Superresolution imaging for neuroscience

Experimental Neurology 242 (2013) 33–40

Contents lists available at SciVerse ScienceDirect

Experimental Neurology journal homepage: www.elsevier.com/locate/yexnr

Review

Superresolution imaging for neuroscience Jan Tønnesen, U. Valentin Nägerl ⁎ Université Bordeaux Segalen, Interdisciplinary Institute for Neuroscience, UMR 5297, 146 rue Léo Saignat, 33077 Bordeaux, France CNRS, Interdisciplinary Institute for Neuroscience, UMR 5297, 146 rue Léo Saignat, 33077 Bordeaux, France

a r t i c l e

i n f o

Article history: Received 17 October 2011 Revised 4 September 2012 Accepted 4 October 2012 Available online 11 October 2012 Keywords: Live-cell superresolution microscopy Diffraction limit Fluorescence nanoscopy Dendritic spines Synaptic plasticity STED PALM STORM SIM

a b s t r a c t The advent of superresolution fluorescence microscopy beyond the classic diffraction barrier of optical microscopy is poised to transform cell-biological research. A series of proof-of-principle studies have demonstrated its vast potential for a wide range of applications in neuroscience, including nanoscale imaging of neuronal morphology, cellular organelles, protein distributions and protein trafficking. This review introduces the main incarnations of these new methodologies, including STED, PALM/STORM and SIM, covering basic theoretical and practical aspects concerning their optical principles, technical implementation, scope and limitations. In addition, it highlights several discoveries relating to synapse biology that have been made using these novel approaches to illustrate their appeal for neuroscience research. © 2012 Elsevier Inc. All rights reserved.

Contents Introduction . . . . . . . . . . . . . . . . . . . . . A new wave of imaging . . . . . . . . . . . . . . . . STED microscopy . . . . . . . . . . . . . . . . . . . PALM/STORM . . . . . . . . . . . . . . . . . . . . Structured illumination microscopy (SIM) . . . . . . . Applications in neuroscience . . . . . . . . . . . . . Nanoscale imaging of synaptic proteins and synaptic STED imaging of synapse morphology . . . . . . . PALM imaging of actin dynamics inside synapses . Perspectives . . . . . . . . . . . . . . . . . . . . . Acknowledgments . . . . . . . . . . . . . . . . . . References . . . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . . . . . . . . . . . . . . vesicles . . . . . . . . . . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

Introduction Fluorescence microscopy is one of the most powerful and widely used imaging techniques in neuroscience research, owing to the fact that it allows to visualize dynamic processes inside living cells with exquisite sensitivity and specificity.

⁎ Corresponding author at: Université Bordeaux Segalen, Interdisciplinary Institute for Neuroscience, UMR 5297, 146 rue Léo Saignat, 33077 Bordeaux, France. E-mail address: [email protected] (U.V. Nägerl). 0014-4886/$ – see front matter © 2012 Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.expneurol.2012.10.004

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

. . . . . . . . . . . .

33 34 34 37 37 38 38 39 39 39 39 39

A series of technological developments played together to facilitate the ascendancy of modern fluorescence microscopy, including laser and detector technology, fluorescent probes and molecular biology. As a result it is becoming increasingly possible to study brain function at the single-cell level under realistic conditions inside intact nervous tissue preparations. Milestones were the development of confocal microscopy in the 1980s and two-photon microscopy in the 1990s. Subsequently, the green-fluorescent protein (GFP) revolution allowed for labeling of specific proteins and organelles inside living cells, and increases in computing power helped deal with large sets of imaging data.

