Institutes of Health, Bethesda, Maryland) studies the structural cell biology of exocytosis and endocytosis with advanced imaging methods, including live-cell microscopy, superresolution fluores-cence, and electron microscopy.

best technology to use. For very thin cells, a biologist might pick a widefield method like SIM. Although this method thrives at imaging depths of 5–10 microns, it can also work on samples 20 microns or more thick, including Drosophila ovaries and C. elegans embryos.

Overall, the depth possible with all fluorescent microscopy depends on the sample. The labeling density plus the transparency and refractive proper-ties of the tissue impact the imaging depth. With fluo-rophores of several colors, chromatic and spherical aberrations can occur with SIM beyond 10 microns, unless the biologist uses an immersion oil with an optimal refractive index. For example, one can use a higher refraction immersion oil to adapt a lens to work optimally at a distance of 10 microns from the cover slip. So getting more depth often means adjusting some of the parameters of the system to allow it.

To get even finer resolution, however, a biologist will move from SIM to other methods. For the high-est spatial resolution, researchers turn to single mol-ecule–localization methods such as PALM and STORM, which can resolve structures just 10 nanometers apart. As mentioned above, those methods require more light as well as certain behaviors of the fluorophores when exposed to this light. Most traditional fluores-cent labels cannot endure the high levels of light required for STED, for instance, so the probes must be modified, necessitating further optimization of sample preparation

Advances in various forms of superresolution microscopy, however, improve the spatial continued>IM

AG

E:

(TO

P)

© V

SH

IVK

OV

A/S

HU

TT

ER

ST

OC

K.C

OM


Join AAAS. Get instant access to Science. Support all of the sciences.

When you subscribe to Science, you become part of the American Association for the Advancement of Science (AAAS), a nonprofit community of more than 120,000 members worldwide who believe in the power of science to make the world a better place. AAAS is hard at work promoting science in government, schools, and in the public commons around the globe.

AAAS’s award-winning journal Science offers the top peer-reviewed research across multiple disciplines.With your subscription, you'll get:

§ 51 weeks of home delivery of Science

§ Instant online retrieval of every Science article ever published, dating back to 1880

§ Full access to the Science mobile site and apps

§ Career advice, webinars, blogs and fascinating features exclusively for AAAS members

§ Members-only newsletters, and much more

With increasing public skepticism about science–and public funding for research more uncertain than ever–our work has never been more important. Join hands with us today!

Visit promo.aaas.org/joinaaas. Together we can make a difference.


resolution of other techniques. Although current SIM provides resolution of about 100 nanometers, for example, resolution will eventually drop even lower. Recent work in Betzig’s lab imaged live cells at about 60-nanometer resolution. In comparing this work with PALM, RESOLFT, and STED, Betzig said, “If you do a side-by-side comparison, what you find is that SIM, with this extended resolution, basically gives you true spatial resolution comparable to these other methods, but with the ability also to do live imaging for a period of time—not as long as you can do with traditional SIM at 100 nanometers because you do still have to throw in more light.”

All about the biologyAs a biologist thinks about superresolution micros-

copy, the need for compromise must be remembered. In many cases, traditional techniques will be far easier to use and provide the needed information for a par-ticular experiment. When superresolution microscopy is required, many factors must be considered. The fastest speed comes from widefield techniques like SIM, but if resolution matters more than speed, then the single-molecule techniques must take over.

The playing field for superresolution microscopy keeps changing as various versions improve and gain new capabilities, however. “A lot of the methods are starting to converge, but it’s getting extremely exciting to be able to actually observe functioning cells [at a] resolution that was never possible before,” Taraska said.

Superresolution microscopy depends on an entire system of technologies. So advances in fluorescent markers and oils used in imaging also improve the capabilities of the various superresolution methods. Other tools also impact the possibilities with sophisti-cated imaging research. As an example, the function of cells will be studied even more closely as biologists combine the advanced forms of superresolution im-aging with methods of modifying genes. For instance, clustered regularly interspaced short palindromic repeats (CRISPR)/CRISPR-associated systems (Cas) make it easy for any biologists to edit genes. Then, the resulting changes can be imaged in live cells at nanoscale resolution.

Ultimately, the biology matters the most. That means carefully interpreting results, because the sam-ple preparation and the microscope create some dis-tortions. “Any biologist who wants to interpret these

images,” said Betzig, has to “think about what kind of distortions may have gone between the specimen and the image.” Only then can the results truly teach us more about the life sciences. Likewise, researchers must know as much as they can about the cells being imaged. According to Das, “You must know the biol-ogy behind your experiments.” A strong combination of biological knowledge, careful sample preparation, and advanced superresolution imaging—perhaps cou-pled with molecular-modification techniques—prom-ises to give researchers views of life in action at a once unimaginable level of detail.

In the future, imaging experts hope for even more information. “It would be really nice to be able to get the diffraction-limited or superresolution imaging of dozens to hundreds of proteins at the same time in the same cell,” stated Betzig. This is not yet a reality, but is possibly something we’ll see in the not-too-distant future.

Mike May is a publishing consultant for science and technology.

Superresolution survey Today’s catalog of methods for superresolution microscopy includes 20–30 acronyms, but the major techniques use three general approaches, which include various subforms:

1) Structured illumination microscopy (SIM): This widefield method illuminates a sample with a high-frequency pattern of stripes, and the resulting emission from fluroescent labels can be restructured for twice the resoltuion of the diffraction limit or better. Saturated structured illumination microscopy (SSIM) adds nonlinear optics to further improve the resolution, but it also re-quires more images, which is a limitation for live cells. Combining SIM with total internal reflection fluorescence (SIM TIRF) provides high-resolution imaging of structures near the cell membrane.

2) Single-molecule localization: This approach, including pho-toactivated localization microscopy (PALM) and stochastic opti-cal reconstruction microscopy (STORM), turns photoactive mol-ecules on and off, and this can locate molecules to a resolution of about 10 nanometers. Fluorescence can be added to PALM (FPALM). Adding carbocyanine dyes to STORM creates dSTORM, providing 20-nanometer resolution even in live-cell samples.

3) Point-scanning: Reversible saturable (or switchable) optical fluorescence transitions (RESOLFT) also uses switchable fluoro-phores and requires two overlapping laser beams. This method forms the basis of several forms of superresolution microscopy, including ground state depletion (GSD) and stimulated emission depletion (STED). These techniques provide resolutions of about 30–80 nanometers.

IMA

GE

: (T

OP

) ©

VSH

IVK

OV

A/S

HU

TT

ER

ST

OC

K.C

OM

SECTION TWO | Q AND A

18 sciencemag.org SCIENCE 19SCIENCE sciencemag.org


JS: The only thing I would add is the standard provison, which is that just because two things are colocalized at some resolution limit does not mean that they are actually molecularly interacting. So the question needs to be decorated with: what are you trying to achieve in that colocalization experiment? Are you just trying to say both of these signals are in the same organelle? Or that these two molecules are interacting? Those are two very different questions.

Q: In terms of confocality, what is the best distance between each acquisition of a series for fixed samples? Is getting more images always a better choice? PG: No, there’s photodamage and photobleaching being caused, even in a fixed sample, if you oversample in the z axis, and there’s a consequence in the data that you acquire. I take a look at the numerical aperture of the lens I’m using. From that I make an estimate of my expected depth of field or z resolution, and try to sample that distance about twice. So in a high NA, 1.4 NA objective, the z resolution is about 600 nm, so I would try to sample about every 300 nm.

JM: I would agree with that. I would say that as Paul said there’s a very clear theoretical answer to that question. The problem is that in biological applications, the amount of photobleaching prevents you from getting a useful set of images if you actually follow the theoretical limits.

Q: What do you think will be the next big revolution in microscopy? Or what you would like to see that would help your research? JS: I’m very excited about a lot of the new live cell imaging technologies. We’ve briefly mentioned superresolution technologies applied to live cell imaging. Seeing not only better spatial resolution, but also better temporal resolution, I think that’s very important.

to-noise ratio, and the rejection of background as necessary.

PG: If I’m looking at single molecules on a coverslip, I think widefield alone, or maybe TIRF, would be sufficient. But as I get more and more complex samples, I have to use the more complicated tools in my tool chest in order to get to an answer.

Q: Are confocal microscopes less affected by spherical aberrations then widefield systems? PG: I think John actually mentioned this in his presentation. There’s actual-ly a greater sensitivity to aberration in a confocal microscope and the same is also true in multiphoton. You need to make sure you don’t have spherical aberration, because that robs so much of the light from both the detection and the illumination of the specimen. Those sorts of aberrations actually have a big effect on all the imaging that we do and ideally need to be corrected at the beginning of the ex-periment.

Q: Have you tried to obtain deconvoluted or 3D-SIM images in real time?JS: First examples of that were published by Mats Gustafsson and colleagues. You have to speed up generation of the pattern and need very, very fast detectors. Commercial realizations of that system are now available and we’re very excited about it too.

Q: What is the best parameter to represent colocalization of two fluorophores in confocal imaging? JM: The best parameter, is overlap of intensities; and one must do the appropriate controls. Make sure that you’re not pushing your interpretation of the localization beyond the scale at which one is likely to get reasonable results, i.e., check the chromatic aberration and make sure that, on a specimen that you know is labeled with two fluorophores, those two images do superimpose on one another.

Jason Swedlow, Ph.D.University of DundeeDundee, Scotland

John Murray, M.D., Ph.D.Indiana UniversityBloomington, IN

widefield microscope than it would be to image a piece of glass. But typically we’re looking at groups of cells grown on a coverslip under the microscope. We can easily get through 20 to 30 microns. Beyond that, in more complex samples where there’ll be a lot of fluorescence light, we may be more limited to only a few microns.

Dr. John Murray: It is certainly speci-men-dependent and one needs to try all the available methods, but I think Paul’s limits are a good guideline.

Q: What happens if the sample thickness of the specimen you’re looking at varies across the sample? How do you deal with that?JM: That means we will probably get a better image in the thin spots than in the thick spots. But that can’t be helped, that’s biology, it’s going to vary.

Q: Which microscopic techniques are best for live cell imaging and which is the best choice: confocal versus widefield versus deconvolution?JS: The bottom line is that imaging, especially fluorescent imaging of biological samples, is quite difficult. The cells themselves are intrinsically sensitive to light. When you add fluorophores to them, they become even more so. So the best method is whichever method uses the least amount of light to achieve the signal-to-noise ratio you need in order to see the sample.

Different imaging methods use different illumination strategies. A widefield microscope usually uses a lamp or an LED light source, which usually use less illumination than laser illumination in confocal or multi-photon. However, it makes absolutely no sense to try to use a widefield microscope to image thick samples with a lot of out-of-focus fluorescence as John showed in his presentation. It’s a series of compromises and tradeoffs to reduce the amount of light you’re using, achieve the signal-

Microscopy in focus: The art and science of image qualityScience/GE webinar March 7, 2013Text edited for brevity and clarity

Q: How can a researcher be sure that a deconvolution algorithm is quantitative and not just creating artifacts or misleading data? Dr. Jason Swedlow: This is a very, very important question. I showed you an example of the validation that we did, which was a lot of work. We wanted to ensure that the methods we were using could be used for making relative intensity measure-ments. I think the general answer is that you have to test the different

methods on the kinds of samples that you use with the kind of imaging that you do.

We published a comparison back in 2002 [J. R. Swedlow et al., Proc. Natl. Acad. Sci. U. S. A. 99, 2014 (2002)].

Mr. Paul Goodwin: What Jason says is absolutely correct in that you want to make measurements with your samples versus different techniques and I think that’s true for any tech-nique in science. You need to validate, whether you’re quantitating DNA in a gel or whether you’re imaging with a confocal or with deconvolution. You want to make sure that you can make the statements that you hope to make in your paper by running care-ful controls and doing the calibration methods yourself‡.

Q: What is the depth or the thickness limits for widefield imaging plus deconvolution versus confocal imaging? PG: As John Murray showed, it’s not so much the depth as the complexity of the sample. You can imagine it would be a much more difficult to image a brick with a

Paul C. Goodwin, M.ScGE HealthcareIssaquah, WA

Speakers

cont.>

questions & answersWebinar

To view this webinar, go to:bitly.com/artandsciencewebinar

‡At GE Healthcare we use a deconvolution algorithm originally created by David Agard in his work with John Sedat at the University of California, San Francisco. In this algorithm we carefully measure the optical performance of the lens and use that empirical measurement of the lens performance.

http://webinar.sciencemag.org/webinar/archive/microscopy-focus




What’s coming out of all of that work, however, is this increasing complexity and difficulty understanding what we’re seeing. We have absolutely stunning movies now of tissues forming, for instance, and now we need the tools to turn those very complex images into some sort of understanding. PG: I think the next big revolution is in various software applications. We’re particularly interested in a number of methods that would allow us to work with noisier images, in other words to start off using raw data with lower signal-to-noise ratio and still be able to achieve the same quality of image in the end. Everything that we can do to be able to reduce that amount of light is going to be critical to reduce photodamage. There are new opportunities coming out of signal theory that I think will result in huge advances over the next couple of years in being able to reduce the light load on a sample.

JM: Having worked through the intro-duction of confocal microscopy, and having seen the totally unpredictable techniques that arose as a result, I predict that the introduction of su-perresolution microscopy into bio-logical imaging will have equally un-predictable effects. And I have no idea what will come! But I’m looking forward to it with great excitement.

about using the most sensitive instrument possible, meaning you don’t have to have long exposure times. In our GE IN Cell 6000 system, you can dial in confocality when you need it, but you don’t have to have it running in all different channels. That means you’re not blasting all of your cells in all of the channels with more light than they need to get you the data.

LT: The point is less is more. If you can get away with fewer time points, less z slices, it’s better. But you can’t avoid photobleaching completely. You just have to make sure that you have some measurement of how healthy your cells are and stop be-fore they’re no longer healthy.

yourself whether you have automatic systems or not. We keep everything hot so we never turn off the live cell environmental chambers and we keep the microscope, the objectives, and the whole box at 37℃ or 38℃ so that we don’t have metal expanding and contracting. We keep the oil and spare objectives inside the box to keep them warm. Also, focal drift during experiments maybe due to evaporation, so we keep the environmental chambers full of beakers of water and soggy tissues to keep the evaporation to a minimum.

Q: How can photobleaching be avoided and is phototoxicity observable at all wavelengths? Is it possible to use different wavelengths to reduce the amount of phototoxicity? EC: For photo bleaching, a big advancement has been EMCCD [electron multiplying charge couple device] cameras, which really work. They have a function called In Chip gain, which allows you to collect all of the imaginable data using tricks of physics. Another factor is the plat-form you’re using. We talked about the difference between confocals and widefield deconvolution micro-scope. There’s an article by John Murray, a true microscopy expert, who has shown that the photon cap-ture efficiency of a deconvolution microscope is better than many other systems. Essentially, deconvolution “restores” light to its origin before scattering. So you can get by with less exposure and this is the key to reducing photobleaching.

To get to the viewer’s question, cells are very much like us in that the light that affects them the most is UV. There are filters to prevent the UV light from getting to your sample, but they will sometimes preclude the use of CFP. But if I had to keep my cells alive, I would use GFP and Cherry or YFP and Cherry because they’re further away from that UV spectrum of light.

NT: From an HCA perspective it’s

Live cell imaging: The future for discoveriesScience/GE webinar May 22, 2013Text edited for brevity and clarity

Q: One of the key issues for live cell imaging of any duration seem to be maintaining focus in z. Do you have any tips and tricks you can share? Dr. Edward Campbell: One new technology is a laser guided system that detects the distance between your lens and your specimen. They are invaluable especially in long-duration imaging experiments.

Dr. Nick Thomas: The issues with HCA are somewhat different. Instru-mentation for HCA is set up to do imaging automatically. Generally, you have hardware autofocus, so you need to ensure that’s working prop-erly and use the right plates. A lot of people don’t take enough care about selecting their plates. If you’re do-ing live cell imaging, it’s even more important to have the right size of plates, make sure that they’re clean and that you don’t get dust in them as that might throw out the focus.

Dr. Lynne Turnbull: There are a lot of things you can do just to help

EC: I think it’s relevant to think that it’s not an either/or. So we’ve not really found a far red fluorescent protein that we enjoy, but a lot of the conjugatable dyes work very well in the Cy5 range.

Q: Could you talk about some of the applications of this work in terms of health care, public health, and disease research?NT: We spend a lot of time looking at how we can use these new technolo-gies to determine both the safety and the efficacy of drugs. In the past couple of years, a game changer has been people recognizing the poten-tial of stem cell-derived models for looking at drug toxicity and efficacy. Companies making those available have allowed researchers to conceive of experiments that they couldn’t do before. For cardiotoxicity, we use both fixed and live cell assays, and we integrate that data from both with other analytical platforms. So you get what I call a holistic surveillance of what that drug is doing to a cell. An advantage of live cell imaging is that often you can distinguish between the mode of action of two drugs that look the same. There’s really no such thing as a completely safe drug, but what we don’t want to do is fail drugs because we don’t have a clear under-standing of the mode of action. Be-cause then the general population is not getting the best drugs that they could have. So that’s what drives me in what I do.

LT: With live cell imaging, you’re ob-serving processes as they happen, which gives you a better understand-ing of the stop points where you might be able to intervene, for in-stance in the infection process. Ad-ditionally, live cell imaging allows one to examine how a new drug works and which part of the process it impacts, which you don’t always get from fixed imaging. This can be very informative.

Edward M. Campbell, Ph.D.Loyola UniversityChicago, IL

Nick Thomas, Ph.D.GE HealthcareCardiff, Wales

Speakers

“Live cell imaging allows one to examine how a new drug works and which part of the process it impacts, which you don’t always get from fixed imaging.”

To view this webinar, go to:bitly.com/livecellwebinar

Lynne Turnbull, Ph.D. University of TechnologySydney, Australia

Webinar questions & answers continued

Q: What are the pros and cons of fluorescent chemical probes versus fluorescent proteins?NT: Some of those GFP experiments I showed were done 10 years ago. Those techniques work, so, “if ain’t broke, don’t fix it”! Nowadays there are more advanced fluorescent proteins giving a wider choice of colors and more stability. Leaving the phototoxicity issue aside for now, there are now tagging systems that allow you to put a tag on any protein and deliver a fluor into a cell. I like GFP—there’s a certain essence about it and you have all of the literature as precedent. And with the increase in sensitivity of instrumentation, you can actually cut down on the amount of protein you need to express in a cell.

http://webinar.sciencemag.org/webinar/archive/live-cell-imaging




live imaging, which is where I think it shines the best.

Q: What are some of the particular challenges of live-cell imaging?RD: The main thing to take into ac-count when imaging live cells and tissues is that you’re shining a lot of light onto these cells, causing phototoxicity. As Dr. Betzig said, SIM seems to be the best compromise when you consider speed and tem-poral resolution, and even depth of view and phototoxicity.

way down from the diffraction limit, lower and lower, until you get what you want.

RD: This is a great question. I found myself in a very similar situation not too long ago. I would say it always depends on the experiments you’re doing, but I’m guessing that most people would be doing fixed tissue or cell imaging and may be using antibodies and several different fluorophores. As Dr. Betzig suggested, use the easier-to-access superresolution technologies first, such as SIM, as you can use the same dyes you would use for widefield.

Q: If somebody is looking for the best resolution, what compromises are they going to need to make and what should they be cognizant of as they’re setting up their experiment?JT: The trade-off you have when you want to get super-high resolution is that you’re taking thousands, sometimes millions, of images. That generally requires the cells to be dead and fixed, and that you’re exposing the sample to large amounts of light. That being said, to do live-cell imaging, you want to expose the cell to as little light as possible, which means reducing the resolution to what is possible under those conditions.

EB: Clearly, you can’t have it all. If you want higher spatial resolution, your image has to have more pixels, which means you’re taking more measurements, which takes more time and more potentially damaging light on the specimen. The absolute highest resolution is probably a single molecule localization method like PALM. To see single molecules, you have to use light that’s 10,000 times brighter than the light that cells evolved with. It’s not necessarily a really great tool for live imaging either, but for high resolution structural imaging, it’s great. I developed PALM, but I’m a fanatic about SIM as a tool particularly for

really have to change the way we do things at all, so my advice to you would be to just go ahead with your normal staining procedure.

EB: It all comes down to the labeling and how good your labeling is, how dense it is, how bright it is. One of the primary advantages of SIM is that it doesn’t require specialized labels like the other methods do. It’s therefore much easier for the biologists to di-rectly transition their sample prepara-tion protocols. With superresolution, the more one wants in terms of reso-lution or speed, the more one has to contribute in terms of optimizing the sample preparation protocols. There are all sorts of trade-offs between the different methods.

Q: For someone just starting out in this field, what advice could you give them with respect to what technologies to consider and also what pitfalls to avoid?EB: I think that there is a perception that one should use the highest resolution method out of the gate, and that’s misplaced. There’s always a trade-off and the more resolution you ask for, the more pain you can expect. So you want to use the method that has the poorest resolution that still gets you the biological answer you’re looking for. If you can do that with a standard widefield or confocal microscope, stop there. You have to first know how to use your diffraction- limited tools to their capability before moving on to superresolution. Beyond that, the next technique is probably SIM. It is wonderful, particularly in a live context because it’s so much faster and less invasive. But it is really limited in how thick of a specimen you can look. Then you start getting into nonlinear SIM, PALM, RESOLFT, and STED. These are typically very difficult methods for biologists to use because while in theory it’s true that STED can use any label, in practice, most of the labels just can’t take the beating that STED delivers. So you will have to optimize labels for that. So, in short, work your

Q: What do you see as some of the most interesting developments in the world of superresolution microscopy in the past two decades? Dr. Raman Das: From the perspective of a biologist who’s really interested in what cells do, what’s made the most difference to me is the applica-tion of superresolution microscopy to biology.

In many ways, this has been made possible by commercialization of many people’s hard work. It’s great that we’ve gotten to the point where someone can go to a microscope, turn a key, and get some beautiful images. That’s what has really made a difference for most biologists.

Dr. Eric Betzig: What Raman said was really critical. However, in my mind, the promise of superresolu-tion is still not completely fulfilled in

part because commercialization is essential, but incomplete. Biologists need a turnkey instrument that is not a struggle to work with. The develop-ment of fluorescence for superreso-lution, and particularly its application to biology, is critical. There’s been an exponential increase in techniques that address different niches, but it’s equally as much been a story about the chemistry, the labels, delivery, fluor attachment strategies… essen-tially the interdisciplinary synergy of all the physical sciences.

Dr. Justin Taraska: The addition of computational tools and a number of other things that have allowed researchers to analyze the images in a way that provides insight into how the biology works has had a major influence in the development of superresolution methods. For me, it’s two different sides of the same coin. Researchers have gotten down to nanometer-scale resolution, to individual proteins, but then doing that again in live cells. That the field has gone in two different directions I think has been the most exciting advance.

Q: What about the application of superresolution to formalin-fixed, paraffin-embedded sections? RD: I think it really depends on which superresolution technology you’re thinking of using; we use SIM. Of course it’s great because we don’t

good as traditional widefield SIM, but it’s a bit beyond the diffraction limit and capable of imaging thicker samples in 3-D for long periods of time without damage.

Another technology, which isn’t true superresolution, but has been classed with this and is probably the greatest method in terms of producing real biology, in my opinion, is single particle tracking PALM, where you’re able to photoactivate subsets of molecules and watch their diffusion. Because you’re not trying to activate anywhere near as many molecules, but are just following tracks, you can throw much less light at the sample.

Q: Could you talk about the impact of the fluorescent dye on imaging and resolution?JT: On the very high resolution side, the organic dyes are the best dyes that you could possibly use. The very, very bright dyes like Alexa 647 and Alexa 750 produce the highest resolution images for us, but they require that you label your cells with something that you can attach those dyes to, whether that’s an antibody or a nanobody or some sort of chemical tag like a SNAP Tag or Halo Tag. Fluorescent proteins are improving, getting brighter and brighter, but they pale in comparison to the organic dyes. The higher resolution you want, the brighter the dye you should use.

EB: The only caveat I would add is distinguishing localization precision from resolution. The brightness of the dye means that you can determine the position of these molecules much more accurately than you could with a dimmer fluorescent protein. But how you attach the fluorophore to the target of interest is as important as making the fluorophores better. If you don’t have enough density of labeling then it doesn’t matter even if you can look at each molecule to, say, one nanometer. If they’re spaced by 100 nanometers, it still just cont.>

Webinar questions & answers continued

“With superresolution, the more one wants in terms of resolution or speed, the more one has to contribute in terms of optimizing the sample preparation protocols.”

EB: I totally agree. The superresolution methods that are used predominantly today are SIM, single molecule localization like PALM, and point-scanning methods like STED and RESOLFT. The light intensities that cells evolved under is a tenth of a watt per square centimeter. PALM requires about four orders of magnitude higher than that. RESOLFT requires about four to five orders of magnitude higher than that. STED requires seven to eight orders of magnitude higher again. In my opinion, there’s been a frustrating lack of control in the literature regarding the amount of light used and what this light is doing to the cells. One solution is a combination of different technologies, called plane illumination. This combines SIM with illuminating just a single focal plane at a time. The resolution is not as

Generating the best superresolution microscopy data: Finding the right tool for the right jobScience/GE webinar July 29, 2015Text edited for brevity and clarity



Microscopy in focusThe art and science of image quality

Live cell imagingThe future for discoveries

To view this webinar, go to: bitly.com/livecellwebinar


looks like a sprinkling of dots without sufficiently decorating the underlying feature.

RD: The photostability of the dye is also almost equally important, especially when you’re using more widefield approaches like SIM, which take multiple images and then use mathematical algorithms to reconstruct these into superresolution images. In my experience, if you don’t have a photostable dye, you see a drastic bleaching of your sample and that results in failure to reconstruct your data.

Q: How deep is it practical to image and what are some of the challenges of deep-tissue imaging techniques?EB: The challenge in going deeper than a few cell layers is that bio-logical tissue is heterogeneous. The light gets scrambled as you go deeper. Superresolution methods in particular are exquisitely sensitive to the scrambling issue and eventually even confocal or two-photon microscopes also succumb. Right now, there is no good commercial solution. In our lab, we use tricks like adaptive optics to unscramble the light beams as they go in and out, but it’s still very much a work in progress. This is one area where point scanning has an advantage over widefield methods, because you can easily optical section by discriminating out-of-focus light using a pinhole. That normally means that STED and RESOLFT would have the edge at great depth, but, again, they require a really good quality focus in order to work, and without adaptive optics, they fail. I think the best methods that go beyond SIM at great depth are still the confocal microscope.

Q: A final question: what do you see is the next big thing in the superresolution world?JT: Currently, we’re able to localize proteins very precisely, but I think

one of the missing pieces is the structure of those proteins. We know where everyone is, but we don’t know what state those proteins are in. The development of conformationally sensitive antibodies or nanobodies, applied to FRET or other methods, would really get at not only where proteins are, but how they’re behaving and whether or not they’re active. This would be a very exciting future development for superresolution techniques.

RD: I think really what we need to do is to develop better tools to do live imaging and deep tissue. I’m particularly excited about something that Dr. Betzig said that they’re developing, adaptive optics for deep tissue imaging. I think this is what is needed to enable many exciting experiments. It’s just incredible to think about.

EB: At a practical level, the next big thing is the hard work being done to actually get these technologies to work well—all the chemistry, all the biology, all the control experiments—to make this as real a tool as the electron microscopy is on a daily basis. Beyond that, my own wish list includes getting rid of fluorescent labels and finding a label-free way of doing protein-specific contrast with superresolution. Finally, it would be really nice to be able to get the diffraction-limited or superresolution images of dozens to hundreds of proteins at the same time in the same cell, not just two or three.

Eric Betzig, Ph.D.HHMI Janelia Research CampusAshburn, VA

Raman Das, Ph.D.University of DundeeDundee, UK

Justin Taraska, Ph.D.National Institutes of HealthBethesda, MD

Speakers

To view this webinar, go to:bitly.com/TheRightTool

Generating the best superresolution microscopy dataFinding the right tool for the right job

To view this webinar, go to: bitly.com/TheRightTool

To view this webinar, go to: bitly.com/artandsciencewebinar

Webinar Series

Produced by the Science/AAAS Custom

Publishing OfficeSponsored by

http://webinar.sciencemag.org/webinar/archive/generating-best-superresolution-microscopy-data

bitly.com/artandsciencewebinar

http://webinar.sciencemag.org/webinar/archive/live-cell-imaging

http://webinar.sciencemag.org/webinar/archive/microscopy-focus

www.gelifesciences.com/deltavisionGE, GE monogram, and DeltaVision are trademarks of General Electric Company. DeltaVision products are for research use only- not for diagnostic use. © 2015 General Electric Company. First published Aug. 2015All goods and services are sold subject to the terms and conditions of sale of the company within GE Healthcare which supplies them. A copy of these terms and conditions is available on request. Contact your local GE Healthcare representative for the most current information.

GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, SwedenGE Healthcare Europe, GmbH, Munzinger Strasse 5, D-79111 Freiburg, GermanyGE Healthcare Bio-Sciences Corp., 800 Centennial Avenue, P.O. Box 1327, Piscataway, NJ 08855-1327, USAGE Healthcare Japan Corporation, Sanken Bldg., 3-25-1, Hyakunincho, Shinjuku-ku, Tokyo 169-0073, JapanFor local office contact information, visit www.gelifesciences.com/contact

29167968 AA 08/2015

GE Healthcare UK Limited Amersham Place Little Chalfont Buckinghamshire, HP7 9NA UK

DeltaVision OMX SR super-resolution microscope DeltaVision OMX SR (Fig 2) is a compact super-resolution microscope system optimized to incorporate the power of proven structured illumination microscopy (SIM) technology into a stable live-cell imaging platform.

See more Many cellular structures are below the spatial resolution limit of standard or confocal microscopes and events happen on sub-second time-scales, putting them beyond our ability to investigate in greater detail using traditional methods. DeltaVision OMX SR is designed for biologists wishing to advance their research beyond what is possible with confocal microscopes, enabling previously unseen detail to be obtained from highly resolved images of small, dim, and live samples. The exclusive Blaze 3D-SIM module provides an eight-fold increase in volume resolution by improving lateral resolution to approximately 120 nm and axial resolution to approximately 340 nm (wavelength dependent) allowing physiologically relevant live-cell imaging.

Super-resolution microscopy doesn’t need to be complicatedUnlike many super-resolution systems, there is minimal change required for your sample preparation or preferred buffers and fluorophores, meaning you can be confident that your results are physiologically relevant from day one.

DeltaVision OMX SR does not require a darkroom and can be installed ready-to-run in a standard lab within days. Full training and post-installation support are provided; giving you the time to focus on your research and contribute to the insight SIM technology has brought to our understanding of cellular processes and structures. Check out some of the hundreds of articles published by biologists already using SIM based super-resolution microscopes for their research.

Advanced imaging capabilities expand your research optionsIn addition to 2D and 3D SIM super-resolution imaging, the DeltaVision OMX SR is also capable of widefield, time-lapse imaging, deconvolution, and TIRF illumination using the same techniques and fluorophores utilized on confocal imagers. For added flexibility, a ring TIRF option provides even illumination and artifact-free TIRF imaging enabling the system to be used for TIRF, localization microscopy, and photokinetic applications.

The latest from the DeltaVision rangeThe DeltaVision OMX SR is the latest addition to the high-performing DeltaVision range (including DeltaVision Elite and DeltaVision OMX), which has earned a reputation for delivering outstanding images particularly for small, dim, and live samples.

Fig 2. DeltaVision OMX SR is a compact super-resolution microscope providing outstanding images of small, dim, and live samples.

References1. Eswaramoorthy, P. et al. Asymmetric Division and Differential

Gene Expression during a Bacterial Developmental Program Requires DivIVA. PLoS Genet. 10(8), (2014).

2. Klutstein, M. et al. The telomere bouquet regulates meiotic centromere assembly; Nat. Cell Biol. 17(4), 458–469 (2015).

3. Revach, O.Y. et al. Mechanical interplay between invadopodia and the nucleus in cultured cancer cells; Sci. Rep. 5, 9466 (2015).

* For a more complete list of references search DeltaVision and Deconvolution at the Highwire site.

www.gelifesciences.com/deltavisionGE, GE monogram, and DeltaVision are trademarks of General Electric Company. DeltaVision products are for research use only- not for diagnostic use. © 2015 General Electric Company. First published Aug. 2015All goods and services are sold subject to the terms and conditions of sale of the company within GE Healthcare which supplies them. A copy of these terms and conditions is available on request. Contact your local GE Healthcare representative for the most current information.

GE Healthcare Bio-Sciences AB, Björkgatan 30, 751 84 Uppsala, SwedenGE Healthcare Europe, GmbH, Munzinger Strasse 5, D-79111 Freiburg, GermanyGE Healthcare Bio-Sciences Corp., 800 Centennial Avenue, P.O. Box 1327, Piscataway, NJ 08855-1327, USAGE Healthcare Japan Corporation, Sanken Bldg., 3-25-1, Hyakunincho, Shinjuku-ku, Tokyo 169-0073, JapanFor local office contact information, visit www.gelifesciences.com/contact

29167968 AA 08/2015

GE Healthcare UK Limited Amersham Place Little Chalfont Buckinghamshire, HP7 9NA UK

DeltaVision OMX SR super-resolution microscope DeltaVision OMX SR (Fig 2) is a compact super-resolution microscope system optimized to incorporate the power of proven structured illumination microscopy (SIM) technology into a stable live-cell imaging platform.

See more Many cellular structures are below the spatial resolution limit of standard or confocal microscopes and events happen on sub-second time-scales, putting them beyond our ability to investigate in greater detail using traditional methods. DeltaVision OMX SR is designed for biologists wishing to advance their research beyond what is possible with confocal microscopes, enabling previously unseen detail to be obtained from highly resolved images of small, dim, and live samples. The exclusive Blaze 3D-SIM module provides an eight-fold increase in volume resolution by improving lateral resolution to approximately 120 nm and axial resolution to approximately 340 nm (wavelength dependent) allowing physiologically relevant live-cell imaging.

Super-resolution microscopy doesn’t need to be complicatedUnlike many super-resolution systems, there is minimal change required for your sample preparation or preferred buffers and fluorophores, meaning you can be confident that your results are physiologically relevant from day one.

DeltaVision OMX SR does not require a darkroom and can be installed ready-to-run in a standard lab within days. Full training and post-installation support are provided; giving you the time to focus on your research and contribute to the insight SIM technology has brought to our understanding of cellular processes and structures. Check out some of the hundreds of articles published by biologists already using SIM based super-resolution microscopes for their research.

Advanced imaging capabilities expand your research optionsIn addition to 2D and 3D SIM super-resolution imaging, the DeltaVision OMX SR is also capable of widefield, time-lapse imaging, deconvolution, and TIRF illumination using the same techniques and fluorophores utilized on confocal imagers. For added flexibility, a ring TIRF option provides even illumination and artifact-free TIRF imaging enabling the system to be used for TIRF, localization microscopy, and photokinetic applications.

The latest from the DeltaVision rangeThe DeltaVision OMX SR is the latest addition to the high-performing DeltaVision range (including DeltaVision Elite and DeltaVision OMX), which has earned a reputation for delivering outstanding images particularly for small, dim, and live samples.

Fig 2. DeltaVision OMX SR is a compact super-resolution microscope providing outstanding images of small, dim, and live samples.

References1. Eswaramoorthy, P. et al. Asymmetric Division and Differential

Gene Expression during a Bacterial Developmental Program Requires DivIVA. PLoS Genet. 10(8), (2014).

2. Klutstein, M. et al. The telomere bouquet regulates meiotic centromere assembly; Nat. Cell Biol. 17(4), 458–469 (2015).

3. Revach, O.Y. et al. Mechanical interplay between invadopodia and the nucleus in cultured cancer cells; Sci. Rep. 5, 9466 (2015).

* For a more complete list of references search DeltaVision and Deconvolution at the Highwire site.

GE Healthcare

High- and super-resolution imaging

DeltaVision™ Microscopy Systems

Fig 1. DeltaVision Elite high-resolution microscope.

DeltaVision DeltaVision high- and super-resolution microscopes deliver the highest data quality possible particularly in challenging circumstances such as low light imaging for small, dim, and live samples.

• Specifically engineered to minimize interference from out-of-focus light.

• Highly efficient light path ensures the system captures the maximum number of photons possible.

• Oil selection can be optimized to best match unique sample conditions and minimize optical aberrations

• Components are handpicked to work seamlessly together

DeltaVision Elite high-resolution microscopeDeltaVision Elite is a fully integrated and flexible high-resolution imaging system optimized for low light imaging of dim and live-cell samples. Efficient light capture and a stable platform ensure quantifiable, high quality images while minimizing damage to the sample due to photobleaching and phototoxicity (Fig 1).

• Improved contrast for better image quality and detection

• Highly efficient light delivery allows for flexibility during exposure condition optimization

• Fully integrated system for seamless operation minimizes the time required for experiment design

• Modular platform design enables future upgrades to expand capabilities for new applications as imaging needs evolve

Maximum flexibility for your imaging needsDeltaVision Elite can handle most imaging applications including:

• Widefield fluorescence

• Time-lapse live-cell imaging

• Multipoint cell tracking

• Total internal reflection fluorescence (TIRF)

• Fluorescence resonance energy transfer (FRET)

• Photokinetics

• Differential interference contrast (DIC).

Data file, 29167968 AA

Deconvolution sharpens your imagesDeconvolution improves image resolution and contrast in X, Y, and Z directions by reassigning blurred light to its original point source, considerably improving image contrast without sacrificing light or data integrity. Our exclusive deconvolution algorithm has been used extensively in discovery research papers* (1, 2, 3) and is provided as an integral part of the system.

Outstanding focus stability for long-term imaging experimentsThe UltimateFocus hardware based autofocus module seamlessly detects and maintains the sample focus position regardless of mechanical or thermal changes that can cause focus drift. In addition to focus maintenance, the UltimateFocus module can be used to acquire initial sample focus without using the eyepieces or camera and unnecessarily exposing the sample to excitation light.

Automated high-precision stagePatented control motors ensure precise stage movement and stability with excellent repeatability for multipoint sampling during time-lapse experiments.

2726 SCIENCE sciencemag.orgsciencemag.org SCIENCE

Technical note Technical note

MICROSCOP Y NOW UPDATE: GET TING THE MOST FROM YOUR IMAGING SECTION THREE | ARTICLES:REVIEW

We have an incredibly detailed view of how proteins and lipids interact in-side cells to govern the generation, maintenance, and function of cellular organization, as determined from bio-

chemical and genetic experiments spanning diverse approaches from in vitro reconstitu-tion of cellular processes to atomic resolu-tion structure determination. However, these techniques only provide a static, snapshot view of cells. Being able to observe processes as they happen within the cell by light mi-croscopy adds a vital extra dimension to our understanding of cell function. Perhaps the commonest approach for studying dynamic cellular events is live cell fluorescence micros-copy, and we will discuss this in some detail. However, transmitted light techniques also have an important part to play (1) and not just as an adjunct to fluorescence imaging.

Environmental considerations. Regard-less of the imaging technique to be used, it is crucial to consider the cells’ health on the microscope stage. Cells are sensitive to photo-damage, particularly in the presence of fluo-rophores (which generate free radicals upon photobleaching), and there are many ways of trying to limit light-induced damage. It is also vital to keep the cellular environment constant. There are a number of solutions to this problem, including the control of tem-perature, humidity, and CO

2. Environmental

control ranges from simple heating jackets to Perspex boxes that fully encase a system. The relative importance of each parameter will vary between samples, but the overriding concern for all three is stability. In time-lapse experiments, once a sample is being imaged, the focal plane must remain stable. Autofocus routines are available, which can compensate for focus drift to some extent, but they require additional illumination of the sample. One important, but often overlooked, cause of fo-cus drift is air conditioning units, which can cause cyclic changes in focus as they turn on and off.

