C H A P T E R

14

Super-resolution Microscopy: A Comparisonof Commercially Available OptionsJennifer A. Thorley, Jeremy Pike, Joshua Z. Rappoport

The University of Birmingham, Birmingham, United Kingdom

INTRODUCTION

Fluorescence microscopy is currently one of the mostpowerful and versatile techniques available for biolog-ical studies. Fluorophore-labeled molecules are verybright and readily distinguishable from other back-ground signals, making it easy to obtain high-contrastimages. With the development of genetically encodedfluorescent proteins, it has become possible to imageprotein expression, localization, and activity in livingcells. However, optical microscopes have an inherentlimitation in spatial resolution because of the wavenature of light. In light microscopy, resolution is funda-mentally limited by the properties of light diffraction asfirst described by Ernst Abbe in 1873. This prevents theresolution of structures smaller than approximately halfthe wavelength of light and, as such, causes sharp point-like objects to appear blurry under a microscope.

Resolution can be quantified by analyzing the point-spread function (PSF) of a microscope. The PSF de-scribes how blurry a single point-like emitter (forexample, a single molecule or small fluorescent bead)will appear when diffracted through a microscope andthe full width at half maximum (FWHM) value of thePSF is a simple way to characterize resolution. Impor-tantly, resolution does not refer to the ability of a micro-scope to detect small structures; rather, it denotes theability to distinguish adjacent objects as separate struc-tures rather than as a single object (Figure 14.1). Howev-er, due to the same limitations the true size of objectssmaller than the PSF cannot be readily determined. Asmany structures in biological samples are smallerand/or closer together than the FWHM of the PSF, im-ages from a fluorescent microscope do not alwaysprovide a true representation of the sample beingvisualized.

Practically speaking, the resolution limit of the lightmicroscope depends on two main factors, the wave-length of light (l) and the numerical aperture (NA) ofthe objective lens. Visible light ranges between ultravio-let (<400 nm) and infrared (>800 nm), and the NA of anobjective refers to the light-collecting ability of the objec-tive lens. Most commercial objective lenses have a NAaround 1, and the practical upper limit for a NA isapproximately 1.5. Thus, applying a simplified equationbased on Abbe’s work (where the size of the finest detailthat can be resolved is d ¼ l/2NA), 500-nm light and anNA of 1 give d¼ 250 nm. This equation was later refinedby Lord Rayleigh in 1896 to give the Rayleigh criterion,defined as the shortest distance at which two point emit-ters can be distinguished as separate objects: R ¼ 0.61l/NA. The Rayleigh criterion is a commonly used measureof the width of the PSF and is shown in Figure 14.1. Thismeans that light diffraction limits the resolution of anoptical microscope to approximately half the wave-length of the light used, usually around 250 nm, andtherefore many fine cellular structures cannot beresolved.

As the de Broglie wavelength of an electron ismuch shorter than visible light, electron microscopyhas a much higher resolution than optical microscopyand has long been relied on to visualize cellular struc-tures smaller and/or closer together than 250 nm.However, fixation, dehydration, and thin sectioningare required during sample preparation for electronmicroscopy, making it technically challenging, proneto artefacts, and incompatible with live-cell imaging.Therefore, microscopic techniques that combine thenondestructive nature of optical microscopy and thenanometer resolution of electron microscopy havebeen the focus of much research and development inrecent years.

199Fluorescence Microscopy

Copyright © 2014 Elsevier Inc. All rights reserved.http://dx.doi.org/10.1016/B978-0-12-409513-7.00014-2

In optical microscopy, the combination of a high-magnification objective lens (e.g., 100�) and acharge-coupled device (CCD) camera with a small pixelsize (e.g., 6 mm) can provide a situation where singlediffraction-limited structures can be visualized acrossmultiple pixels (e.g., with an effective final pixel sizeof 60 nm in this example). Thus, through curve fitting,the center of these objects can be identified with accu-racy greater than the diffraction limit (Figure 14.1).However, with densely packed, brightly labeled diffrac-tion limited structures, often the case in biological sys-tems, this may not be possible, particularly if multiplefluorophores share the same pixel space. That beingsaid, this type of centroid identification remains a criticalstep in some super-resolution approaches (see SingleMolecule Localization Microscopy section, below).

Recent advances in fluorescence microscopy haveresulted in the development of a series of differentsuper-resolution techniques for breaking the diffractionbarrier inherent in light microscopy. This has beenachieved by choosing contexts in which Abbe’s lawdoes not apply, for example exploiting nearfield effects,selectively switching fluorescent dyes between on andoff states, and by localizing the centers of single fluoro-phores with high precision. Any microscopy techniquethat overcomes the resolution limit of conventional lightmicroscopy by at least a factor of two is generally consid-ered to be a super-resolution technique. The recent tech-nical innovation of super-resolution microscopy hasimproved the limits of optical resolution up to nearly

tenfold.1 Different super-resolution technologies areavailable, but they are generally built around conven-tional confocal or widefield fluorescence microscopesequipped with lasers and sensitive camerasdequipmentthat has already been used for fluorescence imaging andsingle-particle tracking experiments for decades.

Super-resolution microscopes, however, are notbased on one single technology, and several methodol-ogies have been developed independently over thepast several years for super-resolution fluorescencemicroscopy. These include structured illuminationmicroscopy (SIM), stimulated emission depletion(STED) microscopy, photoactivation localization mi-croscopy (PALM), and stochastic optical reconstruc-tion microscopy (STORM). Each of these techniquesis not only predicated on a different method for over-coming the diffraction limit but also has inherentadvantages and limitations when applied to differentbiological questions.

In this chapter, we emphasize new developments andapplications of super-resolution microscopy. This willprovide information about the physical basis for each so-lution, as well as a consideration of the practical con-cerns and relative benefits and limitations. Finally,variations on each will be described, as well as future di-rections in super-resolution microscopy, includingnewer emerging techniques. As STED microscopy,PALM/STORM, and structured illumination micro-scopy (SIM) instruments are now commercially avail-able, these are the focus of our discussion. Biologists

FIGURE 14.1 Definition of resolution and the limits of the ability to differentiate point sources. Point-like objects can be resolved by anoptical microscope if they are separated by the Rayleigh criterion but will not be resolved if they are closer together than this distance (<250 nm).

14. OPTIONS FOR SUPER-RESOLUTION MICROSCOPY200

who wish to utilize these techniques will need to makeinformed choices, as there are trade-offs between sensi-tivity, resolution, field and depth of view, speed, andprobe versatility within each technique (Table 14.1).