34

J. Tønnesen, U.V. Nägerl / Experimental Neurology 242 (2013) 33–40

More recently, marking a major breakthrough, the classic limit of spatial resolution for fluorescence microscopy, called the diffraction barrier, was overcome. The received wisdom was that the spatial resolution of light microscopy is fundamentally limited by the diffraction of light, and that the smallest structures that could be faithfully resolved were on the order of half the wavelength of the light used in the microscope, i.e. typically around 250 nm. This limit has been enshrined as a de facto physical law for over a hundred years (Abbe, 1873). If the diffraction barrier was indeed a hard limit, the study of cell-biological structures and processes occurring on the “mesoscale” of 10–200 nm would essentially remain out of reach for fluorescence microscopy, including macromolecular complexes, many cellular organelles and signaling inside nanodomains. In contrast to fluorescence microscopy, electron microscopy provides a spatial resolution down to a few nanometers, however, it requires tissue fixation, which is problematic in itself and which impedes understanding dynamic events. In addition, labeling of multiple proteins for electron microscopy is difficult and sampling of cellular volumes (3D reconstruction) is extremely labor-intensive. That is not to say that electron microscopy will now become superfluous, but certainly its future role in neuroscience will be redefined. The new methods to break this resolution barrier are generally referred to as superresolution microscopy or nanoscopy techniques. This review will go over the basic physical principles and practical implementation of their main incarnations, including stimulated emission depletion microscopy (STED), photo-activated localization microscopy (PALM)/stochastic optical reconstruction microscopy (STORM) and structured illumination microscopy (SIM). In addition, we will discuss their potentials and pitfalls for neuroscience research and highlight several recent applications in neurobiology.

A new wave of imaging Because of its wave-nature it is in fact impossible to focus light to an infinitesimally small spot. Rather, the smallest spot size that can be achieved by focusing lenses is limited by diffraction, which refers to the phenomenon whereby a wave tends to spread out as it travels through small openings (Born and Wolf, 1999). However, this does not mean that far-field optical microscopy, i.e. techniques that use focused visible light for image formation, must be limited by diffraction. By exploiting a strong non-linearity between the excitation light and the emitted fluorescence, the superresolution techniques can effectively break the classic diffraction limit (Hell, 2007), without actually getting rid of diffraction. Thanks to the new techniques, it is now possible to resolve details at the nanoscale (well below 100 nm) in biological specimens without forgoing the inherent benefits of fluorescence microscopy, such as live-cell imaging and bio-molecular labeling specificity. STED microscopy was the first concrete concept that broke the diffraction limit (Hell and Wichmann, 1994; Klar et al., 2000). Since then, other powerful techniques have been developed for nano-imaging of fluorescent samples, such as PALM (Betzig et al., 2006; Hess et al., 2006) and STORM (Bates et al., 2007), as well as non-linear SIM (Gustafsson, 2005; Heintzmann et al., 2002). These new superresolution techniques fall into two main categories, those based on single molecule switching and localization (PALM/ STORM), and those based on imaging dense ensembles of molecules using patterned illumination (STED/SIM). Because of differences in design and implementation, the techniques come with specific strengths and weaknesses in terms of temporal resolution, depth penetration, multi-color imaging, instrumentation requirements, practical handling etc. They all have in common that there is in theory no longer a hard resolution limit and it is possible to achieve a spatial resolution as high as a few nanometers under ideal conditions. However, in practice they are limited by signal noise (from drift