Fluorescence imaging

Whilst it is sometimes possible to image en-dogenous cellular molecules such as NAD(P)H (2) by their inherent fluorescence, it is far more common to introduce exogenous fluorescent molecules. The advent of green fluorescent protein (GFP) technology has revolutionized live cell imaging because an autofluorescent molecule can be genetically encoded as a fusion with the cDNA of inter-est (3). The spectral variants of GFP and the unrelated red fluorescent protein (4) make it feasible to perform multicolor imaging of liv-ing cells. The simultaneous study of multiple fluorophores or ratiometric analysis of a sin-gle probe requires spectral separation of both the excitation and emission light. Commercial systems are now available for “spectral unmix-ing” of data, and this allows the use of closely related fluorophores, but, where possible, it is better to use probes with distinct excitation and emission spectra that are separable at the point of image acquisition.

GFP-based biosensors are opening many fields to optical techniques, notably the spa-tio-temporal analysis of signaling events fol-lowing the development of probes for diverse processes including heterotrimeric G protein activity (5) and phosphoinositide signaling using GFP-tagged pleckstrin homology (PH) domain constructs (6). The field of calcium imaging also makes use of GFP-based probes (7), allowing organelle-specific analysis of cal-cium dynamics. GFP-tagging is also being ap-plied to high-throughput analyses to provide further functional annotation of genome se-quences (8).

FlAsH (fluorescent arsenical helix binder) labeling provides another means for fluores-cent labeling of genetically encoded probes (9). It is mediated by engineering a tetracysteine motif into the target protein and then incu-bating cells with a nonfluorescent biarsenical compound that becomes strongly fluorescent upon binding to this tetracysteine motif. A recent development enables multicolor label-ing and photoconversion of diaminobenzidine for correlative electron microscopy (10). Un-fortunately FlAsH compounds can also label endogenous proteins containing similar tet-racysteine motifs (11). In addition, the cyste-ine residues must be in the reduced state for

labeling to occur, and antidotes must be added simultaneously with labeling to avoid toxicity problems.

There are, of course, many other potential probes that can be introduced into cells. Spe-cific fluorescent lipid molecules and organ-elle-specific dyes are often cell permeable and can simply be added to the culture medium (12). Fluorescently labeled proteins can be introduced by microinjection, and the useful-ness of such probes is also continually driv-ing the technology for studying intracellular dynamics. A prime example of this is the de-velopment of fluorescence speckle microscopy (13). Here, introduction of a limited amount of fluorescent protein to a polymeric struc-ture, such as a microtubule or actin filament, results in a “speckled” appearance. These speckles can then be imaged over time and tracked within the cell to provide accurate quantitative analysis of polymer dynamics. Finally, a number of probes can be activated by light, allowing specific detection only after a pulse of illumination (14–16).

Live cell imaging

When selecting which system to use for im-aging living cells, one should consider three things: sensitivity of detection, speed of ac-quisition, and the viability of the specimen. Light microscopy of living versus fixed sam-ples is essentially a trade-off between acquir-ing images with a high signal-to-noise ratio and damaging the sample under observation; this is a particularly critical issue in live cell imaging. Other important questions center on the sample you want to image. Is it thick or thin? Is the process to be observed fast or slow? Do you need to image for seconds, minutes, hours, or days, and at how many dif-ferent wavelengths does the image need to be sampled? How bright is your signal? You also need to consider several further questions: Will you want to use a specialized technique such as photobleaching? Are transmitted light images required, and if so, of what quality? In many cases, no single microscope system will be best, and compromises will have to be made. A good understanding of the pros and cons of different microscopes is needed, and it is also helpful to understand the resolution of the light microscope (17). The basic features of three types of fluorescence microscope sys-tems are illustrated in Fig. 1.

Limiting cell damage. Because illumina-tion of fluorophores causes photobleaching and therefore cell damage, everything pos-sible should be done to limit the duration and intensity of illumination. A minimum requirement is the ability to shut off illumi-nation light when it is not needed; this is in-herent in confocal systems and can easily be achieved for widefield systems that use elec-tronic shutters controlled by computer (which will usually control the image acquisition as well). Care should also be taken to remove

Since the earliest examination of cellular structures, biologists have been fascinated by observing cells using light microscopy. The advent of fluorescent labeling technologies plus the plethora of sophisticated light microscope techniques now available make study-ing dynamic processes in living cells almost commonplace. For anyone new to this area, however, it can be daunting to decide which techniques or equipment to try. Here, we aim to give a brief overview of the main approaches to live cell imaging, with some mention of their pros and cons.

David J. Stephens1 and Victoria J. Allan2*

Light microscopy techniques for live cell imaging

1Department of Biochemistry, University of Bristol, School of Medical Sciences, University Walk, Bristol, BS8 1TD, UK. 2School of Biological Sciences, Oxford Road, University of Manchester, Manchester, M13 9PT, UK.*Corresponding author. Email: [email protected]

unwanted wavelengths of light and not to rely simply on the excitation filters. Reducing the level of oxygen can help reduce photobleach-ing and free radical production. Oxygen can be removed from the medium as long as the cells are in a sealed chamber (13), providing the cells tolerate oxygen withdrawal. Finally,

omitting phenol red and serum from the me-dium (again, if your cells will stand it) will help reduce background fluorescence.

The system must also make best possible use of the light, so high numerical aperture objectives should be used, and there should be as few optical elements in the light path

as possible. The sensitivity of the camera (or photomultiplier tube, if using a confocal) will be vital (18), because the more sensitive the detector, the lower the illumination intensity needed. Using an intensified camera is one way of increasing sensitivity, at the expense of increasing noise in the image. Alternatively, sensitive back-illuminated charge-coupled device (CCD) cameras with thinned chips are available. A new type of camera that ampli-fies the CCD readout signal on the chip (19) offers further possible advantages. Another simple way of increasing camera sensitivity is to combine signals from multiple pixels (called binning), although this process re-duces image resolution.

Speed of acquisition. A key consideration is speed of data acquisition, particularly when multiple fluorophores are imaged simultane-ously or when a single probe is analyzed ra-tiometrically. Switching between laser lines, filters, or output from a monochromator will slow data acquisition (Fig. 1). Monochroma-tor-based systems have the advantage of rapid switching between excitation wavelengths (typically <3 ms) but suffer from reduced il-lumination intensity, principally due to fiber optic coupling to the microscope. Filter wheel configurations usually have higher light throughput but are often slower in switching. Data acquisition rates of conventional scan-ning confocal microscopes are fast enough for rapid imaging if only small regions are sampled. To image very fast processes such as neuronal network activity, one may obtain faster scanning by using resonant galvanome-ters (which are optional on many commercial systems). Another important consideration is that scanning systems acquire data pixel by pixel, whereas CCD cameras acquire a whole field of view at once.

Scanning speed in confocal microscopy can also be improved with the use of multifocal imaging (20). Here, the excitation light beam is split into multiple foci from which data are collected simultaneously with a CCD (Fig. 1). Nipkow disk confocal microscopy is available commercially through a number of suppli-ers and can achieve speeds of 360 frames per second. Sensitivity and data acquisition rates of Nipkow disk systems, like widefield micro-scopes, depend on the quality and the readout time of the detector CCD. No single camera will perform optimally for all tasks, and cor-rect matching of optics and electronics is es-sential for best performance.

Three- and four-dimensional (3- and 4D) imaging. Researchers are often attracted to confocal systems because high-resolution 3D images (Fig. 2) can be acquired simply. How-ever, many experiments, particularly those using live cells, may be better performed us-ing widefield (conventional) systems with subsequent deconvolution of the data series. Widefield microscopes do not exclude light from any plane of focus; they collect it all. The contribution of light from an infinitely

2928 Originally published 4 April 2003 in SCIENCE SCIENCE sciencemag.orgsciencemag.org SCIENCE

Fig. 1. Comparison of widefield, scanning confocal, and spinning disk confocal systems, with schematics of each. All systems are capable of being equipped for 3D and 4D data acquisition. Excitation beams are shown in green; emission beams, in blue. The differences between these systems mean that no single system is suited for every experiment. Typical system configurations are shown, and user modification and options allow great flexibility.

http://www.sciencemag.org/content/300/5616/82.full#ref-1

http://www.sciencemag.org/content/300/5616/82.full#F1






MICROSCOP Y NOW UPDATE: GET TING THE MOST FROM YOUR IMAGING SECTION THREE | ARTICLES:REVIEW

including photobleaching. However, a key ad-vantage of FCS is that it is also applicable to single molecule studies.

Photobleaching and photoactivation ap-proaches. Because scanning confocal mi-croscopy has control over the region of illumination, it is ideal for photobleaching techniques such as fluorescence recov-ery after photobleaching (FRAP) and fluores-cence loss in photobleaching (FLIP) (43, 44). Both are now widely used to measure dif-fusional mobility of GFP-tagged proteins in cells. Increasingly, they are combined with kinetic modeling of cellular processes (45) to study topics as diverse as membrane traffic and nuclear architecture and function (46). Probes that can be light-activated such as photoactivatable-GFP and Kaede (14), both of which show greatly enhanced fluorescence emission following activation at ∼400 nm, al-low selective labeling of subdomains of cells and organelles.

Fluorescence resonance energy transfer (FRET). Intermolecular interactions form the basis of all processes in live cells and can

small point source to a plane of focus some distance away from that point source is de-scribed by the point spread function (PSF) of the objective. Determination of the PSF of a system enables mathematical reassignment of the out-of-focus light back to its point source by deconvolution (21). This approach has been used with great success in both cell and devel-opmental biology, and it can be particularly advantageous in imaging very weakly fluo-rescent structures such as microtubules (Fig. 2) (22). Deconvolution must be applied with great care and accuracy, however, to avoid the generation of artefacts (21). Deconvolution of large 4D data series can now be achieved in minutes to hours with the use of dual proces-sor personal computers.

Most cellular processes occur in three dimensions over time, so to get a complete picture we need to image cells in four dimen-sions. Most confocal systems and epi-illumi-nation microscopes are either provided with or can be simply adapted to include a means of acquiring data series in four dimensions. Perhaps the most important consideration is the speed, accuracy, and reproducibility of thez position change. Piezoelectric ob-jective drives have the edge here, enabling high-speed acquisition of stacks at multiple wavelengths over time. Multidimensional live cell imaging also requires tools for data

analysis (23). Tools are continually being developed for particle track-ing of objects moving inside cells such as transport vesicles (24). Most reconstruction approaches assume a uniform refractive index through the sample, which is not encoun-tered often in reality (25); further developments are needed to address this issue.

Multiphoton approaches to in vivo imaging. Multiphoton confocal sys-tems are now available from several companies. The two-photon effect ex-cites a chromophore not by a single photon but from two photons being absorbed within a femtosecond time scale (26). This enables the use of longer wavelength excitation, which penetrates deeper into samples and reduces photobleaching. Notably, the analysis of intact organisms or tis-sues greatly benefits from this tech-nique, allowing imaging in situ (27). Such systems have been used for im-aging both tumor development (28) and the pathophysiology of Alzheim-er’s (29) by replacing a small part of the skull with a coverslip. Alterna-tively, imaging of neuronal processes through thinned skulls is also pos-sible (27). However, other techniques, using single photon excitation, have also been applied with great success to the imaging of protein interactions (30) and the analysis of gene expres-

sion (31) in living animals. These approaches are likely to be developed toward medically applicable systems for diagnosis and treat-ment of patients, extending the capabilities of existing magnetic resonance imaging and positron emission tomography technologies. Optical imaging is likely to be of great benefit to the application of gene therapy in combina-tion with nontoxic fluorescent reporters and of monitoring cancer progression and treat-ment. Similarly, confocal imaging has recently been coupled with endoscopy (32) with diag-nostic potential.

Other imaging modes

Bright-field imaging. Imaging living cells with transmitted light is often used along with fluorescence microscopy in order to provide information on cell shape, position, and motil-ity. This is absolutely vital when studying pro-cesses such as apoptosis and mitosis, where cells undergo drastic shape changes. Phase contrast and differential interference contrast (DIC, also called Normarski) microscopy are the most commonly used. To switch between transmitted and fluorescence imaging under computer control requires a shutter on the transmitted light path in addition to a fluo-rescence shutter. A complication with DIC is that it needs a polarizing filter (the analyzer)

between the objective and the camera, and if this is left in place during fluorescence im-age capture it will reduce the intensity of the image reaching the camera. There are some systems available to avoid this, which have a motorized analyzer that can be moved out of the light path, or an analyzer that works only at certain wavelengths of light.

Whilst a simple image captured by the mi-croscope camera system will be enough for many experiments, there is huge potential for obtaining detailed insight into cell function by pushing transmitted systems to the limits

by using the best possible optics and specialized image-processing equipment (33). This is clearly il-lustrated in Fig. 3 and associated movies S1 and S2 [see also (1)]. In addition, there are other special-ized light microscopy techniques, such as reflection contrast mi-croscopy (34) and DRIMAPS [Digitally Recorded Interfer-ence Microscopy with Automatic Phase Shifting, (35)], that provide different types of information about cell structure and function. However, all of these advanced transmitted light techniques do require specialist equipment and knowledge, which probably explains their rather limited use at present. In addition, research-ers will often only be interested in a single organelle or structure, which will make fluorescence the method of choice.

Total internal reflection. Many cellular processes occur in specif-ically restricted areas of the cell, such as the plasma membrane. Total internal reflection fluores-cence microscopy (TIRFM) (Fig. 4) provides a means of direct imaging of processes within very close proximity to the coverslip (36). Excitation at a critical angle generates an evanescent field of excitation light that decays rapidly with distance from the coverslip, limiting the depth of excitation to a distance of ∼100 nm. TIRFM of live cells has given insight into the role of actin and dynamin in endocytosis (37) and can also be combined with other techniques such as photobleaching (38) or widefield im-aging. One of the most exciting recent devel-opments has been the ability to image single molecules in living cells (39), and examples of this include growth factor receptor signaling (40) and viral infection (41). Although techni-cally demanding and requiring state-of-the-art equipment, the coupling of these technologies with GFP-based expression strategies is sure to lead to further developments in the near future. Systems for TIRFM are now commer-cially available but, as with most of the recent developments in microscopy, require skilled use and careful interpretation of data.

Fluorescence correlation spectroscopy. Cel-lular processes can also be imaged using a very small area of illumination by fluores-cence correlation spectroscopy (FCS), which is used to quantitatively measure local con-centration and diffusion of particles through a very small volume and can now be applied to live cells (42). Due to its high sensitivity, the technique is prone to imaging artefacts such as intramolecular changes in fluorescence,

be monitored by measuring the proximity of one component to another. In the context of light microscopy, this can be achieved with the use of FRET. FRET oc-curs when two spectrally overlap-ping fluorophores are very close together and in an orientation such that dipole-dipole coupling results in a transfer of energy from one probe to another (47). Because the efficiency of FRET depends on the inverse sixth power of the distance between the donor and acceptor (47), this allows measurement of protein-protein interactions in live cells. Limitations of this approach are that FRET is extremely in-efficient, and many hypotheti-cal FRET pairs do not produce FRET in live cells. The most reli-able and reproducible examples of FRET occur when the donor and acceptor fluorophores are covalently linked to one another (48). Excellent examples of this include the elimination of FRET after caspase cleavage of a linker between donor and acceptor mol-ecules (49) and the application of FRET to biosensors measuring intracellular processes such as calcium flux (7). Further devel-opments of FRET pairs and im-provements in imaging methods (50) will doubtless enhance the applicability of FRET to live cell studies.

Fluorescence lifetime imaging (FLIM). The detection of fluores-cent probes is typically achieved

by counting the number of photons emitted by the excited state of a fluorophore. An al-ternative approach is to measure the lifetime of this excited state with the use of FLIM (51). This provides a means for detecting multiple fluorophores in live cells including spec-trally related molecules such as GFP vari-ants (52), which have different fluorescence lifetimes despite substantially overlapping spectra. FLIM provides an excellent means for measuring FRET because the lifetime of the excited state decreases greatly when FRET is occurring (essentially there is an additional means for decay from the excited state). FLIM measurement of FRET has re-cently been applied to the imaging of kinase activation (53). However, there are a number of limitations of the approach: resolution is limited, and FLIM is extremely difficult to perform on live cells. Despite the advent of commercial add-on packages for confocal microscopes, a key limitation remains that FLIM is technically very demanding and also requires complicated mathematical analysis of results.


Fig. 2. Examples of 3D images obtained by confocal and widefield deconvolution microscopy. (A) A mitotic spindle in a DLD1 cell imaged by single photon confocal microscopy (61). (B and C) A mitotic spindle in a Xenopus XLK2 cell imaged by 3D widefield microscopy [adapted from (21), with permission from Eaton Publishing]. A single plane of a z series without additional processing [(B), original data] and the same data set after restoration by constrained iterative deconvolution [(C), restored] are shown. Scale bar, 2 μm.

Fig. 3. Video-enhanced transmitted light microscopy. Imaging living cells by video-enhanced differential interference contrast (A and B) and phase contrast (C and D) microscopy reveals a wealth of information on organelles including mitochondria (arrows) and the endoplasmic reticulum (ER) (arrowheads). The nucleus (N) and centrosomal area (C) are marked in (A). The ER is more obvious in the associated movie clips of a Vero cell [(A) and (B), from movie S1] and a Xenopus tissue culture (XTC) cell [(C) and (D), from movie S2], imaged as described (61). An immunofluorescence image of the ER in an XTC cell (61) is shown for comparison (E and F). (B), (D), and (F) are enlargements of the boxed areas in (A), (C), and (E). Scale bars in (A), (C), and (E), 2 μm ; in (B), (D), and (F), 1 μm.

Fig. 4. An evanescent field occurs when incident light passes from a medium of high refractive index (glass) to one of low refractive index (water or a cell). Total internal reflection occurs when the angle of inci-dent light exceeds a critical angle α. This field decays rapidly and, there-fore, only illuminates ∼100 nm of the sample closest to the coverslip. This enables specific visualization of only those fluorophores in direct proxim-ity to the plasma membrane (shown in red), not those lying further away (green). This illumination mode can be coupled with conventional widefield microscopy to allow combined imaging of events close to and away from the coverslip.








MICROSCOP Y NOW UPDATE: GET TING THE MOST FROM YOUR IMAGING SECTION THREE | ARTICLES:RESEARCH


Priscila F. Siesser,1* Marta Motolese,2*† Matthew P. Walker,1 Dennis Goldfarb,1, 3

Kelly Gewain,1 Feng Yan,1 Rima M. Kulikauskas,4 Andy J. Chien,4 Linda Wordeman,5 Michael B. Major1‡

INTRODUCTION

The FAM123 gene family comprises three members: FAM123A, which is also known as AMER2; FAM123B, which is also known as WTX,AMER1, and OSCS; and FAM123C. The founding member,

FAM123B (hereinafter referred to as WTX), plays fundamental roles in normal devel-opment and human disease. Mutations in WTX contribute to various diseases, such as Wilms tumor, a pediatric kidney cancer (1, 2), and osteopathia striata congenita with cranial sclerosis (OSCS), an X-linked devel-opmental disorder that causes bone-related defects in females (3) and is lethal in males, often at embryonic or neonatal developmen-

FAM123A binds to microtubules and inhibits the guanine nucleotide exchange factor ARHGEF2to decrease actomyosin contractility

1Department of Cell and Developmental Biology, Lineberger Comprehensive Cancer Center, University of North Carolina at Chapel Hill School of Medicine, Box 7295, Chapel Hill, NC 27599, USA. 2Howard Hughes Medical Institute, Department of Pharmacology, and Institute for Stem Cell and Regenerative Medicine, University of Washington School of Medicine, Box 357370, Seattle, WA 98195, USA. 3Department of Computer Science, University of North Carolina at Chapel Hill, Box 3175, Chapel Hill, NC 27599, USA. 4Division of Dermatology, Department of Medicine, University of Washington School of Medicine, Box 358056, 815 Mercer Street, Seattle, WA 98109, USA. 5Department of Physiology and Biophysics, University of Washington School of Medicine, Box 357370, Seattle, WA 98195, USA.*These authors contributed equally to this work.†Present address: Istituto NeurologicoMediterraneo Neuromed, Pozzilli 86077, Italy.*Corresponding author. E-mail: [email protected]

tal stages. Mice lacking WTX display various developmental abnormalities in tissues of mesenchymal origin, such as increased bone mass and decreased adipose tissue (4). Cel-lular and molecular analyses of these tissues indicate a critical role for WTX in regulating cell differentiation programs in mesenchymal progenitors.

Mass spectrometry (MS)–based proteomic dissection of WTX protein complexes has revealed several core components of the β-catenin–dependent WNT signal transduc-tion pathway, including β-catenin (encoded by CTNNB1), βTrCP2 (encoded by FBXW11), and adenomatous polyposis coli (APC) (5, 6). Sub-sequent functional studies in various cells and organisms demonstrated that WTX inhibits WNT pathway activity (4, 5, 7). In vitro stud-ies and cell-based assays suggest that WTX promotes β-catenin ubiquitination and sub-sequent proteasomal degradation, perhaps by serving as a membrane-bound scaffold for the β-catenin phosphorylation complex (5–8).

Of the FAM123 family members, WTX and FAM123A share greatest homology, par-ticularly in their N termini (6, 9, 10). Two conserved functional domains in WTX and FAM123A have been identified and charac-terized (6, 11). First, both proteins share an N-terminal phosphatidylinositol (4, 5)-bispho-sphate–binding domain that localizes these

proteins to the plasma membrane and is re-quired for WTX- and FAM123A-mediated in-hibition of WNT signal transduction. Second, WTX and FAM123A directly bind to APC and regulate its subcellular distribution, recruit-ing it from the microtubule tip complex to the plasma membrane. Although the functional consequences of this redistribution are not completely understood, the role of APC in microtubule stabilization and maintenance of cell-cell junctions suggests that WTX and FAM123A may influence directional cell mi-gration and polarity (6). Whether the more distantly related family member FAM123C also regulates WNT signaling, localizes to the plasma membrane, or binds APC remains unknown.

In contrast to WTX, the cellular, devel-opmental, and disease contributions of FAM123A and FAM123C remain less under-stood. Thus, we defined and compared the protein-protein interaction networks for each member in the FAM123 family. Functional annotation of the resulting protein interac-tion network and comparative protein dy-namic studies supports both conserved and divergent functions for the FAM123 family members. Here, we report a “family-unique” function for FAM123A in controlling com-munication between the microtubule and actomyosin cytoskeletal networks. We found that FAM123A binds the microtubule plus-end tracking proteins end-binding protein 1 (EB1) and EB3; moves on microtubules; and controls microtubule dynamics, actomyosin organization, and cell migration. We present a model wherein FAM123A binds to and in-hibits guanine nucleotide exchange factor H1 (GEF-H1; encoded by ARHGEF2) to decrease actomyosin contractility.

RESULTS

Comparative proteomics of the WTX family reveals shared and unique interactions

To provide insight into the cellular functions of FAM123A and FAM123C, we defined their protein interaction networks by shotgun liq-uid chromatography–tandem MS of affinity-purified protein complexes. Integration of the resulting protein interaction networks with a previously defined WTX protein interaction network revealed shared and unique protein-protein interactions (Fig. 1A and table S1). Consistent with the homology relationships within the family, FAM123A and WTX shared several common interacting proteins; in con-trast, the FAM123C protein interaction net-work was distinct (Fig. 1B and table S1).

Because of their homology, overlapping protein interaction networks, and shared function as inhibitors of WNT signaling, protein interactions from the WTX and FAM123A networks were validated. FAM123A or WTX protein complexes were isolated by streptavidin-affinity purification from

The FAM123 gene family comprises three members: FAM123A, the tumor suppressor WTX (also known as FAM123B), and FAM123C. WTX is required for normal development and causally contributes to human disease, in part through its regulation of β-catenin–depen-dent WNT signaling. The roles of FAM123A and FAM123C in signaling, cell behavior, and human disease remain less understood. We defined and compared the protein-protein interaction networks for each member of the FAM123 family by affinity purification and mass spectrometry. Protein localization and functional studies suggest that the FAM123 family members have conserved and divergent cellular roles. In contrast to WTX and FAM123C, we found that microtubule-associated proteins were enriched in the FAM123A protein interaction network. FAM123A interacted with and tracked with the plus end of dynamic microtubules. Domain interaction experiments revealed a “SKIP” amino acid motif in FAM123A that mediated interaction with the microtubule tip tracking proteins end-binding protein 1 (EB1) and EB3—and therefore with microtubules. Cells depleted of FAM123A showed compartment-specific effects on microtubule dynamics, increased actomyosin contractility, larger focal adhesions, and decreased cell migration. These effects required binding of FAM123A to and inhibition of the guanine nucleotide exchange factor ARHGEF2, a microtubule-associated activator of RhoA. Together, these data sug-gest that the SKIP motif enables FAM123A, but not the other FAM123 family members, to bind to EB proteins, localize to microtubules, and coordinate microtubule dynamics and actomyosin contractility.

Conclusions and perspectives

The rapid development of live cell micros-copy has required input from biologists, who have provided new fluorescent probes that can be easily adapted to study myriad different proteins, and physicists, who have driven the improvement in microscope sys-tems and software. These contributions have led to unprecedented access to sophisticated imaging technology. For example, many re-searchers who have no previous microscopy experience may now use a departmental con-focal microscope. It is clear that research in specialist laboratories will continue to drive developments in light microscopy. The reso-lution attainable by light microscopy is being enhanced by recent developments in imaging that break the diffraction limit (54), and such approaches can be applied to live cells. Re-cent work using stimulated emission deple-tion (STED) to quench excited fluorophores at the rim of the focal illumination spot has enabled a substantial increase in resolution to below the diffraction limit, giving a spot size of 100 nm (55). 4Pi confocal microscopy, in which two opposing objective lenses are used to sharpen the point spread function of illumination (56), has been used for live cell imaging, and incorporation of STED with 4Pi microscopy has reduced the spot size to 33 nm (57). Computational adaptive optics, widely used by astronomers, can be used to correct for changes in refractive in-dex within thick specimens (58). Alternative approaches to increase attainable resolution include scanning near-field optical micros-copy (SNOM, also known as NSOM). SNOM is similar to atomic force microscopy in that a sharp probe physically scans the surface of the sample; it can also be coupled to fluores-cence imaging (59), where excitation light is guided through this probe, and its applica-tion to living cells is under development.

The field of live cell imaging is also of great interest to pharmaceutical and bio-technology companies (60). Many are now developing high-throughput and high-content screening platforms for automated analysis of intracellula r localization and dynamics. This is paralleled with the increasing devel-opment of fluorescent biosensor assays that provide an optical readout of a physiologi-cal effect, often based on GFP technology or bioluminescence. Clearly, future develop-ments in this field will be of great interest and benefit to both biotechnology and curiosity-driven research.

REFERENCES AND NOTES 1. C. L. Rieder, A. Khodjakov, Science 300, 91

(2003). 2. D. W. Piston, S. M. Knobel, Trends Endocrinol.

Metab. 10, 413 (1999). 3. J. Lippincott-Schwartz, G. H. Patterson,

Science 300, 87 (2003). 4. A. F. Fradkov et al., FEBS Lett. 479, 127

(2000).

5. C. Janetopoulos, T. Jin, P. Devreotes, Science 291, 2408 (2001).

6. T. Balla, P. Varnai, Sci. STKE 2002, pl13 (2002), http://stke.sciencemag.org/cgi/content/full/OC_ sigtrans;2002/125/pl3.

7. A. Miyawaki et al., Nature 388, 882 (1997). 8. J. C. Simpson, V. E. Neubrand, S. Wiemann,

R. Pepperkok, Histochem. Cell Biol. 115, 23 (2001).

9. B. A. Griffin, S. R. Adams, R. Y. Tsien, Science 281, 269 (1998).

10. G. Gaietta et al., Science 296, 503 (2002).11 . K. Stroffekova, C. Proenza, K. G. Beam, Pflug.

Arch. 442, 859 (2001).12. R. P. Haugland, Handbook of Fluorescent

Probes and Research Products (Molecular Probes, Eugene, OR, ed. 9, 2002).

13. C. M. Waterman-Storer, A. Desai, J. C. Bulinski, E. D. Salmon, Curr. Biol. 8, 1227 (1998).

14. G. H. Patterson, J. Lippincott-Schwartz, Science 297, 1873 (2002).

15. T. J. Mitchison, J. Cell Biol. 109, 637 (1989).16. R. Ando, H. Hama, M. Yamamoto-Hino, H.

Mizuno, A. Miyawaki, Proc. Natl. Acad. Sci. U.S.A. 99, 12651 (2002).

17. J. B. Pawley, in The Handbook of Biological Confocal Microscopy (Plenum, New York, ed. 2, 1995).

18. W. B. Amos, in Protein Localization by Fluorescence Microscopy: A practical approach, V. J. Allan, Ed. (Oxford Univ. Press, Oxford, 2000), pp. 67–108.

19. Technical Note 14 (Roper Scientific, Trenton, NJ, 2002), available at www.roperscientific.com/pdfs/ technotes/onchipgain.pdf.

20. M. Straub, P. Lodemann, P. Holroyd, R. Jahn, S. W.Hell, Eur. J. Cell Biol. 79, 726 (2000).

21. W. Wallace, L. H. Schaefer, J. R. Swedlow, Biotechniques 31, 1076 (2001).

22. J. R. Swedlow, K. Hu, P. D. Andrews, D. S. Roos, J. M. Murray, Proc. Natl. Acad. Sci. U.S.A. 99, 2014 (2002).

23. J. R. Swedlow, I. Goldberg, E. Brauner, P. K. Sorger, Science 300, 100 (2003).

24. J. A. Steyer, W. Almers, Biophys. J. 76, 2262 (1999).

25. J. B. Pawley, Scanning 24, 241 (2002).26. W. Denk, J. H. Strickler, W. W. Webb, Science

248, 73 (1990).27. F. Helmchen, M. S. Fee, D. W. Tank, W. Denk,

Neuron 31, 903 (2001).28. R. K. Jain, L. L. Munn, D. Fukumura, Nature

Rev. Cancer 2, 266 (2002).29. R. H. Christie et al., J. Neurosci. 21, 858

(2001).30. P. Ray et al., Proc. Natl. Acad. Sci. U.S.A. 99,

3105 (2002).31. J. C. Wu, M. Inubushi, G. Sundaresan, H. R.

Schelbert, S. S. Gambhir, Circulation 105, 1631 (2002).

32. T. F. Watson, M. A. Neil, R. Juskaitis, R. J. Cook, T. Wilson, J. Microsc. 207, 37 (2002).

33. D. G. Weiss, W. Maile, R. A. Wick, W. Steffen, in Light Microscopy in Biology: A practical approach, A. J. Lacey, Ed. (Oxford Univ. Press, Oxford, 1999), pp. 73–150.

34. J. S. Ploem, F. A. Prins, I. Cornelese-ten Velded, in Light Microscopy in Biology: A practical approach, A. J. Lacey, Ed. (Oxford

Univ. Press, Oxford, 1999), pp. 275–310.35. D. Zicha, E. Genot, G. A. Dunn, I. M. Kramer, J.

Cell Sci. 112, 447 (1999).36. J. A. Steyer, W. Almers, Nature Rev. Mol. Cell

Biol. 2, 268 (2001).37. C. J. Merrifield, M. E. Feldman, L. Wan, W.

Almers, Nature Cell Biol. 4, 691 (2002).38. S. E. Sund, D. Axelrod, Biophys. J. 79, 1655

(2000).39. A. Ishijima, T. Yanagida, Trends Biochem. Sci.

26, 438 (2001).40. Y. Sako, S. Minoghchi, T. Yanagida, Nature Cell

Biol. 2, 168 (2000).41. G. Seisenberger et al., Science 294, 1929

(2001).42. P. Schwille, U. Haupts, S. Maiti, W. W. Webb,

Biophys. J. 77, 2251 (1999).43. J. Lippincott-Schwartz, E. Snapp, A.

Kenworthy, Nature Rev. Mol. Cell Biol. 2, 444 (2001).

44. T. Misteli, Science 291, 843 (2001).45. R. D. Phair, T. Misteli, Nature Rev. Mol. Cell

Biol. 2, 898 (2001).46. K. Hirschberg et al., J. Cell Biol. 143, 1485

(1998).47. L. Stryer, Annu. Rev. Biochem. 47, 819 (1978).48. R. Y. Tsien, A. Miyawaki, Science 280, 1954

(1998).49. X. Xu et al., Nucleic Acids Res. 26, 2034

(1998).50. T. Zimmermann, J. Rietdorf, A. Girod, V.

Georget, R. Pepperkok, FEBS Lett. 531, 245 (2002).

51. J. R. Lakowicz, H. Szmacinski, K. Nowaczyk, K. W. Berndt, M. Johnson, Anal. Biochem. 202, 316 (1992).

52. R. Pepperkok, A. Squire, S. Geley, P. I. Bastiaens, Curr. Biol. 9, 269 (1999).

53. T. Ng et al., Science 283, 2085 (1999).54. E. H. Stelzer, Nature 417, 806 (2002).55. T. A. Klar, S. Jakobs, M. Dyba, A. Egner, S. W.

Hell, Proc. Natl. Acad. Sci. U.S.A. 97, 8206 (2000).

56. H. Kano, S. Jakobs, M. Nagorni, S. W. Hell, Ultramicroscopy 90, 207 (2001).

57. M. Dyba, S. W. Hell, Phys. Rev. Lett. 88, 163901 (2002).

58. M. J. Booth, M. A. Neil, R. Juskaitis, T. Wilson, Proc. Natl. Acad. Sci. U.S.A. 99, 5788 (2002).

59. F. de Lange et al., J. Cell Sci. 114, 4153 (2001).60. R. A. Blake, Curr. Opin. Pharmacol. 1, 533

(2001).61 . Materials and methods are available as

supporting material on Science Online.62. D.J.S. is supported by a Research Career

Development Award from the Medical Research Council (grant G120/617). Figure 2 was produced with the use of equipment provided by the Biotechnology and Biological Sciences Research Council, UK Bioimaging Initiative (grant C11195) and the Wellcome Trust (grant 043846) to V.J.A.

SUPPORTING ONLINE MATERIALwww.sciencemag.org/cgi/contentfull/300/ 5616/82/ DC1Materials and MethodsMovies S1 and S2

Originally published 4 September 2012 in SCIENCE SIGNALING

http://stke.sciencemag.org/content/5/240/ra64.full#ref-1

















http://stke.sciencemag.org/content/5/240/ra64.full#F1


http://www.sciencemag.org/cgi/content/full/300/5616/82/

http://www.sciencemag.org/cgi/content/full/300/5616/82/




human embryonic kidney (HEK) 293T cells, and associated endogenous proteins were detected by Western blot. Whereas both WTX and FAM123A proteins associated with APC and βTrCP1 (encoded by BTRC) and βTrCP2 (encoded by FBXW11), WTX specifically bound β-catenin, confirming the proteomic data (Fig. 1C). These results strengthen models in which WTX and FAM123A functionally regulate β-catenin–dependent WNT signaling at the level of the destruction complex, most likely through interactions with APC, βTrCP1, or βTrCP2.

We chose to further investigate the enrich-ment of microtubule-associated proteins in the FAM123A protein interaction network

(Fig. 1A, green nodes), such as EB1 (encoded by MAPRE1), an obligate component of the microtubule plus-end tip tracking protein complex (12–14). Western blot analysis of affinity-purified protein complexes revealed an interaction between FAM123A and en-dogenous EB1 but not between WTX and EB1 (Fig. 1C). Additionally, several other proteins in the FAM123A network bind to the micro-tubule cytoskeleton: the microtubule affin-ity–regulating kinases MARK2 and MARK3, GEF-H1, EB3 (encoded by MAPRE3), ACF7 (actin-crosslinking factor 7; encoded by MACF1), CLASP2 (cytoplasmic linker pro-tein–associated protein 2), and APC (15–20) (table S1). By affinity purification and mass

spectrometry (AP/MS), these proteins are the frequently observed FAM123A interact-ing proteins, and are largely absent from the WTX protein interaction network (Fig. 1A).

FAM123A moves on growing microtubules

WTX and FAM123A, but not FAM123C, localize predominantly to the cytoplasmic face of the plasma membrane through interactions with phospholipids (fig. S1) (6). Given the relative enrichment of microtubule-associated pro-teins in the FAM123A protein interaction net-work, we hypothesized that a pool of FAM123A might localize to and move on microtubules.

Fig. 1. FAM123A associates with a microtubule-enriched protein interaction network and moves on microtubules. (A) WTX and FAM123A protein interaction networks as defined by AP/MS. Proteins shown were represented by at least two unique peptides in at least two (of four) WTX experiments or at least two (of three) in FAM123A experiments. DB, database. (B) The FAM123C protein interaction network, as defined by proteins that were identified by at least two independent peptides in at least two of three experiments. In (A) and (B), node size and coloring reflect spectral counts and gene ontology, respectively. (C) Streptavidin affinity–purified protein complexes from HEK293T cells stably expressing SBPHA-GFP, SBPHA-WTX, or SBPHA-FAM123A were analyzed by Western blot for the indicated endogenous proteins (SBP, streptavidin binding peptide; HA, hemagglutinin). Data represent two biological replicates. (D) Images from movies of HT1080 cells that were transiently transfected with Venus-tagged FAM123A (movie S1) or WTX (movie S2) constructs. Data are representative of two independent biological replicates. Scale bar, 20 μm. (E) HeLa cells were transfected with EGFP-FAM123A and mCherry-tagged α-tubulin, treated with dimethyl sulfoxide (DMSO) or nocodazole, fixed, and imaged by deconvolution microscopy. The FAM123A image is representative of five independent experiments imaged on five different days. Of 65 imaged cells, 39 exhibited long filamentous structures in cells expressing Venus-FAM123A that colocalized with microtubules. Scale bar, 5 μm.

Fig. 2. Interaction of FAM123A with EB1. (A) EB1 or anti-mouse immunoprecipitates from HEK293T cells stably expressing SBPHA-FAM123A were immunoblotted for the indicated proteins. Data represent four biological replicates. (B) Top: FAM123A mRNA quantitation by quantitative PCR of HEK293T cells transfected with the indicated siRNAs. The mRNA copy number of FAM123A was normalized to 18S ribosomal RNA. Error bars represent SD in the PCR reactions. Bottom: Western blot analysis of FAM123A abundance in HEK293T cells stably expressing SBPHA-FAM123A that were transfected with the indicated siRNAs. (C) Western blot analysis of endogenous FAM123A in HEK293T cells transfected with control or FAM123A-specific siRNA#2. (D) Endogenous EB1 protein complexes from HEK293T cells were immunoblotted for FAM123A. (E) Endogenous EB1 immunoprecipitates from HeLa cells were immunoblotted for FAM123A. (F) HeLa cells transfected with EGFP-FAM123A were fixed, stained for endogenous EB1, and imaged with confocal microscopy. A Z-projection for all captured 0.2-μm slices and a single 0.2-μm slice at the bottom of the cell is shown. The colocalization of EGFP-FAM123A with EB1 was analyzed in two independent experiments imaged on two different days. For 20 EGFP-FAM123A cells analyzed, all 20 showed colocalization with EB1. Scale bar, 10 μm.

Fig. 3. FAM123A binds APC and EB1 through distinct domains. (A) Cells stably expressing SBPHA-FAM123A were transiently transfected with the indicated siRNAs, lysed, subjected to streptavidin-affinity purification (AP), and Western-blotted. (B and C) HEK293T cells transiently transfected with EB1-EGFP (B) or APC (amino acids 1 to 1060) (C) and the indicated SBPHA-FAM123A fragment were lysed, subjected to streptavidin-affinity purification, and Western-blotted. Data represent three biological replicates. (D) Protein domain interaction mapping shown in (B) and (C) and fig. S2.












Imaging of Venus-FAM123A or Venus-WTX in live HT1080 human sarcoma cells (Fig. 1D) indicated that both proteins exhibited mainly membranous distribution, seemed to be en-riched at cell ruffles, and induced cell death when overexpressed. In cells expressing low to moderate amounts, FAM123A (movie S1), but not WTX (movie S2), localized to filamen-tous structures resembling microtubules. This localization pattern was often polarized with respect to the direction of cell movement; of 142 cells with low to moderate expression of FAM123A, 94 cells showed FAM123A decorat-ing filamentous structures polarized in the direction of cell movement. Colocalization of FAM123A and α-tubulin confirmed the mi-crotubule localization suggested by live-cell imaging (Fig. 1E), which was abolished by the addition of the microtubule destabilizing drug nocodazole. By high-resolution live-cell imaging, EGFP (enhanced green fluorescent protein)–FAM123A was observed to predomi-nantly move on segments of microtubules and was also observed as dots that moved on lin-ear trajectories from the cell body toward the cell periphery (movie S3). Thus, both compar-ative proteomics and live-cell imaging demon-strate that FAM123A, but not WTX, associates with and moves on microtubules.