It should be noted that, owing to space constraintsand issues comparing the different commercially avail-able super-resolution solutions, only lateral (x,y) super-resolution is discussed in detail in this chapter.Although axial (z-plane) super-resolution techniqueshave been developed and applied, these are not coveredin any great detail.2e5

TECHNICAL CONSIDERATIONS

As mentioned above, the differences between thesuper-resolution technologies can influence how suitedan approach is to studying a specific biological ques-tion. Some super-resolution microscopes are not al-ways as user friendly as conventional opticalmicroscopes when it comes to imaging biological sam-ples. Currently, there is still no ideal system that offershigh-speed, three-dimensional nanometer spatial reso-lution with multicolor capabilities for live-cell imaging.Thus, the strengths and weaknesses of each techniquemust be considered with reference to the requirementsof the user and the particular sample to be imaged. As

each super-resolution technique has trade-offs in termsof resolution, speed, ease of use, etc., the three tech-niques discussed in this chapter (SIM, STED, andPALM/STORM) are quite likely to establish theirown application niches as they become more widelyavailable. Various factors must be carefully weighedwhen choosing a technique to use. A few importantquestions for those wishing to use super-resolutionmethods include: What resolution is sufficient toobserve the structure in question? How fast is the bio-logical process of interest? Is live-cell imaging requiredor can fixed specimens be used? Additional consider-ations when selecting a super-resolution technique forvisualization of a particular structure include the sizeand density of the structure, its location within thecell (for example, whether it is primarily in the plasmamembrane or internal), whether it can be labeled byexpression of a fluorescent-tagged protein or taggedwith an antibody or dye, and the signal-to-noise ratioachievable for this label. Some of the trade-offs andtechnical considerations are dealt with in more detailbelow.

Acquisition Speed

Many super-resolution techniques obtain increasedresolution at the cost of the speed of image acquisition.The trade-off between speed and resolution is typically

TABLE 14.1 Relative Benefits and Limitations of Different Super-resolution Techniques

TECHNICAL CONSIDERATIONS 201

more pronounced with single-molecule localizationtechniques such as PALM and STORM (see Single-Molecule Localization Microscopy section, below). Thisis because two molecules cannot be turned on withinthe same PSF at any given time, and multiple roundsof switching fluorophores on and off are required togenerate one complete super-resolution picture, limitingthe speed of the molecular read-out. As such, live-cellimaging is largely impossible with the single-moleculetechniques.

The speed of SIM image acquisition is similarlylimited by the need to record multiple frames in orderto generate one super-resolution image. However, phys-ical movement of the grating itself takes little time andlive-cell imaging with SIM is possible in some samples.6

STED is the fastest of the super-resolution techniques.However, because STED is a scanning microscopy tech-nique, its speed depends on the size of the field ofview; it is relatively fast when only small fields of vieware imaged, making it an excellent choice for video-rateimaging of small areas. As resolution increases, scanningofmore and smaller pixels is required, resulting in longerscanning times and high levels of photobleaching,limiting its applicability to live-cell imaging (althoughlive-cell STED is possible).

Fluorescent Probes

Like all fluorescence microscopy techniques, thechoice of fluorescent probes in super-resolution imagingis paramount, and each of the different techniques hasdifferent criteria for what makes an optimal fluorescentprobe. SIM is the only super-resolution technique forwhich no special fluorescent probes are required. Thesame fluorescent proteins, antibodies, and dyes usedfor conventional fluorescence microscopy techniquesare all applicable to the technique; however, photo-bleaching must be considered as multiple intermediateimages are required to generate one super-resolutionimage.

The single-molecule techniques such as PALM andSTORM require fluorescent probes whose state can becontrolled, either by reversibly or irreversibly switch-ing between a light and a dark state, or by changingfrom one wavelength to another. In both cases, theprobes used must be as bright as possible and requirea high contrast ratio between the two states. In theinitial configuration, STORM requires pairs of dyes;7

however, direct STORM (dSTORM) exploits the blink-ing phenomenon exhibited by certain dyes such asAlexa Fluor 647.8

With most commercially available systems, STED isnot achievable with all conventional fluorophores.This is because with a limited selection of depletionwavelengths there will be fluorophores that are actually

excited by the depletion beam and therefore cannot beused. This often means red fluorescent proteins anddyes are impractical, presenting a challenge for dual-color imaging, where green and red fluorophores aretypically used.

Photobleaching

Photobleaching is the light-induced destruction offluorophores, which can be a particular problem for im-aging biological samples using time-lapse studies forlong periods of time or where high laser powers arerequired, as is the case with STED microscopy. When afluorophore absorbs a photon, an electron becomesexcited from the ground state to an excited state.When fluorophores are in an excited state, they aremore likely to react with other molecules. Althoughthe mechanism of photobleaching is not fully under-stood for most molecules, it is generally thought that itinvolves electronic excitations to triplet states becausetriplet states have longer lifetimes and are more reactive.There are anti-photobleaching agents that reduce theamount of oxygen in the sample to prevent reactionswith oxygen, although many of these are toxic to livecells.

Photobleaching can be an issue in PALM/STORM asmultiple frames must be acquired to generate a singlesuper-resolution image and the user must decide thepoint at which to stop imaging after the majority of fluo-rescent molecules have become photobleached. Photo-bleaching also presents problems for SIM and STED,as these two techniques require saturated fluorescenceexcitation and depletion, respectively. For SIM, exces-sive photobleaching will result in loss of signal betweensubsequent images in the set of multiple frames requiredto make one super-resolution image. This can causeproblems in reconstruction of the super-resolution im-age. Some fluorophores are more photostable thanothers (e.g., Alexa Fluor 488 relative to FITC) and soare preferable when imaging times are longer or whenmultiple images must be acquired.

Spatial Resolution

The single-molecule localization techniques such asPALM and STORM offer the greatest improvement inspatial resolution when compared to conventional fluo-rescence microscopy. SIM offers the least improvementin spatial resolution of the super-resolution techniques,often only doubling the resolution achievable. ForSTED microscopy, the achievable resolution is stronglydependent on the photostability of a sample. For biolog-ical samples, a resolution of 50 to 100 nm can be obtainedwithout causing significant damage, which falls some-where between the high spatial resolution offered by

14. OPTIONS FOR SUPER-RESOLUTION MICROSCOPY202

the single molecule localization techniques and thelimited improvement offered by SIM.

SINGLE-MOLECULE LOCALIZATIONMICROSCOPY

Single-molecule localization microscopy (SMLM)techniques such as photoactivation localization micro-scopy (PALM) and stochastic optical reconstruction mi-croscopy (STORM) differ from all other fluorescenceimaging techniques in that an image is built up, as thename suggests, literally molecule by molecule. Asmentioned previously, with sparse diffraction-limitedfluorescent objects, centroid fitting can be employed toprovide more accurate localization than the diffractionlimit would normally allow. However, with multiple flu-orophores sharing the same pixel space, as is typical inlabeled biological samples, this technique breaksdown. Thus, SMLM methods can be used to limit thenumber of active fluorophores in a sample within anysingle image acquisition.

The SMLM techniques work on the principle that,although thew250-nm limit of resolution in light micro-scopy (Figure 14.1) prevents the separation of two ob-jects at distances of less than 250 nm, the centers ofindividual objects can be determined with nanometerprecision.9,10 However, by inhibiting the fluorescenceemission of the majority of the labels at any one time,only single isolated molecules within a PSF are allowedto fluoresce at a given time. By stochastically switchingon and off different single isolated molecules in subse-quent camera recordings, a final image withsubdiffraction-sized spatial resolution is reconstructedfrom the summation of all localized spatial positions.