inherent in samples, particularly in living biological samples, detector noise, etc.) to some tens of nanometers. STED microscopy In confocal and two-photon laser scanning microscopy the excitation light is focused by the microscope's objective to a small focal spot that is systematically moved across the specimen in two or three spatial dimensions. Thus, images are reconstructed one pixel at a time by successive spatial sampling of the fluorescence signal. For high-quality imaging the scanning system must be very accurate and the size of the fluorescence spot used for scanning must be small relative to the specimen features to be visualized. Any jitter in the scanning will blur the image and thus degrade spatial resolution. However, even using a jitter-free scanner and perfectly aligned laser beams, the microscope's objective will not produce an infinitesimally small scanning spot, but rather a blurry intensity distribution, because of diffraction. It is the extent of this blurry spot, called the point-spread function (PSF), which defines the spatial resolution of the microscope. It is typically >250 nm wide in the focal plane (i.e. in x and y) for confocal microscopy and even wider for two-photon microscopy (>350 nm) because of the use of longer wavelength light. The core idea of STED microscopy is to improve the spatial resolution by quenching fluorescence emission on the outer edge of the PSF, so that emission can only occur from a small spot inside, which can be made much smaller than the diffraction limit (Figs. 1A–C). This is achieved by a second laser beam (called the STED beam), which can de-excite fluorescent molecules by stimulated emission at a wavelength that is longer than the fluorescence. By shaping the STED beam like a doughnut in the focal plane, it actively switches off the fluorescence around a circular rim of the PSF and thus only permits fluorescence to occur from the center of the PSF, which coincides with the center of the doughnut (called the null). By saturating the quenching process on the rim of the doughnut, a very steep spatial gradient for molecules that are either ‘on’ or ‘off’ is created, which underlies the gain in resolution for STED microscopy. As of now, a spatial resolution of 5.8 nm has been reported using diamond crystals, which is more than two orders of magnitude smaller than the wavelength of light that was used in the experiments (Rittweger et al., 2009). The classical doughnut-shaped STED PSF does not offer enhanced resolution in the z-axis. However, this can be achieved by shaping the STED beam using another phase mask in a way that delivers STED light above and below the focal plane, squeezing the PSF also along the optical axis (Wildanger et al., 2009). The maximal speed of STED microscopy is both determined by the imaging hardware and the brightness of the fluorescent sample. Much like confocal and two-photon microscopy there is a trade-off between temporal and spatial resolution, e.g. acquisition speeds up to a few kilohertz can be achieved in line-scan mode, while larger images can take up to several seconds. Using a fast scanning system based on a resonant mirror, STED imaging at video-rate could be performed on small scan areas (Westphal et al., 2008). As a rule of thumb the imaging speed for STED is slightly lower than for confocal or two-photon imaging due to the reduction in signal intensity in the center of the doughnut, which can be compensated by longer pixel dwell-time to collect more photons. Furthermore, the increase in resolution means that the pixel size must be decreased (to satisfy the Nyquist sampling theorem), reducing the field of view or increasing the image acquisition time accordingly. For example, given a five-fold reduction in pixel size, e.g. from 100 nm (confocal case) to 20 nm (for STED), the field of view decreases by a factor of 5 × 5 = 25. As STED microscopy uses two separate laser beams, one for fluorescence excitation and another for fluorescence quenching, it is more difficult to incorporate multi-color imaging than for conventional light microscopy. However, several solutions exist for two-color imaging with STED microscopy, relying either on separate lasers for each

J. Tønnesen, U.V. Nägerl / Experimental Neurology 242 (2013) 33–40

A

B

C

S1

D

z

STED

Fluoresc.

STED

Excitation

x,y

35

S0

E

F

G

PALM/STORM

I

III

II

I

J

SIM

H

Fig. 1. Basic principles of superresolution microscopy techniques. (A) Energy state diagram for STED microscopy. The absorption of a photon excites the molecule from the ground state (S0) to the excited state (S1) (blue arrow), from where it either drops back spontaneously to the ground state via fluorescence (green arrow) or is forcefully de-excited via stimulated emission (yellow arrow) by light of a longer wavelength. (B) Light intensity distributions, or point-spread functions (PSF), of the excitation (blue) and STED (orange) lasers in the focal plane (xy) and along the optical axis (z). (C) The doughnut-shaped STED laser quenches the fluorescence on the edge of the excitation PSF, effectively reducing the area from where the fluorescence can be emitted. Because the PSF of the STED beam is like multiple doughnuts piled on top of each other, the excitation PSF is not reduced along the z axis and hence no resolution enhancement is achieved here. (D) Gain in resolution achieved by STED over confocal mode. In STED, superresolved images are acquired optically and do not require any image processing. (E) PALM/STORM are techniques based on photo-switching and localization of individual fluorophores. The sample needs to be labeled with photo-switchable fluorophores. (F) First, a laser pulse (blue) stochastically activates a sparse subset of individual dye molecules, which are imaged onto a sensitive CCD camera as blurry spots. The center position of each spot can be precisely determined by mathematical fitting – well beyond the diffraction barrier. The activation – imaging – localization cycle is repeated thousands of times, each time addressing a different subset of fluorophores. (G) Superresolved images are reconstructed by plotting all fluorophore positions in a single image frame. (H) In structured illumination microscopy (SIM), high spatial frequencies in the fluorescent sample are recovered as beat patterns (or Moiré-fringes) of lower spatial frequencies, which the microscope can capture. (I) The fluorescent sample is excited by a patterned light of a known spatial frequency (stripes). The spatial frequencies in the fluorescence images are the product of the excitation pattern and the fluorophore distribution in the specimen. (J) A superresolved image is computed based on the spatial frequencies extracted from the fluorescence images. Superresolution is optically achieved, but requires image processing to be visualized.