FAM123A interacts with APC and EB1 through distinct domains

MS and Western blot analysis revealed that FAM123A associated with the plus-end tracking protein EB1 (Fig. 1, A and C). Sta-bly expressed hemagglutinin (HA)–tagged FAM123A was detected in protein complexes isolated by immunopurification of endog-enous EB1 (Fig. 2A). To demonstrate the in-teraction between the endogenous proteins, we generated an antibody against FAM123A and confirmed its specificity with two non-overlapping FAM123A-specific small inter-fering RNAs (siRNAs) (Fig. 2, B and C). In both HEK293T and HeLa cells, endogenous FAM123A was detected within endogenous EB1 immunopurified protein complexes (Fig. 2, D and E). Confocal Z-series images of HeLa cells expressing EGFP-FAM123A and immu-nofluorescently labeled for endogenous EB1 demonstrated basal colocalization of these proteins with increased intensity at the mi-crotubule distal ends toward the leading edge (Fig. 2F).

APC directly interacts with EB1 and FAM123A. To determine whether FAM123A indirectly binds EB1 through APC (21), we purified FAM123A from cells transfected with siRNAs targeting either APC or EB1, and found that FAM123A interacts with EB1 in the absence of APC, and similarly inter-acts with APC after silencing of EB1 (Fig. 3A). These results suggest that FAM123A indepen-dently binds APC and EB1. Although WTX as-sociates with APC, it did not affinity-purify with EB1 (Fig. 1C). To map the domains of

FAM123A that interacted with APC and EB1, we generated a series of FAM123A deletion mutants on the basis of homology within the WTX family and of secondary structure pre-diction for FAM123A (http://robetta.baker-lab.org/). Affinity purification of full-length FAM123A protein or the truncated fragments from cells expressing EB1 or APC indicated that FAM123A interacted with EB1 through its extreme C terminus, specifically amino ac-ids 457 to 552 (Fig. 3B) and with APC through an N-terminal domain comprising amino acids 261 to 349 (Fig. 3C). Additionally, FAM123A interacted with βTrCP2 through a C-terminal region encompassing residues 261 to 470, which overlaps with the APC binding domain (Fig. 3D and fig. S2). These results demonstrate that FAM123A independently interacts with EB1 and APC through nonover-lapping domains.

FAM123A interacts with EB proteins through the EB binding motif “Ser-x-Ile-Pro”

Many plus-end tracking proteins contain a characteristic EB-binding motif, which is de-fined by a Ser-x-Ile-Pro (SxIP) consensus (Fig. 4A) (22), and which enables direct binding to EB1 and EB3 and, consequently, localiza-tion to the growing microtubule plus-end. We found a SKIP487–490 and a TKIP518–521 motif within FAM123A, both of which are in the EB1 binding region identified by domain mapping (Fig. 3B). To determine whether the SKIP487–490 motif is required for binding to EB1 and EB3, we created a mutant in which Ile489 and Pro490 residues were changed to alanine (FAM123A-IPAA) (Fig. 4A). Affinity purification and Western blot analysis revealed that in contrast to wild-type FAM123A, the IPAA mutant did not pull down EB1 or EB3 (Fig. 4, B and C). These results demonstrate that the SKIP487–490 motif in FAM123A is necessary for association with EB1 and EB3; whether the TKIP518–521 mo-tif contributes to binding in the presence of the SKIP487–490 motif remains to be tested.

Many plus-end tracking proteins contain-ing the “SxIP motif” bind a coiled-coil do-main within the C terminus of EBs, referred to as the EBH domain (17, 22). To determine whether FAM123A also binds to the EBH do-main of EB1, we generated two EB1 deletion constructs that encoded the N-terminal micro-tubule-binding domain (amino acids 1 to 135) or the C-terminal EBH domain (amino acids 136 to 268). Affinity purification of FAM123A from cells coexpressing these EB1 truncations revealed an association between FAM123A and the EBH domain of EB1 (Fig. 4D). Thus, FAM123A interacts with the C-terminal region of EB1 through the consensus EB1 binding motif SxIP.

The SxIP motif targets functionally and structurally unrelated plus-end tracking proteins to growing microtubule ends in an EB-dependent manner (22). To determine

whether FAM123A microtubule localization and plus-end tracking require EB associa-tion, we compared the protein distribution and dynamics of FAM123A and FAM123A-IPAA by live-cell imaging (Fig. 4E and movies S4 and S5). Wild-type FAM123A coated the distal ends of the microtubule and cotracked with EB3 in the cell body (movie S4). In con-trast, FAM123A-IPAA exhibited decreased EB3-associated plus-end tip tracking and mi-crotubule decoration, coating very few micro-tubule stretches at the cell periphery (movie S5). A C-terminal fragment of FAM123A that contains the SKIP487–490 EB-interaction motif did not decorate the microtubule lattice but rather behaved as a classic plus-end micro-tubule-binding protein showing robust EB3-associated tip tracking (fig. S3 and movie S6). Mutation of the SKIP487–490 motif com-pletely abolished microtubule colocalization and EB3-associated tip tracking (fig. S3 and movie S7).

Despite the autonomous microtubule-binding capability of EB proteins in vitro, FAM123A could influence EB1 loading or distribution on microtubules ex vivo. How-ever, depletion of FAM123A in HeLa cells did not affect the subcellular distribution of en-dogenous EB1 (Fig. 4F). To complement the loss-of-function approach, we localized EB1 in cells stably overexpressing FAM123A (fig. S4). Forced expression of FAM123A, but not of FAM123A-IPAA, relocalized EB1 to the microtubule lattice at the bottom of the cell and to the plasma membrane more apically. Although consistent with its ability to bind EB1, it remains to be seen whether EB1 re-distribution after FAM123A overexpression occurs normally. Together, these data demon-strate that (i) FAM123A is a microtubule-as-sociated protein with EB-dependent plus-end tip tracking capabilities; (ii) FAM123A binds EB proteins through the SKIP487–490 motif; (iii) FAM123A predominantly decorates the microtubule lattice at the cell periphery in a largely EB-dependent fashion; and (iv) FAM123A silencing does not affect EB1 sub-cellular localization.

EB-protein association is not required for FAM123A regulation of Wnt signaling

Of the three FAM123 family members, only FAM123A contains an SxIP motif, which is consistent with the lack of association be-tween WTX and EB1 (Fig. 1, A and C). Given this unique protein interaction within the FAM123 family, we predicted that the EB1 as-sociation would be dispensable for the com-mon functions of WTX and FAM123A, such as regulation of β-catenin–dependent WNT sig-naling (5, 7, 11). Indeed, siRNA-based knock-down of FAM123A or WTX increased the activity of a β-catenin reporter gene (fig. S5A). In a gain-of-function approach, overexpres-sion of FAM123A or FAM123A-IPAA reduced WNT3A-dependent reporter activation in a

concentration-dependent manner (fig. S5, B and C). These data suggest that the ability of FAM123A to modulate the Wnt–β-catenin pathway is independent of its interaction with EB1 and microtubules.

FAM123A controls microtubule organization and growth rates

Because FAM123A bound EB proteins and moved on microtubules, we tested whether its loss affected microtubule organization and dynamics. siRNA-mediated silencing of FAM123A in HeLa cells induced disorganiza-tion of the microtubule network with exces-sively curved microtubules and increased

microtubule density (Fig. 5, A and B). In contrast, siRNA-mediated silencing of WTX yielded a distinct microtubule organization (Fig. 5A). We used total internal reflective flu-orescence (TIRF) microscopy to image GFP-tagged EB3 in siRNA-transfected HeLa cells, and found that FAM123A promotes microtu-bule growth within the cell body because cells lacking FAM123A had significantly slower microtubule polymerization (Fig. 5, C and D). In contrast, in the absence of FAM123A, mi-crotubules demonstrated less dynamic move-ment near F-actin bundles, particularly near adhesion complexes (Fig. 5D). Thus, FAM123A regulates microtubule dynamics and the over-all organization of the microtubule network.

FAM123A inhibits actin contractility by suppressing the GEF-H1–RhoA–ROCK– MLC pathway

In addition to the altered microtubule orga-nization, we noticed that FAM123A-depleted cells had phase dark cortical membranes (Fig. 6A). These observations and the increase in cortical F-actin detected by TIRF microscopy (Fig. 5D) suggested involvement of the signal-ing pathway that mediates cross talk between the microtubule and actomyosin cytoskeletal networks (15, 23). Specifically, microtubule depolymerization induces actin stress fiber formation and cell contractility through activation of the Rho-specific exchange fac-tor GEF-H1, which subsequently activates

Fig. 4. EB1 association is required for FAM123A microtubule localization. (A) Protein sequence alignment of FAM123A and the SxIP domains of various plus-end tracking proteins. (B and C) Western blot analysis of streptavidin-affinity pulldowns from HEK293T transfected with EB1-EGFP (B) or EB3-EGFP (C) and the indicated SBPHA-FAM123A constructs. Data represent three biological replicates. (D) Western blot analysis of streptavidin-affinity pulldown assays followed by Western blot analysis of SBPHA-FAM123A cells transfected with EGFP, EB1-EGFP, EGFP-EB1-N, or EGFP-EB1-C. Data represent two biological replicates. (E) Images from live-cell imaging of HeLa cells expressing EGFP-FAM123A (movie S4) or EGFP-FAM123A-IPAA (movie S5). The colocalization of EGFP-FAM123A was compared to EGFP-FAM123A-IPAA in three replicate experiments imaged on three separate days. For EGFP-FAM123A, a total of 37 cells were imaged live in conjunction with mRFP-EB3. All cells exhibited greater than or equal to 75% localization of EGFP-FAM123A with mRFP-EB3 in thresholded images. Of 15 EGFP-FAM123A-IPAA cells, no cell exhibited more than 5% colocalization of EGFP-FAM123A-IPAA with mRFP-EB3. Scale bar, 5 μm. (F) HeLa cells transfected with the indicated siRNAs were costained for EB1 and 4′,6-diamidino-2-phenylindole (DAPI). EB1 staining is representative of two biological replicates in which 46 FAM123A-depleted cells and 33 control siRNA–transfected cells were imaged. Scale bar, 20 μm.











http://robetta.bakerlab.org/

http://robetta.bakerlab.org/































(Fig. 6K). Together, these data demonstrate that FAM123A decreases actin contractility by inhibiting the GEF-H1 pathway and that,

epistatically, FAM123A functions upstream of GEF-H1.

Because cellular morphology and cytoskel-

etal organization are particularly sensitive to transfection and siRNA-based off-target ef-fects, we further validated the specificity of

RhoA, Rho kinase (ROCK), and myosin light chain (MLC) (Fig. 6B) (24, 25). In FAM123A-silenced HeLa cells, thick and short actin stress fibers were disposed in a nonparallel arrangement, and microtubule depolymer-ization induced by nocodazole treatment resulted in several morphological changes, including thick bundles of contracted actin stress fibers frequently localized at the cell center, kidney-shaped nuclei shifted to the cell periphery, and cytoplasmic pools of free tubulin confined to a small area close to the nucleus or near the cell center (or both) (Fig. 6D). Thus, FAM123A-depleted cells display an altered microtubule network and increased actomyosin contractility that is exacerbated with nocodazole treatment.

We used multiple experimental approaches to determine whether FAM123A-induced

actomyosin contractility requires the GEF-H1–RhoA–ROCK–MLC2 pathway. First, treat-ment of cells with the ROCK inhibitor Y-27632 (26) abolished actin stress fiber formation and cell contractility induced by FAM123A knockdown (Fig. 6C). Second, siRNAs tar-geting GEF-H1 reversed the actomyosin con-tractility induced by FAM123A depletion but did not rescue the disordered microtubule phenotype induced by FAM123A silencing (Fig. 6D). By blinded quantitation, 78% of FAM123A-depleted cells were scored as dis-playing increased actomyosin contractility; in the absence of GEF-H1, this was reduced to 38% of cells (Fig. 6E). As was previously reported, silencing of GEF-H1 resulted in de-creased actin bundling (fig. S6) (24). Third, activation of RhoA was higher in cells de-pleted of FAM123A than in control siRNA–

transfected cells, an effect that required GEF-H1 (Fig. 6F). Fourth, phosphorylation of myosin was increased in FAM123A-silenced cells but not in cells in which both FAM123A and GEF-H1 were depleted (Fig. 6, G to I). By immunofluorescence, phosphorylated MLC localized primarily to stress fibers near the cell periphery after FAM123A loss (Fig. 6I), compared with the more uniform distri-bution seen in control siRNA–transfected cells. Finally, we used a serum response fac-tor (SRF) transcriptional activity assay as an indirect readout for RhoA activity and actomyosin contractility (27). FAM123A silencing induced SRF reporter activity in a GEF-H1–dependent fashion (Fig. 6J). FAM123A overexpression repressed SRF-mediated transcription induced by both nocodazole and GEF-H1 overexpression

Fig. 5. FAM123A depletion results in altered microtubule organization and increased actomyosin contractility. (A) HeLa cells were transfected with the indicated siRNAs and stained with an anti–α-tubulin antibody. Images are representative of three independent biological replicates. (B) Plots of fluorescence intensity of α-tubulin staining in cells from (A). The integrated intensity was measured within a 5-μm2 box at 5-μm distance from the cell periphery, in three different regions per cell for 23 control and FAM123A siRNA–transfected cells (*P < 0.0001, Welch-corrected t test). (C) Microtubule polymerization rates in siRNA-transfected HeLa cells. Ten microtubules were measured in each cell. Nine cells were analyzed for each siRNA (*P = 0.0129, Welch-corrected t test). (D) TIRF movies of siRNA-transfected HeLa cells expressing EB3-GFP (left) and RFP-Utr (middle). Left: RGB-colorized 5-s Z-projections of EB3-GFP–labeled growing microtubule ends. Growing filaments are labeled either red, green, or blue and overlapping stationary microtubule ends are white. Regions in cells lacking FAM123 where the microtubule ends are stationary (not assembling) are indicated (white arrows). Middle: Expressed RFP-Utr shows the filamentous actin near the cell substratum. Right: A 15-frame (75 s) Z-stack of EB3-GFP (green) and RFP-Utr (red) over time. Stationary microtubule ends (white arrows) are in close apposition to actin bundles at the cell periphery (red open arrows).

Fig. 6. FAM123A regulates actomyosin contractility through GEF-H1. (A) HeLa cells transfected with siRNAs were imaged by phase-contrast microscopy. Images represent more than three biological replicates. Cell morphologies were confirmed with three different FAM123A siRNAs and two different WTX siRNAs. Scale bar, 50 μm. (B) Major signaling proteins connecting microtubule destabilization to actin contractility. (C) HeLa cells were transfected with control or FAM123A siRNA. Where indicated, cells were treated with the indicated drug before costaining with an anti–α-tubulin antibody, Alexa Fluor 647–conjugated phalloidin, and DAPI. Images are representative of three independent biological replicate experiments. Scale bar, 20 μm. (D) siRNA-transfected HeLa cells were treated with nocodazole and costained with an anti–α-tubulin antibody, Alexa Fluor 647–conjugated phalloidin, and DAPI. Images are representative of three biological replicates. (E) Quantification of cell phenotypes in (D). Samples were scored in a blinded fashion by three independent investigators. n, number of cells scored. (F) siRNA-transfected HeLa cells were serum-starved, treated with nocodazole, and analyzed for RhoA activity (P = 0.002, Student’s t test of biological triplicate experiments). (G) siRNA-transfected HeLa cells were immunoblotted for the indicated proteins. Data are representative of five independent biological experiments. (H) Total MLC2 and phosphorylation of MLC2 at Ser19 from biological triplicate experiments were quantitated by LI-COR and plotted (*P = 0.0159, Student’s t test). (I) siRNA-transfected HeLa cells were stained with phospho-MLC2 (Ser19) antibody, Alexa Fluor 647–conjugated phalloidin, and DAPI. Images are representative of three biological replicates. (J) Quantitation of luciferase activity in HeLa cells expressing an SRF firefly luciferase reporter and a cytomegalovirus (CMV)–driven Renilla luciferase reporter. Cells were treated with DMSO or nocodazole before luciferase quantitation. (K) Quantitation of luciferase activity in HeLa cells that were transiently transfected with the indicated expression construct(s), SRF firefly luciferase reporter, and a CMV-driven Renilla luciferase reporter (*P < 0.05, Welch-corrected t test of biological triplicates).



















the FAM123A phenotype. HeLa cells express-ing one of two siRNAs that targeted the 3′ untranslated region of FAM123A (fig. S7A) showed an exacerbated actomyosin contrac-tility phenotype after nocodazole treatment, similar to the open reading frame–directed FAM123A#1 and #2 siRNAs. This pheno-type was rescued by forced expression of a FAM123A-GFP fusion protein (fig. S7B). To-gether, these data confirm that FAM123A de-creases actomyosin contractility.

FAM123A binds to and inhibits GEF-H1 activity

Proteomic analysis of the FAM123A protein complex identified GEF-H1 as a potential interacting protein, suggesting a molecu-lar mechanism by which FAM123A controls actomyosin contractility (Fig. 1A). First, we confirmed the GEF-H1–FAM123A interaction by Western blot analysis of affinity-purified FAM123A protein complexes or immunopu-rified GEF-H1 protein complexes (Fig. 7, A and B). Endogenous GEF-H1 was detected in affinity-purified FAM123A protein com-plexes (Fig. 7A), and FAM123A immunopu-rified with endogenous GEF-H1 (Fig. 7B). The FAM123A-IPAA mutant, which lacks the ability to bind EB proteins, also associated with GEF-H1 (Fig. 7, A and B). Additionally, we detected the endogenous forms of the microtubule-associated proteins CLASP2 and MARK2 in affinity-purified FAM123A protein complexes, as was suggested by the FAM123A proteomic analyses (Fig. 7A).

As a first test of the functional relation-ship between FAM123A and GEF-H1, we assessed the activation status of GEF-H1 in FAM123A-depleted cells by purifying ac-tive GEF-H1 with a nucleotide-free mutant of RhoA, RhoA-G17A coupled to glutathi-one S-transferase (GST) (28). Western blot analysis of RhoA-G17A pulldowns revealed a significant increase in active GEF-H1 after FAM123A depletion (Fig. 7C). We attempted to examine the subcellular localization of GEF-H1 after FAM123A overexpression or knockdown; however, we were unable to vi-sualize microtubule localization of GEF-H1 with various commercially available anti-bodies and fixation techniques. We reasoned that an effect of FAM123A on GEF-H1 activity could be indirectly assessed by global inter-rogation of the GEF-H1 protein interaction network. For this approach, we used SILAC (stable isotope labeling with amino acids in cell culture)–based quantitative proteomics of immunopurified endogenous GEF-H1 pro-tein complexes from parental cells or cells overexpressing FAM123A. Averaging the SI-LAC ratios from biological replicates identi-fied GEF-H1 protein interactions that were increased by FAM123A expression, such as the association between GEF-H1 and DSTN, an F-actin–depolymerizing protein (29, 30), and those that were decreased by FAM123A

expression, such as the association between GEF-H1 and SLK, a kinase involved in actin and microtubule organization (31–33). To-gether, these data demonstrate that FAM123A controls GEF-H1 activity and GEF-H1 protein-protein interactions.

FAM123A regulates cell adhesion and cell migration

Actin stress fibers and focal adhesions are physically and functionally tethered; increases in cellular contractility or application of exter-nal mechanical force induces simultaneous growth of focal adhesions and the attached ac-tin stress fibers (34, 35). To determine whether actomyosin contractility induced by FAM123A loss affected focal adhesions, we visualized vinculin as a marker of focal adhesions (Fig. 8A). Consistent with higher contractility, FAM123A-depleted cells had larger adhesions than control cells (Fig. 8A). To assess cell ad-hesion and spreading assay, we performed real-time quantitation of electrical impedance during cell spreading, which revealed no effect of FAM123A silencing during the early stages of cell attachment and spreading (Fig. 8B). However, FAM123A-depleted cells had higher impedance values than control cells after the first hour of cell plating, indicating increased adhesion (Fig. 8C), a phenotype that required GEF-H1 (Fig. 8, B and C). We also examined the formation of adhesion complexes by im-munostaining during cell spreading (Fig. 8D and fig. S8). Consistent with the results from the cell spreading assay, FAM123A-depleted and control cells showed phenotypically simi-lar formation of adhesion and actin stress fibers in the early stages of cell spreading. However, after the first hour, the adhesions in FAM123A-depleted cells were larger and had thicker actin stress fibers, effects that were GEF-H1–dependent because cells depleted of both FAM123A and GEF-H1 or GEF-H1 alone resembled the control siRNA–transfected cells. Given these cytoskeletal and adhesion pheno-types, we tested whether FAM123A silencing affected cell migration. HeLa cells transfected with either control or FAM123A siRNAs were plated to confluency, scratched, and imaged by live-cell microscopy. Compared with control siRNA–transfected cells, FAM123A depletion resulted in a 20% decrease in cell migration, which may have resulted from increased adhe-sion (Fig. 8E).

DISCUSSIONWe performed unbiased protein-protein in-teraction screens to discover new functions for members of the FAM123 protein family. Here, we have characterized a family-unique physical and functional relationship between FAM123A and the microtubule and acto-myosin cytoskeletal networks. We found that FAM123A binds EB proteins and interacts with dynamic microtubules through its SKIP487–490 motif. FAM123A knockdown resulted in com-

partment-specific effects on microtubule dy-namics, a globally disorganized microtubule network, increased GEF-H1 activity, increased actomyosin contractility, increased cell adhe-sion, and decreased cell migration.

Members of the FAM123 gene family control β-catenin–dependent WNT signaling

FAM123A inhibits β-catenin–dependent WNT signaling (11). Our data both confirm this findings and provide additional mechanistic insight. First, comparative protein-protein interaction studies of WTX and FAM123A re-vealed a robust association between β-catenin and WTX but not between β-catenin and FAM123A. Although a low affinity or transient interaction between FAM123A and β-catenin may occur, we interpret our data to suggest that association with β-catenin is not required for regulation of WNT signaling by WTX (or FAM123A). Consistent with this notion, the REA repeats in WTX that are responsible for mediating direct interaction with β-catenin are found only in mammalian orthologs of WTX, although WTX inhibits WNT signal-ing in zebrafish and Xenopus (7). On the ba-sis of our findings, it is likely that WTX and FAM123A inhibit WNT signaling through interactions with APC or βTrCP1/2 (or both), which associate with both WTX and FAM123A.

Our data demonstrate that the FAM123A-EB interaction is dispensable for WNT regula-tion, at least with respect to nonpolarized cells grown in two dimensions. In contrast to WTX, which is uniformly distributed across human tissues, FAM123A is largely restricted to neu-ronal tissues (9). These disparities in distribu-tion may provide an explanation as to why the ability to control cytoskeletal dynamics is specific to FAM123A, which may have evolved to remodel the cytoskeleton during neuronal migration (11). Because the FAM123C protein interaction network did not share common protein associations with FAM123A or WTX, our data suggest that FAM123C lacks regula-tory functions over WNT signaling, although this remains to be formally demonstrated.

FAM123A tracks growing microtubules

Plus-end tracking proteins comprise a struc-turally and functionally diverse group of microtubule-associated proteins that accu-mulate at the ends of growing microtubules (15, 17, 19, 36). Many plus-end tracking pro-teins have a conserved SxIP motif that di-rectly associates with EB proteins and thus mediates localization to the microtubule plus-end (22). In at least two ways, however, our protein dynamic studies also differenti-ate FAM123A from other plus-end tracking proteins.

First, unlike EB1 and many plus-end track-ing proteins that localize to microtubule tips throughout the cell body, FAM123A pre-

dominantly decorates the distal ends of mi-crotubules oriented in the direction of cell movement, a polarized distribution also seen for CLASPs, APC, and CDK5RAP2 (37–40). Through asymmetric distribution to the leading edge, these plus-end tracking pro-teins modulate microtubule dynamics and consequently promote cell polarization and directional migration (41, 42). We found that FAM123A loss differentially affects microtu-bule dynamics in different subcellular com-partments (Fig. 5). Therefore, it is possible that FAM123A functions to regulate spatially confined microtubule stability, which presum-ably contributes to the establishment, main-tenance, or modulation of asymmetric cell behavior.

Second, although they bind EB proteins for plus-end tracking, many plus-end track-ing proteins can also directly associate with microtubules. By contrast, FAM123A micro-tubule tracking and localization was largely attenuated in the absence of EB association; one possibility is that the remaining microtu-bule localization is due to bridging proteins

that tether the N terminus of FAM123A to the microtubule, such as APC. Moreover, the sub-cellular distribution of full-length FAM123A and the C-terminal fragment are substantially different, although both bind EB proteins and track with assembling microtubule ends. Whereas the C terminus colocalizes with EB comets, full-length FAM123A exhibits slower microtubule tracking near the cortex, fre-quently coating the microtubule lattice. These data suggest that FAM123A may be involved in coupling assembling microtubules to mem-brane signaling pathways.

FAM123A regulates microtubule dynamics and organization

Like many other EB-dependent plus-end tracking proteins, we found that FAM123A de-pletion results in a disorganized microtubule network. Whereas FAM123A asymmetrically localized to microtubules in the direction of cell movement, its depletion affected micro-tubule architecture throughout the entire cell body. FAM123A depletion resulted in de-

creased microtubule polymeriza-tion rates within the cell body and less dynamic and stabilized microtubules in apposition to ac-tin adhesion complexes. Although untested, it is possible that the decreased microtubule polymer-ization rates occur secondarily to the increased cortical stability because of reprogramming over-all tubulin homeostasis. For ex-ample, decreases in free tubulin dimers as a result of increased stability in some microtubules may result in global effects on microtubule dynamics (43). It also remains possible that acto-myosin contractility induced by FAM123A silencing functions in a feed-forward loop to globally influence microtubule dynamics (44). Although our data do not provide an exact mechanism, we propose that FAM123A may con-trol microtubule dynamics by reg-ulating cortical capture through the tethering of microtubules to the plasma membrane. That is, FAM123A complexes with several plus-end tracking proteins that are thought to stabilize microtu-bules at the leading edge, such as EB1, EB3, CLASPs, APC, and ACF7 (15). It is possible that in associa-tion with these plus-end tracking proteins, FAM123A asymmetri-cally tracks microtubules in the direction of cell movement and promotes both cortical microtu-bule stabilization and polarized cargo delivery to the cortical membrane. Because WTX and

FAM123A bind phospholipids and localize predominantly to the cytoplasmic membrane, an elaboration on this model might have FAM123A “jumping” off microtubules onto a phospholipid landing pad, perhaps deliver-ing its associated proteins, such as GEF-H1 and APC, to the leading edge. In agreement with this model, FAM123A recruits APC to the plasma membrane and drives EB1 to the mem-brane when overexpressed (fig. S4) (6, 11).

FAM123A inhibits actomyosin contractility, thereby regulating adhesion and cell migration

Our data suggest that FAM123A binds to and inhibits GEF-H1, a guanine nucleotide ex-change factor that localizes to microtubules, is activated by microtubule depolymerization, and activates RhoA-dependent actomyosin contraction (24, 25). We show that FAM123A depletion results in GEF-H1–dependent in-creased actomyosin contractility, enlarged fo-cal adhesions, increased cellular adhesion, and decreased cell migration (Fig. 8). These obser-

Fig. 7. FAM123A binds to GEF-H1 and regulates its activity. (A) Streptavidin-affinity pulldowns of HEK293T cells stably expressing SBPHA-tagged CCDC94, FAM123A, or FAM123A-IPAA were immunoblotted for endogenous associated proteins. (B) With the HEK293T stable cell lines described in (A), endogenous GEF-H1 was immunoprecipitated and protein complexes were analyzed by Western blot. (C) Active GEF-H1 and total cellular GEF-H1 were determined by densitometry (*P < 0.05, paired Student’s t test of six biological replicate experiments; error bars represent SE). (D) The GEF-H1 protein interaction network, illustrating the FAM123A-sensitive interactions as determined by quantitative IP/MS. The average SILAC ratios from biological duplicate experiments are shown below the network.















































vations demonstrate that FAM123A is simi-lar to p21-activated protein kinase 1 (PAK1), PAK4, calpain-6, MARK2, and extracellular signal–regulated kinase (ERK), all of which regulate GEF-H1 to control spatiotemporal regulation of the actin cytoskeleton (45–54). That said, several important questions remain regarding the precise molecular mechanism by which FAM123A regulates GEF-H1.

First, our data demonstrate that FAM123A is epistatically upstream of GEF-H1 with respect to RhoA, ROCK, MLC2, actomyosin contractil-ity, and cell adhesion but not to microtubule organization (Fig. 6). Therefore, actomyosin contractility likely underlies the FAM123A-dependent cell adhesion phenotype, rather than the altered microtubule dynamics. Given

the limitations of interpreting cause and effect from siRNA-based experiments, additional studies are needed to more precisely define the role of FAM123A in microtubule dynam-ics and to understand how that effect is com-municated through GEF-H1 to the actomyosin network. Second, as mentioned, the precise mechanism by which FAM123A inhibits GEF-H1 remains to be established. Although we show that ectopic FAM123A binds endogenous GEF-H1 (and vice versa), it remains to be de-termined whether the endogenous proteins associate. Moreover, we were unable to dem-onstrate GEF-H1 subcellular colocalization with FAM123A. Third, we found that FAM123A inhibits GEF-H1 in the presence and absence of microtubules. Specifically, the IPAA mutant

form of FAM123A, which does not bind mi-crotubules, associates with GEF-H1 (Fig. 7), and FAM123A knockdown exacerbates actin contraction in the absence of microtubules (Fig. 6). These data suggest that FAM123A functionally affects GEF-H1 independently of microtubule polymerization status, similar to tumor necrosis factor–α–mediated GEF-H1 activation in tubular epithelia (48, 49). Finally, we found that FAM123A altered the GEF-H1 protein interaction network, which we inter-pret as further corroborative evidence that FAM123A controls GEF-H1 activity or subcel-lular localization. The biological implications of these GEF-H1 interactions and their control by FAM123A await further study.

In summary, we found, using comparative

proteomic analyses of the FAM123 family, that FAM123A associates with numerous micro-tubule-binding proteins. Subsequent protein dynamic studies and functional interrogation revealed that FAM123A controls microtubule polymerization rates, actomyosin contractil-ity, and, consequently, cell adhesion and cell migration.

MATERIALS AND METHODS

Constructs

FAM123A isoform 2 complementary DNA (cDNA) was obtained by polymerase chain reaction (PCR) amplification from clone BCO32653 (Open Biosystems). The mutant FAM123A-IPAA (Ile489-Pro490 mutagenized to Ala489-Ala490) was created by standard PCR-based mutagenesis. The WTX constructs were previously described (5). The EB3-pEG-FPN1 and EB3-RFP (red fluorescent protein) constructs were provided by A. Akhmanova (Erasmus Medical Center, Rotterdam, Neth-erlands). The EB1-pEGFPN1 construct was provided by L. Cassimeris (Lehigh University, Bethlehem, PA). EB1-pEGFPN2 and the dele-tions were obtained by amplifying EB1 by PCR from pEGFPN1-EB1 construct. The mCherry-tubulin plasmid was made by replacing GFP at the Bsr GI and Bam HI sites in pEGFP-Tub with mCherry. The β-catenin–activated firefly reporter (pBAR) and pcDNA3.1-Flag-APC (amino acids 1 to 1060) were previ-ously described (5). The SRF-RE reporter pGL4.34[luc2P/SRF-RE/Hygro] was obtained from Promega. GFP-GEF-H1 was kindly pro-vided by R. Garcia-Mata (UNC-Chapel Hill).

Reagents

Wnt3A and control conditioned media were produced with mouse fibroblasts (L cells) ac-cording to the American Type Culture Collec-tion protocol. Nocodazole and Y-27632 were purchased from Sigma-Aldrich Corporation (catalog nos. M1404 and Y0503).

Tissue culture, transfections, and reporter assays

All cells were grown in Dulbecco’s modified Eagle’s medium (DMEM) supplemented with 10% fetal bovine serum (FBS) and 1% penicil-lin/streptomycin, in a 37°C humidified incu-bator with 5% CO

2. Transient transfections of

siRNAs were performed with Lipofectamine RNAiMAX, as directed by the manufacturer (Invitrogen). All siRNAs were used at a final concentration of 20 nM and for 72 hours un-less otherwise stated. siRNAs targeting human forms of WTX (siRNA#1 and #2), APC, axin1/2, and β-catenin have been previously published (5). The following siRNAs were used: Stealth M2 (Invitrogen) as control; EB1, AAGUGAAAU-UCCAAGCUAAGC (42);FAM123A siRNA#1, GCCGGCUCUGUCUAAAAAG[dT ][dT ]; FAM123A siRNA#2, GAGAUAUUAUUG-

CAGACCAAGAGG; FAM123A siRNA#3, CCUACGUUCAGUUGUUAGAUAUGCA; FAM123A siRNA#4, CCUCUCAAGAU-AAGUCCCUGAGAAU; GEF-H1 siRNA#1, AGCAGGCCACGGAACUGGCAUUACU; GEF-H1 siRNA#2, UCCAUACACGCUUC-CUCAGCCAGCU; WTX siRNA#3, AAAGGCA-GUCAUCUCCAGGUGGAGA. All siRNAs were synthesized by Invitrogen as Stealths, except for EB1 and FAM123A#1. Expression constructs were transiently transfected into HEK293T cells with Lipofectamine 2000 (In-vitrogen) or into HeLa cells with FuGENE 6-HD (Roche), as directed by the manufac-turer. For gain-of-function reporter assays, cells were seeded in 48-well plates before transfection with a firefly luciferase reporter, Renilla luciferase control reporter, and effec-tor plasmids. Cells stably engineered with these reporters were used for the loss-of-function experiments. Luciferase activity was quantitated with the Dual-Glo Luciferase As-say System (Promega).

Affinity purification and MS

Tandem affinity purification of FAM123A (isoform 2) was performed as previously de-scribed (55, 56), with minor modifications. Protein complexes were eluted from the streptavidin beads with 5 mM biotin in the absence of TEV protease. PPS Silent Surfac-tant (0.1%; Protein Discovery) was included in the final elution from the calmodulin beads. Before MS, PPS was acid-cleaved at 37°C for 30 min. Flag-based AP/MS of FAM123C was performed in triplicate as pre-viously described (57). Quantitative immuno-precipitation and mass spectrometry (IP/MS) of endogenous GEF-H1 was performed in du-plicate via SILAC (K6/R10) labeling. Briefly, protein lysates (~150 mg) from the following HEK293T-derived cell lines were compared at low confluency: parental cells (light) and Flag-FAM123A (heavy) or GFP (light) and HA-FAM123A (heavy). Cells were lysed in 50 mM Hepes-NaOH (pH 8.0), 150 mM NaCl, 10% glycerol, 0.1% NP-40, 2 mM EDTA, 2 mM dithiothreitol, and protease and phosphatase inhibitor cocktails (Roche) and subjected to immunoprecipitation with 10 μg of GEF-H1 antibody (A301-929A, Bethyl Labs). Follow-ing an on-beads digestion with FASP Protein Digestion Kit (Protein Discovery), tryptic peptides were cleaned up with C18 spin col-umn (Thermo Scientific), then separated by reversed-phase nano–high-performance liquid chromatography with a nanoAquity UPLC system (Waters Corp.). Peptides were first trapped in a 2-cm trapping column [75-μm inside diameter (ID), Michrom Magic C18 beads of 5.0-μm particle size, 200-Å pore size] and then separated on a self-packed 25-cm column (75-μm ID, Michrom Magic C18 beads of 3.0-μm particle size, 100-Å pore size) at room temperature. The flow rate was 200 nl/min over a gradient of 1% buffer B (0.1%

formic acid in acetonitrile) to 30% buffer B in 180 min. Then, a following wash raised buffer B to 70%. The identity of the eluted peptides was determined with an in-line LTQ-Orbitrap Velos mass spectrometer (Thermo Scientific). The ion source was operated at 2.0 to 2.4 kV with the ion transfer tube temperature set at 275°C. Full MS scan [300 to 2000 mass/charge ratio (m/z)] was acquired in Orbitrap at 60,000 resolution setting; data-dependent MS2 spectra were acquired in LTQ by colli-sion-induced dissociation with the 20 most intense ions. Precursor ions were selected on the basis of charge states (1, 2, or 3) and intensity thresholds (above 2000) from the full scan; dynamic exclusion (one repeat dur-ing 30 s, with a 60-s exclusion time window) was also taken into account. The polysiloxane lock mass of 445.120030 was used throughout spectral acquisition.

Protein identification and quantification

All raw data were converted to mzXML for-mat before a search of the resultant spectra with Sorcerer-SEQUEST (build 4.0.4, Sage-N Research) and the Trans-Proteomic Pipeline (TPP v4.3.1). Data were searched against the human UniProtKB/Swiss-Prot sequence database (Release 2011_08) supplemented with common contaminants, such as porcine (Swiss-Prot P00761) and bovine (P00760) trypsin, and further concatenated with its reversed copy as a decoy (40,494 total se-quences). Search parameters used were a pre-cursor mass between 400 and 4500 atomic mass units (amu), up to 2 missed cleavages, a precursor-ion tolerance of 3 amu, accurate mass binning within PeptideProphet (58), semi-tryptic digestion, a static carbamido-methyl cysteine modification, variable me-thionine oxidation, and additional variable modifications of R10 and K6 for SILAC ex-periments. SILAC ratios were calculated with TPP’s XPRESS (59), and protein abundance ratios were first normalized by the bait’s ratio, then combined from replicate experiments by taking a weighted average using the number of quantified spectra for the protein in each replicate. False discovery rates (FDRs) were determined by ProteinProphet (58), and mini-mum protein probability cutoffs resulting in a 1% FDR were selected individually for each experiment. Further filtering of identified proteins was accomplished using the follow-ing criteria: at least two unique peptides were identified for the protein in each of at least two (of four) WTX experiments, at least two (of three) for proteins in FAM123C experi-ments, two (of two) for ARHGEF2, and two (of three) for FAM123A. Common contami-nants were removed at the authors’ discretion on the basis of previous experiments, such as keratins, ribosomal, and DEAD box proteins. Unfiltered data are provided in table S1 and may be downloaded from ProteomeCom-

Fig. 8. FAM123A regulates cell adhesion and migration. (A) siRNA-transfected HeLa cells were fixed and stained with an anti-vinculin antibody, Alexa Fluor 594–phalloidin, and DAPI. Data represent four biological replicates. (B) Adhesion quantification of siRNA-transfected HeLa cells that were subjected to cell attachment with real-time acquisition of electrode impedance. (C) Histogram of relative impedance index 180 min after cell plating. Error bars represent SE across the biological replicates. (*P = 0.005 and **P = 0.005, paired Student’s t test; n, biological replicates). (D) siRNA-transfected HeLa cells were allowed to attach to fibronectin-coated coverslips for the indicated times before fixation and immunostaining with an anti-vinculin antibody. FN, fibronectin. (E) HeLa cells transfected with control or FAM123A siRNA were scratch-wounded and imaged over 24 hours. (*P = 0.03 and **P = 0.018 by Student’s t test). n = 3 biological replicates.




















mons.org Tranche using the following hash: SG+Lb+X7PAOmf9pdlyQREvVmA7sOZX/iwK-CV7zQBvmofGaBNxvXg8D9dZkUs5CgbiDMIB-p6JlNW1fh0E geWb8SOaaxMAAAAAAAADcA==.

Bioinformatics

PeptideProphet/ProteinProphet results for each AP/MS experiment were stored in a lo-cal ProHits database (60). ProHits performed the mapping of UniProtKB accession identi-fiers to Entrez Gene IDs. These results were then imported into Cytoscape v2.8.2 (61) for network visualization. Gene Ontology annota-tions were imported from the National Center for Biotechnology Information Entrez Gene through Cytoscape. Known protein-protein interactions were extracted from the follow-ing databases: BIND, BioGRID, DIP, HPRD, IntAct, MINT, and Reactome—downloaded on 15 August 2010.