Cells can therefore be imaged at nanometer resolutionby determining the exact location of each fluorophoreone by one. All single-molecule localization microscopytechniques rely on this temporal separation of fluores-cence emission, which is achieved either by switchingbetween a dark and fluorescent state or by consecutivebinding of individual fluorophores to the structure.This is achieved by the application of photoconvertibleor photoactivatable dyes and proteins. The emissionwavelength of photoconvertible dyes or fluorescent pro-teins (FPs) can be optically converted from one wave-length to another or fluorescence can be turned on andoff in the case of photoactivatable proteins. This princi-ple has been published by two independent groupswho named the technique PALM and STORM,respectively.

Both PALM1 and fluorescent PALM (fPALM)11 usegenetically expressed photoactivatable fluorescentprobes. Under normal conditions, EosFP, the first fluo-rescent protein to be imaged by PALM, emits green

fluorescence at 516 nm.12 Upon illumination at around400 nm, a photo-induced break in the peptide backboneadjacent to the chromophore occurs, causing the fluoro-phore to emit in the yellow region. If the number of con-verted fluorophores is small, the proteins emitting in theyellow region will, on average, be well separated andcan be imaged with high resolution. When this first sub-set of EosFPs becomes photobleached, another subset ofEosFPs can be converted and imaged, and this process iscycled thousands of times until the entire population offluorescent proteins becomes photobleached and acqui-sition is stopped.

Developed in the laboratory of Jennifer Lippincott-Schwartz at the National Institutes of Health, fPALMwas originally predicated on the development of thephotoactivatable green fluorescent protein (PA-GFP).13

Although a detailed description of GFP and the specificphysiochemical properties of PA-GFP are outside thescope of this chapter, both originate from the jellyfishAquaria victoria and, as the names imply, fluoresce greenwhen illuminated with blue light. In brief, when initiallyexpressed PA-GFP is essentially nonfluorescent. How-ever, when illuminated with an ultraviolet, or nea-ultraviolet (e.g., 405 nm) pulse, PA-GFP enters a brightstate in which it can be excited, as with normal GFP,with 488-nm light. Through the precise calibration ofthe activation pulse and the excitation illumination,the user can cycle a small proportion of the PA-GFP ina sample from dark to bright to photobleached as inPALM. The photobleaching step is technically not essen-tial but aids in preventing the same fluorophores frombeing imaged multiple times. This reduces both errorin fluorophore number measurements and the numberof frames that will ultimately have to be acquired togenerate the final image.

Because only a few fluorophores in each frame arevisualized, large numbers of images are required.Thus, the iterative nature of PALM means that, evenwith a high-speed electron multiplier CCD (EM-CCD),it can often take several minutes to acquire all the infor-mation required to reconstruct a single image. Whenconsidering that subdiffraction limit localization is thegoal of PALM, this introduces a significant potentialfor artefacts generated by even small amounts of sampledrift during acquisition. For this reason, fiducialmarkers, such as small fluorescent beads, are oftenadded to samples to be imaged by PALM or otherSMLM techniques, and a final image registration algo-rithm is employed during post-acquisition processing.

Many other photoconvertible, photoactivatable, andphotoswitchable fluorescent proteins have been devel-oped. Photoactivatable FPs can be activated from adark state to a bright state using ultraviolet light, andphotoswitchable FPs such as Dronpa can be cycled be-tween light and dark states with specific illumination

SINGLE-MOLECULE LOCALIZATION MICROSCOPY 203

wavelengths.14 A probe that switches only once is ad-vantageous when quantifying the absolute number offluorescent molecules present within a sample, as eachwill only be counted once. However, in practice, mostorganic dyes can be reversibly photoswitched, resultingin the localization of individual molecules occurringmultiple times within an acquisition. In addition, a sin-gle fluorescent protein will also appear as a cluster ofslightly different localizations owing to variable inter-vals of blinking before irreversible photobleaching even-tually occurs.

Another similar SMLM method is STORM, whichwas developed by Xiaowei Zhuang of the HowardHughes Medical Institute (HHMI) at Harvard Univer-sity.7 In theory and practice, there are many similaritiesbetween PALM and STORM, although STORM useschemical fluorophores rather than fluorescent proteins.These are typically photoswitchable pairs of cyanine(Cy) dyes coupled to antibodies to act as reporter andactivator pairs in order to cycle multiple times betweenthe dark and light states (e.g., Cy3eCy5 pairs). In directSTORM (dSTORM),8 several synthetic dyes, such asAlexa Fluor 647, can also be used in a blinking modeto achieve a similar effect. In the original version ofSTORM, pairs of fluorophores are used which can beswitched on and off by different laser lines. The physio-chemical nature of this phenomenon is not easilyunderstood and will not be detailed in this chapter.Otherwise, the iterative nature of the STORM imageacquisition process is quite similar to that of PALMand, generally speaking, microscopes that have beendeveloped for PALM can be used for STORM and viceversa.

Despite the technical differences between PALM andSTORM, they can both localize molecules of interestwith w20 nm resolution, depending on the number of

photons collected, and are often known as the single-molecule techniques. In both techniques, subsets of flu-orophores are switched on with a brief laser pulse,which is so weak that only a few molecules are stochas-tically switched on at a time, resulting in such a low den-sity of activated molecules that overlap withindiffraction limited resolution is unlikely. Imaging ofthe on fluorophores is performed until all activated mol-ecules are bleached, and the process is repeated untilseveral hundreds of thousands of molecules are imaged.This process is summarized in Figure 14.2. Once the po-sitions of all fluorophore molecules are identified, asuper-resolution image with resolution of up to w20nm can be reconstructed.

Benefits and Limitations

The chemical fluorophores employed in STORMmust be posttranslationally linked to proteins of interestusually through indirect methods such as immunocyto-chemistry. This introduces other potential concerns suchas those regarding sample preparation (see below).However, distances between the target protein and thefluorophore can also raise problems. Antibodies arenearly 15 nm in length, and considering that in indirectimmunofluorescence a primary antibody specific to theantigen protein of interest and a fluorophore-tagged sec-ondary antibody are employed, this can result in a rela-tively large distance between the fluorophore and target.In comparison, most fluorescent proteins have acompact barrel-shaped structure (GFP has a diameterof approximately only 4 nm). Thus, even if short peptidelinkers are placed between the protein of interest and thefluorescent protein tag for PALM imaging, antibodiesused in STORM can potentially introduce a significantlocalization error. Furthermore, the requirement of a

FIGURE 14.2 The process of super-resolution image generation by SMLM techniques. Through iterative identification of the centroids ofindividual fluorophores, super-resolution images can be reconstructed.

14. OPTIONS FOR SUPER-RESOLUTION MICROSCOPY204

dual fluorophore probe for conventional STORM addsother technical concerns. However, the technique ofdSTORM, which is generally considered to be photo-chemically equivalent to the phenomenon of groundstate depletion (GSD), more recently has been gainingprominence. In this case, a single fluorophore (usuallyAlexa Fluor 647) is shifted between dark and brightstates, thus simplifying methodological concerns(Figure 14.3).

In PALM, a cDNA construct of a chimeric fusion pro-tein directly tagging the protein of interest must be intro-duced into the cell to be imaged. Although fluorescentproteins are generally innocuous tags that do not affecttarget protein structure, function, or localization, theseare potential concerns. In addition, possible overexpres-sion artefacts much be considered, as in conventionalmicroscopy studies.