fluorophore (Donnert et al., 2007; Meyer et al., 2008), measuring fluorescence life-times (Bückers et al., 2011), or photo-switchable fluorescent proteins (Willig et al., 2011). In addition, we have recently developed a simple method for two-color STED microscopy based on spectral unmixing that permits the use of popular green-yellow fluorescent labels such as YFP and GFP or Alexa Fluor 488 at the same time (Fig. 2D) (Tønnesen et al., 2011). As with any other light microscopy technique, tissue penetration of STED microscopy is limited by scattering and spherical aberrations. Several strategies exist to reduce both effects, which otherwise degrade spatial resolution. The use of two-photon light for excitation reduces scattering of the excitation spot and it was recently shown to be compatible with STED microscopy (Ding et al., 2009; Li et al., 2009; Moneron and Hell, 2009). Additionally, adaptive optics can be used to reduce aberrations, to ensure a high-quality STED doughnut. Using a glycerine-immersion objective with a correction ring to reduce aberrations, we demonstrated recently that it is possible to achieve sub-diffraction spatial resolution at imaging depths of around 100 μm below the surface of living brain slices

(Fig. 2F) (Urban et al., 2011). Recently, STED microscopy has been applied to obtain images of spines from the brain of an intact animal, illustrating the potential of the technique for in vivo applications (Berning et al., 2012). By definition, stimulated emission is used in a STED microscope to quench the fluorescence. However, other processes can be used as well to switch off the fluorescence on the rim of the doughnut and hence achieve superresolution. For instance, it was shown that photoswitching between excitable and non-excitable states also works, achieving comparable gains in spatial resolution (Grotjohann et al., 2011; Schwentker et al., 2007). The more general concept behind STED microscopy is called RESOLFT (reversible saturable optical fluorescence transition) (Hofmann et al., 2005), which states that basically any type of photo-switching can be used for superresolution microscopy. Similarly, the quenching beam does not have to have a doughnut intensity distribution. Other patterns of illumination, such as stripes or multiple doughnuts, can be used as well. However, they would require the use of a camera for image acquisition.

36

J. Tønnesen, U.V. Nägerl / Experimental Neurology 242 (2013) 33–40

2D View 3D-STORM

A

“Side” View

Bassoon Homer1

B

“Face” View

C

200 nm

D

E

STED in living brain slices

130 nm

F

1 µm

500 nm

G PALM _ fixed cels

H a

b c

1 µm

J

PALM _ live cels

I

200 nm

J. Tønnesen, U.V. Nägerl / Experimental Neurology 242 (2013) 33–40

PALM/STORM PALM/STORM refers to methods that are based on stochastic on/off switching of single fluorescent molecules and their computational localization in wide-field illumination. They rely on the same optical principle to achieve the gain in resolution; the difference is that PALM was initially developed using fluorescent proteins (Betzig et al., 2006; Hess et al., 2006), such as photo-activatable GFP (PA-GFP, Patterson and Lippincott-Schwartz, 2002), while STORM was developed using organic dyes such as cyanine (Rust et al., 2006). Either fluorophore type has unique advantages, e.g. fluorescent proteins can be genetically targeted in living cells, while the spectroscopic properties of organic dyes can be powerfully tweaked by organic chemistry. In PALM/STORM a superresolved image is constructed out of a large number of conventional wide-field images, each containing the positional information of different subsets of dispersed single fluorescent molecules. The imaging procedure thus follows these basic steps: 1) activation of a sparse set of dye molecules by low-intensity illumination using light of suitable wavelength; 2) excitation of activated fluorophores and localization of individual molecules by determining the centers of their images on a sensitive CCD camera; 3) deactivation of this set of dye molecules, usually by bleaching or photo-switching; 4) repetition of 100 to 10,000 times (Fig. 1F). In this manner the superposition of many pointillist maps yields a superresolved image of the sample. The basic idea and optical implementation are straightforward, but heavy imaging processing is needed to produce superresolved images. Even though these wide-field images just contain blurry diffractionlimited PSFs, the position of the individual fluorophores can be very precisely determined by mathematically estimating the center of each PSF, as long as they do not overlap with each other and can be discriminated. Thus, to acquire high-quality superresolved images the localization procedure must be very accurate and the density of localized fluorophores must adequately sample the features of the specimen. Due to the single molecule nature, PALM/STORM sets the current record in spatial resolution for fluorescence microscopy, achieving in biological samples a spatial resolution of 10–30 nm in the x–y plane. In STORM an axial resolution of around 35 nm has been achieved at the same time, using a cylindrical lens to produce an astigmatism, which provides information on the axial location of a given fluorophore (Huang et al., 2008). While temporal resolution used to be fairly low, advances in fluorophores and image processing algorithms have reduced acquisition times for 2D images down to a few seconds (Izeddin et al., 2011; Jones et al., 2011), improving PALM/STORM's suitability for imaging dynamical events inside cells. In addition, it has also become possible to rapidly (kilohertz) map out the movements of many (hundreds) of particles by combining single-particle tracking with PALM (sptPALM) (Manley et al., 2008), where a succession of superresolved images are created in a timelapse fashion. However, this is not imaging in the true sense of the word anymore, since only the trajectories of a limited number of particles are determined, rather than producing an entire image. As originally proposed using cyanine dyes, STORM is well-suited for multi-color imaging (Bates et al., 2007), because two cyanine dye molecules can be coupled together, where one acts as an activator of the other, so that it can be excited. This makes it possible to get color discrimination not just by using different activator dyes but also by