Affinity pulldowns, immunoprecipitation, and Western blotting

In all biochemical experiments, cells were lysed in a buffer containing 50 mM tris-HCl (pH 8.0), 150 mM NaCl, 10% glycerol, 1% Tri-ton X-100, 2 mM EDTA, and protease and phosphatase inhibitor cocktails (Roche). For streptavidin-affinity purification, cleared ly-sates were incubated for 1 hour with strepta-vidin resin (GE Healthcare) and subsequently washed and eluted. Immunoprecipitation of endogenous GEF-H1 was performed with anti–GEF-H1 antibodies (A301-929A, Bethyl Labs). Detection of proteins by Western blot was performed with the following antibod-ies: anti–Flag M2 monoclonal (Sigma-Aldrich Corporation), anti-HA polyclonal (1867423, Roche), anti-GFP polyclonal (ab290, Abcam), anti–β-catenin polyclonal (9562, Cell Signal-ing Technology), anti-βTrCP1 monoclonal (37-3400, Invitrogen), anti–β-tubulin mono-clonal (T7816, Sigma-Aldrich Corporation), anti–GEF-H1 (A301-928A, Bethyl Labs), anti-MARK2 (Roche), anti-CLASP2 (A302-155A, Bethyl Labs), anti-APC polyclonal (A300-180A, A300-981A, Bethyl Labs), anti-EB1 monoclo-nal (610534, BD Transduction Laboratories), anti-MLC2 (3672, Cell Signaling Technology), and anti–phospho-MLC2 (Ser19) (no. 3671, Cell Signaling Technology). For detection of endogenous FAM123A, a custom monoclonal antibody was produced (Abmart).

For phospho-MLC detection, HeLa cells were transfected with siRNA at 10 nM for 48 hours. Cells were lysed in the well with UPX sample buffer (Protein Discovery) supple-mented with protease and phosphatase in-hibitors. Samples were boiled for 3 min before the addition of 4× sample buffer and sonica-tion. Proteins were detected by Western blot and quantified with Odyssey Imager and Soft-ware by LI-COR Biosciences.

Immunofluorescence

For immunofluorescence, HeLa sarcoma cells were plated on fibronectin (10 μg/ml)–coated coverslips in DMEM supplemented with 10% FBS and allowed to attach and spread. Cells were fixed in 4% paraformaldehyde (PFA; Electron Microscopy Sciences) in cytoskeletal buffer [5 mM Pipes (pH 6), 137 mM NaCl, 5 mM KCl, 1.1 mM sodium phosphate buffer, 0.4 mM KH

2PO

4, 0.4 mM MgCl

2, 0.4 mM

NaHCO3, 2 mM EGTA, 50 mM glucose] for 15

min and permeabilized with 0.1% Triton in phosphate-buffered saline (PBS) for 5 min. After being blocked in 1% bovine serum al-bumin/PBS for 1 hour, cells were incubated with primary antibodies at 4°C overnight, followed by incubation with appropriate sec-ondary antibodies, RRX-conjugated donkey anti-mouse immunoglobulin G (IgG) and fluorescein isothiocyanate–conjugated don-key anti-mouse IgG (Jackson ImmunoRe-search Laboratories), at room temperature for 1 hour. For staining of endogenous EB1, cells were fixed in methanol at −20°C for 5 min. Staining of proteins was performed with the following antibodies: monoclonal anti–α-tubulin (clone DM1A, T9026, Sigma-Aldrich Corporation), monoclonal anti-vinculin (clone nVIN-1, V9131, Sigma-Aldrich Corporation), and anti-EB1 monoclonal (610534, BD Trans-duction Laboratories). Actin was stained with Alexa Fluor 647–phalloidin or Alexa Fluor 594–phalloidin (Invitrogen). Coverslips were mounted to slides with the ProLong Gold Antifade reagent (Invitrogen). Staining was analyzed with an Olympus IX 81-ZDC inverted fluorescence microscope (Olympus Corporation of the Americas) equipped with a 60×/1.42 Oil PlanApo objective lens and a Hamamatsu C10600-10B camera (OrcaR2, Hamamatsu Photonics Ltd.).

For subcellular localization of FAM123A, HeLa cells were grown on 18-mm glass cov-erslips coated with poly-L-lysine (Sigma-Aldrich Corporation) and cotransfected with EGFP-FAM123A and mCherry-tubulin. Cells were in fixed in 1% PFA in pure −20°C methanol for 2 min and mounted with Pro-Long Gold Antifade with DAPI (Invitrogen). The imaging was performed on a DeltaVision deconvolution microscope (Applied Preci-sion). For costaining of FAM123A-GFP and EB1, cells were fixed similarly and EB1 was stained with anti-EB1 antibody (BD Trans-duction Laboratories), followed by incubation with donkey anti-mouse secondary antibody (Jackson ImmunoResearch Laboratories). Fixed cells were imaged with a complete Z-series in 0.2-μm sections, deconvolved, and projected with Adobe Photoshop CS and Im-ageJ software.

For comparing the distribution of FAM123A family members, HEK293T transiently ex-pressing fluorescent-tagged proteins were plated on fibronectin-coated coverslips, fixed, and mounted as described above and imaged

with a Zeiss LSM5 Pascal confocal laser scan-ning microscope equipped with a 63×/1.42 Oil PlanApo objective lens.

Live-cell imaging

For low-resolution imaging analysis, HT1080 sarcoma cells were plated onto fibronectin (5 μg/ml)–coated MatTek dishes (MatTek Corporation) in DMEM supplemented with 10% FBS and allowed to attach and spread. Cells were transfected with Venus-FAM123A or Venus-WTX with FuGENE HD (Promega) according to the manufacturer’s instructions. Cells were imaged the next day in contrast phase and GFP fluorescence on a Nikon Bio-station (every 5 min for 10 hours), with a 20× objective, at 37°C and 5% CO

2.

For dynamic analysis of fluorescence-tagged proteins at higher magnification, HT1080 were plated onto fibronectin (5 μg/ml)–coated Delta T dishes (Bioptechs Inc.) in DMEM supplemented with 10% FBS and transfected with EGFP-FAM123A with Fu-GENE HD (Promega). The next day, cells were imaged at 37°C and 5% CO

2 with an Olym-

pus IX 81-ZDC inverted fluorescence micro-scope (Olympus Corporation of the Americas) equipped with a Delta T Open Dish System (Bioptechs Inc.) with a heated lid. Time-lapse images were captured every 10 s with a heated 60×/1.42 Oil PlanApo objective lens and a Hamamatsu C10600-10B camera (OrcaR2, Hamamatsu Photonics Ltd.). Data acquisition was carried out with Velocity (version 5.5.1, PerkinElmer), and image processing was per-formed with ImageJ and Adobe Photoshop CS software.

HeLa cells were transfected with either EGFP-FAM123A, EGFP–C-terminal FAM123A, or EGFP–C-terminal FAM123A-IPAA and in some cases cotransfected with EB3-RFP with Nucleofector II (Amaxa Biosystems) and plated onto poly-D-lysine–coated MatTek dishes. After 24 hours of expression, the du-ally transfected cells were imaged at 37°C at a rate of 1 frame/5 s on a DeltaVision RT micro-scope (Applied Precision). Cells transfected with only one construct (movies S6 and S7) were imaged at a rate of 1 frame/2 s.

For microtubule dynamic studies, HeLa cells were transfected with EB3-GFP and RFP-Utr expressing DNA constructs and plated onto fibronectin-coated MatTek film-ing dishes; RFP-Utr encodes an RFP-fused calponin homology domain of the utrophin protein (62). The cells were transfected for 48 hours with either control siRNA or siRNA directed against FAM123A. The cells were then imaged with a Personal DeltaVision mi-croscope custom-outfitted with TIRF light paths and a 60× Olympus TIRF objective. Images were collected at 5-s intervals. Z-pro-jections were made using successive 5-frame intervals for a total of 15 frames. Each 5-frame projection was colored red, green, or blue to visualize the assembling microtubule

ends. With this regimen, stationary microtu-bule ends consisting of RGB overlap are de-picted as white.

For quantitation of cell migration, HeLa cells were transfected with 10 nM siRNA and plated onto a 12-well plate (4 × 105 cells per well) in complete growth medium 48 hours after transfections. After 12 hours, the mono-layer of cells was wounded by manual scratch-ing with a pipette tip, washed with PBS, and placed into complete growth medium. Time-lapse phase-contrast images were acquired ev-ery 7 min for 18 hours with an Olympus IX70 inverted fluorescence microscope, enclosed within an environmental chamber controlled for temperature, relative humidity, and CO

2,

and equipped with a 4× 0.13 Uplan FL PhL lens and a Hamamatsu Orca C4742-95 charge-coupled device camera. Data were acquired with Velocity (version 5.5.1, PerkinElmer). For calculation of relative migration, the scratch area was determined from movie-derived im-ages at time 0 and at 6 hours. The difference in area between t= 0 and t = 6 at three loca-tions along the scratch for each of the biologi-cal triplicate experiments was used to plot the relative rate of wound closure.

RhoA and GEF-H1 activity assays

Purified GST-RhoA(G17A) was provided by C. Guilluy and K. Burridge (UNC-Chapel Hill). Affinity precipitation of exchange factors with the nucleotide-free RhoA mutant (G17A) was performed as previously described (28). For quantitation, Western blot data from six bio-logical replicate experiments were analyzed by densitometry. Data were first normalized to a loading control and then to 1 before aver-aging and plotting. Paired Student’s t test was calculated using raw data. For RhoA activity quantitation, the G-LISA RhoA Activation As-say Biochem Kit (Cytoskeleton Inc.) was used. Specifically, 4 × 105 HeLa cells were plated in 60-mm plates. Cells were transfected over-night with siRNA and allowed to recover in 10% FBS/DMEM for 8 hours. Cells were then washed with PBS and starved in DMEM for 48 hours with one medium change at 24 hours. After 1-hour treatment with nocodazole, cells were lysed in 140 μl of cell lysis buffer. Plates were read on a Synergy HT microplate reader from BioTek.

Cell adhesion and spreading assay

Cell adhesion and spreading were measured as changes in impedance with the RT-CES sys-tem (ACEA Biosciences Inc.). The 16-well E-plates were coated with fibronectin (10 μg/ml) for 1 hour at 37°C. HeLa cells were transfected with 10 nM siRNAs and subjected to cell spreading assay about 65 hours after transfec-tion. For measurements, the same number of cells (4 × 103) was added to each well. The E-plates containing HeLa cells were incubated at room temperature for 10 min before being

placed on the device station in the incubator for continuous recording of impedance-based cell index (every 30 s, for 3 hours). Addition-ally, siRNA-transfected cells were plated on fibronectin-coated coverslips, allowed to at-tach and spread for various periods, fixed, and stained for vinculin according to the methods described earlier.

REFERENCES 1. M. N. Rivera, W. J. Kim, J. Wells, D. R. Driscoll,

B. W. Brannigan, M. Han, J. C. Kim, A. P. Feinberg, W. L. Gerald, S. O. Vargas, L. Chin, A. J. Lafrate, D. W. Bell, D. A. Haber, Science 315, 642–645 (2007).

2. E. C. Ruteshouser, S. M. Robinson, V. Huff, Genes Chromosomes Cancer 47, 461–470 (2008).

3. Z. A. Jenkins, M. van Kogelenberg, T. Morgan, A. Jeffs, R. Fukuzawa, E. Pearl, C. Thaller, A. V. Hing, M. E. Porteous, S. Garcia-Miñaur, A. Bohring, D. Lacombe, F. Stewart, T. Fiskerstrand, L. Bindoff, S. Berland, L. C. Adès, M. Tchan, A. David, L. C. Wilson, R. C. Hennekam, D. Donnai, S. Mansour, V. Cormier-Daire, S. P. Robertson, Nat. Genet. 41, 95–100 (2009).

4. A. Moisan, M. N. Rivera, S. Lotinun, S. Akhavanfard, E. J.Coffman, E. B. Cook, S. Stoykova, S. Mukherjee, J. A. Schoonmaker, A. Burger, W. J. Kim, H. M. Kronenberg, R. Baron, D. A. Haber, N. Bardeesy, Dev. Cell 20, 583–596 (2011).

5. M. B. Major, N. D. Camp, J. D. Berndt, X. Yi, S. J. Goldenberg, C. Hubbert, T. L. Biechele, A. C. Gingras, N. Zheng, M. J. Maccoss, S. Angers, R. T. Moon, Science 316, 1043–1046 (2007).

6. A. Grohmann, K. Tanneberger, A. Alzner, J. Schneikert, J. Behrens, J. Cell Sci. 120, 3738–3747 (2007).

7. K. Tanneberger, A. S. Pfister, V. Kriz, V. Bryja, A. Schambony, J. Behrens, J. Biol. Chem. 286, 19204–19214 (2011).

8. K. Tanneberger, A. S. Pfister, K. Brauburger, J. Schneikert, M. V. Hadjihannas, V. Kriz, G. Schulte, V. Bryja, J. Behrens, EMBO J. 30, 1433–1443 (2011).

9. A. Boutet, G. Comai, A. Schedl, The WTX/AMER1 gene family: Evolution, signature and function. BMC Evol. Biol. 10, 280 (2010).

10. G. Comai, A. Boutet, Y. Neirijnck, A. Schedl, Dev. Dyn. 239, 1867–1878 (2010).

11. A. S. Pfister, K. Tanneberger, A. Schambony, J. Behrens, J. Biol. Chem. 287, 1734–1741 (2012).

12. K. T. Vaughan, J. Cell Biol. 171, 197–200 (2005).

13. P. Bieling, S. Kandels-Lewis, I. A. Telley, J. van Dijk, C. Janke, T. Surrey, J. Cell Biol. 183, 1223–1233 (2008).

14. B. Vitre, F. M. Coquelle, C. Heichette, C. Garnier, D. Chrétien, I. Arnal, Nat. Cell Biol. 10, 415–421 (2008).

15. I. Kaverina, A. Straube, Semin. Cell Dev. Biol. 22, 968–974 (2011).

16. B. van der Vaart, A. Akhmanova, A. Straube, Biochem. Soc. Trans. 37, 1007–1013 (2009).

17. A. Akhmanova, M. O. Steinmetz, Nat. Rev. Mol. Cell Biol. 9, 309–322 (2008).

18. X. Wu, A. Kodama, E. Fuchs, Cell 135, 137–148 (2008).

19. A. Akhmanova, M. O. Steinmetz, J. Cell Sci. 123, 3415–3419 (2010).

20. A. Kodama, I. Karakesisoglou, E. Wong, A. Vaezi, E. Fuchs, Cell 115, 343–354 (2003).

21. L. K. Su, M. Burrell, D. E. Hill, J. Gyuris, R. Brent, R. Wiltshire, J. Trent, B. Vogelstein, K. W. Kinzler, Cancer Res. 55, 2972–2977

(1995).22. S. Honnappa, S. M. Gouveia, A. Weisbrich,

F. F. Damberger, N. S. Bhavesh, H. Jawhari, I. Grigoriev, F. J. van Rijssel, R. M. Buey, A. Lawera, I. Jelesarov, F. K. Winkler, K. Wüthrich, A. Akhmanova, M. O. Steinmetz, Cell 138, 366–376 (2009).

23. R. Li, G. G. Gundersen, Nat. Rev. Mol. Cell Biol. 9, 860–873 (2008).

24. Y. C. Chang, P. Nalbant, J. Birkenfeld, Z. F. Chang, G. M. Bokoch, Cell 19, 2147–2153 (2008).

25. M. Krendel, F. T. Zenke, G. M. Bokoch, Nat. Cell Biol. 4, 294–301 (2002).

26. M. Uehata, T. Ishizaki, H. Satoh, T. Ono, T. Kawahara, T. Morishita, H. Tamakawa, K. Yamagami, J. Inui, M. Maekawa, S. Narumiya, Nature 389, 990–994 (1997).

27. C. S. Hill, J. Wynne, R. Treisman, Cell 81, 1159–1170 (1995).

28. R. García-Mata, K. Wennerberg, W. T. Arthur, N. K. Noren, S. M. Ellerbroek, K. Burridge, Methods Enzymol. 406, 425–437 (2006).

29. H. Hatanaka, K. Ogura, K. Moriyama, S. Ichikawa, I. Yahara, F. Inagaki, Cell 85, 1047–1055 (1996).

30. K. Moriyama, E. Nishida, N. Yonezawa, H. Sakai, S. Matsumoto, K. Iida, I. Yahara, J. Biol. Chem. 265, 5768–5773 (1990).

31. C. J. Storbeck, S. Wagner, P. O’Reilly, M. McKay, R. J. Parks, H. Westphal, L. A. Sabourin, Mol. Biol. Cell 20, 4174–4182 (2009).

32. A. V. Burakov, O. N. Zhapparova, O. V. Kovalenko, L. A. Zinovkina, E. S. Potekhina,

N. A. Shanina, D. G. Weiss, S. A. Kuznetsov, E. S. Nadezhdina, Mol. Biol. Cell 19, 1952–1961 (2008).

33. S. Wagner, T. A. Flood, P. O’Reilly, K. Hume, L. A. Sabourin, J. Biol. Chem. 277, 37685–37692 (2002).

34. D. Leopoldt, H. F. Yee Jr., E. Rozengurt, J. Cell. Physiol. 188, 106–119 (2001).

35. D. Riveline, E. Zamir, N. Q. Balaban, U. S. Schwarz, T. Ishizaki, S. Narumiya, Z. Kam,

B. Geiger, A. D. Bershadsky, J. Cell Biol. 153, 1175–1186 (2001).

36. S. C. Schuyler, D. Pellman, Cell 105, 421–424 (2001).

37. A. Akhmanova, C. C. Hoogenraad, K. Drabek, T. Stepanova, B. Dortland, T. Verkerk, W. Vermeulen, B. M. Burgering, C. I. De Zeeuw, F. Grosveld, N. Galjart, Cell 104, 923–935 (2001).

38. I. S. Näthke, C. L. Adams, P. Polakis, J. H. Sellin, W. J. Nelson, J. Cell Biol. 134, 165–179 (1996).

39. M. Bienz, Nat. Rev. Mol. Cell Biol. 3, 328–338 (2002).

40. K. W. Fong, S. Y. Hau, Y. S. Kho, Y. Jia, L. He, R. Z. Qi, Mol. Biol. Cell 20, 3660–3670 (2009).








SECTION THREE | ARTICLES:REPORT

41. K. Drabek, M. van Ham, T. Stepanova, K. Draegestein, R. van Horssen, C. L. Sayas,

A. Akhmanova, T. Ten Hagen, R. Smits, R. Fodde, F. Grosveld, N. Galjart, Curr. Biol. 16, 2259–2264 (2006).

42. Y. Wen, C. H. Eng, J. Schmoranzer, N. Cabrera-Poch, E. J. Morris, M. Chen, B. J. Wallar, A. S. Alberts, G. G. Gundersen, Nat. Cell Biol. 6, 820–830 (2004).

43. J. S. Logue, J. L. Whiting, B. Tunquist, D. B. Sacks, L. K. Langeberg, L. Wordeman, J. D. Scott, J. Biol. Chem. 286, 39269–39281 (2011).

44. K. A. Myers, K. T. Applegate, G. Danuser, R. S. Fischer, C. M. Waterman, J. Cell Biol. 192, 321–334 (2011).

45. J. Birkenfeld, P. Nalbant, B. P. Bohl, O. Pertz, K. M. Hahn, G. M. Bokoch, Dev. Cell 12, 699–712 (2007).

46. M. G. Callow, S. Zozulya, M. L. Gishizky, B. Jallal, T. Smeal, Cell Sci. 118, 1861–1872

(2005).47. S. H. Fujishiro, S. Tanimura, S. Mure,

Y. Kashimoto, K. Watanabe, M. Kohno, Biochem. Biophys. Res. Commun. 368, 162–167 (2008).

48. E. Kakiashvili, Q. Dan, M. Vandermeer, Y. Zhang, F. Waheed, M. Pham, K. Szászi, J. Biol. Chem. 286, 9268–9279 (2011).

49. E. Kakiashvili, P. Speight, F. Waheed, R. Seth, M. Lodyga, S. Tanimura, M. Kohno,

O. D. Rotstein, A. Kapus, K. Szászi, J. Biol. Chem. 284, 11454–11466 (2009).

50. T. Matsuzawa, A. Kuwae, S. Yoshida, C. Sasakawa, A. Abe, EMBO J. 23, 3570–3582 (2004).

51. K. Tonami, Y. Kurihara, S. Arima, K. Nishiyama, Y. Uchijima, T. Asano, H. Sorimachi, H. Kurihara, J. Cell Sci. 124, 1214–1223 (2011).

52. Y. Yamahashi, Y. Saito, N. Murata-Kamiya, M. Hatakeyama, J. Biol. Chem. 286, 44576–44584 (2011).

53. Y. Yoshimura, H. Miki, Biochem. Biophys. Res. Commun. 408, 322–328 (2011).

54. F. T. Zenke, M. Krendel, C. DerMardirossian, C. C. King, B. P. Bohl, G. M. Bokoch, J. Biol. Chem. 279, 18392–18400 (2004).

55. S. Angers, Methods Mol. Biol. 468, 223–230 (2008).

56. S. Angers, C. J. Thorpe, T. L. Biechele, S. J. Goldenberg, N. Zheng, M. J. MacCoss,

R. T. Moon, Nat. Cell Biol. 8, 348–357 (2006).57. G. I. Chen, A. C. Gingras, Methods 42,

298–305 (2007).58. A. Keller, A. I. Nesvizhskii, E. Kolker, R.

Aebersold, Anal. Chem. 74, 5383–5392 (2002).

59. D. K. Han, J. Eng, H. Zhou, R. Aebersold, Nat. Biotechnol. 19, 946–951 (2001).

60. G. Liu, J. Zhang, B. Larsen, C. Stark, A. Breitkreutz, Z. Y. Lin, B. J. Breitkreutz, Y. Ding, K. Colwill, A. Pasculescu, T. Pawson, J. L. Wrana, A. I. Nesvizhskii, B. Raught, M. Tyers, A. C. Gingras, Nat. Biotechnol. 28, 1015–1017 (2010).

61 . M. E. Smoot, K. Ono, J. Ruscheinski, P. L. Wang, T. Ideker, Bioinformatics 27, 431–432 (2011).

62. B. M. Burkel, G. von Dassow, W. M. Bement, Cell Motil. Cytoskeleton 64, 822–832 (2007).

ACKNOWLEDGMENTSWe thank R. T. Moon and the members of the Moon laboratory and Major laboratory for their invalu-able assistance. We also thank R. T. Moon and the Howard Hughes Medical Institute for providing space and funding for M.M. during the early stages of this work. We thank F. Nicoletti for financial sup-port to M.M. We thank A. Akhmanova, C. Guilluy, and N. Mitin for helpful advice. Funding: M.B.M. is supported by the NIH through the NIH Director’s New Innovator Award, 1-DP2-OD007149-01, and a Scholar Award from the Sidney Kimmel Cancer Foundation. L.W. is supported by GM69429/GM/NIGMS from the NIH. Author contributions: M.B.M., P.F.S., and M.M. designed the experiments and analyzed the data; these and all other authors performed theexperiments. M.B.M., F.Y., and D.G. performed the MS analysis. P.F.S., M.M., and M.B.M.wrote the manuscript. Competing interests: The authors declare that they have no compet-ing interests. Data and materials availability: The data associated with this manuscript may be downloaded from ProteomeCommons.

org Tranche using the following hash: SG+Lb+ X7PAOmf9pdlyQREvVmA7sOZX/ iwKCV7zQBv-mofGaBNxvXg8D9dZkUs5CgbiDMIBp6JlNW1f-h0EgeWb8SOaaxMAAAAAAAADcA==.

SUPPLEMENTARY MATERIALSwww.sciencesignaling.org/cgi/content/full/5/240/ra64/DC1Fig. S1. FAM123C does not localize to the cytoplasmic membrane.Fig. S2. FAM123A binds FBXW11.Fig. S3. Mutation of the SKIP motif in FAM123A abrogates its microtubule localization and tracking.Fig. S4. FAM123A overexpression alters the subcellular localization of EB1.Fig. S5. EB1 and microtubule association is dispensable for the FAM123A control of β-catenin–dependent WNT signaling.Fig. S6. GEF-H1 silencing alters microtubule organization and actomyosin contractility.Fig. S7. The increased actin contractility observed in FAM123A siRNA–transfected cells is due to the specific silencing of FAM123A.Fig. S8. FAM123A controls maturation of focal adhesions during cell spreading.Table S1. Affinity purification and MS data.Movie S1. Venus-fused FAM123A localizes to the cell membrane and to filamentous structures in the cell lamellipodium.Movie S2. Venus-fused WTX localizes predominantly to the cell membrane.Movie S3. EGFP-fused FAM123A moves on presumed microtubules toward the cell cortex.Movie S4. EGFP-fused FAM123A tracks the tips and distal ends of EB3-associated microtubules.Movie S5. FAM123A-IPAA shows decreased EB3-associated plus-end tip tracking and decoration of distal ends of microtubules.Movie S6. The C-terminal fragment of FAM123A, which contains the SKIP EB-interaction motif, behaves as a classic +TIP showing robust microtubule tip tracking, but does not decorate the microtubule lattice.Movie S7. The C-terminal fragment of FAM123A with a mutated SKIP motif does not show colocalization or tip tracking of microtubules.

Newborn neurons detach an apical cell-process from the ventricular surface and then migrate to the lateral neural tube or to form cortical layers within the brain (1, 2). This step is required for the

generation of neuronal and tissue architec-ture (2, 3), and its failure leads to human peri-ventricular heterotopia (4). Down-regulation of N-cadherin is associated with this event (3, 5), as is loss of apical complex proteins (6, 7). The latter may be mediated by down-regulation; protein modification/degradation or relocalization; or loss of apical membrane.

To investigate cell behavior underlying neu-ron birth, we labeled membranes of individu-al cells by mosaic transfection of green fluo-rescent protein– glycosylphosphatidylinositol (pCAGGS-GFP-GPI) into the chick embryonic spinal cord (8). We then monitored neurogen-esis in ex vivo embryo slice cultures (1) using wide-field time-lapse microscopy (8). New-born neurons have a basally located cell body and extend a long, thin cell-process to the api-cal/ventricular surface. Movies of such cells revealed that shedding of the apical-most cell membrane preceded withdrawal of this cell-process (Fig. 1A). This event, which we name apical abscission, takes ~1 hour (56 min, SD = 18 min, n = 21 cells). It begins with formation of a bulb-like “bouton,” followed by subapical constriction, membrane thinning, and even-tual abscission, after which the apical cell-process withdraws (42 abscising cells in 34 embryos; all stages observed in 21 cells) (Fig. 1A, fig. S1, and movies S1 to S3). Abscised par-ticles tracked so far remain at the ventricle.

Using structured illumination microscopy (8) to generate super-resolution images of abscising cells transfected with membrane-lo-

Neural Development Group, Division of Cell and DevelopmentalBiology, College of Life Sciences, University of Dundee,Dundee DD1 5EH, UK*Corresponding author. E-mail: [email protected]

calized Tag–red fluorescent protein–Farnesyl (TagRFP-Farn) revealed a thin membranous connection between apical cell-process and the abscising particle. This confirmed the ex-istence of abscission events in fixed tissue not subject to culture and imaging regimes (n = 5 cells in 3 embryos) (fig. S2 and movie S4). We also observed apical abscission in completely unmanipulated embryos fixed and labeled to reveal the early neuronal marker Tuj1 (class III beta-tubulin). Some Tuj1+ cells with a basally localized nucleus and a ventricle-contacting apical cell-process were found to have a dis-tinct constriction, coincident with subapical actin (n = 31 of 78 cells in 5 embryos) (Fig. 1B and movie S5). To characterize the abscised membrane, we assessed localization of endog-enous apical Par-complex protein, atypical protein kinase C (aPKC) (9) in such Tuj1+ cells; aPKC was confined to the abscising particle (n = 31 of 31 cells in 5 embryos) (Fig. 1B and movie S5). This indicates that differentiating neurons experience rapid loss of apical polar-ity. It is also consistent with the absence of Par-complex proteins from withdrawing cell-processes (6, 7), which, now liberated from apical-junctional complexes, extend transient membrane protrusions (18 cells in 9 embryos) (e.g., see movies S1 and S2). Similar apical constrictions were visible in Tuj1+ ventricle-contacting cells in mouse spinal cord (22 of 40 cells in 4 embryos) (Fig. 1C and movie S6), with aPKC confined to the abscising particle (22 of 22 cells). This demonstrates that api-cal abscission is conserved across species. In chick, we further characterized cells poised to abscise as indicated by a basally located nucleus and ventricle-contacting apical cell-process revealed by TagRFP-Farn labeling and found a similar localization of actin and aPKC (21 of 21 cells in 6 embryos) (Fig. 1D and mov-ie S7). Many such TagRFP-Farn–labeled cells with this morphology also express low levels of the early neuronal marker, NeuroM (26 of

29 cells in 5 embryos) (Fig. 1, E to G). These NeuroM-positive cells were further found to express the interneuron marker Lim1/2 (23 of 23 cells) but not the later neuronal marker HuC/D (0 of 12 cells) nor the postmitotic cell marker Cdk-inhibitor p27/Kip1 (0 of 11 cells) [(10) and see (7)], identifying these cells as immature neurons that have yet to commit to cell cycle exit (Fig. 1, E to G).

Neuroepithelial cells contain a subapical actin cable that mediates normal cell con-striction at the ventricular surface. To inves-tigate whether apical abscission involves ac-tin dynamics, we cotransfected GFP-GPI and Actin-TagRFP vectors into chick neural tube and monitored protein localization. Subapical actin was visible in cells poised to abscise and coincided with the region of constriction be-fore abscission (Fig. 2A and movie S8). As ab-scission began, Actin-TagRFP signal intensity increased (8), reaching a maximum shortly before abscission completion (Fig. 2B); actin was then depleted from the withdrawing cell-process tip (n = 24 abscising cells in 18 em-bryos) (Fig. 2A, fig. S3, and movies S8 to S10). This local actin increase raised the possibility that actin-myosin contraction mediates apical abscission

We therefore next surveyed myosin local-ization using a myosin regulatory light chain 2 GFP construct (MRLC2-GFP); this revealed strong subapical localization and diffuse cy-tosolic distribution in all cells (fig. S4). Be-cause myosin phosphorylation is essential for actin-mediated apical constriction, we next discriminated sites of myosin activity by monitoring MRLC2T18DS19D-GFP, a constitu-tively active form of MRLC2. To increase the incidence of neuronal differentiation, we co-transfected cells with a plasmid encoding the proneural gene Neurog2 [pCAGGS-Neurog2-IRES-nucGFP (pCIG-Neurog2)], which pro-motes neuronal differentiation (10). In such cells, also coexpressing TagRFP-Farn to label cell membranes, active MRLC2 localized sub-apically until shortly after abscission (8) (n = 12 cells in 9 embryos) (Fig. 2, C and D, and movies S11 to S13). Thus, actin is localized and myosin is active in the subapical region of the abscising neuron.

To investigate the requirement for myosin activity, cells were transfected with GFP-GPI and pCIG-Neurog2, and slices were cultured in medium containing blebbistatin (inhibitor of myosin motor function), ML-7 (inhibitor of myosin light chain kinase MLCK, which phos-phorylates Myosin II) (see fig. S5), or dimethyl sulfoxide (DMSO) control. Although few cells in control slices failed to abscise and retract their cell-processes within an 8-hour period (n = 4 of 33 cells in 5 embryos) (Fig. 2E and movies S14 to S16), the majority of cells ex-posed to blebbistatin (n = 33 of 36 cells in 6 embryos) (Fig. 2F and movies S17 to S19; api-cal surface definition, fig. S6) or ML7 (n = 68 of 83 cells in 15 embryos; Fig. 2G and movies S20 to S22; apical surface definition, fig. S6)

Raman M. Das, Kate G. Storey*

Apical abscission alters cell polarity and dismantles the primary cilium during neurogenesis

Withdrawal of differentiating cells from proliferative tissue is critical for embryonic development and adult tissue homeostasis; however, the mechanisms that control this cell behavior are poorly understood. Using high-resolution live-cell imaging in chick neural tube, we uncover a form of cell subdivision that abscises apical cell membrane and mediates neuron detachment from the ventricle. This mechanism operates in chick and mouse, is dependent on actin-myosin contraction, and results in loss of apical cell polarity. Apical abscission also dismantles the primary cilium, known to transduce sonic-hedgehog signals, and is required for expression of cell-cycle-exit gene p27/Kip1. We further show that N-cadherin levels, regulated by neuronal-differentiation factor Neurog2, determine cilium disassembly and final abscission. This cell-biological mechanism may mediate such cell transitions in other epithelia in normal and cancerous conditions.

Originally published 10 January 2014 in SCIENCE

http://www.sciencesignaling.org/cgi/content/full/5/240/ra64/DC1

http://www.sciencesignaling.org/cgi/content/full/5/240/ra64/DC1







































http://www.sciencemag.org/search?author1=Raman+M.+Das&sortspec=date&submit=Submit

http://www.sciencemag.org/search?author1=Kate+G.+Storey&sortspec=date&submit=Submit

http://www.sciencemag.org/content/343/6167/200.full#corresp-1





remained attached at the ventricle. Further-more, coexpression of active MRLC2 in the presence of ML-7 rescued its effects, with most cells now abscising within 8 hours (n = 14 of 18 cells in 7 embryos) (Fig. 2H and movies S23 to S25). Misexpression of active MRLC2 alone, however, did not increase neuron num-bers, so the potential to increase actin-myosin contraction is by itself insufficient to promote apical abscission (fig. S7). These data indicate that myosin activity is required, but not suf-ficient, for apical abscission.

Neuroepithelial cells lining the ventricle possess a primary cilium. This projects from the basal body/centrosome located at the apical pole. While this cilium plays a key role in transducing sonic hedgehog (Shh) (and possibly other) signals that maintain neuroepithelial cells in a proliferative state (11), the centrosome is further implicated in positioning axon outgrowth (12). To observe the effect of apical abscission on the primary cilium, we transfected GFP-GPI and pCIG-Neurog2 into the neural tube together

with a construct containing a pericentrin-AKAP450 centrosomal targeting (PACT) domain sequence that confers centrosomal localization fused to TagRFP (PACT-TagRFP). As abscission began, the centrosome localized to the withdrawing cell-process (n = 45 cells in 15 embryos) (Fig. 3Aand movies S26 to S28). Conversely, the primary cilium, identified with ciliary membrane–associated Arl13b-TagRFP, remained attached to the abscised apical membrane (n = 21 cells in 7 embryos) (Fig. 3B and movies S29 to S31). During

Fig. 1. Apical abscis-sion during neuronal differentiation. (A) Time-lapse sequence of a cell undergoing distinct stages of api-cal abscission [movie S1; these are addi-tional frames of a cell shown in supplemen-tary movie 2 in (6)]. (B to D) Maximum intensity projections of constricting abscis-sion site (white ar-rowheads) visible in Tuj1+ventricle-contact-ing cells in chick (B) and mouse (C) embry-os and TagRFP-Farn–labeled cell in chick (D); the abscising par-ticle is distal to actin and contains the apical Par-complex, marked by aPKC [three-dimen-sional reconstructions of (B), (C), and (D) in movies S4, S5, and S6, respectively]. (E to G) Cells poised to abscise express NeuroM and Lim1/2 (E) but not HuC/D (F) nor p27 (G) (magenta arrows). Ab-scission site (white ar-rowheads), withdraw-ing apical cell-process (white arrows), ab-scised particle (yellow arrows), and apical surface (white dashed line) here and in all figures. Scale bars, 10 μm; enlarged regions, 2 μm.

Fig. 2. Apical abscission depends on actin-myosin activity. (A and B) Time-lapse sequence showing actin localization (A) (movie S8) and (B) quantification of Actin-TagRFP intensity during apical abscission (average normalized values for four cells; error bars, mean ± SEM). (C and D) Active myosin (MRLC2T18DS19D-GFP) (C, green at cell-process tip; movie S11) localiz-es to abscission site (D) MRLC2T18DS19D-GFP intensity during apical abscission

(average normalized values for five cells error bars, mean ± SEM). (E to H) Cells exposed to control DMSO undergo abscission (E) (movie S14), but not in the presence of blebbistatin (F) (movie S17) or ML-7 (G) (movie S20). ML-7 abscission inhibition is rescued by expression of MRLC2T18DS19D-GFP (H) (movie S23). For definition of apical surfaces, N-cadherin, and aPKC localization, see fig. S6. [(B) and (D)] Membrane thinning, black arrow; abscission complete, black arrowhead. Scale bars, 10 μm; enlarged regions, 2 μm.

Fig. 3. Apical abscission dismantles the primary cilium. (A and B) Time-lapse sequences showing centrosome release into the apical cell-process (A) (movie S26), while Arl13b-labeled cilium is retained at the apical membrane (B) (movie S29). (C) Widefield and (C′) structured illumination imaging (white dotted outline) (movie S32) of TagRFP-Farn labeled apical cell-process and abscised particle containing Arl13b-GFP–labeled cilium. (D and E) Tuj1+ cells with ventricle-contacting apical cell-processes exhibit Smo (D) (movie S33) and Gli2 accumulation (E) (movie S34) (empty arrowheads) in their primary cilium [identified with Arl13b-GFP or Ift88 (intraflagellar-transport-protein 88), respectively]. Scale bars, 10 μm; enlarged regions and C and C′, 2 μm.










apical abscission, the Arl13b-labeled cilium also shortened [two-fold reduction in cilium length; n = 5 abscising cells (8)]. We further used structured illumination microscopy to confirm the presence of Arl13b-GFP in particles abscised from TagRFP-Farn–labeled cells in tissue not subject to culture and live imaging (n = 5 cells) (Fig. 3C and movie S32).

Active Shh signaling is indicated by accu-mulation of the Shh transducer Smoothened (Smo) and its key pathway effector Gli2 in the primary cilium (13, 14). Shh signaling is high-est in the ventral half of the neural tube, and we therefore assessed localization of endog-enous Smo and Gli2 in Tuj1+ cells with ventri-cle-contacting cell-processes in this region. This revealed many cells with ciliary accu-mulation of Smo (n = 38 of 43 cells in 4 chick embryos) (Fig. 3D and movie S33) or Gli2 (n = 33 of 35 cells in 3 mouse embryos) (Fig. 3E and movie S34). This localization of pro-teins suggests that cells poised to abscise are responding to Shh signals and predicts that disjunction of the centrosome and Arl13b-la-beled cilium during apical abscission curtails Shh signaling.

Onset of neuronal differentiation is char-acterized by down-regulation of N-cadherin (3, 5), which forms subapical adherens junc-tions between neuroepithelial cells (15), and abnormal persistence of N-cadherin inhibits apical cell-process withdrawal (3). Cadherins are connected intracellularly to the contrac-tile actin cable, and this serves to maintain tension at apical junctions and cell-cell adhe-sion (15). Declining N-cadherin levels within the prospective neuron may therefore trigger apical abscission by loosening cell-cell junc-tions and connection with the intracellular actin-myosin cable. Because the centrosome is localized in the withdrawing cell-process, it must be released from the cilium before final abscission. To determine how persis-tent N-cadherin affects apical abscission, we misexpressed N-cadherin-YFP (yellow fluo-rescent protein) together with GFP-GPI andPACT-TagRFP. Increased N-cadherin blocked cell-process withdrawal, and the centrosome remained at the apical pole (16 cells in 10 embryos) (Fig. 4A and movies S35 to S37). This indicates that N-cadherin down-regulation is required for centrosome release from the apical surface, as well as for final

abscission of apical membrane.One consequence of failure to undergo

N-cadherin down-regulation and apical ab-scission might therefore be maintenance of Shh signaling and therefore inhibition of cell cycle exit. To assess the relationship between cell cycle regulation and apical abscission, we next determined the effect of persistent N-cadherin on expression of p27/Kip1, which normally begins after apical cell-process withdrawal [Figs. 1G and 4B′ and see (7)]. N-cadherin misexpressing cells lacked p27/Kip1 after 24 hours [N-cad-YFP+TagRFP-Farn misexpressing cells 3% p27/Kip1 posi-tive (25 of 689 cells in 4 embryos) (Fig. 4B); control TagRFP-Farn only expressing cells 13% p27/Kip1 positive (76 of 649 cells in 4 embryos) (Fig. 4B′]. These findings therefore place N-cadherin loss and apical abscission, including cilium disassembly, upstream of cell cycle exit as defined by p27/Kip1 expres-sion. Furthermore, driving premature cell cycle exit by p27/Kip1 misexpression did not promote apical abscission or neuronal differ-entiation (fig. S8), consistent with proneural genes promoting expression of Cdk inhibi-tors, which then act in concert with other proneural targets to orchestrate neuronal differentiation (16).