In SMLM techniques, it is possible to achieve opticalresolution so high that the labeling density, the numberof fluorophores per target protein, becomes the limitingfactor in improving resolution. In fact, labeling density iscritical to the ultimate resolution achievable and shouldnot be too great or too sparse. Moreover, as the opticalresolution continues to improve toward around 10 nmand beyond, the physical size of the antibodies or fluo-rescent proteins may start to become the limiting factorfor resolution.

The trade-off between speed and resolution in super-resolution microscopy is typically most pronouncedwith the SMLM techniques. The speed of acquisition is

limited because two molecules cannot be turned onwithin the same PSF at any given time. This makesimaging live cells extremely challenging, as the struc-tures within live cells move and gain and lose moleculesover time. As such, live-cell, single-molecule super-reso-lution imaging requires that both spatial and temporalsampling are fast enough to avoid the image becomingblurred.

Additionally, as the SMLM techniques build up im-ages literally molecule by molecule, the user mustconsider how many of these fluorescent molecules aremissed or even artificially added during image acquisi-tion. In addition, the user must decide when to stop theacquisition by judging when a significant proportion offluorophores are bleached, which can introduce error.Moreover, in PALM, an unknown fraction of the fluores-cent proteins are actually photoswitchable and the levelof replacement of endogenous protein with taggedproteins influence the fraction of fluorescent proteinspresent.

Single-molecule localization microscopy also facesthe problem that if too many fluorescent molecules aremissed because a low number of photons are emitted,the resulting image will be incomplete and give an incor-rect image reconstruction rather than simply generatinga low quality but accurate image. As such, it is not clearif localization techniques can always be used for accu-rate protein quantification. In contrast, overlabelingcan result in images that are improperly rendered owingto the presence of partially overlapping centroids that

FIGURE 14.3 Example of successful dSTORM imaging. Immunolocalized CD81 labeled with an Alexa Fluor 647 conjugated secondaryantibody was visualized first by TIRF microscopy and then through direct stochastic optical reconstruction microscopy (dSTORM), using a NikonnSTORM microscope.

SINGLE-MOLECULE LOCALIZATION MICROSCOPY 205

cannot be resolved. Thus, although extremely powerful,SMLM is not suitable for all applications, and other tech-niques that are more rapid or that can be used with allfluorophores have been developed.

Data Processing

As described above, once the series of sparsely popu-lated individual frames has been acquired, techniquessuch as Gaussian fitting can be applied to generate amap of fluorophore centroids, although more recentlya significantly faster wavelet approach has also beenapplied.15 Thus, when summed together, these individ-ual processed frames provide a super-resolution image.Localization microscopic techniques generate imagesthat are different to those we are used to seeing as mi-croscopists. Our eyes tend to look for patterns and tofocus on the largest and brightest structures in a micro-scopic image. However, STORM and PALM images aregenerated mathematically, and the size and brightnessof the “dots” in the image represent user-selected pa-rameters. There are currently two mathematical ap-proaches commonly used to describe the distributionof molecules in two dimensions: Ripley’s K functionanalysis compares the actual distribution relative to arandom distribution and assigns a degree of non-randomness to each molecule to create a cluster map,16

whereas pair-correlation analysis determines the proba-bility of finding a molecule at a given distance fromanother molecule compared with the probability ex-pected from a random distribution of molecules.17

Thus, especially when considering that commerciallyavailable software does not provide the user with a greatdeal of detail into the precise algorithms employed, caremust be taken when interpreting and quantifying recon-structed SMLM images.

Application to 3D and Multicolor Imaging

Three-dimensional fPALM has been achieved using atechnique known as biplane detection.18 A beam-splittersplits the fluorescence light into a shorter and longerpath to form two detection planes for axial positiondetermination. Additionally, three-dimensional STORMhas been achieved by introducing astigmatism to theimage using a cylindrical lens. Images above and belowthe focal plane are ellipsoidal, meaning that the positionof the fluorophore can be monitored by examiningthe ellipticity of the circular fluorophore.3 Two-colorPALM has also been implemented. For example,COS-7 cells tagged with transferrin receptor (TfR)-PAm-Cherry1 and PA-green fluorescent protein (PAGFP)-Clathrin light chain (CLC) were alternately imaged at561 nm and 468 nm to excite the red (PAmCherry1)and green (PAGFP) fluorescent labels.19 Multicolor

STORM has also been demonstrated. Microtubuleswere imaged alongside clathrin-coated pits, usingCy2eAlexa Fluor 647 and Cy3eAlexa Fluor 647 to labelthe microtubules and clathrin, respectively.20 Laserwavelengths of 457 and 532 nm were used to selectivelyexcite the two dye pairs, making it possible to image themicrotubules separately from the clathrin-coated pitswith w30-nm spatial resolution. Thus, the inherentpower of SMLM is being extended to a variety of newspecific applications. However, other super-resolutiontechniques have been developed that do not sufferfrom the same limitations as SMLM.

STRUCTURED ILLUMINATIONMICROSCOPY

Developed collaboratively between the laboratoriesof John Sedat at the University of California, San Fran-cisco, and the late Mats Gustafsson at HHMI JaneliaFarms, structured illumination microscopy (SIM) alsorelies upon computational approaches to generate asuper-resolution image from a series of acquisitions(Figure 14.4). However, as the name suggests, ratherthan relying on fluorophore effects, SIM utilizes a struc-tured pattern of illumination light to excitate the wholefield.21 When a grid with high spatial frequency is pro-jected onto a sample, fluorophore emission is blurredwhen detected. When this excitation pattern mixeswith the spatial pattern of the sample, an interferencepattern called a moire fringe is produced, which ismuch coarser than either pattern alone. Moire fringesare simply interference patterns created when lightpasses through regular structures and patterns (shownin Figure 14.4).

If the pattern is moved across the specimen, a charac-teristic signal variation in the fluorescence response canbe observed as a function of time and grid position. Ifone of the patterns is an unknown structure (the mole-cules in the sample being imaged) and the other is aknown pattern (the grid), the corresponding moirefringes will contain information about the unknown

FIGURE 14.4 Generation of moire fringes by overlaying grids

with different patterns. The rotation of two diffraction gratings gen-erates the moire fringes involved in structure illumination microscopy(SIM).

14. OPTIONS FOR SUPER-RESOLUTION MICROSCOPY206

structure. When the known patterns have higher spatialfrequencies than the unknown pattern, the techniquewill offer an improvement in spatial resolution. Thismeans that, by analyzing the signal variations betweenimages, structural features can be resolved that wouldnot be visible by regular microscopy.21 Because struc-tures that have a parallel orientation to the grid wouldnot benefit from this effect, the grid must be not onlyshifted but also rotated between images to generate a se-ries of several images of the same sample that havedifferent moire fringes. Thus, in order to restore the sub-resolution information from the moire fringes, it isnecessary to acquire several image frames, each usinga different illumination pattern. SIM typically recordsframes at three different pattern orientations and threeto five different positions, generating a total of 9 to 15frames per final super-resolution SIM image.22 As theillumination pattern is known, this information can beused to generate the super-resolution image.