N-cadherin down-regulation in prospec-tive neurons is mediated by the transcrip-tion factor FoxP2/4, expression of which is promoted by the proneural gene Neurog2 (3, 10). To determine whether Neurog2 mis-expression is sufficient to overcome excess N-cadherin, we cotransfected constructs en-coding these two genes into the neural tube. Despite excess N-cadherin, cells with excess Neurog2 dismantled their cilium and under-went abscission and cell-process withdrawal (16 cells in 7 embryos) (Fig. 4C and movies S38 to S40). In this context, N-cadherin-TagRFP was localized to the abscission site and then lost before abscission (n = 19 cells in 8 embryos) (Fig. 4D and movies S41 to S43). This indicates that localization and regula-tion of N-cadherin protein [(as well as tran-scriptional down-regulation of endogenous N-cadherin (3)] is directed by factors down-stream of Neurog2.

These findings uncover a cell biological mechanism, apical abscission, that takes place downstream of N-cadherin loss and in-volves actin-myosin–dependent cell constric-tion and dismantling of the primary cilium. This abscission event detaches newborn neu-rons from the ventricular surface and results in loss of apical-complex–containing cell membrane and therefore apical polarity. By separating centrosome from Arl13b-labeled cilium, apical abscission may curtail active Shh signaling, as indicated by ciliary accu-mulation of Smo and Gli2 in cells poised to abscise. Consistent with a loss of mitogenic Shh, abscission is also required for expres-sion of cell-cycle exit gene p27/Kip1 (fig. S9). Apical abscission is thus a decisive event in

Fig. 4. N-cadherin misexpression blocks apical abscission. (A) Cells misexpressing N-cad-YFP constrict (white arrowheads) but do not abscise and the centrosome (PACT-TagRFP, magenta arrows) remains at the apical cell pole (movie S35). (B) N-cad-YFP misexpressing cells do not express p27 (B′), which is normally detected after apical cell-process detachment (cell nuclei; empty arrowheads). (C) Neurog2 misexpression rescues centrosome release and abscission in N-cad-YFP–expressing cells (movie S38). (D) Neurog2 misexpression decreases subapical N-cad-TagRFP levels (magenta arrows), followed by abscission and cell-process withdrawal. A second underlying cell has yet to withdraw (movie S41). Scale bars, 10 μm; enlarged regions, 2 μm.

the neuronal differentiation program, which triggers reorganization of the newborn neu-ron and its withdrawal from the ventricular environment. Abscising the apical mem-brane and leaving this, at least initially, at the apical surface may also help to maintain tissue integrity. During mitosis, cilia are re-sorbed or partially internalized (17) rather than shed, and regulated cilium shedding has only been reported in the alga Chlam-ydomonas (18). Loss of apical complex pro-teins also characterizes cells undergoing an epithelial to mesenchymal transition, includ-ing tumor cell metastasis (19), and some cancers exhibit cilia loss (20). Investigation of apical abscission in normal and also onco-genic epithelia may therefore provide insight into mechanisms that direct critical cell state transitions.

REFERENCES AND NOTES 1. A. C. Wilcock, J. R. Swedlow, K. G. Storey,

Development 134, 1943–1954 (2007). 2. S. C. Noctor, V. Martínez-Cerdeño, L. Ivic, A. R.

Kriegstein, Nat. Neurosci. 7, 136–144 (2004). 3. D. L. Rousso et al., Neuron 74, 314–330

(2012). 4. V. L. Sheen, Scientifica 2012, 480129 (2012). 5. G. K. W. Wong, M.-L. Baudet, C. Norden, L.

Leung, W. A. Harris, J. Neurosci. 32, 223–228 (2012).

6. R. M. Das, K. G. Storey, EMBO Rep. 13, 448–454 (2012).

7. S. Ghosh et al., Proc. Natl. Acad. Sci. U.S.A. 105, 335–340 (2008).

8. Materials and methods are available as supplementary material on Science Online.

9. S. Ohno, Curr. Opin. Cell Biol. 13, 641–648 (2001).

10. N. Bertrand, D. S. Castro, F. Guillemot, Nat. Rev. Neurosci. 3, 517–530 (2002).

11. A. Louvi, E. A. Grove, Neuron 69, 1046–1060 (2011).12. M. Kuijpers, C. C. Hoogenraad, Mol. Cell.

Neurosci. 48, 349–358 (2011).13. K. C. Corbit et al., Nature 437, 1018–1021

(2005).14. J. Kim, M. Kato, P. A. Beachy, Proc. Natl. Acad.

Sci. U.S.A. 106, 21666–21671 (2009).15. A. Ratheesh, A. S. Yap, Nat. Rev. Mol. Cell Biol.

13, 673–679 (2012).16. H. Gui, S. Li, M. P. Matise, Dev. Biol. 301, 14–26 (2007).17. J. T. M. L. Paridaen, M. Wilsch-Bräuninger, W.

B. Huttner, Cell 155, 333–344 (2013).18. L. M. Quarmby, in Int. Rev. Cytol. (Academic

Press, 2004), vol. 233, pp. 47–91.19. B. Xue, K. Krishnamurthy, D. C. Allred, S. K.

Muthuswamy, Nat. Cell Biol. 15, 189–200

(2013).20. N. B. Hassounah, T. A. Bunch, K. M.

McDermott, Clin. Cancer Res. 18, 2429–2435 (2012).

ACKNOWLEDGMENTSWe thank C. Weijer, A. Muller, and J. Januschke for comments; J. Swedlow for imaging advice; and S. Swift and C. Thomson in the College of Life Sciences Light Microscopy Facility (LMF) for tech-nical support. Structuredillumination microscopy was carried out with assistance of M. Posch (LMF) and L. Ferrand (GE Healthcare) and supported by the Medical Research Council Next Generation Optical Microscopy Award (MR/K015869/1). R.M.D. and K.G.S. are funded by Wellcome Trust program grant 083611/Z/07/Z.

SUPPLEMENTARY MATERIALSwww.sciencemag.org/content/343/6167/200 suppl/DC1Materials and MethodsFigs. S1 to S9Movies S1 to S43References (21–30)22 October 2013; accepted 2 December 201310.1126/science.1247521
































Light microscopy is a key technology in modern cell biology and, in combination with immunofluorescence, fluores-cent protein fusions, or in situ hybridiza-

tion, allows the specific localization of nearly all cellular components. A fundamental limi-tation of optical microscopy is its low resolu-tion relative to the scale of subcellular struc-tures. This limitation occurs because light traveling through a lens cannot be focused

to a point but only to an Airy disk (1) with a diameter of about half the wavelength of the emitted light (2, 3). Because the wavelengths of visible light range from 400 to 700 nm, ob-jects closer than 200 to 350 nm apart cannot be resolved but appear merged into one.

Improving resolution beyond the 200-nm diffraction limit while retaining the advan-tages of light microscopy and the specificity of molecular imaging has been a long-standing

goal. Here, we present results demonstrating that this goal can be achieved with the use of a microscope system that implements three-di-mensional structured illumination microsco-py (3D-SIM) (4) in an easy-to-use system that makes no extra demands on experimental pro-cedures. Structured illumination microscopy (SIM) resolves objects beyond the diffraction limit by illuminating with multiple interfer-ing beams of light (5). The emitted light then contains higher-resolution image informa-tion, encoded by a shift in reciprocal (Fourier, or frequency) space into observable modula-tions of the image, in a manner similar to the formation of Moiré patterns (fig. S1). This ex-tra information can be decoded to reconstruct fine details, resulting in an image with twice the resolution of a conventional image taken on the same microscope (Fig. 1 and fig. S2). The 3D-SIM method extends previous 2D SIM methods by using three beams of interfering

Fluorescence light microscopy allows multicolor visualization of cellular components with high specificity, but its utility has until recently been constrained by the intrinsic limit of spatial resolution. We applied three-dimensional structured illumination microscopy (3D-SIM) to circumvent this limit and to study the mammalian nucleus. By simultane-ously imaging chromatin, nuclear lamina, and the nuclear pore complex (NPC), we observed several features that escape detection by conventional microscopy. We could resolve single NPCs that colocalized with channels in the lamin network and peripheral heterochromatin. We could differentially localize distinct NPC components and detect double-layered invaginations of the nuclear envelope in prophase as previously seen only by electron microscopy. Multicolor 3D-SIM opens new and facile possibilities to analyze subcellular structures beyond the diffraction limit of the emitted light.

Lothar Schermelleh,1* Peter M. Carlton,2* Sebastian Haase,2,4 Lin Shao,2

Lukman Winoto,2 Peter Kner,2 Brian Burke,3 M. Cristina Cardoso,4 David A. Agard,2 Mats G. L. Gustafsson,5 Heinrich Leonhardt,1*† John W. Sedat2*†

Subdiffraction multicolor imagingof the nuclear periphery with 3Dstructured illumination microscopy

1Center for Integrated Protein Science, Department of Biology, Ludwig Maximilians University Munich, 82152 Planegg-Martinsried, Germany. 2Department of Biochemistry and Biophysics, University of California, San Francisco, CA 94143, USA. 3Department of Anatomy and Cell Biology, University of Florida, Gainesville, FL 32610, USA. 4Max Delbrück Center for Molecular Medicine, 13125 Berlin, Germany. 5Department of Physiology and Program in Bioengineering, University of California, San Francisco, CA 94143, USA.*These authors contributed equally to this work.†Corresponding authors. E-mail: [email protected] (H.L.); [email protected] (J.W.S.)

light, which generate a pattern along the axial (z) direction as well as the lateral (x and y) directions. We implemented 3D-SIM in a cus-tom-built microscope designed for ease of use so that slides prepared for conventionalmi-croscopes can be imaged without any further treatment, and operation of the microscope is similar to any modern commercial system. Although several subdiffraction-resolution optical microscopy methods have been devel-oped in the past decade (6–8) and have been used to address specific biological questions (9, 10), they still present limitations, such as the restriction of the resolution enhancement to either the lateral or the axial direction or to the near or evanescent field, limited mul-ticolor and 3D sectioning abilities, and the requirement of unusual dyes. 3D-SIM is cur-rently the only subdiffraction-resolution im-aging technique that allows detection of three (and potentially more) wavelengths in the same sample, using standard fluorescent dyes, with 3D optical sectioning and enhancement of resolution in both lateral (x and y) and axial (z) directions.

We used 3D-SIM to probe higher-order chromatin structure and the relative local-ization of nuclear pores, the nuclear lamina, and chromatin at the nuclear periphery in mammalian tissue culture cells (11). The ver-tebrate nuclear pore complex (NPC) is a ∼120 MD protein complex (12) that mediates com-munication and selective exchange between the nucleoplasm and cytoplasm. The relation between chromatin, the NPC, and other com-

ponents of the nuclear envelope, such as the nuclear lamina, has been extensively studied (13, 14). Electron microscopy (EM) has been instrumental in determining the fine struc-ture of the NPC (12, 15–17), but it cannot pro-vide a global 3D view of the entire nucleus with specific labeling of the molecular compo-nents. 3D-SIM can bridge this technology gap and shed light on the nuclear periphery.

We first tested the ability of 3D-SIM to resolve the fine structure of interphase chro-matin in three-dimensionally preserved form-aldehyde-fixed mouse C2C12 myoblast cells stained with 4′,6-diamidino-2-phenylindole (DAPI). For comparison we recorded a refer-ence image stack with conventional wide-field epiflu orescence microscopy and applied con-strained iterative deconvolution (18) to re-duce out-of-focus blur (Fig. 2). In the 3D-SIM image, chromatin shows a more evidently fibrous substructure, and a brighter rim of heterochromatin staining is visible near the nuclear envelope as observed by EM. These chromatindense regions are surrounded by chromatin-poor regions, consistent with the interchromatin compartment observed by combined fluorescence and electron micro-scopic studies (19).

Unexpectedly, in 3D-SIM images of the nuclear periphery, we observed thousands of well-defined holes in DAPI staining, which could not be observed in the corresponding wide-field epifluorescence images (Fig. 2B). The size, number, and position of these holes suggested that they represented the

exclusion of DNA from NPCs. To test this hypothesis and the potential of 3D-SIM for simultaneous multicolor 3D imaging of vari-ous nuclear structures, we co-immunostained these cells with NPC-specific antibodies and antibodies against lamin B. For comparison we recorded 3D image stacks of cells from the same sample with state-of-the-art confocal la-ser scanning microscopy (CLSM) and applied deconvolution (Fig. 3, A and B). The interme-diate filament protein lamin B is a major com-ponent of the nuclear lamina that lines and stabilizes the nuclear envelope (20). The NPC antibodies used here (referred to as αNPC) are directed against the FG-repeat domain common to several nuclear pore proteins and mainly detect Nup62, Nup214, and Nup358, which are located in the center and at the cy-toplasmic side of the NPC (21) with a minor signal from Nup153. In lateral cross sections of 3D-SIM image stacks, we consistently ob-served the peripheral heterochromatin rim outlined by a fine heterogeneous layer of the nuclear lamina and nuclear pore signals further above the lamina. This clear triple-layered organization was not resolvable with CLSM (Fig. 3A). For comparison, we also used a monoclonal antibody specific for Nup153, which is located on the nucleoplasmic side of the NPC (21) and obtained a pore signal in the same plane as the lamin B signal, dem-onstrating the potential of 3D-SIM to resolve subtle differences of epitope locations within the NPC (Fig. 3C). In apical cross sections, we found that NPC foci strictly localize within

Subdiffraction Multicolor Imagingof the Nuclear Periphery with 3DStructured Illumination MicroscopyLothar Schermelleh,1* Peter M. Carlton,2* Sebastian Haase,2,4 Lin Shao,2Lukman Winoto,2 Peter Kner,2 Brian Burke,3 M. Cristina Cardoso,4 David A. Agard,2Mats G. L. Gustafsson,5 Heinrich Leonhardt,1*† John W. Sedat2*†

Fluorescence light microscopy allows multicolor visualization of cellular components with highspecificity, but its utility has until recently been constrained by the intrinsic limit of spatialresolution. We applied three-dimensional structured illumination microscopy (3D-SIM) tocircumvent this limit and to study the mammalian nucleus. By simultaneously imaging chromatin,nuclear lamina, and the nuclear pore complex (NPC), we observed several features that escapedetection by conventional microscopy. We could resolve single NPCs that colocalized with channelsin the lamin network and peripheral heterochromatin. We could differentially localize distinct NPCcomponents and detect double-layered invaginations of the nuclear envelope in prophase aspreviously seen only by electron microscopy. Multicolor 3D-SIM opens new and facile possibilitiesto analyze subcellular structures beyond the diffraction limit of the emitted light.

Light microscopy is a key technology inmodern cell biology and, in combinationwith immunofluorescence, fluorescent pro-

tein fusions, or in situ hybridization, allows thespecific localization of nearly all cellular compo-nents. A fundamental limitation of optical micros-copy is its low resolution relative to the scale ofsubcellular structures. This limitation occurs be-cause light traveling through a lens cannot be fo-cused to a point but only to an Airy disk (1) with

a diameter of about half the wavelength of theemitted light (2, 3). Because the wavelengths ofvisible light range from 400 to 700 nm, objectscloser than 200 to 350 nm apart cannot be re-solved but appear merged into one.

Improving resolution beyond the 200-nm dif-fraction limit while retaining the advantages oflight microscopy and the specificity of molecularimaging has been a long-standing goal. Here, wepresent results demonstrating that this goal can be

achieved with the use of a microscope systemthat implements three-dimensional structured il-lumination microscopy (3D-SIM) (4) in an easy-to-use system that makes no extra demands onexperimental procedures. Structured illuminationmicroscopy (SIM) resolves objects beyond thediffraction limit by illuminating with multiple in-terfering beams of light (5). The emitted lightthen contains higher-resolution image informa-tion, encoded by a shift in reciprocal (Fourier, orfrequency) space into observable modulations ofthe image, in a manner similar to the formationof Moiré patterns (fig. S1). This extra informationcan be decoded to reconstruct fine details, re-sulting in an image with twice the resolution of aconventional image taken on the same micro-scope (Fig. 1 and fig. S2). The 3D-SIM methodextends previous 2D SIM methods by usingthree beams of interfering light, which generate apattern along the axial (z) direction as well as thelateral (x and y) directions. We implemented 3D-SIM in a custom-built microscope designed for

1Center for Integrated Protein Science, Department of Biol-ogy, Ludwig Maximilians University Munich, 82152 Planegg-Martinsried, Germany. 2Department of Biochemistry andBiophysics, University of California, San Francisco, CA 94143,USA. 3Department of Anatomy and Cell Biology, University ofFlorida, Gainesville, FL 32610, USA. 4Max Delbrück Center forMolecular Medicine, 13125 Berlin, Germany. 5Department ofPhysiology and Program in Bioengineering, University of Cal-ifornia, San Francisco, CA 94143, USA.

*These authors contributed equally to this work.†To whom correspondence should be addressed. E-mail:[email protected] (H.L.); [email protected] (J.W.S.)

Fig. 1. Subdiffraction resolution imaging with 3D-SIM. (Aand B) Cross section through a DAPI-stained C2C12 cellnucleus acquired with conventional wide-field illumination(A) and with structured illumination (B), showing the stripedinterference pattern (inset). The renderings to the rightillustrate the respective support of detection in frequencyspace. The axes kx, ky, and kz indicate spatial frequenciesalong the x, y, and z directions. The surfaces of therenderings represent the corresponding resolution limit. Thedepression of the frequency support (“missing cone”) in zdirection in (A) indicates the restriction in axial resolution ofconventional wide-field microscopy. With 3D-SIM, the axialsupport is extended but remains within the resolution limit.(C) Five phases of the sine wave pattern are recorded ateach z position, allowing the shifted components to beseparated and returned to their proper location in frequencyspace. Three image stacks are recorded with the diffractiongrating sequentially rotated into three positions 60° apart,resulting in nearly rotationally symmetric support over alarger region of frequency space. (D) The same cross sectionof the reconstructed 3D-SIM image shows enhanced imagedetails compared with the original image (insets). Theincrease in resolution is shown in frequency space on theright, with the coverage extending two times farther fromthe origin. Scale bars indicate 5 mm.

6 JUNE 2008 VOL 320 SCIENCE www.sciencemag.org1332

REPORTS

on

Febr

uary

6, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

6, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

6, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

6, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

6, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

Fig. 1. Subdiffraction resolution imaging with 3D-SIM. (A and B) Cross section through a DAPI-stained C2C12 cell nucleus acquired with conventional wide-field illumi-nation (A) and with structured illumination (B), showing the striped interference pattern (inset). The renderings to the right illustrate the respective support of detection in frequency space. The axes kx, ky, and kz indicate spatial frequencies along the x, y, and z directions. The surfaces of the renderings represent the corresponding resolution limit. The depression of the frequency support (“missing cone”) in z direction in (A) indicates the restriction in ax-ial resolution of conventional wide-field microscopy. With 3D-SIM, the axial support is extended but remains within the resolution limit. (C) Five phases of the sine wave pat-tern are recorded at each z position, allowing the shifted components to be separated and returned to their proper location in frequency space. Three image stacks are re-corded with the diffraction grating sequentially rotated into three positions 60° apart, resulting in nearly rotation-ally symmetric support over a larger region of frequency space. (D) The same cross section of the reconstructed 3D-SIM image shows enhanced image details compared with the original image (insets). The increase in resolu-tion is shown in frequency space on the right, with the coverage extending two times farther from the origin. Scale bars indicate 5 μm.

ease of use so that slides prepared for conven-tional microscopes can be imaged without anyfurther treatment, and operation of the microscopeis similar to any modern commercial system. Al-though several subdiffraction-resolution opticalmicroscopy methods have been developed inthe past decade (6–8) and have been used to ad-dress specific biological questions (9, 10), theystill present limitations, such as the restriction ofthe resolution enhancement to either the lateralor the axial direction or to the near or evanes-cent field, limited multicolor and 3D sectioningabilities, and the requirement of unusual dyes. 3D-SIM is currently the only subdiffraction-resolutionimaging technique that allows detection of three(and potentially more) wavelengths in the samesample, using standard fluorescent dyes, with3D optical sectioning and enhancement of res-olution in both lateral (x and y) and axial (z)directions.

We used 3D-SIM to probe higher-order chro-matin structure and the relative localization ofnuclear pores, the nuclear lamina, and chromatinat the nuclear periphery in mammalian tissueculture cells (11). The vertebrate nuclear porecomplex (NPC) is a ~120 MD protein complex(12) that mediates communication and selectiveexchange between the nucleoplasm and cyto-plasm. The relation between chromatin, the NPC,and other components of the nuclear envelope,such as the nuclear lamina, has been extensivelystudied (13, 14). Electron microscopy (EM) hasbeen instrumental in determining the fine struc-ture of the NPC (12, 15–17), but it cannot pro-

vide a global 3D view of the entire nucleus withspecific labeling of the molecular components.3D-SIM can bridge this technology gap and shedlight on the nuclear periphery.

We first tested the ability of 3D-SIM to re-solve the fine structure of interphase chromatinin three-dimensionally preserved formaldehyde-fixed mouse C2C12 myoblast cells stained with4′,6-diamidino-2-phenylindole (DAPI). For com-parison we recorded a reference image stack withconventional wide-field epifluorescence micros-copy and applied constrained iterative deconvo-lution (18) to reduce out-of-focus blur (Fig. 2). Inthe 3D-SIM image, chromatin shows a more evi-dently fibrous substructure, and a brighter rim ofheterochromatin staining is visible near the nuclearenvelope as observed by EM. These chromatin-dense regions are surrounded by chromatin-poorregions, consistent with the interchromatin com-partment observed by combined fluorescenceand electron microscopic studies (19).

Unexpectedly, in 3D-SIM images of the nu-clear periphery, we observed thousands of well-defined holes in DAPI staining, which could notbe observed in the corresponding wide-field epi-fluorescence images (Fig. 2B). The size, number,and position of these holes suggested that theyrepresented the exclusion of DNA from NPCs.To test this hypothesis and the potential of 3D-SIM for simultaneous multicolor 3D imaging ofvarious nuclear structures, we co-immunostainedthese cells with NPC-specific antibodies andantibodies against lamin B. For comparisonwe recorded 3D image stacks of cells from the

same sample with state-of-the-art confocal laserscanning microscopy (CLSM) and applied de-convolution (Fig. 3, A and B). The intermediatefilament protein lamin B is a major componentof the nuclear lamina that lines and stabilizes thenuclear envelope (20). The NPC antibodies usedhere (referred to as aNPC) are directed againstthe FG-repeat domain common to several nu-clear pore proteins and mainly detect Nup62,Nup214, and Nup358, which are located in thecenter and at the cytoplasmic side of the NPC(21) with a minor signal from Nup153. In lateralcross sections of 3D-SIM image stacks, we con-sistently observed the peripheral heterochroma-tin rim outlined by a fine heterogeneous layer ofthe nuclear lamina and nuclear pore signals fur-ther above the lamina. This clear triple-layeredorganization was not resolvable with CLSM (Fig.3A). For comparison, we also used a monoclonalantibody specific for Nup153, which is locatedon the nucleoplasmic side of the NPC (21) andobtained a pore signal in the same plane as thelamin B signal, demonstrating the potential of 3D-SIM to resolve subtle differences of epitope loca-tions within the NPC (Fig. 3C). In apical crosssections, we found that NPC foci strictly localizewithin DAPI voids (Fig. 3B and movie S1). Withvery few exceptions, every DAPI void containeda focus of NPC staining and vice versa, suggest-ing that most if not all NPCs exclude chromatinfrom their immediate vicinity.

We calculated the density of NPC foci in 3D-SIM and confocal images (Fig. 3 and fig. S3) byusing automatic detection with identical criteria.

Fig. 2. Comparison of wide-field imaging and 3D-SIM in resolving interphase chromatin fine struc-ture. 3D image stacks of the same DAPI-stainedC2C12 cell nucleus were recorded with con-ventional wide-field illumination (left) and with3D-SIM (right). Deconvolution was applied to thewide-field data set (middle). Scale bar, 5 mm.Arrowheads indicate the position of the respectivecutting planes. (A) Mid cross section shows brightlystained chromocenters of clustered (peri-)centromericheterochromatin. (Insets) Higher-detail informa-tion of chromatin substructures when recorded with3D-SIM. Arrow points to a less-bright chromatinstructure that has been spuriously eroded by thedeconvolution procedure. (B) Projection of fourapical sections (corresponding to a thickness of0.5 mm) taken from the surface of the nuclearenvelope closest to the coverslip. Whereas the rawimage shows diffuse DAPI staining, the decon-volved image shows more pronounced variationsin fluorescence intensities. Image data with 3D-SIM extended-resolution information reveal apunctuated pattern of regions devoid of DAPIstaining. (C) Orthogonal cross-section throughthe entire 3D image stack demonstrates the lowsectioning capability of conventional wide-fieldmicroscopy, which cannot be mitigated solely bydeconvolution. In contrast, clear layers of pe-ripheral heterochromatin can be resolved with3D-SIM (arrows).

www.sciencemag.org SCIENCE VOL 320 6 JUNE 2008 1333

REPORTS

Fig. 2. Comparison of wide-field imaging and 3D-SIM in resolving interphase chromatin fine structure. 3D image stacks of the same DAPI-stained C2C12 cell nucleus were recorded with conventional wide-field illumination (left) and with 3D-SIM (right). Deconvolution was ap-plied to the wide-field data set (middle). Scale bar, 5 μm. Arrowheads indicate the position of the respective cutting planes. (A) Mid cross sec-tion shows brightly stained chromocenters of clustered (peri-) centromeric heterochromatin. (Insets) Higher-detail information of chromatin substructures when recorded with 3D-SIM. Ar-row points to a less-bright chromatin structure that has been spuriously eroded by the decon-volution procedure. (B) Projection of four api-cal sections (corresponding to a thickness of 0.5 μm) taken from the surface of the nuclear envelope closest to the coverslip. Whereas the raw image shows diffuse DAPI staining, the de-convolved image shows more pronounced varia-tions in fluorescence intensities. Image data with 3D-SIM extended-resolution information reveal a punctuated pattern of regions devoid of DAPI staining. (C) Orthogonal cross-section through the entire 3D image stack demonstrates the low sectioning capability of conventional wide-field microscopy, which cannot be mitigated solely by deconvolution. In contrast, clear layers of pe-ripheral heterochromatin can be resolved with 3D-SIM (arrows).

Originally published 6 June 2008 in SCIENCE








mailto:[email protected]


























DAPI voids (Fig. 3B and movie S1). With very few exceptions, every DAPI void contained a focus of NPC staining and vice versa, suggest-ing that most if not all NPCs exclude chroma-tin from their immediate vicinity.

We calculated the density of NPC foci in 3D-SIM and confocal images (Fig. 3 and fig. S3) by using automatic detection with identi-cal criteria. The 3D-SIM image showed an av-erage (±SD) of 5.6 ± 3.3 foci per μm2, whereas only 2.8 ± 1.1 foci per μm2 were detected in the confocal image. By comparison, 12 ± 1.8 NPCs per μm2 have been observed in mouse liver nuclei by freeze-fracture EM (22). Although the numbers cannot be directly compared because of cell type and cell cycle–dependent variations (23), they still indicate that 3D-SIM

detects and resolves most NPCs and outper-forms CLSM.

Although the discrete chromatin voids at NPCs are not visible in confocal or stan-dard wide-field images, the intensity of DAPI staining should still show fluctuations that anticorrelate with NPC foci. We also rea-soned that, even though the lamina does not show a striking one-to-one exclusion from NPCs under any of our methods, it should still be anticorrelated with the NPCs. To pur-sue these questions, we examined the aver-age environment of an NPC focus in our im-ages (fig. S4). Sub-images were cropped from the data set centered on the peak intensities of NPC foci, which were automatically de-tected in 3D. Intensity profiles through the

center of each sub-image show a drop in intensity in DAPI staining, centered on the NPC focus, in all confocal and 3D-SIM im-ages. A similar drop in intensity of the lamin signal is seen only in the case of 3D-SIM. In composites of the sub-images, a central hole can be seen in the DAPI channel. This indi-cates that the expected intensity fluctuations are present in all cases. The composite lam-ina image recorded by 3D-SIM also shows a hole, which was not detected by confocal microscopy. The width of the NPC signal was determined from intensity profiles by mea-suring the full width at half maximum. With CLSM this width was ∼200 nm (192 ± 17 and 192 ± 11 nm before and after deconvolution, respectively), which essentially reflects the

point spread function (PSF) of CLSM. In con-trast, the measured width of NPC signal re-corded with 3D-SIM was 120 ± 3 nm, which is in good agreement with EM measurements (15).

Many of the nuclei we imaged contained invaginations of the nuclear envelope that are especially prominent near the centro-some during prophase. These invaginations have been shown by EM studies to be tubular extensions of both nuclear envelopes (24), but because their width is generally smaller than the diffraction limit they appear as a single line by conventional microscopy. We inves-tigated these nuclear invaginations in pro-phase nuclei by CLSM and 3D-SIM (Fig. 4) and could resolve the bilaminar structure of

these invagination only with 3D-SIM (movie S2). This detailed substructure of nuclear in-vaginations has so far only been detected by transmission electron microscopy (TEM) (fig. S5). We also quantified the width of the lam-in B signal in interphase nuclei in lateral and axial directions by fitting Gaussian curves to the measured intensities. Because the thick-ness of the lamina is in the range of 20 to 50 nm, the obtained values should reflect the resolution limit of either of the applied meth-ods. With 3D-SIM we determined a width of threads in the lamin network with an upper limit of 98 ± 12 nm laterally and 299 ± 22 nm axially, whereas confocal images showed more than twofold higher values (243 ± 30 nm laterally and 736 ± 225 nm axially). Sub-

sequent deconvolution of CLSM data/images did not improve the lateral (231 ± 28 nm) but did improve the axial (418 ± 26 nm) resolu-tion, which is mostly due to an increased sig-nal-to-noise ratio by removing out-of-focus blur and suppression of background (2). This 3D imaging of complex biological structures demonstrates about twofold enhanced reso-lution of 3D-SIM over conventional fluores-cence imaging techniques in lateral and axial directions.

Many cellular structures and macromo-lecular complexes, including the nuclear envelope and its pores, fall just below the diffraction limit of conventional light mi-croscopy, preventing quantitative analysis. In doubling the resolution of conventional

The 3D-SIM image showed an average (±SD)of 5.6 ± 3.3 foci per mm2, whereas only 2.8 ±1.1 foci per mm2 were detected in the confocalimage. By comparison, 12 ± 1.8 NPCs per mm2

have been observed in mouse liver nuclei byfreeze-fracture EM (22). Although the numberscannot be directly compared because of cell typeand cell cycle–dependent variations (23), theystill indicate that 3D-SIM detects and resolvesmost NPCs and outperforms CLSM.

Although the discrete chromatin voids at NPCsare not visible in confocal or standard wide-fieldimages, the intensity of DAPI staining shouldstill show fluctuations that anticorrelate with NPCfoci. We also reasoned that, even though the lam-ina does not show a striking one-to-one exclu-sion from NPCs under any of our methods, itshould still be anticorrelated with the NPCs. Topursue these questions, we examined the aver-

age environment of an NPC focus in our images(fig. S4). Sub-images were cropped from thedata set centered on the peak intensities of NPCfoci, which were automatically detected in 3D.Intensity profiles through the center of each sub-image show a drop in intensity in DAPI stain-ing, centered on the NPC focus, in all confocaland 3D-SIM images. A similar drop in intensityof the lamin signal is seen only in the case of 3D-SIM. In composites of the sub-images, a centralhole can be seen in the DAPI channel. This in-dicates that the expected intensity fluctuationsare present in all cases. The composite laminaimage recorded by 3D-SIM also shows a hole,which was not detected by confocal microscopy.The width of the NPC signal was determinedfrom intensity profiles by measuring the fullwidth at half maximum. With CLSM this widthwas ~200 nm (192 ± 17 and 192 ± 11 nm before

and after deconvolution, respectively), whichessentially reflects the point spread function(PSF) of CLSM. In contrast, the measured widthof NPC signal recorded with 3D-SIM was 120 ±3 nm, which is in good agreement with EMmeasurements (15).

Many of the nuclei we imaged contained in-vaginations of the nuclear envelope that are es-pecially prominent near the centrosome duringprophase. These invaginations have been shownby EM studies to be tubular extensions of bothnuclear envelopes (24), but because their widthis generally smaller than the diffraction limit theyappear as a single line by conventional micros-copy. We investigated these nuclear invagina-tions in prophase nuclei by CLSM and 3D-SIM(Fig. 4) and could resolve the bilaminar struc-ture of these invagination only with 3D-SIM(movie S2). This detailed substructure of nuclear

Fig. 3. Simultaneous imaging of DNA, nuclear lamina, and NPC epitopes by 3D-SIM. C2C12 cells are immu-nostained with antibodies against lamin B (green) and antibodies that recognize different NPC epitopes (red).DNA is counterstained with DAPI (blue). (A) Central cross sections. (B) Projections of four apical sections (cor-responding to a thickness of 0.5 mm). Boxed regions are shown below at 4× magnification; scale bars indicate5 mm and 1 mm, respectively. (A) CLSM and deconvolution still show partially overlapping signals. In contrast, with 3D-SIM the spatial separation of NPC, lamina, andchromatin and chromatin-free channels underneath nuclear pores are clearly visible. (B) Top view on the nuclear envelope. Whereas CLSM fails to resolve close nuclearpores, 3D-SIM shows clearly separated NPC signals at voids of peripheral chromatin and surrounded by an irregular network of nuclear lamina. (C) Mid sectionscomparing stainings with an antibody that mainly reacts with Nup214, Nup358, and Nup62 (aNPC) and one specifically recognizing Nup153 (aNup153). The aNPCsignal is above the lamina (140 ± 8 nm), whereas the aNup153 pore signal is at the same level as the lamina (–15 ± 20 nm). Scale bars 1 mm. (D) Schematic outlineof the NPC, showing the relative position of Nup proteins and surrounding structures. ONM, outer nuclear membrane; INM, inner nuclear membrane.

6 JUNE 2008 VOL 320 SCIENCE www.sciencemag.org1334

REPORTS

Fig. 3. Simultaneous imaging of DNA, nuclear lamina, and NPC epitopes by 3D-SIM. C2C12 cells are immunostained with antibodies against lamin B (green) and antibodies that recognize dif-ferent NPC epitopes (red). DNA is counterstained with DAPI (blue). (A) Central cross sections. (B) Projections of four apical sections (corresponding to a thickness of 0.5 μm). Boxed regions are shown below at 4× magnification; scale bars indicate 5 μm and 1 μm, respectively. (A) CLSM and deconvolu-tion still show partially overlapping signals. In contrast, with 3D-SIM the spatial separation of NPC, lamina, and chromatin and chromatin-free channels underneath nuclear pores are clearly visible. (B) Top view on the nuclear envelope. Whereas CLSM fails to resolve close nuclear pores, 3D-SIM shows clearly separated NPC signals at voids of peripheral chromatin and surrounded by an irregular network of nuclear lamina. (C) Mid sections comparing stainings with an antibody that mainly reacts with Nup214, Nup358, and Nup62 (αNPC) and one specifically recognizing Nup153 (αNup153). The αNPC signal is above the lamina (140 ± 8 nm), whereas the αNup153 pore signal is at the same level as the lamina (–15 ± 20 nm). Scale bars 1 μm. (D) Schematic outline of the NPC, showing the relative position of Nup proteins and surrounding structures. ONM, outer nuclear membrane; INM, inner nuclear membrane.

invaginations has so far only been detected bytransmission electron microscopy (TEM) (fig.S5). We also quantified the width of the lamin Bsignal in interphase nuclei in lateral and axialdirections by fitting Gaussian curves to the mea-sured intensities. Because the thickness of thelamina is in the range of 20 to 50 nm, the ob-tained values should reflect the resolution limitof either of the applied methods. With 3D-SIM we determined a width of threads in thelamin network with an upper limit of 98 ± 12 nmlaterally and 299 ± 22 nm axially, whereasconfocal images showed more than twofoldhigher values (243 ± 30 nm laterally and 736 ±225 nm axially). Subsequent deconvolution ofCLSM data/images did not improve the lateral(231 ± 28 nm) but did improve the axial (418 ±26 nm) resolution, which is mostly due to anincreased signal-to-noise ratio by removing out-of-focus blur and suppression of background

(2). This 3D imaging of complex biological struc-tures demonstrates about twofold enhancedresolution of 3D-SIM over conventional fluo-rescence imaging techniques in lateral and axialdirections.

Many cellular structures and macromolecularcomplexes, including the nuclear envelope andits pores, fall just below the diffraction limit ofconventional light microscopy, preventing quan-titative analysis. In doubling the resolution ofconventional microscopy in three dimensions,3D-SIM was able to resolve individual nuclearpores, detect and measure the exclusion of chro-matin and the nuclear lamina from nuclear pores,and accurately image invaginations of the nuclearenvelope caused by the formation of the mitoticspindle. We have demonstrated both an increasein quantitative precision of measurement and thedetection of novel cytological features, by imag-ing to a resolution approaching 100 nm. Although

this level of resolution is less than that affordedby other techniques such as stimulated emissiondepletion (STED), photoactivated localization(PALM), or stochastic optical reconstruction(STORM) microscopy (25–29), 3D-SIM is cur-rently the only subdiffraction-resolution imag-ing technique that can produce multicolor 3Dimages of whole cells with enhancement ofresolution in both lateral and axial directions.Notably, these results were obtained with stan-dard cytological methods, without the need forunconventional fluorescent dyes or coverslips, andon amicroscope platform designed to be nomoredifficult to use than a conventional commercialmicroscope. The possibility of using 3D-SIMwith well-established standard labeling tech-niques and to simultaneously locate differentmolecules or structures in the 3D cellular contextopens interesting new perspectives for molecularcell biology.

Fig. 4. Invaginations of the nuclear envelope in mitotic prophase. C2C12 cells are immunostained with antibodies againstlamin B (green), and the DNA is counterstained with DAPI (magenta). (A) Maximum intensity projections, (B) lateral, and(C) orthogonal cross sections. Scale bars, 5 mm. Inset boxes are shown below at 4× magnification (scale bars, 1 mm). 3D-SIM reveals the globular substructure of condensed chromosomes and the fine-structured fibrillar network of the nuclearlamina (movie S2). (D) Lateral and axial line profiles of dashed lines in (B) and (C), respectively. Whereas the lateral peakwidths are only slightly decreased in the lateral direction (1 versus 1′), the axial profiles show clearly decreased peak widthsafter deconvolution, indicating a substantial improvement of axial resolution (2 versus 2′). With 3D-SIM, the peak widths ofthe lamin B signal are about halved with respect to the deconvolved confocal image in lateral and axial directions, and doublepeaks are resolved where only single peaks are seen in the confocal profiles. Similarly, DAPI staining shows multiple smallpeaks, which again reflects the increase of image details in 3D-SIM.

www.sciencemag.org SCIENCE VOL 320 6 JUNE 2008 1335

REPORTS

Fig. 4. Invaginations of the nuclear envelope in mitotic prophase. C2C12 cells are immunostained with antibod-ies against lamin B (green), and the DNA is counterstained with DAPI (magenta). (A) Maximum intensity projec-tions, (B) lateral, and (C) orthogonal cross sections. Scale bars, 5 mm. Inset boxes are shown below at 4× magnification (scale bars, 1 μm). 3DSIM reveals the globular substructure of condensed chromosomes and the fine-structured fibrillar network of the nuclear lamina (movie S2). (D) Lateral and axial line profiles of dashed lines in (B) and (C), respectively. Whereas the lateral peak widths are only slightly decreased in the lateral direction (1 versus 1′), the axial profiles show clearly decreased peak widths after deconvolution, indicating a substantial improvement of axial resolution (2 versus 2′). With 3D-SIM, the peak widths of the lamin B signal are about halved with respect to the deconvolved confocal image in lateral and axial directions, and double peaks are resolved where only single peaks are seen in the confocal profiles. Similarly, DAPI staining shows multiple small peaks, which again reflects the increase of image details in 3D-SIM.












microscopy in three dimensions, 3D-SIM was able to resolve individual nuclear pores, de-tect and measure the exclusion of chromatin and the nuclear lamina from nuclear pores, and accurately image invaginations of the nuclear envelope caused by the formation of the mitotic spindle. We have demonstrated both an increase in quantitative precision of measurement and the detection of novel cy-tological features, by imaging to a resolution approaching 100 nm. Although this level of resolution is less than that afforded by other techniques such as stimulated emission de-pletion (STED), photoactivated localization (PALM), or stochastic optical reconstruction (STORM) microscopy (25–29), 3D-SIM is cur-rently the only subdiffraction-resolution im-aging technique that can produce multicolor 3D images of whole cells with enhancement of resolution in both lateral and axial direc-tions. Notably, these results were obtained with standard cytological methods, without the need for unconventional fluorescent dyes or coverslips, and on a microscope platform designed to be no more difficult to use than a conventional commercial microscope. The possibility of using 3D-SIM with well-estab-lished standard labeling techniques and to simultaneously locate different molecules or structures in the 3D cellular context opens interesting new perspectives for molecular cell biology.