Data Processing

This method of sequential data collection requiresthat the final super-resolution image be reconstructedfrom the raw frames after data collection. Typically, areconstruction algorithm working in the Fourier domainis used to distinguish sharp image components fromsignal oscillations to generate a super-resolution imagefrom the frames taken.21 This algorithm requires thatthe images acquired have high-contrast elements, mean-ing that significant levels of out-of-focus fluorescencecan result in a lack of information from which to recon-struct a true super-resolution image. This issue dependson the thickness of the sample being imaged and thelocalization of the molecules within the sample. Thiscan represent a particular problem for overexpressedproteins tagged with fluorescent proteins such asGFP, where high levels of out-of-focus light are oftenobserved.

This effect can be minimized through the utilizationof illumination via total internal reflection fluorescence(TIRF) microscopy. Although a detailed description ofTIRF is outside the context of this chapter, TIRF usesoblique illumination to selectively excite fluorophoreswithin a thin (approximately 200-nm) plane above thecoverslip to which the sample is adhered.23 SMLM canalso utilize TIRF illumination, and the flexibility of beingable to use either TIRF or epi-illumination is one of theshared benefits of these two techniques. However, inthe case of SIM, if the target protein is not within200 nm of the coverslip and is surrounded above andbelow by fluorophores, proper reconstruction of asuper-resolution image can be difficult to achieve.

Once generated, the super-resolution nature of a SIMimage can be directly validated by inspection of the

Fourier transform (Figure 14.5). Thus, although artefactscan be difficult to discern in SMLM, in SIM clear quanti-tative criteria can be employed to determine whetherreconstruction was successful. However, it should benoted that most commercial SIM and SMLM systemsprovide the ability to acquire corresponding conven-tional non-super-resolution images which can permitdirect comparisons and serve as a guide to validatethat the reconstructions generated reasonablelocalizations.

Benefits and Limitations

SIM is arguably one of the most user friendly super-resolution techniques, particularly in terms of samplepreparation, as any label used in conventional fluores-cence microscopy can be applied to the technique. Inaddition, conventional excitation routines are applied,resulting in less photobleaching than some of the othertechniques (although multiple images of the same sam-ple must be acquired to generate a super-resolution im-age). SIM is typically a widefield approach, meaningthat fast CCD cameras can be used. However, the imag-ing rate is typically slower than that of conventional mi-croscopy because of the need to acquire several imagesfrom different grating positions. SIM allows resolutionsof around 100 nm in x, y, and z when biological materialis used,5 and as such is the least powerful super-resolution technique regarding gained resolution over

FIGURE 14.5 Examples of successful and unsuccessful SIM im-

aging. Immunolocalized CD81 labeled with an Alexa Fluor 488 con-jugated secondary antibody was successfully visualized by structuredillumination microscopy (SIM) and verified by the width and charac-teristic “petal” shaped of the Fourier transform. GFP-tagged FGFRwasnot successfully visualized by SIM, as demonstrated by the smallerwidth and conventional shape of the Fourier transform.

STRUCTURED ILLUMINATION MICROSCOPY 207

conventional fluorescence microscopy. This is because,unfortunately, the spatial frequencies that are opticallycreated in SIM are themselves limited by diffraction.As such, SIM is generally considered to be limited tothis factor-of-two improvement, as it is limited by thePSF of conventional microscopy.21

Application to Live-Cell, 3D, and MulticolorImaging

Multicolor SIM can be easily realized as long as appro-priate lasers and correctly aligned diffraction gratings areavailable. Furthermore, SIM is currently the most widelyused super-resolution technique for live-cell imaging andhas been applied to a broad range of biological studies,including, for example, live-cell 2D SIM imaging ofmicro-tubules.6 A difficulty encountered with live-cell SIM isthat 9 to 15 recorded frames are required to image thesample. If the object in question moves even slightly dur-ing image acquisition, artefacts are created that canprevent reconstruction of the final super-resolution im-age from the individual frames. 3D-SIM imaging is notlimited to regions of interest at the coverslip and canimage as far as 10 mm into the sample, and 3D SIM imag-ing of the nucleus has been used to demonstrate that nu-clear pore complexes are adjoined by channels in both thenuclear lamin and peripheral heterochromatin.5 Howev-er, as stated above, fluorescence signal above or below theplane of focus can limit the ability to successfully recon-struct a SIM image.

STIMULATED EMISSION DEPLETIONMICROSCOPY

Both SMLM and SIM employ widefield (either epi orTIRF) illumination and collection through a CCD(or more commonly an EM-CCD), but stimulated emis-sion depletion (STED) microscopy is based on confocallaser scanning microscopy (CLSM) (Figure 14.6). In alaser-scanning confocal microscope, a laser beam isfocused by an objective lens into a small focal spotwithin a specimen. An image is then acquired pointby point, by scanning the specimen, using pinhole op-tics to ensure that only a single focal plane is visualizedat a time.

The physical basis for STED microscopy is the gener-ation of an illumination beam with an effective diameterthat is smaller than the diffraction limit.24 The techniqueuses a conventional focused laser beam to stimulate anarea of fluorescent molecules that can be several hun-dred nanometers in diameter. Before the spontaneousemission of fluorescence occurs (within only a few nano-seconds), a second red-shifted doughnut-shaped beamilluminates the sample, depleting the emission of the

fluorophore outside the central region. Because thedepletion beam could also be capable of exciting the flu-orophore, it should not be too close to the absorptionband. This second beam, known as the STED, or deple-tion, beam (which in most commercial systems is at 592nm), forces probemolecules from their excited electronicstate back to their ground state by stimulating the emis-sion of a photon of the same wavelength. A bandpass fil-ter excludes these photons, meaning that only theshorter wavelength photons from molecules within thecenter of the doughnut (which have not been quenched)will be collected. Both the excitation and depletionbeams are pulsed lasers, and although they are synchro-nized, the depletion laser pulse is temporally extendedrelative to the excitation beam, thus generating “de-excitation” outside the central spot. By overlapping thetwo beams, fluorescence is only allowed from the centerof the spot, essentially switching off a subset of fluoro-phores and therefore generating a much smaller excitingfocal spot, as shown in Figure 14.6. As a smaller effectivePSF is generated, this causes less blurring and results inhigher resolution. Generally, 100% depletion of thespontaneous emission by the STED beam is not achiev-able, but 90 to 95% depletion will produce an imagewith an acceptable contrast ratio. Using this principle,up to 10-fold increases in resolution in one dimensionhave been achieved.

Increasing the intensity of the doughnut-shapedSTED beam dramatically improves the resolutionachievable by this technique. This switches off fluoro-phores even at the inner ring of the doughnut, furthersharpening the center fluorescent spot to a size muchsmaller than the diffraction limit. Typically this re-quires laser powers around 1000 times that used in con-ventional confocal microscopy. Although this meansthat there is no theoretical resolution limit to STED im-ages, in practice photo-damage caused to biologicalsamples by high laser intensities usually sets the inten-sity limit on the depletion beam for biologicalapplications.