REFERENCES AND NOTES 1. M. Born, E. Wolf, Eds., Principle of Optics (Cambridge Univ. Press, Cambridge, 1998). 2. J. B. Pawley, Ed., Handbook of Biological Confocal Microscopy (Springer, New York, ed. 3, 2006). 3. E. Abbe, Arch. Mikrosk. Anat. 9, 413 (1873). 4. M. G. Gustafsson et al., Biophys. J., in press; published online 7 March 2008 (10.1529 biophysj.107.120345). 5. M. G. Gustafsson, J. Microsc. 198, 82 (2000). 6. R. Heintzmann, G. Ficz, Brief. Funct. Genomics Proteomics 5, 289 (2006). 7. S. W. Hell, Nat. Biotechnol. 21, 1347 (2003). 8. S. W. Hell, Science 316, 1153 (2007). 9. R. J. Kittel et al., Science 312, 1051 (2006); published online 13 April 2006 (10.1126/science.1126308).10. K. I. Willig, S. O. Rizzoli, V. Westphal, R. Jahn, S. W. Hell, Nature 440, 935 (2006).11. Material and methods are available on Science Online.12. R. Reichelt et al., J. Cell Biol. 110, 883 (1990).13. B. Burke, C. L. Stewart, Nat. Rev. Mol. Cell Biol. 3, 575 (2002).14. M. R. Paddy, A. S. Belmont, H. Saumweber, D. A. Agard, J. W. Sedat, Cell 62, 89 (1990).15. M. Beck et al., Science 306, 1387 (2004); published online 28 October 2004 (10.1126/science.1104808).16. M. Beck, V. Lucic, F. Forster, W. Baumeister, O. Medalia, Nature 449, 611 (2007).17. D. Stoffler et al., J. Mol. Biol. 328, 119 (2003).18. D. A. Agard, Y. Hiraoka, P. Shaw, J. W. Sedat, Methods Cell Biol. 30, 353 (1989).19. H. Albiez et al., Chromosome Res. 14, 707 (2006).20. L. Gerace, A. Blum, G. Blobel, J. Cell Biol. 79, 546 (1978).21. B. Fahrenkrog et al., J. Struct. Biol. 140, 254 (2002).22. R. Tonini, F. Grohovaz, C. A. Laporta, M. Mazzanti, FASEB J. 13, 1395 (1999).23. M. Winey, D. Yarar, T. H. Giddings Jr., D. N. Mastronarde, Mol. Biol. Cell 8, 2119 (1997).24. M. Fricker, M. Hollinshead, N. White, D. Vaux, J. Cell Biol. 136, 531 (1997).25. M. Bates, B. Huang, G. T. Dempsey, X. Zhuang, Science

317, 1749 (2007); published online 15 August 2007 (10.1126/science.1146598).26. E. Betzig et al., Science 313, 1642 (2006); published online 9 August 2006 (10.1126/science.1127344).27. G. Donnert et al., Proc. Natl. Acad. Sci. U.S.A. 103, 11440 (2006).28. B. Huang, W. Wang, M. Bates, X. Zhuang, Science 319, 810 (2008); published online 2 January 2008 (10.1126/science.1153529).29. M. J. Rust, M. Bates, X. Zhuang, Nat. Methods 3, 793 (2006).

ACKNOWLEDGMENTSThis work was supported by grants from the Bavaria California Technology Center, the Center for NanoScience, the Nanosystems Initiative Munich, and the Deutsche Forschungsgemeinschaft to L. Schermelleh, M.C.C., and H.L.; by NIH grant GM-2501–25 to J.W.S.; by the David and Lucile Packard Foundation; and by NSF through the Center for Biophotonics, an NSF Science and Technology Center managed by the University of California, Davis, under cooperative agree-ment no. PHY 0120999. P.M.C. is partially supported by the Keck Laboratory for Advanced Microscopy at the University of California, San Francisco. We thank A. Čopïč, K. Weis, and F. Spada for comments on the manuscript and helpful discussions. P.M.C., L. Shao, L.W., and P.K. have performed limited paid consulting for Applied Precision, which is planning a commercial microscope system using three-dimensional structured illumination. The University of California holds patents for structured illumination microscopy.

SUPPORTING ONLINE MATERIALwww.sciencemag.org/cgi/content/full/320/5881/1332/DC1 Materials and MethodsFigs. S1 to S6References and Notes Movies S1 and S2

25 February 2008; accepted 13 May 2008 10.1126/science.1156947.

Transfected cells expressing fluorescent proteins (1) contain information that is accurate at the molecular level about the spatial organization of the tar-get proteins to which they are bound.

However, the best resolution that can be obtained by diffraction-limited conventional optical techniques is coarser than the molec-ular level by two orders of magnitude. Great progress has been made with superresolution

We introduce a method for optically imaging intracellular proteins at nanometer spatial resolution. Numerous sparse subsets of photoactivatable fluorescent protein molecules were activated, localized (to ∼2 to 25 nanometers), and then bleached. The aggregate po-sition information from all subsets was then assembled into a superresolution image. We used this method—termed photoactivated localization microscopy—to image specif-ic target proteins in thin sections of lysosomes and mitochondria; in fixed whole cells, we imaged vinculin at focal adhesions, actin within a lamellipodium, and the distribution of the retroviral protein Gag at the plasma membrane.

Eric Betzig,1,2*† George H. Patterson,3 Rachid Sougrat,3 O. Wolf Lindwasser,3 Scott Olenych,4

Juan S. Bonifacino,3 Michael W. Davidson,4 Jennifer Lippincott-Schwartz,3 Harald F. Hess5*

Imaging intracellular fluorescentproteins at nanometer resolution

methods that penetrate beyond this limit, such as near field (2), stimulated emission depletion (3), structured illumination (4, 5), and revers-ible saturable optical fluorescence transitions microscopy (6), but the goal remains a fluores-cence technique capable of achieving resolu-tion closer to the molecular scale.

Early results (7) in single-molecule mi-croscopy (8) and the spatiospectral isolation of individual exciton recombination sites in

a semiconductor quantum well (9) led to a proposal for a means of molecular resolution fluorescence microscopy a decade ago (10). In brief, individual molecules densely packed within the resolution limit of a given instru-ment [as defined by its point-spread function (PSF)] are first isolated from one another on the basis of one or more distinguishing optical characteristics. Each molecule is then local-ized to much higher precision by determining its center of fluorescence emission through a statistical fit of the ideal PSF to its measured photon distribution (Fig. 1). When the back-ground noise is negligible compared with the molecular signal, the error in the fitted posi-tion is σ

x, y ≈ s/(N½), where s is the standard

deviation of a Gaussian approximating the

true PSF (≈200 nm for light of wavelength λ = 500 nm) and N is the total number of detected photons (11, 12). Given that it is possible to detect many more than 104 photons from a single fluorophore before it bleaches, single-molecule localization to nearly 1-nm preci-sion has already been demonstrated (13–15) and applied to studies of molecular motor dynamics (13).

Multiple emitters within a single diffrac-tion-limited region (DLR) have been isolated from one another by either spectral (15, 16)

or temporal means, the latter exploiting the photobleaching (14, 17) or blinking (18) of the emitters. However, the number of emit-ters isolated per DLR (typically 2 to 5) has been too small to give resolution within the DLR that is comparable to existing super-resolution techniques, and it is far from the molecular level. Here, we developed a method for isolation of single molecules at high den-sities (up to ∼105/μm2) based on the serial photoactivation and subsequent bleaching of numerous sparse subsets of photoactivat-able fluorescent protein (PA-FP) molecules (19–24) within a sample. We then applied the method to image specific target proteins in thin (∼50- to 80-nm) sections and near the surfaces of fixed cultured cells, resolving the most precisely localized molecules therein at separations (∼10 nm) approaching the mo-lecular level.

The method and typical data subsets are shown in Fig. 1. Cultured mammalian cells expressing PA-FP–tagged target proteins were prepared by transient transfection, fixed, and processed on cover slips either as whole cells or in cryosections cut from a centrifuged pellet of cells (25). Such cover slips were then placed

in a custom microscope chamber (fig. S1) de-signed to minimize thermal and mechanical drift (fig. S2) (25). They were continuously excited by a laser at a wavelength (λ

exc = 561

nm) near the excitation maximum of the ac-tivated form of the expressed PA-FPs. Finally, to minimize both autofluorescence and detec-tor noise, they were imaged by total internal reflection fluorescence (TIRF) microscopy (13, 26) onto an electron-multiplying charge cou-pled device (EMCCD) camera that can detect single photons.

Initial image frames typically consisted of sparse fields of individually resolvable single molecules on a weaker background presumably dominated by the much larger population of PA-FP molecules still in the inactivated state. When necessary, excitation and thus bleaching was maintained until such sparse fields were obtained. Additional image frames were then captured until single-molecule bleaching resulted in a mean molecular separation considerably larger than that required for isolation (Fig. 1, A and C). At that point, we applied a light pulse from a second laser at a wavelength (λ

act =

405 nm) capable of activating the remaining

the frames in which it appears. This result (Fig.

1G, left) is then fitted using a robust nonlinear

least squares algorithm to an assumed Gaussian

PSF of free center coordinates xo, y

o(Fig. 1G,

center) (25), yielding coordinates xm, y

mfor

the location of the molecule, with a position

uncertainty (sx, y

)m. Finally, each molecule is

rendered in a new xy frame as a Gaussian of

standard deviation (sx, y

)m

(rather than the

much larger standard deviation s of the original

PSF), centered at xm, y

m(Fig. 1, G, right, and

A¶ to D¶) and normalized to unit strength when

integrated over all xy space. Thus, the super-

resolution image obtained by summing the

rendered Gaussians associated with all localized

molecules in the original image stack (Fig. 1, E¶

and F¶) provides a probability density map

where brightness is proportional to the likeli-

hood that a PA-FP molecule can be found at a

given location.

This technique, termed photoactivated lo-

calization microscopy (PALM), is capable of

resolving the most precisely localized mole-

cules at separations of a few nanometers. These

represent the very brightest emitters (the much

larger population of all isolated molecules ex-

hibits a much broader range of photon counts;

fig. S4). Thus, when rendering PALM images,

a fundamental trade-off exists: Including fewer,

but brighter, molecules results in higher local-

ization and crisper images, but at a reduced

molecular density giving less complete infor-

mation about the spatial distribution of the

target protein (fig. S5). Both parameters—

localization precision and the density of rendered

molecules—are key to defining performance in

PALM. Their specific values for the images in

Figs. 2 to 4 are given in table S1.

This performance is largely dictated by the

photophysical characteristics of the PA-FPs.

Longer photobleaching half-life leads to more

photons per molecule, but for a given excitation

intensity, it also requires longer data acquisition

times between activation pulses to maintain

an appropriate density of individually resolv-

able molecules. Higher excitation cross-sectional

s and fluorescence quantum efficiency F can

speed this process of signal extraction and

bleaching, with the added benefit of increasing

the molecular contrast relative to the autofluo-

rescence background. Also vital is the contrast

C(lexc

) 0 (sF)act/(sF)

inactbetween the PA-FP

in its activated and inactivated state at lexc

,

because this dictates the maximum molecular

density beyond which the background from

many weakly emitting inactivated molecules in

a DLR dominates the signal from a single ac-

tivated one. PA-FPs that remain activated until

bleached ensure that all possible photons are

extracted. Finally, PA-FPs less prone to blinking

are desirable, given that it can be difficult to

distinguish a single blinking molecule from

multiple molecules that are serially activated

and bleached in the same DLR (25).

Although we have demonstrated isolation

and localization with both green Ephotoactivatablegreen fluorescent protein (PA-GFP) and Dronpa^

Fig. 2. Comparative summed-molecule TIRF (A) and PALM (B) images of the same region within a cryo-prepared thin section from a COS-7 cell expressing the lysosomal transmembrane protein CD63 taggedwith the PA-FP Kaede. The larger boxed region in (B), when viewed at higher magnification (C) revealssmaller associated membranes that may represent interacting lysosomes or late endosomes that are notresolvable by TIRF. In a region where the section is nearly orthogonal to the lysosomal membrane, the mosthighly localized molecules fall on a line of width È10 nm (inset). In an obliquely cut region [(D), from thesmaller boxed region in (B)], the distribution of CD63 within the membrane plane can be discerned.

Fig. 1. The principle behind PALM. A sparse subsetof PA-FP molecules that are attached to proteins ofinterest and then fixed within a cell are activated (Aand B) with a brief laser pulse at lact 0 405 mmand then imaged at lexc 0 561 mm until most arebleached (C). This process is repeated many times(C and D) until the population of inactivated,unbleached molecules is depleted. Summing themolecular images across all frames results in adiffraction-limited image (E and F). However, ifthe location of each molecule is first determinedby fitting the expected molecular image given bythe PSF of the microscope [(G), center] to theactual molecular image [(G), left], the moleculecan be plotted [(G), right] as a Gaussian that hasa standard deviation equal to the uncertaintysx,y in the fitted position. Repeating with allmolecules across all frames (A¶ through D¶) andsumming the results yields a superresolutionimage (E¶ and F¶) in which resolution is dictatedby the uncertainties sx,y as well as by the densityof localized molecules. Scale: 1 � 1 mm in (F) and(F¶), 4 � 4 mm elsewhere.

www.sciencemag.org SCIENCE VOL 313 15 SEPTEMBER 2006 1643

REPORTS

the frames in which it appears. This result (Fig.

1G, left) is then fitted using a robust nonlinear

least squares algorithm to an assumed Gaussian

PSF of free center coordinates xo, y

o(Fig. 1G,

center) (25), yielding coordinates xm, y

mfor

the location of the molecule, with a position

uncertainty (sx, y

)m. Finally, each molecule is

rendered in a new xy frame as a Gaussian of

standard deviation (sx, y

)m

(rather than the

much larger standard deviation s of the original

PSF), centered at xm, y

m(Fig. 1, G, right, and

A¶ to D¶) and normalized to unit strength when

integrated over all xy space. Thus, the super-

resolution image obtained by summing the

rendered Gaussians associated with all localized

molecules in the original image stack (Fig. 1, E¶

and F¶) provides a probability density map

where brightness is proportional to the likeli-

hood that a PA-FP molecule can be found at a

given location.

This technique, termed photoactivated lo-

calization microscopy (PALM), is capable of

resolving the most precisely localized mole-

cules at separations of a few nanometers. These

represent the very brightest emitters (the much

larger population of all isolated molecules ex-

hibits a much broader range of photon counts;

fig. S4). Thus, when rendering PALM images,

a fundamental trade-off exists: Including fewer,

but brighter, molecules results in higher local-

ization and crisper images, but at a reduced

molecular density giving less complete infor-

mation about the spatial distribution of the

target protein (fig. S5). Both parameters—

localization precision and the density of rendered

molecules—are key to defining performance in

PALM. Their specific values for the images in

Figs. 2 to 4 are given in table S1.

This performance is largely dictated by the

photophysical characteristics of the PA-FPs.

Longer photobleaching half-life leads to more

photons per molecule, but for a given excitation

intensity, it also requires longer data acquisition

times between activation pulses to maintain

an appropriate density of individually resolv-

able molecules. Higher excitation cross-sectional

s and fluorescence quantum efficiency F can

speed this process of signal extraction and

bleaching, with the added benefit of increasing

the molecular contrast relative to the autofluo-

rescence background. Also vital is the contrast

C(lexc

) 0 (sF)act/(sF)

inactbetween the PA-FP

in its activated and inactivated state at lexc

,

because this dictates the maximum molecular

density beyond which the background from

many weakly emitting inactivated molecules in

a DLR dominates the signal from a single ac-

tivated one. PA-FPs that remain activated until

bleached ensure that all possible photons are

extracted. Finally, PA-FPs less prone to blinking

are desirable, given that it can be difficult to

distinguish a single blinking molecule from

multiple molecules that are serially activated

and bleached in the same DLR (25).

Although we have demonstrated isolation

and localization with both green Ephotoactivatablegreen fluorescent protein (PA-GFP) and Dronpa^

Fig. 2. Comparative summed-molecule TIRF (A) and PALM (B) images of the same region within a cryo-prepared thin section from a COS-7 cell expressing the lysosomal transmembrane protein CD63 taggedwith the PA-FP Kaede. The larger boxed region in (B), when viewed at higher magnification (C) revealssmaller associated membranes that may represent interacting lysosomes or late endosomes that are notresolvable by TIRF. In a region where the section is nearly orthogonal to the lysosomal membrane, the mosthighly localized molecules fall on a line of width È10 nm (inset). In an obliquely cut region [(D), from thesmaller boxed region in (B)], the distribution of CD63 within the membrane plane can be discerned.

Fig. 1. The principle behind PALM. A sparse subsetof PA-FP molecules that are attached to proteins ofinterest and then fixed within a cell are activated (Aand B) with a brief laser pulse at lact 0 405 mmand then imaged at lexc 0 561 mm until most arebleached (C). This process is repeated many times(C and D) until the population of inactivated,unbleached molecules is depleted. Summing themolecular images across all frames results in adiffraction-limited image (E and F). However, ifthe location of each molecule is first determinedby fitting the expected molecular image given bythe PSF of the microscope [(G), center] to theactual molecular image [(G), left], the moleculecan be plotted [(G), right] as a Gaussian that hasa standard deviation equal to the uncertaintysx,y in the fitted position. Repeating with allmolecules across all frames (A¶ through D¶) andsumming the results yields a superresolutionimage (E¶ and F¶) in which resolution is dictatedby the uncertainties sx,y as well as by the densityof localized molecules. Scale: 1 � 1 mm in (F) and(F¶), 4 � 4 mm elsewhere.


REPORTS

Fig. 1. The principle behind PALM. A sparse subset of PA-FP molecules that are attached to proteins of interest and then fixed within a cell are activated (A and B) with a brief laser pulse at λact = 405 mm and then imaged at λexc = 561 mm until most are bleached (C). This process is repeated many times (C and D) until the popula-tion of inactivated, unbleached molecules is de-pleted. Summing the molecular images across all frames results in a diffraction-limited image (E and F). However, if the location of each mol-ecule is first determined by fitting the expected molecular image given by the PSF of the micro-scope [(G), center] to the actual molecular im-age [(G), left], the molecule can be plotted [(G), right] as a Gaussian that has a standard de-viation equal to the uncertainty σx,y in the fitted position. Repeating with all molecules across all frames (A′ through D′) and summing the results yields a superresolution image (E′ and F′) in which resolution is dictated by the uncer-tainties σx,y as well as by the density of localized molecules. Scale: 1 × 1 μm in (F) and (F′), 4 × 4 μm elsewhere.

Fig. 2. Comparative summed-molecule TIRF (A) and PALM (B) images of the same region within a cryo-prepared thin section from a COS-7 cell expressing the lysosomal transmembrane pro-tein CD63 tagged with the PA-FP Kaede. The larger boxed region in (B), when viewed at higher magnification (C) reveals smaller associated membranes that may represent interacting lysosomes or late endosomes that are not resolvable by TIRF. In a region where the section is nearly orthogonal to the lysosomal membrane, the most highly localized molecules fall on a line of width ∼10 nm (inset). In an obliquely cut region [(D), from the smaller boxed region in (B)], the distribution of CD63 within the membrane plane can be discerned.

1Howard Hughes Medical Institute, Janelia Farm Research Campus, Ashburn, VA 20147, USA. 2New Millennium Research, LLC, Okemos, MI 48864, USA. 3Cell Biology and Metabolism Branch, National Institute of Child Health and Human Development (NICHD), Bethesda, MD 20892, USA.4National High Magnetic Field Laboratory, Florida State University, Tallahassee, FL 32310, USA. 5NuQuest Research, LLC, La Jolla, CA 92037, USA.*These authors contributed equally to this work.†Corresponding author. E-mail: [email protected]

Originally published 15 September 2006 in SCIENCE




http://www.sciencemag.org/cgi/content/full/320/5881/1332/DC1


































inactive PA-FPs, at a duration and intensity chosen so that the overall density of activated PA-FPs was increased back to a higher, but still resolvable, level (Fig. 1, B and D). This process of photoactivation, measurement, and bleaching was then repeated (movie S1) for many cycles over ∼104 to >105 image frames (depending on the expression level and spatial distribution of the PA-FPs) until the population of inactivated, unbleached molecules was depleted. At typical frame rates of ∼0.5 to 1.0 s, between 2 and 12 hours were required to acquire a complete image stack that could be distilled to a single superresolution image containing ∼105 to >106 localized molecules. We continued to explore methods (such as brighter molecules, higher excitation power, and higher activation density) to speed this process.

When the xy frames from any such image stack are summed across time t, the molecu-lar signals overlap to produce a diffraction-limited image (Fig. 1, E and F) similar to that obtained by conventional TIRF, in which all molecules emit simultaneously (fig. S3). How-ever, when the data are plotted in a multi-dimensional volume xyt (Fig. 1, center), the signal from each molecule m is uniquely iso-lated and can be summed at each pixel and across all of the frames in which it appears. This result (Fig. 1G, left) is then fitted using a robust nonlinear least squares algorithm to an assumed Gaussian PSF of free center coordinates x

o, y

o (Fig. 1G, center) (25), yield-

ing coordinates xm, y

m for the location of the

molecule, with a position uncertainty (σx,y

)m.

Finally, each molecule is rendered in a new xy frame as a Gaussian of standard deviation (σ

x,y)

m (rather than the much larger standard

deviation s of the original PSF), centered at x

m, y

m (Fig. 1, G, right, and A′ to D′) and nor-

malized to unit strength when integrated over all xy space. Thus, the superresolution image obtained by summing the rendered Gauss-ians associated with all localized molecules in the original image stack (Fig. 1, E′ and F′) provides a probability density map where brightness is proportional to the likelihood that a PA-FP molecule can be found at a given location.

This technique, termed photoactivated localization microscopy (PALM), is capable of resolving the most precisely localized mol-ecules at separations of a few nanometers. These represent the very brightest emitters (the much larger population of all isolated molecules exhibits a much broader range of photon counts; fig. S4). Thus, when render-ing PALM images, a fundamental trade-off exists: Including fewer, but brighter, mol-ecules results in higher localization and crisper images, but at a reduced molecular density giving less complete information about the spatial distribution of the target protein (fig. S5). Both parameters—local-ization precision and the density of ren-dered molecules—are key to defining per-formance in PALM. Their specific values

for the images in Figs. 2 to 4 are given in table S1.

This performance is largely dictated by the photophysical characteristics of the PA-FPs. Longer photobleaching half-life leads to more photons per molecule, but for a given excitation intensity, it also requires longer data acquisition times between activation pulses to maintain an appropriate density of individually resolvable molecules. Higher excitation cross-sectional σ and fluores-cence quantum efficiency Φ can speed this process of signal extraction and bleaching, with the added benefit of increasing the mo-lecular contrast relative to the autofluores-cence background. Also vital is the contrast C(λ

exc) = (σΦ)

act/(σΦ)

inact between the PA-FP in

its activated and inactivated state at λexc

, be-cause this dictates the maximum molecular density beyond which the background from many weakly emitting inactivated molecules in a DLR dominates the signal from a single activated one. PA-FPs that remain activated until bleached ensure that all possible pho-tons are extracted. Finally, PA-FPs less prone to blinking are desirable, given that it can be difficult to distinguish a single blinking molecule from multiple molecules that are serially activated and bleached in the same DLR (25).

Although we have demonstrated isola-tion and localization with both green [photoactivatable green fluorescent protein (PA-GFP) and Dronpa] and yellow [Kaede,

Kikume Green-Red (KikGR), and Eos Fluo-rescent Protein (EosFP)] excitable PA-FPs, for imaging cellular structures we focused on tetrameric Kaede and the various oligo-mers of EosFP—the former for its somewhat higher brightness and the latter for their less perturbative effect on cellular struc-ture and function. Each also exhibits high contrast relative to both the inactivated state and autofluorescence background at λ

exc = 561 nm.

We used PALM imaging to view intracel-lular structures in thin cryosections (25), akin to those used in transmission electron microscopy (TEM) but imaged under ambi-ent conditions (Figs. 2 and 3). In Fig. 2, lyso-somes in a COS-7 cell are visualized through expression of the lysosomal transmembrane protein CD63 fused to Kaede. Localization to the lysosome membrane was confirmed by comparative immunofluorescence label-ing in similarly prepared samples (fig. S6). A TIRF image shows the outlines of the limit-ing membrane (Fig. 2A) but only hints at the intricate structure that is resolved by PALM, such as smaller associated membranes that may represent interacting lysosomes or late endosomes (Fig. 2, B and C). Indeed, in re-gions where the section plane is nearly or-thogonal to the membrane, the most highly localized molecules fall on a line with an

apparent width of ∼10 nm (inset, Fig. 2C), demonstrating that they are indeed fixed and that sample drift has been successfully mitigated (25). In other regions of the cryo-section where the cut is more oblique to the lysosome, a wider, yet still sharply defined, swath of membrane is projected onto the image plane (Fig. 2D), permitting detailed investigation of the distribution of CD63 within the membrane plane.

In Fig. 3, PALM images of dEosFP-tagged cytochrome-C oxidase import sequence lo-calized within the matrix of mitochondria in a COS-7 cell are compared with TEM images of the same mitochondria. The high degree of correlation between the two data sets validates the PALM imaging principle, and the sharpness of the mitochondrial edges (Fig. 3H) as viewed by PALM is far closer to that seen by TEM than that observed by diffraction-limited TIRF (Fig. 3A). Such com-parative PALM/TEM imaging permits the nanometer-scale distribution of a specified protein to be determined in relation to the rest of the cellular ultrastructure at much higher molecular density than in immunola-beled TEM—more than 5500 molecules are localized in Fig. 3E, compared with the 20 or so particles typical in immunogold label-ing of the mitochondrial matrix. Superposi-tion of the PALM and TEM images (Fig. 3, D

and G) also reveals that the matrix reporter molecules extend up to, but not into, the ∼20-nm outer mitochondrial membrane, under-scoring the resolution capability of the tech-nique. Correlated PALM/TEM does not have the added preparation steps and specificity issues associated with exogenous labels for combined fluorescence/EM such as fluores-cein or resorufin arsenical helix binder (27). Finally, efforts are underway to establish dual-labeled PALM or PALM fluorescence resonance energy transfer, which would per-mit the relative distribution or regions of interaction between multiple proteins to be discerned at the nanometer level.

Thin sections are advantageous for PALM because they exhibit less autofluorescence than bulk samples, ensure that the PA-FPs are immobile, and permit the study of in-tracellular organelles that are inaccessible under TIRF excitation. However, demon-stration of PALM on fixed cultured cells in phosphate-buffered saline (Fig. 4) is also notable both as a means to study proteins at or near the plasma membrane under minimally invasive conditions and as a pre-cursor to eventual three-dimensional (3D) PALM imaging.

Confirmation that the nanometer-level resolution of PALM is retained under such conditions is given by the comparison of TIRF (Fig. 4A) and PALM (Fig. 4B) images of dEos-fused vinculin at focal adhesion re-gions (fig. S7) of a fox lung fibroblast (FoLu) cell to a cover slip. PALM reveals the hetero-geneity within a selected attachment (box, Fig. 4A) and, in one subregion, suggests the partial assembly of a vinculin network (ar-rows, Fig. 4B). Similarly, a TIRF image (Fig. 4C) of tandem-dimer EosFP-fused actin in a cultured FoLu cell (fig. S8) shows both large cytoskeletal stress fibers and a lamellipo-dium, whereas PALM within the latter (Fig. 4D) reveals an increased concentration of actin at the leading edge. Under even high-er magnification (inset, Fig. 4D), numerous short filaments are observed. These may be independent structures fixed in the process of assembly, or they may be part of a larger, continuous 3D network only partially re-vealed by the short extent of the evanescent excitation field.

In whole cells, PALM with TIRF excitation is well suited to studies of proteins bound to the plasma membrane, such as the dEosFP-fused Gag protein of human immunodefi-ciency virus 1 imaged by TIRF and PALM in Fig. 4, E and F, respectively. Gag, a retroviral protein that mediates the assembly of virus-like particles (VLPs), is revealed by PALM in various stages of organization: voids (arrows marked V), one high-density region (arrow R), and several tight clusters probably indica-tive of budding VLPs (arrows marked P, and magnified inset of Fig. 4F).

In the future, PALM should benefit from improvements in and additions to the

and yellow EKaede, Kikume Green-Red (KikGR),and Eos Fluorescent Protein (EosFP)^ excitablePA-FPs, for imaging cellular structures we

focused on tetrameric Kaede and the various

oligomers of EosFP—the former for its some-

what higher brightness and the latter for their

less perturbative effect on cellular structure and

function. Each also exhibits high contrast rela-

tive to both the inactivated state and autoflu-

orescence background at lexc

0 561 nm.

We used PALM imaging to view intra-

cellular structures in thin cryosections (25), akin

to those used in transmission electron microsco-

py (TEM) but imaged under ambient conditions

(Figs. 2 and 3). In Fig. 2, lysosomes in a COS-7

cell are visualized through expression of the

lysosomal transmembrane protein CD63 fused

to Kaede. Localization to the lysosome mem-

brane was confirmed by comparative immuno-

fluorescence labeling in similarly prepared

samples (fig. S6). A TIRF image shows the

outlines of the limiting membrane (Fig. 2A) but

only hints at the intricate structure that is re-

solved by PALM, such as smaller associated

membranes that may represent interacting lyso-

somes or late endosomes (Fig. 2, B and C).

Indeed, in regions where the section plane is

nearly orthogonal to the membrane, the most

highly localized molecules fall on a line with an

apparent width of È10 nm (inset, Fig. 2C),

demonstrating that they are indeed fixed and

that sample drift has been successfully miti-

gated (25). In other regions of the cryosection

where the cut is more oblique to the lysosome,

a wider, yet still sharply defined, swath of mem-

brane is projected onto the image plane (Fig.

2D), permitting detailed investigation of the

distribution of CD63 within the membrane

plane.

In Fig. 3, PALM images of dEosFP-tagged

cytochrome-C oxidase import sequence local-

ized within the matrix of mitochondria in a

COS-7 cell are compared with TEM images of

the same mitochondria. The high degree of

correlation between the two data sets validates

the PALM imaging principle, and the sharpness

of the mitochondrial edges (Fig. 3H) as viewed

by PALM is far closer to that seen by TEM

than that observed by diffraction-limited TIRF

(Fig. 3A). Such comparative PALM/TEM im-

aging permits the nanometer-scale distribution

of a specified protein to be determined in

relation to the rest of the cellular ultrastructure

at much higher molecular density than in im-

munolabeled TEM—more than 5500 molecules

are localized in Fig. 3E, compared with the 20

or so particles typical in immunogold labeling

of the mitochondrial matrix. Superposition of

the PALM and TEM images (Fig. 3, D and G)

also reveals that the matrix reporter molecules

extend up to, but not into, the È20-nm outer

mitochondrial membrane, underscoring the

resolution capability of the technique. Corre-

lated PALM/TEM does not have the added

preparation steps and specificity issues asso-

ciated with exogenous labels for combined

fluorescence/EM such as fluorescein or resoru-

fin arsenical helix binder (27). Finally, efforts

are underway to establish dual-labeled PALM

or PALM fluorescence resonance energy trans-

fer, which would permit the relative distribution

or regions of interaction between multiple

proteins to be discerned at the nanometer level.

Thin sections are advantageous for PALM be-

cause they exhibit less autofluorescence than bulk

samples, ensure that the PA-FPs are immobile,

and permit the study of intracellular organelles

that are inaccessible under TIRF excitation. How-

ever, demonstration of PALM on fixed cultured

cells in phosphate-buffered saline (Fig. 4) is also

notable both as a means to study proteins at or

near the plasma membrane under minimally

invasive conditions and as a precursor to eventual

three-dimensional (3D) PALM imaging.

Confirmation that the nanometer-level resolu-

tion of PALM is retained under such conditions is

given by the comparison of TIRF (Fig. 4A) and

PALM (Fig. 4B) images of dEos-fused vinculin at

focal adhesion regions (fig. S7) of a fox lung

fibroblast (FoLu) cell to a cover slip. PALM

reveals the heterogeneity within a selected attach-

ment (box, Fig. 4A) and, in one subregion, sug-

gests the partial assembly of a vinculin network

(arrows, Fig. 4B). Similarly, a TIRF image (Fig.

4C) of tandem-dimer EosFP-fused actin in a

cultured FoLu cell (fig. S8) shows both large

cytoskeletal stress fibers and a lamellipodium,

whereas PALM within the latter (Fig. 4D) reveals

an increased concentration of actin at the leading

Fig. 3. Comparative summed-molecule TIRF (A), PALM (B), TEM (C), andPALM/TEM overlay (D) images of mitochondria in a cryo-prepared thinsection from a COS-7 cell expressing dEosFP-tagged cytochrome-C ox-idase import sequence. Higher magnification PALM (E), TEM (F), andoverlay (G) images within the box in (B) reveal that these matrix re-

porter molecules extend up to, but not into, the È20-nm outer mito-chondrial membrane. The molecular distribution across two mitochon-dria along lines 1 and 2 in PALM image (E) are compared in (H) to theTEM signal along lines 3 and 4 in (F) across the same mitochondria.Scale bars: 1.0 mm in (A) to (D); 0.2 mm in (E) to (G).

15 SEPTEMBER 2006 VOL 313 SCIENCE www.sciencemag.org1644

REPORTS

Fig. 3. Comparative summed-molecule TIRF (A), PALM (B), TEM (C), and PALM/TEM overlay (D) images of mitochondria in a cryo-prepared thin section from a COS-7 cell expressing dEosFP-tagged cytochrome-C oxidase import sequence. Higher magnification PALM (E), TEM (F), and overlay (G) images within the box in (B) reveal that these matrix reporter molecules extend up to, but not into, the ∼20-nm outer mitochondrial membrane. The molecular distribution across two mitochondria along lines 1 and 2 in PALM image (E) are compared in (H) to the TEM signal along lines 3 and 4 in (F) across the same mitochondria. Scale bars: 1.0 μm in (A) to (D); 0.2 μm in (E) to (G).

edge. Under even higher magnification (inset, Fig.

4D), numerous short filaments are observed.

These may be independent structures fixed in the

process of assembly, or they may be part of a

larger, continuous 3D network only partially

revealed by the short extent of the evanescent

excitation field.

In whole cells, PALM with TIRF excitation is

well suited to studies of proteins bound to the

plasma membrane, such as the dEosFP-fused Gag

protein of human immunodeficiency virus 1

imaged by TIRF and PALM in Fig. 4, E and F,

respectively. Gag, a retroviral protein that medi-

ates the assembly of virus-like particles (VLPs), is

revealed by PALM in various stages of organi-

zation: voids (arrows marked V), one high-density

region (arrow R), and several tight clusters

probably indicative of budding VLPs (arrows

marked P, and magnified inset of Fig. 4F).

In the future, PALM should benefit from im-

provements in and additions to the palette of

available PA-FPs, as well as from the discovery of

means to modify the PA-FP environment to en-

hance photostability (13) and suppress blinking.

Recently, we demonstrated photoactivation in

PALM through ultraviolet-induced uncaging

(28) of fluorophores (fig. S10) which, when

combined with immunolabeling or other devel-

oping methods to achieve high-specificity intra-

cellular protein labeling (27, 29), might offer a

different avenue to improved localization pre-

cision and faster frame rates, given that a broad

spectrum of high-brightness caged fluorophores

is potentially available.

Algorithmically, additional well-localized mol-

ecules might be mined from the data if better

means are found to unambiguously collate the

multiple photon bursts from blinking molecules.

Possible improvements to the fitting algorithm to

achieve higher localization accuracy should also

be explored. Although most of the observed mol-

ecules are well represented by a circularly sym-

metric Gaussian PSF, possible systematic position

errors due to chromophore orientation, pixel non-

uniformity, and chromatic aberration deserve

closer attention. Perhaps most importantly, position

error due to background nonuniformity within the

molecular fitting window needs to be addressed,

particularly when the number of inactivated mole-

cules contributing to this background is high.

Experimentally, multiple angles and polariza-

tions of TIRF excitation may eventually permit the

precise determination of the xyz position and

dipole orientation for fixed PA-FP molecules

within the evanescent field. Standing wave TIRF

could provide an excitation PSF of widthÈ lexc/6,

improving localization precision for a given

photon count. Bulk cellular autofluorescence

complicates the extension of PALM to 3D, but

the improved single-molecule sensitivity predicted

for a proposed optical lattice microscope (30) may

help. However, the most promising path to 3D

may involve cryogenic PALM of vitrified cells,

due to the narrow molecular line widths, large

cross-sections, and improved stability expected (8).

On the other hand, the ambient, TIRF-based

PALM system demonstrated here has the ad-

vantage of simplicity, requiring only a TIRF-

capable microscope with appropriate lasers, filters,

and EMCCD camera, as well as basic acquisition,

localization, and image rendering software. As

such, it could be widely adopted in short order for

the near-molecular resolution imaging of specified

proteins for in vitro preparations and fixed cells.

References and Notes1. J. Lippincott-Schwartz, G. H. Patterson, Science 300, 87

(2003).2. E. Betzig, J. K. Trautman, Science 257, 189 (1992).3. K. I. Willig, S. O. Rizzoli, V. Westphal, R. Jahn, S. W. Hell,

Nature 440, 935 (2006).4. M. G. L. Gustafsson, J. Microsc. 198, 82 (2000).5. M. G. L. Gustafsson, Proc. Natl. Acad. Sci. U.S.A. 102,

13081 (2005).6. M. Hofmann, C. Eggeling, S. Jakobs, S. W. Hell, Proc.

Natl. Acad. Sci. U.S.A. 102, 17565 (2005).7. E. Betzig, R. J. Chichester, Science 262, 1422 (1993).8. W. E. Moerner, J. Phys. Chem. B 106, 910 (2002).9. H. F. Hess, E. Betzig, T. D. Harris, L. N. Pfeiffer,

K. W. West, Science 264, 1740 (1994).10. E. Betzig, Opt. Lett. 20, 237 (1995).11. M. K. Cheezum, W. F. Walker, W. H. Guilford, Biophys. J.

81, 2378 (2001).12. R. E. Thompson, D. R. Larson, W. W. Webb, Biophys. J. 82,

2775 (2002).13. A. Yildiz et al., Science 300, 2061 (2003); published

online 5 June 2003 (10.1126/science.1084398).14. X. Qu, D. Wu, L. Mets, N. F. Scherer, Proc. Natl. Acad. Sci.

U.S.A. 101, 11298 (2004).15. L. S. Churchman, Z. Okten, R. S. Rock, J. F. Dawson,

J. A. Spudich, Proc. Natl. Acad. Sci. U.S.A. 102, 1419 (2005).16. A. M. van Oijen, J. Kohler, J. Schmidt, M. Muller,

G. J. Brackenhoff, J. Opt. Soc. Am. A 16, 909 (1999).17. M. P. Gordon, T. Ha, P. R. Selvin, Proc. Natl. Acad. Sci.

U.S.A. 101, 6462 (2004).18. K. A. Lidke, B. Rieger, T. M. Jovin, R. Heintzmann, Opt.

Express 13, 7052 (2005).19. G. H. Patterson, J. Lippincott-Schwartz, Science 297,

1873 (2002).20. R. Ando, H. Hama, M. Yamamoto-Hino, H. Mizuno,

A. Miyawaki, Proc. Natl. Acad. Sci. U.S.A. 99, 12651 (2002).21. J. Wiedenmann et al., Proc. Natl. Acad. Sci. U.S.A. 101,

15905 (2004).22. R. Ando, H. Mizuno, A. Miyawaki, Science 306, 1370 (2004).23. H. Tsutsui, S. Karasawa, H. Shimizu, N. Nukina, A. Miyawaki,

EMBO Rep. 6, 233 (2005).24. K. A. Lukyanov, D. M. Chudakov, S. Lukyanov, V. V.

Verkhusha, Nat. Rev. Mol. Cell Biol. 6, 885 (2005).25. Materials and methods are available as supporting

material on Science Online.26. D. Axelrod, Methods Cell Biol. 30, 245 (1989).27. G. Gaietta et al., Science 296, 503 (2002).28. J. C. Politz, Trends Cell Biol. 9, 284 (1999).29. I. Chen, A. Y. Ting, Curr. Opin. Biotechnol. 16, 35 (2005).30. E. Betzig, Opt. Express 13, 3021 (2005).31. This project was supported in part by the Intramural

Program of NICHD and the NIH Intramural AIDS TargetedAntiviral Program (IATAP). We thank J. Wiedenmann andU. Nienhaus of the University of Ulm for the gift of acDNA encoding the EosFP.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/1127344/DC1Materials and MethodsFigs. S1 to S10Table S1Movie S1References

13 March 2006; accepted 2 August 2006Published online 10 August 2006;10.1126/science.1127344Include this information when citing this paper.

Fig. 4. Examples of PALM imaging near the surfaces of whole, fixed cells. (A) A summed-molecule TIRFimage of focal adhesions for a FoLu cell expressing dEos-tagged vinculin. (B) A magnified PALM view ofthe structure within a single adhesion over the region indicated by the box in (A), including apparentassembly of vinculin in a partial network (arrows). (C) A summed-molecule TIRF image near the peripheryof a FoLu cell expressing tdEos-tagged actin. (D) A magnified PALM view of the actin distribution withinthe portion of the lamellipodium outlined by the box in (C). Inset, a further magnified view near theleading edge over the region indicated by the smaller box. (E and F) Summed-molecule TIRF and PALMimages, respectively, of a COS-7 cell expressing the retroviral protein Gag tagged with dEos. The PALMimage highlights voids (arrows labeled V), a higher density region (arrow R), and probable condensationat several points (arrows labeled P) into VLPs of È100- to 150-nm size (inset).


REPORTS

Fig. 4. Examples of PALM imaging near the surfaces of whole, fixed cells. (A) A summed-molecule TIRF image of focal adhesions for a FoLu cell expressing dEos-tagged vinculin. (B) A magnified PALM view of the structure within a single adhesion over the region indicated by the box in (A), including apparent assembly of vinculin in a partial network (arrows). (C) A summed-molecule TIRF image near the periphery of a FoLu cell expressing tdEos-tagged actin. (D) A magnified PALM view of the actin distribution within the portion of the lamellipodium outlined by the box in (C). Inset, a further magnified view near the leading edge over the region indicated by the smaller box. (E and F) Summed-molecule TIRF and PALM images, respectively, of a COS-7 cell expressing the retroviral protein Gag tagged with dEos. The PALM image highlights voids (arrows labeled V), a higher density region (arrow R), and probable condensation at several points (arrows labeled P) into VLPs of ∼100- to 150-nm size (inset).










































palette of available PA-FPs, as well as from the discovery of means to modify the PA-FP environment to enhance photostability (13) and suppress blinking. Recently, we demon-strated photoactivation in PALM through ultraviolet-induced uncaging (28) of fluoro-phores (fig. S10) which, when combined with immunolabeling or other developing meth-ods to achieve high-specificity intracellular protein labeling (27, 29), might offer a differ-ent avenue to improved localization precision and faster frame rates, given that a broad spectrum of high-brightness caged fluoro-phores is potentially available.

Algorithmically, additional well-localized molecules might be mined from the data if better means are found to unambiguously col-late the multiple photon bursts from blinking molecules. Possible improvements to the fit-ting algorithm to achieve higher localization accuracy should also be explored. Although most of the observed molecules are well rep-resented by a circularly symmetric Gaussian PSF, possible systematic position errors due to chromophore orientation, pixel nonunifor-mity, and chromatic aberration deserve clos-er attention. Perhaps most importantly, posi-tion error due to background nonuniformity within the molecular fitting window needs to be addressed, particularly when the numb er of inactivated molecules contributing to this background is high.

Experimentally, multiple angles and po-larizations of TIRF excitation may eventu-ally permit the precise determination of the xyz position and dipole orientation for fixed PA-FP molecules within the evanescent field. Standing wave TIRF could provide an excita-tion PSF of width ∼λ

exc/6, improving localiza-

tion precision for a given photon count. Bulk

cellular autofluorescence complicates the extension of PALM to 3D, but the improved single-molecule sensitivity predicted for a proposed optical lattice microscope (30) may help. However, the most promising path to 3D may involve cryogenic PALM of vitrified cells, due to the narrow molecular line widths, large cross-sections, and improved stability expected (8). On the other hand, the ambi-ent, TIRF-based PALM system demonstrated here has the advantage of simplicity, requir-ing only a TIRF-capable microscope with ap-propriate lasers, filters, and EMCCD camera, as well as basic acquisition, localization, and image rendering software. As such, it could be widely adopted in short order for the near-molecular resolution imaging of speci-fied proteins for in vitro preparations and fixed cells.

REFERENCES AND NOTES 1. J. Lippincott-Schwartz, G. H. Patterson, Science 300, 87 (2003). 2. E. Betzig, J. K. Trautman, Science 257, 189 (1992). 3. K. I. Willig, S. O. Rizzoli, V. Westphal, R. Jahn, S. W. Hell, Nature 440, 935 (2006). 4. M. G. L. Gustafsson, J. Microsc. 198, 82 (2000). 5. M. G. L. Gustafsson, Proc. Natl. Acad. Sci. U.S.A. 102, 13081 (2005). 6. M. Hofmann, C. Eggeling, S. Jakobs, S. W. Hell, Proc. Natl. Acad. Sci. U.S.A. 102, 17565 (2005). 7. E. Betzig, R. J. Chichester, Science 262, 1422 (1993). 8. W. E. Moerner, J. Phys. Chem. B 106, 910 (2002). 9. H. F. Hess, E. Betzig, T. D. Harris, L. N. Pfeiffer, K. W. West, Science 264, 1740 (1994).10. E. Betzig, Opt. Lett. 20, 237 (1995).11. M. K. Cheezum, W. F. Walker, W. H. Guilford, Biophys. J. 81, 2378 (2001).12. R. E. Thompson, D. R. Larson, W. W. Webb, Biophys. J. 82, 2775 (2002).13. A. Yildiz et al., Science 300, 2061 (2003); published online 5 June 2003 (10.1126/science.1084398).14. X. Qu, D. Wu, L. Mets, N. F. Scherer, Proc. Natl. Acad. Sci. U.S.A. 101, 11298 (2004).

15. L. S. Churchman, Z. Ökten, R. S. Rock, J. F. Dawson, J. A. Spudich, Proc. Natl. Acad. Sci. U.S.A. 102, 1419 (2005).16. A. M. van Oijen, J. KÖ hler, J. Schmidt, M. MÜ ller, G. J. Brackenhoff, J. Opt. Soc. Am. A 16, 909 (1999).17. M. P. Gordon, T. Ha, P. R. Selvin, Proc. Natl. Acad. Sci. U.S.A. 101, 6462 (2004).18. K. A. Lidke, B. Rieger, T. M. Jovin, R. Heintzmann, Opt. Express 13, 7052 (2005).19. G. H. Patterson, J. Lippincott-Schwartz, Science 297, 1873 (2002).20. R. Ando, H. Hama, M. Yamamoto-Hino, H. Mizuno, A. Miyawaki, Proc. Natl. Acad. Sci. U.S.A. 99, 12651 (2002).21. J. Wiedenmann et al., Proc. Natl. Acad. Sci. U.S.A. 101, 15905 (2004).22. R. Ando, H. Mizuno, A. Miyawaki, Science 306, 1370 (2004).23. H. Tsutsui, S. Karasawa, H. Shimizu, N. Nukina, A. Miyawaki, EMBO Rep. 6, 233 (2005).24. K. A. Lukyanov, D. M. Chudakov, S. Lukyanov, V. V. Verkhusha, Nat. Rev. Mol. Cell Biol. 6, 885 (2005).25. Materials and methods are available as supporting material on Science Online.26. D. Axelrod, Methods Cell Biol. 30, 245 (1989).27. G. Gaietta et al., Science 296, 503 (2002).28. J. C. Politz, Trends Cell Biol. 9, 284 (1999).29. I. Chen, A. Y. Ting, Curr. Opin. Biotechnol. 16, 35 (2005).30. E. Betzig, Opt. Express 13, 3021 (2005).

ACKNOWLEDGEMENTSThis project was supported in part by the IntramuralProgram of NICHD and the NIH Intramural AIDS Targeted Antiviral Program (IATAP). We thank J. Wiedenmann andU. Nienhaus of the University of Ulm for the gift of a cDNA encoding the EosFP.

SUPPORTING ONLINE MATERIALwww.sciencemag.org/cgi/content/full/1127344/DC1Materials and MethodsFigs. S1 to S10Table S1Movie S1References

13 March 2006; accepted 2 August 2006Published online 10 August 2006;10.1126/science.1127344Include this information when citing this paper.

INTRODUCTION: In vivo imaging pro-vides a window into the spatially complex, rapidly evolving physiology of the cell that structural imaging alone cannot. However, observing this physiology directly involves inevitable tradeoffs of spatial resolution, temporal resolution, and phototoxicity. This is especially true when imaging in three di-mensions, which is essential to obtain a com-plete picture of many dynamic subcellular processes. Although traditional in vivo imag-ing tools, such as widefield and confocal mi-croscopy, and newer ones, such as light-sheet microscopy, can image in three dimensions, they sacrifice substantial spatiotemporal reso-lution to do so and, even then, can often be used for only very limited durations before al-tering the physiological state of the specimen.

RATIONALE: To address these limitations, we developed a new microscope using ul-trathin light sheets derived from twodimen-sional (2D) optical lattices. These are scanned

Bi-Chang Chen, Wesley R. Legant, Kai Wang, Lin Shao, Daniel E. Milkie, Michael W. Davidson, Chris Janetopoulos, Xufeng S. Wu, John A. Hammer III, Zhe Liu, Brian P. English, Yuko Mimori-Kiyosue, Daniel P. Romero, Alex T. Ritter, Jennifer Lippincott-Schwartz, Lillian Fritz-Laylin, R. Dyche Mullins, Diana M. Mitchell, Joshua N. Bembenek, Anne-Cecile Reymann, Ralph Böhme, Stephan W. Grill, Jennifer T. Wang, Geraldine Seydoux, U. Serdar Tulu, Daniel P. Kiehart, Eric Betzig*

Lattice light-sheet microscopy: Imaging molecules to embryos at high spatiotemporal resolution

plane-by-plane through the specimen to gen-erate a 3D image. The thinness of the sheet leads to high axial resolution and negligible photobleaching and background outside of the focal plane, while its simultaneous illumi-nation of the entire field of view permits im-aging at hundreds of planes per second even at extremely low peak excitation intensities. By implementing either superresolution struc-tured illumination or by dithering the lattice to create a uniform light sheet, we imaged cells and small embryos in three dimensions, often at subsecond intervals, for hundreds to thousands of time points at the diffraction limit and beyond.

RESULTS: We demonstrated the technique on 20 different biological processes spanning four orders of magnitude in space and time, including the binding kinetics of single Sox2 transcription factor molecules, 3D superreso-lution photoactivated localization microscopy of nuclear lamins, dynamic organelle rear-

rangements and 3D tracking of microtubule plus ends during mitosis, neutrophil motility in a collagen mesh, and subcellular protein localization and dynamics during embryogen-esis in Caenorhabditis elegans and Drosophila melanogaster. Throughout, we established the performance advantages of lattice light-sheet microscopy compared with previous techniques and highlighted phenomena that, when seen at increased spatiotemporal detail, may hint at previously unknown biological mechanisms.

CONCLUSION: Photobleaching and pho-totoxicity are typically reduced by one to two orders of magnitude relative to that seen with a 1D scanned Bessel beam or the point array scanned excitation of spinning disk confocal microscopy. This suggests that the instanta-neous peak power delivered to the specimen

may be an even more important metric of cell health than the total photon dose and should enable extended 3D ob-servation of endogenous levels of even sparsely

expressed proteins produced by genome ed-iting. Improvements of similar magnitude in imaging speed and a twofold gain in axial res-olution relative to confocal microscopy yield 4D spatiotemporal resolution high enough to follow fast, nanoscale dynamic processes that would otherwise be obscured by poor resolu-tion along one or more axes of spacetime. Last, the negligible background makes lattice light-sheet microscopy a promising platform for the extension of all methods of superresolution to

larger and more densely fluores-cent specimens and enables the study of signaling, transport, and stochastic self-assembly in com-plex environments with single-molecule sensitivity.

The list of author affiliations is available in the full article online. *Corresponding author. E-mail: [email protected] Cite this article as B.-C. Chen et al., Sci-ence 346, 1257998 (2014). DOI: 10.1126/ science.1257998

24 OCTOBER 2014 • VOL 346 ISSUE 6208 439SCIENCE sciencemag.org

INTRODUCTION: In vivo imaging provides

a window into the spatially complex, rap-

idly evolving physiology of the cell that

structural imaging alone cannot. However,

observing this physiology directly involves

inevitable tradeoffs of spatial resolution,

temporal resolution, and phototoxicity. This

is especially true when imaging in three

dimensions, which is essential to obtain a

complete picture of many dynamic subcel-

lular processes. Although traditional in vivo

imaging tools, such as widefield and con-

focal microscopy, and newer ones, such as

light-sheet microscopy, can image in three

dimensions, they sacrifice substantial spa-

tiotemporal resolution to do so and, even


ADVANCED IMAGING

Bi-Chang Chen, Wesley R. Legant, Kai Wang, Lin Shao, Daniel E. Milkie,

Michael W. Davidson, Chris Janetopoulos, Xufeng S. Wu, John A. Hammer III,

Zhe Liu, Brian P. English, Yuko Mimori-Kiyosue, Daniel P. Romero, Alex T. Ritter,

Jennifer Lippincott-Schwartz, Lillian Fritz-Laylin, R. Dyche Mullins, Diana M. Mitchell,

Joshua N. Bembenek, Anne-Cecile Reymann, Ralph Böhme, Stephan W. Grill,

Jennifer T. Wang, Geraldine Seydoux, U. Serdar Tulu, Daniel P. Kiehart, Eric Betzig*

RESEARCH ARTICLE SUMMARY

Lattice light-sheet microscopy. An ultrathin structured light sheet (blue-green, center)

excites f uorescence (orange) in successive planes as it sweeps through a specimen (gray) to

generate a 3D image. The speed, noninvasiveness, and high spatial resolution of this approach

make it a promising tool for in vivo 3D imaging of fast dynamic processes in cells and embryos,

as shown here in f ve surrounding examples.

then, can often be used for only very limited

durations before altering the physiological

state of the specimen.

RATIONALE: To address these limitations,

we developed a new microscope using

ultrathin light sheets derived from two-

dimensional (2D) optical lattices. These are

scanned plane-by-plane through the speci-

men to generate a 3D image. The thinness of

the sheet leads to high axial resolution and

negligible photobleaching and background

outside of the focal plane, while its simulta-

neous illumination of the entire field of view

permits imaging at hundreds of planes per

second even at extremely low peak excitation

intensities. By implementing either superres-

olution structured illumination or by dither-

ing the lattice to create a uniform light sheet,

we imaged cells and small embryos in three

dimensions, often at subsecond intervals, for

hundreds to thousands of time points at the

diffraction limit and beyond.

RESULTS: We demonstrated the technique

on 20 different biological processes span-

ning four orders of magnitude in space

and time, including the binding kinetics of

single Sox2 transcription factor molecules,

3D superresolution photoactivated localiza-

tion microscopy of nuclear lamins, dynamic

organelle rearrangements and 3D tracking

of microtubule plus ends during mitosis,

neutrophil motility in a collagen mesh, and

subcellular protein localization and dynam-

ics during embryogenesis in Caenorhabditis

elegans and Drosophila

melanogaster. Through-

out, we established the

performance advantages

of lattice light-sheet mi-

croscopy compared with

previous techniques and

highlighted phenomena that, when seen at

increased spatiotemporal detail, may hint at

previously unknown biological mechanisms.

CONCLUSION: Photobleaching and photo-

toxicity are typically reduced by one to two

orders of magnitude relative to that seen

with a 1D scanned Bessel beam or the point

array scanned excitation of spinning disk

confocal microscopy. This suggests that the

instantaneous peak power delivered to the

specimen may be an even more important

metric of cell health than the total pho-

ton dose and should enable extended 3D

observation of endogenous levels of even

sparsely expressed proteins produced by

genome editing. Improvements of similar

magnitude in imaging speed and a twofold

gain in axial resolution relative to confocal

microscopy yield 4D spatiotemporal resolu-

tion high enough to follow fast, nanoscale

dynamic processes that would otherwise be

obscured by poor resolution along one or

more axes of spacetime. Last, the negligible

background makes lattice light-sheet mi-

croscopy a promising platform for the ex-

tension of all methods of superresolution

to larger and more densely fluorescent

specimens and enables the study of signal-

ing, transport, and stochastic self-assembly

in complex environments with single-

molecule sensitivity. ■

The list of author affiliations is available in the full article online.

*Corresponding author. E-mail: [email protected] this article as B.-C. Chen et al., Science 346, 1257998 (2014). DOI: 10.1126/science.1257998

Read the full article at http://dx.doi.org/10.1126/science.1257998

ON OUR WEB SITE

Published by AAAS

Corrected 30 January 2015; see full text.

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

Lattice light-sheet microscopy. An ultrathin structured light sheet (blue-green, center) excites fluorescence (orange) in successive planes as it sweeps through a specimen (gray) to generate a 3D image. The speed, noninvasiveness, and high spatial resolution of this approach make it a promising tool for in vivo 3D imaging of fast dynamic processes in cells and embryos, as shown here in five surrounding examples.

24 OCTOBER 2014 • VOL 346 ISSUE 6208 439SCIENCE sciencemag.org

INTRODUCTION: In vivo imaging provides

a window into the spatially complex, rap-

idly evolving physiology of the cell that

structural imaging alone cannot. However,

observing this physiology directly involves

inevitable tradeoffs of spatial resolution,

temporal resolution, and phototoxicity. This

is especially true when imaging in three

dimensions, which is essential to obtain a

complete picture of many dynamic subcel-

lular processes. Although traditional in vivo

imaging tools, such as widefield and con-

focal microscopy, and newer ones, such as

light-sheet microscopy, can image in three

dimensions, they sacrifice substantial spa-

tiotemporal resolution to do so and, even


ADVANCED IMAGING

Bi-Chang Chen, Wesley R. Legant, Kai Wang, Lin Shao, Daniel E. Milkie,

Michael W. Davidson, Chris Janetopoulos, Xufeng S. Wu, John A. Hammer III,

Zhe Liu, Brian P. English, Yuko Mimori-Kiyosue, Daniel P. Romero, Alex T. Ritter,

Jennifer Lippincott-Schwartz, Lillian Fritz-Laylin, R. Dyche Mullins, Diana M. Mitchell,

Joshua N. Bembenek, Anne-Cecile Reymann, Ralph Böhme, Stephan W. Grill,

Jennifer T. Wang, Geraldine Seydoux, U. Serdar Tulu, Daniel P. Kiehart, Eric Betzig*

RESEARCH ARTICLE SUMMARY

Lattice light-sheet microscopy. An ultrathin structured light sheet (blue-green, center)

excites f uorescence (orange) in successive planes as it sweeps through a specimen (gray) to

generate a 3D image. The speed, noninvasiveness, and high spatial resolution of this approach

make it a promising tool for in vivo 3D imaging of fast dynamic processes in cells and embryos,

as shown here in f ve surrounding examples.

then, can often be used for only very limited

durations before altering the physiological

state of the specimen.

RATIONALE: To address these limitations,

we developed a new microscope using

ultrathin light sheets derived from two-

dimensional (2D) optical lattices. These are

scanned plane-by-plane through the speci-

men to generate a 3D image. The thinness of

the sheet leads to high axial resolution and

negligible photobleaching and background

outside of the focal plane, while its simulta-

neous illumination of the entire field of view

permits imaging at hundreds of planes per

second even at extremely low peak excitation

intensities. By implementing either superres-

olution structured illumination or by dither-

ing the lattice to create a uniform light sheet,

we imaged cells and small embryos in three

dimensions, often at subsecond intervals, for

hundreds to thousands of time points at the

diffraction limit and beyond.

RESULTS: We demonstrated the technique

on 20 different biological processes span-

ning four orders of magnitude in space

and time, including the binding kinetics of

single Sox2 transcription factor molecules,

3D superresolution photoactivated localiza-

tion microscopy of nuclear lamins, dynamic

organelle rearrangements and 3D tracking

of microtubule plus ends during mitosis,

neutrophil motility in a collagen mesh, and

subcellular protein localization and dynam-

ics during embryogenesis in Caenorhabditis

elegans and Drosophila

melanogaster. Through-

out, we established the

performance advantages

of lattice light-sheet mi-

croscopy compared with

previous techniques and

highlighted phenomena that, when seen at

increased spatiotemporal detail, may hint at

previously unknown biological mechanisms.

CONCLUSION: Photobleaching and photo-

toxicity are typically reduced by one to two

orders of magnitude relative to that seen

with a 1D scanned Bessel beam or the point

array scanned excitation of spinning disk

confocal microscopy. This suggests that the

instantaneous peak power delivered to the

specimen may be an even more important

metric of cell health than the total pho-

ton dose and should enable extended 3D

observation of endogenous levels of even

sparsely expressed proteins produced by

genome editing. Improvements of similar

magnitude in imaging speed and a twofold

gain in axial resolution relative to confocal

microscopy yield 4D spatiotemporal resolu-

tion high enough to follow fast, nanoscale

dynamic processes that would otherwise be

obscured by poor resolution along one or

more axes of spacetime. Last, the negligible

background makes lattice light-sheet mi-

croscopy a promising platform for the ex-

tension of all methods of superresolution

to larger and more densely fluorescent

specimens and enables the study of signal-

ing, transport, and stochastic self-assembly

in complex environments with single-

molecule sensitivity. ■

The list of author affiliations is available in the full article online.

*Corresponding author. E-mail: [email protected] this article as B.-C. Chen et al., Science 346, 1257998 (2014). DOI: 10.1126/science.1257998

Read the full article at http://dx.doi.org/10.1126/science.1257998

ON OUR WEB SITE

Published by AAAS

Corrected 30 January 2015; see full text.

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

o

n Fe

brua

ry 1

2, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Febr

uary

12,

201

5w

ww

.sci

ence

mag

.org

Dow

nloa

ded

from

Originally published 24 October 2014 in SCIENCE







http://www.sciencemag.org/cgi/content/full/1127344/DC1




Cells, the functional units of life, are best studied in vivo. This is particu-larly true for neurons, which perform their basic function of information processing by connecting with their

neighbors. Therefore, unraveling the inner workings of the brain requires the imaging of neurons in the living animal. Although confo-cal and multiphoton microscopy can visualize neurons tagged with fluorescent proteins in transgenic living systems, they cannot dis-cern features closer than half the wavelength of light (200 to 300 nm) (1). By causing such features to fluoresce sequentially, stimulated emission depletion (STED) microscopy and other emerging superresolution techniques have now overcome this barrier (2). Whereas these techniques have been applied through-out the life sciences, in vivo nanoscale imag-ing of cells in higher animals has remained elusive. We used STED microscopy to super-resolve neurons and their subtle dynamics in the cerebral cortex of a living mouse.

We developed an upright scanning STED fluorescence microscope with a 1.3–numerical aperture (NA) lens focusing a 80-MHz train of 70-ps excitation pulses of 488-nm wavelength on the animal brain (Fig. 1A). To visualize neu-rons, we used heterozygous TgN(Thy1-EYFP) mice expressing enhanced yellow fluorescent protein (EYFP) as a nonfusion protein in neu-ronal cytoplasm (3), which is under the control of the regulatory element from the thy1 gene. The focused excitation pulses were coaligned

Sebastian Berning,1 Katrin I. Willig,1* Heinz Steffens,1 Payam Dibaj,2 Stefan W. Hell1

Nanoscopy in a living mouse brain

and synchronized with doughnut-shaped 592-nm STED pulses of 300-ps duration and 25-mW focal power for silencing the EYFP. The fluorescence was imaged onto a confocal de-tector so that 600-nm-thick layers inside the brain could be discriminated. Optical access was provided by a cover glass–sealed hole in the skull, exposing the mouse’s somatosen-sory cortex. The anaesthetized mouse was ar-tificially ventilated and controlled for its vital functions, such as body temperature, ventila-tion, blood oxygenation, and heart function [by recording the electrocardiogram]. Rigid construction and keeping the optical paths short protected our setup from external vi-brations. Thus, the images could be recorded without active vibration compensation or numerical image processing. Cardiovascular and respiratory motion was suppressed by an optimized surgical preparation procedure, which was particularly important for record-ing potential movements of the dendritic spines. Although such movements have been observed in hippocampal organotypical slices of 5- to 7-day-old mice (4, 5) and during de novo growth in the developing cortex (6), un-til now it has remained unclear whether they can occur in the adult animal brain.

The STED image in Fig. 1B shows a dendrit-ic process within the molecular layer of the somatosensory cortex of a TgN(Thy1-EYFP) mouse, located 10 to 15 μm below the surface The STED images show structures of <70 nm in size (Fig. 1D), indicating that the resolution is at least of that order. Recording images ev-ery 7 to 8 min revealed that adult dendritic spines can undergo morphologic changes and movements (Fig. 1C, 0 to 30 min, and movie S1) on the time scale of minutes. These movements were repeatedly observed in all six individuals imaged. The mice were aged

between 66 and 205 days. To exclude random defocus from being mistaken for movement, we rendered each image by a maximum inten-sity projection of a stack of five images with 600-nm-depth spacing. Although the dendrite of origin retained largely the same shape throughout the experiments, morphological changes were found at the head and neck re-gions of the dendritic spines, potentially re-flecting alterations in the connectivity of the neural network, as is seen in the immature brain. Apparently, dendritic spines can alter their morphology in the intact somatosensory cortex of the adult mouse.

Although absorption of the 592-nm STED beam by EYFP is negligible, adverse absorp-tion in the tissue may occur at this wave-length. Indeed, slight local swellings have oc-casionally been observed for relatively thick dendritic processes featuring many mitochon-dria. However, we did not observe degrada-tion or disaggregation of dendritic processes as known to occur right after cell death or exitus. A remedy should be the use of red fluorescent proteins calling for STED beams of wavelengths >700 nm, where absorption by mitochondrial metabolites is negligible. Another option is to use transgenic animals tagged with reversibly switchable fluorescent proteins, enabling the STED-like nanoscopy called RESOLFT (reversible saturable optical fluorescence transitions) that requires far low-er light intensities than STED (7). Future long-term in vivo studies should solve central as-pects of brain development and, using mouse moadels, also of brain disease. Altogether, we expect in vivo optical nanoscopy to assume a major role in the quest for deciphering our primary organ.

1Department of NanoBiophotonics, Max Planck Institute (MPI) for Biophysical Chemistry, Am Fassberg 11, 37077 Göttingen, Germany. 2MPI for Experimental Medicine, Hermann-Rein-Strasse 3, 37075 Göttingen, Germany.*Corresponding authors: E-mail: [email protected] (K.I.W.); [email protected] (S.W.H.)

Nanoscopy in a Living Mouse BrainSebastian Berning,1 Katrin I. Willig,1* Heinz Steffens,1 Payam Dibaj,2 Stefan W. Hell1*

Cells, the functional units of life, are beststudied in vivo. This is particularly truefor neurons, which perform their basic

function of information processing by connect-ing with their neighbors. Therefore, unravelingthe inner workings of the brain requires the im-aging of neurons in the living animal. Althoughconfocal and multiphoton microscopy can vi-sualize neurons tagged with fluorescent proteinsin transgenic living systems, they cannot discernfeatures closer than half the wavelength of light(200 to 300 nm) (1). By causing such features tofluoresce sequentially, stimulated emission de-pletion (STED) microscopy and other emergingsuperresolution techniques have now overcomethis barrier (2). Whereas these techniques havebeen applied throughout the life sciences, in vivonanoscale imaging of cells in higher animals hasremained elusive. We used STED microscopy tosuperresolve neurons and their subtle dynamicsin the cerebral cortex of a living mouse.

We developed an upright scanning STEDfluorescence microscope with a 1.3–numericalaperture (NA) lens focusing a 80-MHz train of70-ps excitation pulses of 488-nm wavelengthon the animal brain (Fig. 1A). To visualize neu-rons, we used heterozygous TgN(Thy1-EYFP)mice expressing enhanced yellow fluorescentprotein (EYFP) as a nonfusion protein in neuro-nal cytoplasm (3), which is under the control ofthe regulatory element from the thy1 gene. Thefocused excitation pulses were coaligned andsynchronized with doughnut-shaped 592-nmSTED pulses of 300-ps duration and 25-mW fo-cal power for silencing the EYFP. The fluores-cence was imaged onto a confocal detector sothat 600-nm-thick layers inside the brain could

be discriminated. Optical access was providedby a cover glass–sealed hole in the skull, ex-posing the mouse’s somatosensory cortex. Theanaesthetized mouse was artificially ventilatedand controlled for its vital functions, such as bodytemperature, ventilation, blood oxygenation, andheart function [by recording the electrocardio-gram]. Rigid construction and keeping the opti-cal paths short protected our setup from externalvibrations. Thus, the images could be recordedwithout active vibration compensation or numer-ical image processing. Cardiovascular and respi-ratory motion was suppressed by an optimizedsurgical preparation procedure, which was par-ticularly important for recording potential move-ments of the dendritic spines. Although suchmovements have been observed in hippocampalorganotypical slices of 5- to 7-day-old mice (4, 5)and during de novo growth in the developing cor-tex (6), until now it has remained unclear whetherthey can occur in the adult animal brain.

The STED image in Fig. 1B shows a den-dritic process within the molecular layer of thesomatosensory cortex of a TgN(Thy1-EYFP)mouse, located 10 to 15 mm below the surface.The STED images show structures of <70 nmin size (Fig. 1D), indicating that the resolutionis at least of that order. Recording images every7 to 8 min revealed that adult dendritic spinescan undergo morphologic changes and move-ments (Fig. 1C, 0 to 30 min, and movie S1) onthe time scale of minutes. These movements wererepeatedly observed in all six individuals imaged.The mice were aged between 66 and 205 days.To exclude random defocus from being mistakenfor movement, we rendered each image by amaximum intensity projection of a stack of five

images with 600-nm-depth spacing. Althoughthe dendrite of origin retained largely the sameshape throughout the experiments, morphologi-cal changes were found at the head and neckregions of the dendritic spines, potentially reflect-ing alterations in the connectivity of the neuralnetwork, as is seen in the immature brain. Ap-parently, dendritic spines can alter their mor-phology in the intact somatosensory cortex ofthe adult mouse.

Although absorption of the 592-nm STEDbeam by EYFP is negligible, adverse absorptionin the tissue may occur at this wavelength. In-deed, slight local swellings have occasionallybeen observed for relatively thick dendritic pro-cesses featuring many mitochondria. However,we did not observe degradation or disaggrega-tion of dendritic processes as known to occurright after cell death or exitus. A remedy shouldbe the use of red fluorescent proteins calling forSTED beams of wavelengths >700 nm, whereabsorption by mitochondrial metabolites is neg-ligible. Another option is to use transgenic ani-mals tagged with reversibly switchable fluorescentproteins, enabling the STED-like nanoscopy calledRESOLFT (reversible saturable optical fluores-cence transitions) that requires far lower light in-tensities than STED (7). Future long-term in vivostudies should solve central aspects of brain de-velopment and, using mouse models, also of braindisease. Altogether, we expect in vivo optical nano-scopy to assume a major role in the quest for de-ciphering our primary organ.

References and Notes1. T. Misgeld, M. Kerschensteiner, F. M. Bareyre, R. W. Burgess,

J. W. Lichtman, Nat. Methods 4, 559 (2007).2. S. W. Hell, Science 316, 1153 (2007).3. G. P. Feng et al., Neuron 28, 41 (2000).4. U. V. Nägerl, K. I. Willig, B. Hein, S. W. Hell, T. Bonhoeffer,

Proc. Natl. Acad. Sci. U.S.A. 105, 18982 (2008).5. N. T. Urban, K. I. Willig, S. W. Hell, U. V. Nägerl,

Biophys. J. 101, 1277 (2011).6. H. B. Kwon, B. L. Sabatini, Nature 474, 100 (2011).7. T. Grotjohann et al., Nature 478, 204 (2011).

Acknowledgments: The experiments were performed accordingto the ethics guidelines of national law regarding animalprotection procedures and were authorized by the ethicscommittee of the Max Planck Institute for Biophysical Chemistryand by the responsible authorities, the NiedersächsischesLandesamt für Verbraucherschutz. We thank A. Schönle for thesoftware Imspector. S.W.H. holds patents on STED microscopy:German patent DE 4416 558 C2 and U.S. Patent US5731588A.

Supporting Online Materialwww.sciencemag.org/cgi/content/full/335/6068/551/DC1Materials and MethodsFigs. S1 and S2Movie S1

17 October 2011; accepted 15 December 201110.1126/science.1215369

BREVIA

1Department of NanoBiophotonics, Max Planck Institute (MPI)for Biophysical Chemistry, Am Fassberg 11, 37077 Göttingen,Germany. 2MPI for Experimental Medicine, Hermann-Rein-Strasse 3, 37075 Göttingen, Germany.

*To whom correspondence should be addressed. E-mail:[email protected] (K.I.W.); [email protected] (S.W.H.)

B

t=0 7min 15min 22min 30min30minCD

67nm67nm

A

Fig. 1. STEDmicroscopy in themolecular layer of the somatosensory cortex of a mouse with EYFP-labeledneurons. (A) Anesthetizedmouse under the objective lens (63×, NA 1.3, glycerol immersion) with trachealtube. (B) Projected volumes of dendritic and axonal structures reveal (C) temporal dynamics of spinemorphology with (D) an approximately fourfold improved resolution compared with diffraction-limitedimaging. Curve is a three-pixel-wide line profile fitted to raw data with a Gaussian. Scale bars, 1 mm.

www.sciencemag.org SCIENCE VOL 335 3 FEBRUARY 2012 551

on

Aug

ust 1

0, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

Fig. 1. STED microscopy in themolecular layer of the somatosensory cortex of a mouse with EYFP-labeled neurons. (A) Anesthetizedmouse under the objective lens (63×, NA 1.3, glycerol im-mersion) with tracheal tube. (B) Projected volumes of dendritic and axonal structures reveal (C) tem-poral dynamics of spine morphology with (D) an approximately fourfold improved resolution com-pared with diffraction-limited imaging. Curve is a three-pixel-wide line profile fitted to raw data with a Gaussian. Scale bars, 1 mm.

REFERENCES AND NOTES 1. T. Misgeld, M. Kerschensteiner, F. M. Bareyre, R. W. Burgess, J. W. Lichtman, Nat. Methods 4, 559 (2007). 2. S. W. Hell, Science 316, 1153 (2007). 3. G. P. Feng et al., Neuron 28, 41 (2000). 4. U. V. Nägerl, K. I. Willig, B. Hein, S. W. Hell, T. Bonhoeffer, Proc. Natl. Acad. Sci. U.S.A. 105, 18982 (2008). 5. N. T. Urban, K. I. Willig, S. W. Hell, U. V. Nägerl, Biophys. J. 101, 1277 (2011). 6. H. B. Kwon, B. L. Sabatini, Nature 474, 100 (2011). 7. T. Grotjohann et al., Nature 478, 204 (2011).

ACKNOWLEDGMENTSThe experiments were performed according to the ethics guidelines of national law regarding animal protection procedures and were authorized by the ethicscommittee of the Max Planck Institute for Biophysical Chemistry and by the responsible authorities, the Niedersächsisches Landesamt für Verbraucherschutz. We thank A. Schönle for the software Imspector. S.W.H. holds patents on STED microscopy: German patent DE 4416 558 C2 and U.S. Patent US5731588A.

SUPPORTING ONLINE MATERIALWww.sciencemag.org/cgi/content/full/335/6068/551/DC1

MATERIALS AND METHODSFigs. S1 and S2Movie S1

17 October 2011; accepted 15 December 201110.1126/science.1215369

The quantitative organization of cellular pathways is not well understood. One well-researched membrane trafficking pathway, synaptic vesicle recycling, occupies its own compartment, the

synaptic bouton, and can therefore be studied in isolation. It is a relatively simple pathway, comprising only a few steps (1–3). First, neu-rotransmitter-filled synaptic vesicles dock to the release site (active zone), are primed for release, and then fuse with the plasma mem-brane (exocytosis). The vesicle molecules are later sorted and retrieved from the plasma membrane (endocytosis). An additional sort-ing step in an early endosome (3–5) may take place before the vesicle refills with neu-rotransmitter.

To quantify the organization of synaptic vesicle recycling, we first purified synaptic boutons (synaptosomes) from the cellular layers of the cortex and cerebellum of adult rats, using a modified version (6) of a classi-cal brain fractionation protocol (7) (Fig. 1A). The different cellular components were sepa-rated by Ficoll density gradients, resulting in a heterogeneous sample, which we first ana-lyzed by electron microscopy. About 58.5% of all organelles were resealed, vesicle-loaded

Synaptic vesicle recycling has long served as a model for the general mechanisms of cel-lular trafficking. We used an integrative approach, combining quantitative immunob-lotting and mass spectrometry to determine protein numbers; electron microscopy to measure organelle numbers, sizes, and positions; and super-resolution fluorescence microscopy to localize the proteins. Using these data, we generated a three-dimensional model of an “average” synapse, displaying 300,000 proteins in atomic detail. The copy numbers of proteins involved in the same step of synaptic vesicle recycling correlated closely. In contrast, copy numbers varied over more than three orders of magnitude between steps, from about 150 copies for the endosomal fusion proteins to more than 20,000 for the exocytotic ones.

Benjamin G. Wilhelm,1,2 Sunit Mandad,3* Sven Truckenbrodt,1,5* Katharina Kröhnert,1

Christina Schäfer,1 Burkhard Rammner,1 Seong Joo Koo,6 Gala A. Claßen,6

Michael Krauss,6 Volker Haucke,6 Henning Urlaub,3,4 Silvio O. Rizzoli1†

Composition of isolated synapticboutons reveals the amounts ofvesicle trafficking proteins

synaptosomes (fig. S1). Most of the remaining organelles, such as mitochondria (~20%) and myelin (8%) (fig. S1), contained few proteins relevant to synaptic vesicle recycling and thus did not bias synaptic protein quantification. The electron microscopy analysis of the syn-aptosomes also provided their spatial param-eters (size, surface, and volume), which are critical in understanding protein concentra-tions (Fig. 1, B and C).

Before proceeding to investigate the synap-tic protein copy numbers, we tested whether the synaptosomes lost a significant propor-tion of their proteins during the purification procedure. We compared the amounts of 27 soluble proteins and 2 transmembrane pro-teins in synaptosomes and in undisturbed synapses from brain slices, using fluorescence microscopy (fig. S2, A and B). The large ma-jority of the proteins exhibited no signifi-cant changes after synaptosome purification (fig. S2C).