FIGURE 14.6 Generation of an effective PSF smaller than the

diffraction limit using a STED depletion beam. Combination of theco-centered excitation and depletion beams creates the phenomenon ofstimulated emission depletion (STED), which effectively images onlythe region inside of the doughnut-shaped depletion beam.

14. OPTIONS FOR SUPER-RESOLUTION MICROSCOPY208

Benefits and Limitations

STED is generally realized as an addition to conven-tional CLSM, increasing the user friendliness to thosealready familiar with this widespread technique. Thescan times for STEDmicroscopy are quite similar to thoseof CLSM, apart from an increase proportional to thesmaller effective spot size of the scanning laser. If a con-ventional CLSM can scan a 512 � 512 image in approxi-mately 1 second with a diffraction limited laser spot, aSTED beam tuned to approximately 60 nm would takeonly about 10 seconds to generate an image. However,as with SMLM, there are fluorophore limitations withSTED microscopy (see below), and, as with conventionalCLSM, a great deal of emitted light is rejected and neverreaches the detector. Thus, STEDmicroscopy can be diffi-cult to employ with weakly fluorescent samples and/orfluorophores that are not particularly photostable.

Although compatible with a large number of fluo-rescent proteins and chemical dyes, not all fluoro-phores will successfully display STED. In order to bedepleted, a dye should have significant emission inthe range of the depletion beam wavelength (e.g.,592 nm). Furthermore, a critical consideration is thatthe fluorophore is not excited by the depletion beam(Figure 14.7). Thus, as most commercial systems aresupplied with a 592-nm depletion beam, many conven-tional red fluorophores cannot be used for STED imag-ing. This can create particular issues with multicolor

studies, as a combination of green and red fluorophoresis often employed. However, many alternative fluoro-phores have been identified that are compatible withmulticolor STED.

Further concerns about the addition of a STED laserarise from the increased photobleaching of the fluores-cent label. This derives from the fact that the STED laseracts on the excited fluorescent label. Light from thedepletion beam is not absorbed by the label and doesnot produce any photoreactive and thus phototoxic spe-cies. Rather, stimulated emission shortens the time thefluorophore spends in its excited electronic state andcan reduce photoreactions such as photobleaching.Although excited-state absorption of STED light can un-fortunately cause severe photobleaching,25,26 it wasshown that fast scanning and the right choice of STEDwavelength can minimize these effects.26e29

The use of fast beam scanners has established STEDmicroscopy as the fastest super-resolution imaging tech-nique available, with recording times of up to 60 to 80frames per second for observation areas of a fewmicronsin size.30

One key advantage of this technique is that STEDmicroscopy gives instant gratification with a what-you-see-is-what-you-get image similar to conventionalconfocal microscopy, and it requires no data process-ing after acquisition. However, as STED is a scanningmicroscopy technique, its speed depends on the sizeof the field of view. The technique is relatively fastwhen only small fields of view are imaged, making itan excellent choice for video-rate imaging of smallareas. As resolution increases, scanning of more andsmaller pixels is required, meaning longer scanningtimes and high levels of photobleaching, limiting itsapplicability to live-cell imaging. However, live-cellSTED has been achieved in several cases as detailedbelow.

Application to Live-Cell, 3D, and MulticolorImaging

The potential phototoxic effects of the added STEDbeam were long believed to be incompatible with thestudy of living cells. (Although the excitation laser po-wer is comparable to that used in conventionalconfocal microscopy, the STED laser power can bearound 1000 times higher.) This has introduced theneed to pay attention to heat absorption and potentiallight-induced toxic reactions. However, live-cell video-rate imaging of synaptic vesicles (28 Hz) has beensuccessfully demonstrated by streamlining theinstrumentation and limiting the area imaged to justw2.5 � 1.8 mm.30

Stimulated emission depletion microscopy can beapplied to three-dimensional as well as multicolor

FIGURE 14.7 Examples of successful and unsuccessful STED

imaging. Immunolocalized caveolin1 labeled with an Alexa Fluor 488conjugated secondary antibody was successfully visualized by stim-ulated emission depletion (STED) microscopy. Immunolocalized CD81labeled with an Alexa Fluor 568 conjugated secondary antibody wasnot successfully visualized by STED microscopy due to excitation ofthe fluorophore by the depletion beam.

STIMULATED EMISSION DEPLETION MICROSCOPY 209

imaging. The doughnut-shaped STED beam used intwo-dimensional STED works well to deplete the spon-taneous fluorescence emission in the lateral direction,but it offers no improvement in resolution in the axial di-rection. It is possible to combine the improved lateralresolution described above with improved axial resolu-tion if an alternatively shaped STED beam is employed.This STED beam includes axially shifted lobes thatquench the axial extension of the PSF.2,4

The complexity of the STED system limits fluoro-phore choices and can make multicolor imaging diffi-cult. Multicolor imaging represents a challenge toSTED microscopy because two laser wavelengths arerequired for each dye: a conventional excitation laserbeam and a red-shifted doughnut-shaped STED beam.This means that for dual-color imaging, four laser beamswith four different wavelengths would be needed,which is not only technically challenging but couldalso easily produce undesired interference and createeven greater photobleaching issues. Nevertheless, multi-color STED imaging has been achieved in somecases.27,31 To simplify two-color imaging, combiningone fluorophore with a second fluorophore that has asimilar emission spectrum but a large Stokes shift be-tween excitation and emission can allow single STEDlaser to be used for both dyes, which simplifies the tech-nique somewhat.32 This approach has also been used forlive-cell STED microscopy.33 Moreover, several recentpublications have shown great promise in simplifyingthe instrument and increasing the availability of addi-tional laser lines.28,34,35

The small effective spot size of STED microscopy alsocan be exploited to measure nanoscale dynamics in fluo-rescence correlation spectroscopy (FCS). This hasallowed the direct observation that sphingolipids andglycosylphosphatidylinositol-anchored proteins aretrapped in cholesterol-mediated sub-20-nm complexesin plasma membranes, supporting the existence of lipidrafts or other microdomains.36

SUMMARY

Among the three main commercially available super-resolution techniques, various relative benefits and lim-itations exist (Table 14.1), although all require significantattention to issues such as labeling and system align-ment. Although SMLM provides the highest spatial res-olution, it also takes the longest time to acquire thenecessary data, is conducted in fixed cells, and is onlycompatible with certain fluorophores. SIM can be usedwith any fluorophore and is relatively rapid, making ituseful for live-cell imaging studies, but it does not pro-vide the same resolution gains as SMLM or STED.Furthermore, SIM can be difficult to perform

successfully with samples that have fluorophores aboveand below the plane of focus. SMLM and SIM share thebenefits of being widefield techniques that are compat-ible with both epi and TIRF illumination, but they alsoshare the limitation that image reconstruction via appli-cation of computational algorithms is required. Thus, asin some deconvolution techniques, direct image acquisi-tion is not possible, creating a concern for potential arte-facts that cannot easily be discerned. STED is a variant ofCLSM and is therefore subject to some of the same ben-efits and limitations of that technique. STED generallyprovides intermediate resolution gains between that ofSMLM and SIM and is not compatible with all fluoro-phores; however, STED can be employed, as CLSM, tooptically section samples with high fluorophore densitywithin the sample volume. Furthermore, STED is rela-tively rapid as only a single image is acquired, and asa direct visualization technique no reconstruction isrequired.