Having verified that the purification pro-cedure maintains the protein composition of the synaptic bouton, we used quantitative immunoblotting to determine the amount of protein of interest per microgram of synapto-somes for 62 synaptic proteins (Fig. 1, D and E). To transform this value into copy numbers per synaptosome, we determined the number of particles in the synaptosome preparation by fluorescence microscopy (~17 million) (fig. S3) and the fraction represented by synapto-somes by electron microscopy (fig. S1, A and B) and by immunostaining for synaptic mark-ers (fig. S1B). Both measurements indicate that ~58% of all particles are synaptosomes, ~9.95 million synaptosomes per microgram.

The results we obtained for all proteins tested are included in table S1. Despite the heterogeneous preparation we started with, our results are very close to synaptic vesicles purified to more than 95% (8), taking into ac-

1Department of Neuro- and Sensory Physiology, University ofGöttingen Medical Center, European Neuroscience Institute,Cluster of Excellence Nanoscale Microscopy and MolecularPhysiology of the Brain, Göttingen, Germany. 2International Max Planck Research School Neurosciences, 37077Göttingen, Germany. 3Bioanalytical Mass Spectrometry Group, Max-Planck-Institute for Biophysical Chemistry, 37077 Göttingen, Germany. 4Bioanalytics, Department of Clinical Chemistry, University Medical Center Göttingen, 37075 Göttingen, Germany. 5International Max Planck Research School Molecular Biology, 37077 Göttingen, Germany. 6Leibniz Institut für Molekulare Pharmakologie, Department of Molecular Pharmacology and Cell Biology, Robert-Rössle-Strasse 10, 13125 Berlin, Germany.*These authors contributed equally to this work. †Corresponding author. E-mail: [email protected]

count the known fractions of these proteins on the synaptic plasma membrane (9, 10) (Fig. 1F). We only detected a sizeable difference for synaptic vesicle 2 (SV2) [12 copies per synap-tic vesicle in our study, versus 1.7 for (8)]. A more recent study, using an antibody-based approach that is likely to underestimate the copy numbers of abundant synaptic vesicle proteins, found about five SV2 molecules per vesicle (11).

The immunoblot analysis also provided the total mass of each protein per microgram of synaptosome preparation, which could be translated to percentage of the total protein in the preparation. Our quantification of syn-aptic proteins addressed ~23% of the total protein in the preparation. Because the syn-aptosomes make ~58% of the preparation, our quantification thus addressed ~40.5% of the total protein in synaptosomes (without pre-synaptic mitochondria). To test and extend these values, we turned to quantitative mass spectrometry, using a label-free approach, in-tensity-based absolute quantification (iBAQ) (12). iBAQ estimates the abundance of par-ticular proteins by summing the intensities of all peptides derived from them and then nor-malizing to the total possible number of pep-tides. We compared the peptides derived from recombinant synaptic proteins (same as those used for quantitative immunoblotting) from human Universal Protein Standards (UPS2) and finally from synaptosomes, using a hybrid mass spectrometer. iBAQ values were then calculated using MaxQuant (13) and the An-dromeda search engine (14), and the amounts of proteins present in synaptosomes were de-termined by linear regression. The estimates obtained by iBAQ correlated well with the immunoblotting results (fig. S4). The iBAQ approach generated abundance estimates for ~1100 additional proteins in the preparation (see table S2 for a number of well-known pro-teins relevant to synaptic activity; see table S3 for all other proteins). All quantified proteins (iBAQ and immunoblot analysis) added up to ~88.4% of the protein weight of the entire syn-aptosome preparation (obtained by summing the percentages indicated in tables S1 to S3).

The members of heteromultimeric com-plexes, such as the vesicular adenosine tri-phosphatase (vATPase), were present in the correct (expected) stoichiometries (Fig. 1F), verifying the accuracy of our quantification procedure. The copy numbers of proteins known to be involved in a particular step of synaptic vesicle recycling correlated remark-ably well. This observation applied to the exocytotic fusion proteins [SNAREs (fig. S5B), whose abundance was only matched by actin and tubulin (fig. S5M)], to proteins involved in fusion regulation [SNARE-binding or prim-ing proteins (fig. S5C)], to proteins of the clathrin-mediated endocytosis pathway (fig. S5E), to endosomal or constitutive fusion pro-teins (fig. S5D), to structural vesicle cluster proteins (fig. S5F), to active zone proteins (fig.

Originally published 3 February 2012 in SCIENCE Originally published 30 May 2014 in SCIENCE





































S5G), to major synaptic vesicle constituents (fig. S5H), or to adhesion proteins (fig. S5I). Proteins involved in membrane trafficking pathways unrelated to synaptic vesicle recy-cling, such as the exocyst pathway (fig. S5J), were not abundant. There was no correlation between structurally similar proteins, such as those of the Rab or septin families (fig. S5, K and L). Protein copy numbers are high in some steps of the vesicle recycling pathway but much lower in other steps. For exam-ple, the exocytotic SNAREs were present in 20,100 to 26,000 copies, despite the fact that one vesicle fusion event requires the forma-

tion of only one to three SNARE complexes, which contain one copy of each of the three SNAREs (15–17). SNARE-interacting proteins were found at copy numbers of one to several thousands (Munc13a, Munc18a, complexin I, and complexin II) (fig. S5C). In contrast, only ~4000 clathrin molecules and 2300 dy-namin molecules were present in the average synapse. Because at least 150 to 180 copies of clathrin are needed for one recycling vesicle (18, 19), the entire clathrin complement of the synapse would be sufficient for the simultane-ous endocytosis of only 7% of all vesicles. The dynamin complement of the synapse was only

sufficient for 11% of the vesicles, taking into account that at least 52 copies, corresponding to two adjacent dynamin rings, are needed for one pinch-off event (20). Finally, the endosom-al SNAREs, which form tetrameric complexes containing one copy of each SNARE (4,6), were even less abundant (50 to 150 copies) than the endocytotic cofactors.

For some proteins, a strong enrichment in the location where they function may com-pensate for their low copy numbers. Con-versely, abundant proteins may be scattered throughout the synaptic space, which would render their concentrations fairly low at

known fractions of these proteins on the syn-aptic plasmamembrane (9, 10) (Fig. 1F). We onlydetected a sizeable difference for synaptic vesicle2 (SV2) [12 copies per synaptic vesicle in ourstudy, versus 1.7 for (8)]. A more recent study,using an antibody-based approach that is likelyto underestimate the copy numbers of abun-dant synaptic vesicle proteins, found about fiveSV2 molecules per vesicle (11).The immunoblot analysis also provided the

total mass of each protein per microgram ofsynaptosome preparation, which could be trans-lated to percentage of the total protein in thepreparation. Our quantification of synaptic pro-teins addressed ~23% of the total protein in the

preparation. Because the synaptosomes make~58% of the preparation, our quantification thusaddressed ~40.5% of the total protein in synap-tosomes (without presynaptic mitochondria). Totest and extend these values, we turned to quan-titative mass spectrometry, using a label-free ap-proach, intensity-based absolute quantification(iBAQ) (12). iBAQ estimates the abundance ofparticular proteins by summing the intensitiesof all peptides derived from them and thennormalizing to the total possible number ofpeptides. We compared the peptides derivedfrom recombinant synaptic proteins (same asthose used for quantitative immunoblotting) fromhuman Universal Protein Standards (UPS2) and

finally from synaptosomes, using a hybrid massspectrometer. iBAQ values were then calculatedusing MaxQuant (13) and the Andromeda searchengine (14), and the amounts of proteins presentin synaptosomes were determined by linear re-gression. The estimates obtained by iBAQ cor-related well with the immunoblotting results(fig. S4). The iBAQ approach generated abun-dance estimates for ~1100 additional proteinsin the preparation (see table S2 for a number ofwell-known proteins relevant to synaptic activ-ity; see table S3 for all other proteins). All quan-tified proteins (iBAQ and immunoblot analysis)added up to ~88.4% of the protein weight ofthe entire synaptosome preparation (obtained

Fig. 1. Physical characteristics of the averagesynaptosome. (A) Schema illustrating the purifi-cation of synaptosomes. See the supplementarymaterials for details. (B) Serial electronmicrographsof purified synaptosomes were used to reconstructentire synapses.The plasma membrane is depictedin light beige, the active zone in red, synaptic ves-icles in dark beige, larger organelles in dark gray,and mitochondria in purple. This synaptosome re-sembles the average physical parameters (C) andwas used to model the average presynaptic ter-minal (Fig. 3). (C) Table listing the average physicalparameters of synaptosomes. The values repre-sent mean T SEM of 65 reconstructions from fourindependent synaptosome preparations. (D) Quantitativeimmunoblots of the three synaptic SNARE proteins (SNAP 25,syntaxin 1, and VAMP 2).The lanes on the left represent increasing amountsof the purified protein of interest, forming a standard curve (protein amountversus band intensity). The different synaptosome samples are depicted inthe four lanes on the right. (E) Standard curves of the three SNARE proteinsobtained from the immunoblots depicted in (D). Linear regression was usedto determine the absolute amount of the protein of interest in the synap-tosomes. (F) (Left) The copy numbers for eight major synaptic vesicle pro-teins, normalized to the number of synaptic vesicles per synaptosome, arecompared with the numbers obtained in a previous quantification of syn-aptic vesicles (8). The red line represents identity. (Middle) The model shows

the eight compared proteins in correct copy numbers on an average ves-icle. (Right) Correlation between the copy numbers of different vATPasesubunits (highlighted in different colors in the vATPasemodel, above the graph).The immunoblot quantification of the a1 subunit (green; only the trans-membrane part is shown) suggests the presence of 742 vATPase complexesper bouton. The copy numbers of the B, C, E, and F subunits (derived fromiBAQ mass spectrometry) are plotted against their expected stoichiome-tries for 742 complexes. The stoichiometry of the different vATPasesubunits was obtained from (34). The black line represents identity. Alldata represent means T SEM from four independent preparations.

RESEARCH | REPORTS

1024 30 MAY 2014 • VOL 344 ISSUE 6187 sciencemag.org SCIENCE

Fig. 1. Physical characteristics of the average synaptosome. (A) Schema illustrating the puri-fication of synaptosomes. See the supplementa-ry materials for details. (B) Serial electron micro-graphs of purified synaptosomes were used to reconstruct entire synapses. The plasma mem-brane is depicted in light beige, the active zone in red, synaptic vesicles in dark beige, larger or-ganelles in dark gray, and mitochondria in purple. This synaptosome resembles the average physi-cal parameters (C) and was used to model the average presynaptic terminal (Fig. 3). (C) Table listing the average physical parameters of syn-aptosomes. The values represent mean ± SEM of 65 reconstructions from four independent synaptosome preparations. (D) Quantitative im-

munoblots of the three synaptic SNARE proteins (SNAP 25, syntaxin 1, and VAMP 2). The lanes on the left represent increasing amounts of the purified protein of interest, forming a standard curve (protein amount versus band intensity). The different synaptosome samples are depicted in the four lanes on the right. (E) Standard curves of the three SNARE proteins obtained from the immunoblots depicted in (D). Linear regression was used to determine the absolute amount of the protein of interest in the synaptosomes. (F) (Left) The copy numbers for eight major synaptic vesicle proteins, normalized to the number of synaptic vesicles per synaptosome, are compared with the numbers obtained in a previous quantification of synaptic vesicles (8). The red line represents identity. (Middle) The model shows the eight compared proteins in correct copy numbers on an average vesicle. (Right) Correlation between the copy numbers of different vATPase subunits (highlighted in different colors in the vATPase model, above the graph). The immunoblot quantification of the a1 subunit (green; only the transmembrane part is shown) suggests the presence of 742 vATPase complexes per bouton. The copy numbers of the B, C, E, and F subunits (derived from iBAQ mass spectrometry) are plotted against their expected stoichiometries for 742 complexes. The stoichiometry of the different vATPase subunits was obtained from (34). The black line represents identity. All data represent means ± SEM from four independent preparations.

Fig. 2. Presynaptic proteinorganization. (A) Proteinorganization in synaptosomes. Thescheme indicates an overviewof the preparation. AZ, active zone;ves, synaptic vesicles. Purifiedsynaptosomes were immunostainedin parallel for the protein ofinterest, VAMP 2 (red, STEDresolution), for an active zone marker,bassoon (blue, confocal resolution), and for a vesicle marker, synaptophysin(green, confocal resolution).The fourth panel shows the relative spatial distributionof VAMP 2 as obtained from average images (several hundred synapses from twoindependent experiments; see the supplementary materials for further details).Theputative outline of the synapse is indicated by thewhite line, the active zonebythe black circle; the relative spatial abundance is color-coded (see color bar). Scalebarsare500nm(imagepanels) and200nm(fourthpanel).The last twopanelsonthe right are density distributions for two additional presynaptic proteins,amphiphysin and syntaxin 16. Scale bar is 200 nm. (B) Protein organizationin hippocampal cultures. Details as in (A). Scale bars are 2 mm and 200 nm,respectively. (C) Protein organization in the mouse neuromuscular junction.Instead of immunostaining for bassoon, the active zone position was obtained bylabelingpostsynaptic acetylcholine receptorswithbungarotoxin.All otherdetails asin (A). Scale bars are 2 mmand500nm, respectively. Imagingdata for all the otherproteins areprovided in fig. S6. (D)Different spatial parametersweremeasured foreach of the 62 proteins we imaged, as indicated by the labeling of the rows.

Parameter values were normalized to the maximum (100%). All values areindicated according to the color scale (right).The proteins are grouped accord-ing to functional categories: active zone proteins (bassoon, piccolo, andRIM1), synaptic vesicle proteins (synaptophysin,VGlut 1/2,VAMP2,VAMP 1, SV2A/B, synapsin I/II, and synaptogyrin 1), calcium sensor proteins (synaptotagmin2, synaptotagmin 1, synaptotagmin 7, doc 2A/B, and calmodulin), SNAREcofactors (CSP, Munc13a, Munc18a, NSF, a-SNAP, and complexin 1/2), smallguanosine triphosphatases (GTPases) (Rab3, Rab5, and Rab7), disease-relatedproteins (a/b-synuclein, APP, and b-secretase), mitochondrial proteins (VDAC),endocytosis proteins (AP-2 mu2, SGIP1, synaptojanin, epsin 1, clathrin heavychain, clathrin light chain, dynamin 1,2,3, endophilin I,II,III, amphiphysin, Hsc70,intersectin 1, PIPK Ig, AP 180, and syndapin 1), endosomal SNAREs (syntaxin 13,syntaxin 16, syntaxin 7, syntaxin 6, Vti1a, and VAMP4), plasma membraneSNAREs (syntaxin 1, SNAP23,SNAP25, andSNAP29), general secretory proteins(CAPS, SCAMP 1, SGTa, and vATPase a1), calcium buffer proteins (calbindin,calretinin, and parvalbumin), and cytoskeletal proteins (actin, septin 5, and tubulin).

RESEARCH | REPORTS

SCIENCE sciencemag.org 30 MAY 2014 • VOL 344 ISSUE 6187 1025

Fig. 2. Presynaptic proteinorganization. (A) Proteinorganization in synaptosomes. The scheme indicates an overview of the preparation. AZ, active zone; ves, synaptic vesicles. Purified synaptosomes were immunostained in parallel for the protein of interest, VAMP 2 (red, STED resolution), for an active zone marker, bassoon (blue,confocal resolution), and for a vesicle marker, synaptophysin (green, confocal resolution). The fourth panel shows the relative spatial distribution of VAMP 2 as obtained from average images (several hundred synapses from two independent experiments; see the supplementary materials for further details). The putative outline of the synapse is indicated by the white line, the active zone by the black circle; the relative spatial abundance is color-coded (see color bar). Scale bars are 500 nm (image panels) and 200 nm (fourth panel). The last two panels on the right are density distributions for two additional presynaptic proteins, amphiphysin and syntaxin 16. Scale bar is 200 nm. (B) Protein organization in hippocampal cultures. Details as in (A). Scale bars are 2 μm and 200 nm, respectively. (C) Protein organization in the mouse neuromuscular junction. Instead of immunostaining for bassoon, the active zone position was obtained by labeling postsynaptic acetylcholine receptors with bungarotoxin. All other details as in (A). Scale bars are 2 μm and 500 nm, respectively. Imaging data for all the other proteins are provided in fig. S6. (D) Different spatial parameters were measured for each of the 62 proteins we imaged, as indicated by the labeling of the rows. Parameter values were normalized to the maximum (100%). All values are indicated according to the color scale (right). The proteins are grouped according to functional categories: active zone proteins (bassoon, piccolo, and RIM1), synaptic vesicle proteins (synaptophysin, VGlut 1/2, VAMP 2, VAMP 1, SV2 A/B, synapsin I/II, and synaptogyrin 1), calcium sensor proteins (synaptotagmin 2, synaptotagmin 1, synaptotagmin 7, doc 2A/B, and calmodulin), SNARE cofactors (CSP, Munc13a, Munc18a, NSF, α-SNAP, and complexin 1/2), small guanosine triphosphatases (GTPases) (Rab3, Rab5, and Rab7), disease-related proteins (α/β-synuclein, APP, and β-secretase), mitochondrial proteins (VDAC), endocytosis proteins (AP-2 mu2, SGIP1, synaptojanin, epsin 1, clathrin heavy chain, clathrin light chain, dynamin 1,2,3, endophilin I,II,III, amphiphysin, Hsc70, intersectin 1, PIPK Iγ, AP 180, and syndapin 1), endosomal SNAREs (syntaxin 13, syntaxin 16, syntaxin 7, syntaxin 6, Vti1a, and VAMP4), plasma membrane SNAREs (syntaxin 1, SNAP 23, SNAP 25, and SNAP 29), general secretory proteins (CAPS, SCAMP 1, SGTα, and vATPase a1), calcium buffer proteins (calbindin, calretinin, and parvalbumin), and cytoskeletal proteins (actin, septin 5, and tubulin).















individual sites. To estimate the influence of protein localization, we selected 62 proteins and analyzed them by immunostaining and fluorescence microscopy. We used stimulated emission depletion (STED) (21), a diffraction-unlimited technique, to investigate protein po-sitions with a resolution of ~40 nm (Fig. 2A). To avoid bias owing to possible artifacts con-nected to the brain homogenization procedure required for generating synaptosomes, we also studied two additional preparations: cultured hippocampal neurons (Fig. 2B) and the leva-tor auris longus neuromuscular junction (Fig. 2C), acutely dissected from adult animals (22).

We analyzed the proteins of interest in rela-tion to the positions of the release site (identi-fied by marking active zone proteins) and of the vesicle cluster (visualized by staining for the protein that is most strongly enriched in purified synaptic vesicles, synaptophysin) (8). We averaged single synapses by overlapping their active zones and rotating the images un-til reaching the best possible alignment of the vesicle cluster and of the protein of interest. This procedure provided an overview of the relative spatial distribution of each protein. Overall, many of the protein distributions were similar (Fig. 2, A to C, and fig. S6). Ac-tive zone proteins were mostly confined to the active zone areas. Most of the other pro-teins could be found throughout the synaptic boutons [albeit they were enriched to differ-ent levels in areas such as the active zone or the vesicle cluster (Fig. 2D); see fig. S7, A to H, for a more detailed analysis of differences between the proteins]. These observations are consistent with the presence of most of the proteins on purified synaptic vesicles (8) and with the fact that the synaptic vesicle cluster occupies much of the synaptic bouton volume (Fig. 1B). Thus, for the majority of proteins, lo-calization does not appear to compensate for low copy numbers.

Although the imaging parameters measured above did not pinpoint actual positions within the synapse, they allowed us to make broad estimates for the organization of each protein (fig. S7I). We used the data to generate a three-dimensional (3D) model containing 60 pro-teins placed within a typical synaptic volume (obtained from an individual electron micros-copy reconstruction whose parameters were close to synaptosome averages) (Fig. 3). The proteins were modeled in atomic detail, ac-cording to their known molecular structures, and were placed in the synaptic space accord-ing to the information provided by the STED images and the literature (Fig. 2 and fig. S6). For example, the SNARE molecules syntaxin 1 and SNAP 25 are shown in clusters with a specific organization (23–25). The hippocam-pal culture images (Fig. 2B) were used to ob-tain an additional set of data, the correlation of protein amounts with synapse size [judged from the amount of vesicles (26) (fig. S6)]. The copy numbers of some proteins increase lin-early with synapse size; others, including most endocytotic proteins, follow an exponential

curve, which implies that small synapses con-tain proportionally larger amounts of these proteins than large synapses.

We used the modeled volumes of the pro-teins to calculate the fraction of the synaptic volume that they occupy. This value, ~7% of the synaptosome volume (excluding mito-chondria), is comparable to the space occu-pied by the synaptic vesicles (~6%, derived from the electron microscopy measurements). These low values could lead to the impres-sion that the synaptic volume is not densely populated by proteinaceous structures. How-ever, the 3D model suggests that the synaptic space is rather crowded, especially inside the vesicle cluster and at the active zone (Fig. 3, A to C, and movie S1). This probably places constraints on both organelle and protein dif-fusion. The high copy numbers of exocytosis-related proteins may have evolved as a mecha-nism to cope with these constraints, to ensure the high speed of neurotransmitter release. In contrast, endocytosis can take place for many tens of seconds after exocytosis. This allows endocytosis to proceed with proportionally lower numbers of cofactor proteins. In prin-ciple, the synaptic boutons could increase the speed of endocytosis by accumulating larger amounts of endocytotic proteins. This, how-ever, would result in an even greater conges-tion of the synaptic space, which presumably might perturb synaptic function. A simpler solution for the problem of balancing rapid release with slow vesicle retrieval appears to have been to maintain a large enough reser-voir of vesicles (22, 27, 28).

Our data reveal a correlation between the copy numbers of proteins involved in the same steps of synaptic vesicle recycling. The mecha-nisms behind this correlation are unclear. A simple hypothesis would be that such proteins either are produced together or are transport-ed to the synapse together. However, these proteins have different lifetimes (29) and are transported from the neuronal cell body on different precursors (30). One possible expla-nation, at least for the soluble cofactor pro-teins, is that the synaptic vesicle cluster regu-lates their number. The vesicles are known to bind to and buffer such proteins (22, 31–33), thereby retaining in the synapse only a defined number of cofactors. Such mechanisms do not apply, however, to transmembrane proteins, whose regulation remains to be determined.

Fig. 3. A 3D model of synaptic architecture. (A) A section through the synaptic bouton, indicating 60 proteins. The proteins are shown in the copy numbersindicated in tables S1 and S2 and in positions determined according to the imaging data (Fig. 2 and fig. S6) and to the literature (see fig. S6 for details). (B) High-zoom view of the active zone area. (C) High-zoom view of one vesicle within the vesicle cluster. (D) High-zoom view of a section of the plasma membrane in thevicinity of the active zone. Clusters of syntaxin (yellow) and SNAP 25 (red) are visible, as well as a recently fused synaptic vesicle (top).The graphical legend indicatesthe different proteins (right). Displayed synaptic vesicles have a diameter of 42 nm.

RESEARCH | REPORTS

1026 30 MAY 2014 • VOL 344 ISSUE 6187 sciencemag.org SCIENCE

Fig. 3. A 3D model of synaptic architecture. (A) A section through the synaptic bouton, indicating 60 proteins. The proteins are shown in the copy num-bers indicated in tables S1 and S2 and in positions determined according to the imaging data (Fig. 2 and fig. S6) and to the literature (see fig. S6 for details). (B) High-zoom view of the active zone area. (C) High-zoom view of one vesicle within the vesicle cluster. (D) High-zoom view of a section of the plasma membrane in the vicinity of the active zone. Clusters of syntaxin (yellow) and SNAP 25 (red) are visible, as well as a recently fused synaptic vesicle (top). The graphical legend indicates the different proteins (right). Displayed synaptic vesicles have a diameter of 42 nm.

REFERENCES AND NOTES 1. V. Haucke, E. Neher, S. J. Sigrist, Nat. Rev. Neurosci. 12, 127–1388 (2011). 2. R. Jahn, D. Fasshauer, Nature 490, 201–207 (2012). 3. T. C. Südhof, Annu. Rev. Neurosci. 27, 509–547 (2004). 4. P. Hoopmann et al., Proc. Natl. Acad. Sci. U.S.A. 107, 19055–19060 (2010). 5. V. Uytterhoeven, S. Kuenen, J. Kasprowicz, K. Miskiewicz, P. Verstreken, Cell 145, 117–132 (2011). 6. S. O. Rizzoli et al., Traffic 7, 1163–1176 (2006). 7. D. G. Nicholls, T. S. Sihra, Nature 321, 772–773 (1986). 8. S. Takamori et al., Cell 127, 831–846 (2006). 9. F. Opazo et al., Traffic 11, 800–812 (2010).10. M. Darna et al., J. Biol. Chem. 284, 4300–4307 (2009).

11. S. A. Mutch et al., J. Neurosci. 31, 1461–1470 (2011).12. B. Schwanhäusser et al., Nature 473, 337–342 (2011).13. J. Cox, M. Mann, Nat. Biotechnol. 26, 1367–1372 (2008).14. J. Cox et al., J. Proteome Res. 10, 1794–1805 (2011).15. R. Mohrmann, H. de Wit, M. Verhage, E. Neher, J. B. Sørensen, Science 330, 502–505 (2010).16. R. Sinha, S. Ahmed, R. Jahn, J. Klingauf, Proc. Natl. Acad. Sci. U.S.A. 108, 14318–14323 (2011).17. G. van den Bogaart et al., Nat. Struct. Mol. Biol. 17, 358–364 (2010).18. Y. Cheng, W. Boll, T. Kirchhausen, S. C. Harrison, T. Walz, J. Mol. Biol. 365, 892–899 (2007).19. H. T. McMahon, E. Boucrot, Nat. Rev. Mol. Cell Biol. 12, 517–533 (2011).20. A. V. Shnyrova et al., Science 339, 1433–1436 (2013).21. K. I. Willig, S. O. Rizzoli, V. Westphal, R. Jahn, S. W. Hell, Nature 440, 935–939 (2006).22. A. Denker et al., Proc. Natl. Acad. Sci. U.S.A. 108, 17177–17182 (2011).23. D. Bar-On et al., J. Biol. Chem. 287, 27158–27167 (2012).24. J. J. Sieber, K. I. Willig, R. Heintzmann, S. W. Hell, T. Lang, Biophys. J. 90, 2843–2851 (2006).25. J. J. Sieber et al., Science 317, 1072–1076 (2007).26. V. N. Murthy, T. Schikorski, C. F. Stevens, Y. Zhu, Neuron 32, 673–682 (2001).27. V. Marra et al., Neuron 76, 579–589 (2012).28. T. Rose, P. Schoenenberger, K. Jezek, T. G. Oertner, Neuron 77, 1109–1121 (2013).29. L. D. Cohen et al., PLOS ONE 8, e63191 (2013).30. D. Bonanomi, F. Benfenati, F. Valtorta, Prog. Neurobiol. 80, 177–217 (2006).31. A. Denker, K. Kröhnert, J. Bückers, E. Neher, S. O. Rizzoli, Proc. Natl. Acad. Sci. U.S.A. 108, 17183–17188 (2011).32. O. Shupliakov, Neuroscience 158, 204–210 (2009). 33. S. O. Rizzoli, EMBO J. 33, 788–822 (2014).34. N. Kitagawa, H. Mazon, A. J. R. Heck, S. Wilkens, J. Biol. Chem. 283, 3329–3337 (2008).

ACKNOWLEDGMENTSWe thank the following collaborators for providing purified proteins and antibodies: R. Jahn, C. Griesinger, B. Shwaller, and A. Roux. We thank H. Martens for technical help and sup-port, B. Rizzoli for helpful comments on the manuscript, and T. Sargeant for providing the carve source code for creation of the voltage- dependent anion channel (VDAC)-based mitochondrial membrane cut-outs. B.G.W. was supported by a Boehringer Ingelheim Fonds PhD Fellowship. S.T. was sup-ported by an Excellence Stipend of the Göttingen Graduate School for Neurosciences, Biophysics, and Molecular Biosciences (GGNB). The work was supported by grants to S.O.R. from the European Research Council (FP7 NANOMAP and ERC-2013-CoG NeuroMolAnatomy) and from the Deutsche Forschungsgemeinschaft (DFG) Cluster of Excellence Nanoscale Microscopy and Molecular Physiology of the Brain, as well as from DFG grants RI 1967 2/1, RI 1967 3/1, and SFB 889/A5. We acknowledge support by the DFG to V.H. (Exc-257-Neurocure and SFB 958/A01), H.U. (SFB 889), and M.K. (SFB 958/A11). Author contributions: B.G.W. prepared the synaptosomes and performed all immunoblot-ting experiments. K.K. performed the electron microscopy imaging and all neuromuscular junction imaging. C.S. per-formed the hippocampal culture imaging. S.T. performed the synaptosome imaging. B.R. generated the synapse model. S.J.K., G.A.C., and M.K. participated in the biochemis-try experiments. S.M. and H.U. designed and performed all mass spectrometry experiments. S.O.R., B.G.W., and V.H. designed the experiments. All authors analyzed the data and contributed to writing the manuscript.

SUPPORTING ONLINE MATERIALSwww.sciencemag.org/content/344/6187/1023/suppl/DC1Materials and MethodsFigs. S1 to S7Tables S1 to S3Movie S1References (35–46)

3 March 2014; accepted 6 May 201410.1126/science.1252884
































RESEARCH ARTICLE SUMMARY◥

ADVANCED IMAGING

Extended-resolution structuredillumination imaging of endocyticand cytoskeletal dynamicsDong Li, Lin Shao, Bi-Chang Chen, Xi Zhang, Mingshu Zhang, Brian Moses,Daniel E. Milkie, Jordan R. Beach, John A. Hammer III, Mithun Pasham,Tomas Kirchhausen, Michelle A. Baird, Michael W. Davidson, Pingyong Xu, Eric Betzig*

INTRODUCTION: Various methods of super-resolution (SR) fluorescence microscopy havethe potential to follow the dynamic nanoscaleinteractions of specific macromolecular assem-blies in living cells. However, this potential isoften left unfulfilled, either owing to themethod’s inability to follow these processesat the speeds dictated by nature or becausethey require intense light that can substan-tially perturb the very physiology one hopes tostudy. An exception is structured illuminationmicroscopy (SIM), which can image live cellsfar faster and with orders of magnitude lesslight than required for other SR approaches.However, SIM’s resolution is usually limitedto only a twofold gain beyond conventionaloptical microscopes, or ~100 nm with visiblelight.

RATIONALE: We endeavored to find ways toextend SIM to the sub-100-nm regime while re-taining, to the greatest extent possible, theadvantages that make it the preferred SRmeth-od for live-cell imaging. Our first solution usedan ultrahigh numerical aperture (NA) lens andtotal internal reflection fluorescence (TIRF) toachieve 84-nm resolution at subsecond acqui-sition speeds over hundreds of time points inmultiple colorsnear thebasalplasmamembrane.Our second exploited the spatially patternedactivation of a recently developed, reversiblyphotoswitchable fluorescent protein to reach45- to 62-nm resolution, also at subsecond ac-quisition, over ∼10 to 40 time points.

RESULTS: We used high-NA TIRF-SIM toimage the dynamic associations of cortical fila-

mentous actin with myosin IIA, paxillin, orclathrin, as well as paxillin with vinculin andclathrin with transferrin receptors. Thanks tothe combination of high spatial and temporalresolution, we were able to measure the sizesof individual clathrin-coated pits through theirinitiation, growth, and internalization.Wewere

also able to relate pit sizeto lifetime, identify andcharacterize localized hotspots of pit generation,and describe the interac-tion of actin with clathrinand its role in accelerat-

ing endocytosis. With nonlinear SIM by use ofpatterned activation (PA NL-SIM), we moni-tored the remodeling of the actin cytoskeletonand the dynamics of caveolae at the cell sur-face. By combining TIRF-SIM and PA NL-SIMfor two-color imaging, we followed the dynamicassociation of actin with a-actinin in expand-ing filopodia and membrane ruffles and char-acterized shape changes in and the transportof early endosomes. Last, by combining PANL-SIMwith lattice light sheetmicroscopy, weobserved, in three dimensions and across theentire volume of whole cells, the dynamics ofthe actin cytoskeleton, the fusion and fissionof mitochondria, and the trafficking of vesi-cles to and from the Golgi apparatus, each ataxial resolution fivefold better than that ofconventional widefield microscopy.In addition, through direct experimental

comparisons, we demonstrated that the reso-lution for our methods is comparable with orbetter than other SR approaches yet allowed

us to image at far higher speeds, andfor far longer durations. To under-stand why this is so, we developed adetailed theoretical model showingthat our methods transmit the infor-mation encoded in spatial frequenciesbeyond the diffraction limitwithmuchgreater strength than do other alter-natives and hence require far fewerphotons emitted from the specimen,using far less intense light.

CONCLUSION: High-NA TIRF-SIMand PA NL-SIM fill an unmet needfor minimally invasive tools to im-age live cells in the gap between the100-nm resolution traditionally as-sociated with SIM and the sub-60-nmregime of protein-specific structuralimaging served by single-moleculelocalization microscopy.▪

RESEARCH

944 28 AUGUST 2015 • VOL 349 ISSUE 6251 sciencemag.org SCIENCE

The list of author affiliations is available in the fullarticle online.*Corresponding author. E-mail: [email protected] this article as D. Li et al., Science 349, aab3500(2015). DOI: 10.1126/science.aab3500

Two approaches for improved live-cell imaging at sub-100-nm resolution. (Left) Association of corticalactin (purple) with clathrin-coated pits (green), the latter seen as rings (inset) at 84-nm resolution via acombination of total internal reflection fluorescence and structured illumination microscopy at ultrahighnumerical aperture (high-NA TIRF-SIM). (Right) Progression of resolution improvement across the actincytoskeleton of a COS-7 cell, from conventional, diffraction-limited TIRF (220-nm resolution), to TIRF-SIM(97-nm resolution), and nonlinear SIM based on the patterned activation of a reversibly photoswitchablefluorescent protein (PA NL-SIM, 62 nm resolution). (Left and right represent single frames from time-lapsemovies over 91 and 30 frames, respectively. Scale bars, 2 mm (left); 3 mm (right).

ON OUR WEB SITE◥

Read the full articleat http://dx.doi.org/10.1126/science.aab3500..................................................

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

Dong Li, Lin Shao, Bi-Chang Chen, Xi Zhang, Mingshu Zhang, Brian Moses, Daniel E. Milkie, Jordan R. Beach, John A. Hammer III, Mithun Pasham, Tomas Kirchhausen, Michelle A. Baird, Michael W. Davidson, Pingyong Xu, Eric Betzig*

INTRODUCTION: Various methods of su-perresolution (SR) fluorescence microscopy have the potential to follow the dynamic na-noscale interactions of specific macromolecu-lar assemblies in living cells. However, this po-tential is often left unfulfilled, either owing to the method’s inability to follow these process-es at the speeds dictated by nature or because they require intense light that can substan-tially perturb the very physiology one hopes to study. An exception is structured illumination microscopy (SIM), which can image live cells far faster and with orders of magnitude less light than required for other SR approaches. However, SIM’s resolution is usually limited to only a twofold gain beyond conventional opti-cal microscopes, or ~100 nm with visible light.

RATIONALE: We endeavored to find ways to extend SIM to the sub-100-nm regime while retaining, to the greatest extent pos-sible, the advantages that make it the pre-ferred SR method for live-cell imaging. Our first solution used an ultrahigh numerical aperture (NA) lens and total internal reflec-tion fluorescence (TIRF) to achieve 84-nm resolution at subsecond acquisition speeds over hundreds of time points in multiple colors near the basal plasma membrane. Our second exploited the spatially patterned ac-tivation of a recently developed, reversibly photoswitchable fluorescent protein to reach 45- to 62-nm resolution, also at subsecond acquisition, over ∼10 to 40 time points.

RESULTS: We used high-NA TIRF-SIM to image the dynamic associations of cor-tical filamentous actin with myosin IIA, paxillin, or clathrin, as well as paxillin with vinculin and clathrin with transfer-rin receptors. Thanks to the combination

of high spatial and temporal resolution, we were able to measure the sizes of individual clathrin-coated pits through their initiation, growth, and internalization. We were also able to relate pit size to lifetime, identify and characterize localized hot spots of pit

generation, and describe the interaction of actin with clathrin and its role in accelerating endocyto-sis. With nonlinear SIM by use of patterned acti-vation (PA NL-SIM), we monitored the remodel-

ing of the actin cytoskeleton and the dynam-ics of caveolae at the cell surface. By combin-ing TIRF-SIM and PA NL-SIM for two-color imaging, we followed the dynamic associa-tion of actin with α-actinin in expanding fi-lopodia and membrane ruffles and charac-terized shape changes in and the transport of early endosomes. Last, by combining PANL-SIM with lattice light sheet microscopy, we observed, in three dimensions and across the entire volume of whole cells, the dynamics of the actin cytoskeleton, the fusion and fission of mitochondria, and the trafficking of vesicles to and from the Golgi apparatus, each at axial resolution fivefold better than that of conventional widefield microscopy.

In addition, through direct experimental comparisons, we demonstrated that the reso-lution for our methods is comparable with or better than other SR approaches yet allowed us to image at far higher speeds, and for far longer durations. To understand why this is so, we developed a detailed theoretical model

showing that our methods transmit the information encoded in spatial frequencies beyond the diffrac-tion limit with much greater strength than do other alternatives and hence require far fewer photons emit-ted from the speci-men, using far less intense light.

C O N C L U S I O N : High-NA TIRF-SIM and PA NL-SIM fill an unmet need for mini-mally invasive tools to image live cells in the gap between the 100-nm resolution traditionally associat-ed with SIM and the sub-60-nm regime of protein-specific struc-tural imaging served by single-molecule localization micros-copy.

The list of author affiliations is available in the fullarticle online.*Corresponding author. E-mail: [email protected] this article as D. Li et al., Science 349, aab3500 (2015). DOI: 10.1126/science.aab3500


ADVANCED IMAGING











RESEARCH




ON OUR WEB SITE◥


on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m


ADVANCED IMAGING











RESEARCH




ON OUR WEB SITE◥


on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

on

Aug

ust 2

7, 2

015

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fro

m

Originally published 28 August 2015 in SCIENCE

Stay focused on what matters when doing live cell imagingEliminate focus drift and extend exposure times for small, dim, and live samples with the DeltaVision™ Elite high-resolution microscope.

Visit us at www.gelifesciences.com/deltavision to discover more.

See more today, know more tomorrow.

Amersham I ÄKTA I Cytell I Biacore I Whatman I Xuri*

www.gelifesciences.com/deltavision

GE and GE monogram are trademarks of General Electric Company.DeltaVision, *Amersham, ÄKTA, Cytell, Biacore, Whatman, and Xuri are trademarks of General Electric Company or one of its subsidiaries.DeltaVision products are for research use only- not for diagnostic use© 2015 General Electric Company—All rights reserved. First published Apr. 2015GE Healthcare UK Ltd, Amersham Place, Little Chalfont, Buckinghamshire, HP7 9NA, UK

K15082 04/2015

Two approaches for improved live-cell imaging at sub-100-nm resolution. (Left) Association of cortical actin (purple) with clathrin-coated pits (green), the latter seen as rings (inset) at 84-nm resolution via a combination of total internal reflection fluorescence and structured illumination microscopy at ultrahigh numerical aperture (high-NA TIRF-SIM). (Right) Progression of resolution improvement across the actin cytoskeleton of a COS-7 cell, from conventional, diffraction-limited TIRF (220-nm resolution), to TIRF-SIM (97-nm resolution), and nonlinear SIM based on the patterned activation of a reversibly photoswitchable fluorescent protein (PA NL-SIM, 62 nm resolution). (Left and right represent single frames from time-lapse movies over 91 and 30 frames, respectively. Scale bars, 2 µm (left); 3 µm (right).

Science Headquarters1200 New York Avenue, NWWashington, DC 20005 USA

http://www.sciencemag.org

microscopy now update: getting the most from your imaging · tips and tweaks for optimal...

Documents