Among the three main commercially available super-resolution solutions, no single technique can be said tobe optimal for all applications. Depending upon the spe-cific questions being addressed and the constraints ofthe experimental system, the user will need to decidewhich is best. Of course, considering the expense associ-ated with super-resolution systems (each can easily costin the vicinity of US$500,000), it may not be practical foreach researcher, or even institution, to invest in all threetypes of systems, so some sort of strategic planningneeds to be performed before a system is selected forpurchase.

FUTURE DIRECTIONS AND EMERGINGTECHNIQUES

Although SMLM, SIM, and STED microscopy repre-sent the main commercially available super-resolutionsolutions, several other techniques, or variants of thesetechniques, have been developed. Furthermore, special-ized applications of these techniques have recentlybegun to emerge, and the future will certainly see theemergence of novel microscopy technologies employinginnovative methods to break the diffraction barrier.Additionally, in certain cases, three-dimensional andmulticolor versions of the existing technologies havehad to make use of significant developments to be real-ized. Other variants of these techniques exist, including,as mentioned above, live-cell TIRF-based SIM and morerecently STED compatible with deep penetration intotissue, even in living mice.37 Furthermore, recent devel-opments in computational approaches can providemorerapid SIM reconstruction from fewer images.38 Thiscould mean that reconstruction could be performed dur-ing and not after acquisition. Several innovative SMLM

14. OPTIONS FOR SUPER-RESOLUTION MICROSCOPY210

variants have been developed that employ alternativestrategies to permit iterative illumination of selectivepopulations of fluorophores. These have includedreversible pH quenching and the newly developed tech-nique of universal point accumulation imaging in nano-scale topography (uPAINT), which involves theaddition of fluorophores to the sample while imaging.39

uPAINT is particularly useful with fluorescent ligandsthat will bind to extracellular domains of plasma mem-brane proteins in living cells; however, uPAINT requirescareful fluorophore titration.

Other types of SMLMexist, including relatively simpleoptions that make use of certain inherent properties ofparticular fluorophores. These include techniques suchasBayesian analysis of blinking and bleaching (3B),whichsimply tracks these events that can be rare enough tooccur only once per pixel during a specific time frame.40

However, this can be relatively inefficient and, like otherSMLM techniques, subject to labeling density issues.Furthermore, this raises a concern regarding otherSMLM techniques such as PALM and STORM, asinherent blinking of a single activated or converted fluo-rophores could be scored asmultiple independent fluoro-phores, leading to reconstruction and/or fluorophorecounting artefacts. Therefore, some SMLM algorithmshave been developed with blinking correction functions.

Additionally, single particle tracking (SPT) in live cellscan be realized with SMLM.41 In this case, a small num-ber of fluorophores are selectively visualized throughphotoactivation or photoconversion and then trackedutilizing very high speed acquisition. This has thebenefit over conventional SPT that each molecule of in-terest can be tracked and can be verified to be a singlefluorophore without extensive validation procedures.Furthermore, the entire procedure can be repeatedly per-formed on individual cells. However, individual trackscan be short owing to fluorophore photobleaching.

Finally, entirely new techniques for achieving super-resolution are being generated. One exciting solutioncurrently in development under the commercial nameBioaxial involvesCLSMwith the illumination laserpassedthrough a series of polarizing Pockels cells and then into athin biaxial crystal. The effect of this is, in some ways, ahybrid between SIM and STED. A CCD is employedupon which an iterative series of three to five differentdiffraction patterns are projected at each point. This tech-nique can generate resolution gains between SIM andSTED, uses very low light levels, can be performedwithout high magnification or high-NA objectives, andis compatible with all fluorophores. However, a largenumber of images is generated (e.g., 5 � 512 � 512) fora single scan. Thus, although with rapid exposuretimes acquisition can be relatively quick, this results in alarge amount of data to be processed, rendering this tech-nique currently computationally challenging.

In conclusion, this is a very exciting, and confusing,time for super-resolution microscopy. Commercial mi-croscope companies have rapidly responded to user in-terest with a series of different solutions, each withparticular benefits and limitations. It can be difficultfor potential adopters to decipher which techniquesand technologies are appropriate for their particularapplications. Concerns such as fluorophore compati-bility, speed of acquisition, and, of course, resolutionmust all be balanced (Table 14.1). Furthermore, for thefacility manager, issues such as cost, flexibility, anduser friendliness must also be considered. Thus, twothings are clear: (1) no one commercially availablesuper-resolution microscope is the ideal solution for allexperimental questions, and (2) more techniques willbe developed. Finally, as with all innovations, limita-tions must be appreciated and not all investigationsrequire super-resolution imaging. Thus, care needs tobe taken to avoid the unnecessary purchase and use ofthese powerful, and often complicated, microscopes.

Acknowledgments

The authors would like to thank Professor Jane McKeating for hercontribution of reagents and expertise. The authors would like toacknowledge assistance from Leica Microscopy, Nikon Microscopy,and the Nikon Imaging Centre at Kings College London in obtainingimages. The authors would also like to thank Dr. Natalie Poulter forassistance with image acquisition, as well as Dr. Patrina Pellett for crit-ically reviewing this chapter.

References

1. Betzig E, Patterson GH, Sougrat R, Lindwasser OW, Olenych S,Bonifacino JS, et al. Imaging intracellular fluorescent proteins atnanometer resolution. Science 2006;313:1642e5.

2. Harke B, Ullal CK, Keller J, Hell SW. Three-dimensional nano-scopy of colloidal crystals. Nano Lett 2008;8(5):1309e13.

3. Huang B, Wang W, Bates M, Zhuang X. Three-dimensionalsuperresolution imaging by stochastic optical reconstructionmicroscopy. Science 2008;319(5864):810e3.

4. Klar TA, Jakobs S, Dyba M, Egner A, Hell SW. Fluorescencemicroscopy with diffraction resolution limit broken by stimu-lated emission. Proc Natl Acad Sci USA 2000;97:8206e10.

5. Schermelleh L, Carlton PM, Haase S, Shao L, Winoto L, Kner P,et al. Subdiffraction multicolor imaging of the nuclear peripherywith 3D structured illumination microscopy. Science 2008;320:1332e6.

6. Kner P, Chhun BB, Griffis ER, Winoto L, Gustafsson MG. Super-resolution video microscopy of live cells by structuredillumination. Nat Methods 2009;6(5):339e42.

7. Rust MJ, Bates M, Zhuang X. Sub-diffraction-limit imaging bystochastic optical reconstruction microscopy (STORM). Nat

Methods 2006;3(10):793e5.8. Heilemann M, van de Linde S, Schuttpelz M, Kasper R,

Seefeldt B, Mukherjee A, Tinnefeld P, Sauer M. Subdiffraction-resolution fluorescence imaging with conventional fluorescentprobes. Angew Chem Int Ed Engl 2008;47(33):6172e6.

FUTURE DIRECTIONS AND EMERGING TECHNIQUES 211

9. Gelles J, Schnapp BJ, Sheetz MP. Tracking kinesin-driven move-ments with nanometre-scale precision. Nature 1988;331(6155):450e3.

10. Yildiz A, Forkey JN, McKinney SA, Ha T, Goldman YE,Selvin PR. Myosin V walks hand-over-hand: single fluorophoreimagingwith 1.5-nm localization. Science 2003;300(5628):2061e5.

11. Hess ST, Girirajan TP, MasonMD. Ultra-high resolution imagingby fluorescence photoactivation localizationmicroscopy. BiophysJ 2006;91(11):4258e72.

12. Wiedenmann J, Ivanchenko S, Oswald F, Schmitt F, Rocker C,Salih A, et al. EosFP, a fluorescent marker protein withUV-inducible green-to-red fluorescence conversion. Proc Natl

Acad Sci USA 2004;101(45):15905e10.13. Patterson GH, Lippincott-Schwartz J. A photoactivatable GFP

for selective photolabeling of proteins and cells. Science 2002;297(5588):1873e7.

14. Habuchi S, Ando R, Dedecker P, Verheijen W, Mizuno H,Miyawaki A, et al. Reversible single-molecule photoswitchingin the GFP-like fluorescent protein Dronpa. Proc Natl Acad Sci

USA 2005;102(27):9511e6.15. Kechkar A, Nair D, Heilemann M, Choquet D, Sibarita JB. Real-

time analysis and visualization for single-molecule basedsuperresolution microscopy. PLoS ONE 2013;8(4):e62918.

16. Hsu CJ, Baumgart T. Spatial association of signaling proteins andF-actin effects on cluster assembly analyzed via photoactivationlocalization microscopy in T cells. PLoS ONE 2011;6(8):e23586.

17. Sengupta P, Jovanovic-Talisman T, Lippincott-Schwartz J. Quan-tifying spatial organization in point-localization superresolutionimages using pair correlation analysis. Nat Protoc 2013;8(2):345e54.

18. Juette MF, Gould TJ, LessardMD,Modzianoski MJ, Nagpure BS,Bennett BT, et al. Three-dimensional sub-100 nm resolution fluo-rescence microscopy of thick samples. Nat Methods 2008;5(6):527e9.

19. Subach FV, Patterson GH, Manley S, Gillette JM, Lippincott-Schwartz J, Verkhusha VV. Photoactivatable mCherry for high-resolution two-color fluorescence microscopy. Nat Methods

2009;6(2):153e9.20. Bates M, Huang B, Dempsey GT, Zhuang X. Multicolor

superresolution imaging with photo-switchable fluorescentprobes. Science 2007;317:1749e53.

21. Gustafsson MG. Surpassing the lateral resolution limit by a fac-tor of two using structured illumination microscopy. J Microsc

2000;198(Pt 2):82e7.22. Gustafsson MGL, Shao L, Carlton PM, Wang CJR,

Golubovskaya IN, Cande WZ, et al. Three-dimensional resolu-tion doubling in wide-field fluorescence microscopy by struc-tured illumination. Biophys J 2008;94:4957e70.

23. Mattheyses AL, Simon SM, Rappoport JZ. Imaging with total in-ternal reflection fluorescence microscopy for the cell biologist. JCell Sci 2010;123(Pt 21):3621e8.

24. Hell SW, Wichmann J. Breaking the diffraction resolution limitby stimulated emission: stimulated-emission-depletion fluores-cence microscopy. Opt Lett 1994;19(11):780e2.

25. Eggeling C, Ringemann C, Medda R, et al. Direct observationof the nanoscale dynamics of membrane lipids in a living cell.Nature 2008;457:1159e62.

26. Hotta J, Fron E, Dedecker P, et al. Spectroscopic rationale for effi-cient stimulated-emission depletion microscopy fluorophores. JAm Chem Soc 2010;132:5021e3.

27. Donnert G, Keller J, Wurm CA, Rizzoli SO, Westphal V,Schonle A, et al. Two-color far-field fluorescence nanoscopy. Bio-phys J 2007;92(8):L67e9.

28. Moneron G, Medda R, Hein B, Giske A, Westphal V, Hell SW.Fast STED microscopy with continuous wave fiber lasers. Opt

Express 2010;18(2):1302e9.29. Rankin BR, Moneron G, Wurm CA, Nelson JC, Walter A,

Schwarzer D, et al. Nanoscopy in a livingmulticellular organismexpressing GFP. Biophys J 2011;100(12):L63e5.

30. Westphal V, Rizzoli SO, Lauterbach MA, Kamin D, Jahn R,Hell SW. Video-rate far-field optical nanoscopy dissects synapticvesicle movement. Science 2008;320(5873):246e9.

31. Meyer L, Wildanger D, Medda R, Punge A, Rizzoli SO,Donnert G, et al. Dual-color STED microscopy at 30-nm focal-plane resolution. Small 2008;4:1095e100.

32. Schmidt R, Wurm CA, Jakobs S, Engelhardt J, Egner A, Hell SW.Spherical nanosized focal spot unravels the interior of cells. Nat

Methods 2008;5(6):539e44.33. Pellett PA, Sun X, Gould TJ, Rothman JE, Xu MQ, Correa Jr IR,

et al. Two-color STED microscopy in living cells. Biomed Opt Ex-

press 2011;2(8):2364e71.34. Rankin BR, Hell SW. STED microscopy with a MHz pulsed

stimulated-Raman-scattering source. Opt Express 2009;17(18):15679e84.

35. Wildanger D, Buckers J, Westphal V, Hell SW, Kastrup L. ASTED microscope aligned by design. Opt Express 2009;17(18):16100e10.

36. Eggeling C, Ringemann C, Medda R, Schwarzmann G,Sandhoff K, Polyakova S, Belov VN, Hein B, vonMiddendorff C, Schonle A, Hell SW. Direct observation of thenanoscale dynamics of membrane lipids in a living cell. Nature

2009;457(7233):1159e62.37. Berning S, Willig KI, Steffens H, Dibaj P, Hell SW. Nanoscopy in

a living mouse brain. Science 2012;335(6068):551.38. Orieux F, Sepulveda E, Loriette V, Dubertret B, Olivo-Marin JC.

Bayesian estimation for optimized structured illuminationmicroscopy. IEEE Trans Image Process 2012;21(2):601e14.

39. Giannone G, Hosy E, Levet F, Constals A, Schulze K,Sobolevsky AI, et al. Dynamic superresolution imaging ofendogenous proteins on living cells at ultra-high density. BiophysJ 2010;99(4):1303e10.

40. Burnette DT, Sengupta P, Dai Y, Lippincott-Schwartz J, Kachar B.Bleaching/blinking assisted localization microscopy for super-resolution imaging using standard fluorescent molecules. ProcNatl Acad Sci USA 2011;108(52):21081e6.

41. Manley S, Gillette JM, Patterson GH, Shroff H, Hess HF,Betzig E, et al. High-density mapping of single-molecule trajec-tories with photoactivated localization microscopy. Nat Methods

2008;5(2):155e7.

14. OPTIONS FOR SUPER-RESOLUTION MICROSCOPY212