37

the timing of their activation, even though the same dye molecule is used for fluorescence labeling. Because of the combinatorial nature, many different colors can in theory be separated. However, it requires careful offline processing and subtraction of channel crosstalk originating from unintentional activation by pulses designated the complementary fluorophore pair, or from direct excitation of the fluorophore by the activation laser, which should not excite the molecule (Bates et al., 2007). Recently, crosstalk has been substantially reduced by performing multicolor imaging using the same activator dye and spectrally well separated reporter fluorophores (Dempsey et al., 2011). The original buffer solutions used in STORM contained relatively high concentrations of reducing agents (e.g. β-mercaptoethanol), which are toxic to cells, restricting the use of this superresolution imaging technique to fixed tissue preparations. However, using a modified buffer solution, STORM was recently successfully used for imaging live cells with superresolution in 3D (Jones et al., 2011). Because they are based on wide-field illumination, the singlemolecule switching techniques PALM/STORM have mostly been used for imaging monolayer cell cultures so far. However, reasonable depth penetration, e.g. imaging a few cell layers below the surface of a brain slice, should in theory be possible, if combined with other imaging modalities such as two-photon laser scanning to improve optical sectioning. In fact, using the method of two-photon temporal focusing, penetration depths of ~10 μm have been reached for PALM (Vaziri et al., 2008).

Structured illumination microscopy (SIM) In structured illumination microscopy (SIM) an image with two times the normal spatial resolution is computed from multiple wide-field raw images containing interference fringes (called Moiré patterns) caused by illuminating the sample with patterned light such as stripes, i.e. spatially non-uniform illumination, as opposed to spatially homogenous illumination normally used in wide-field microscopy (Gustafsson, 2000; Heintzmann and Cremer, 1999). The basic idea can be appreciated by thinking in terms of spatial frequencies representing the unknown sample structure, the high frequencies containing information on fine details of the sample structure. The higher spatial frequencies normally get filtered out by the microscope objective. However, when the specimen is illuminated by spatially varying (‘patterned’ or ‘structured’) excitation light, these spatial frequencies are effectively shifted to lower ones that can be resolved by the imaging system (Figs. 1H–J). This effect is similar to heterodyne detection, which is commonly used in physics. In this technique the signal of interest, which is usually time-varying, is multiplied by a (time-varying) reference signal. The mathematical product contains a new signal at a lower frequency (called beat frequency) that can be more easily detected given the bandwidth of the detector or the noise spectrum. By analogy, in SIM imaging the specimen structure represents the (spatially varying) signal of interest and the patterned illumination serves as the reference signal. The fluorescence signal represents the multiplicatively mixed readout, because it is proportional both to the excitation light intensity and the concentration of dye molecules. The high spatial frequency information of the specimen structure is thus shifted to a lower frequency range, which is covered by the bandwidth of the optical system.

Fig. 2. (A–C) Presynaptic Bassoon and postsynaptic Homer1 protein imaged inside synapses using 3D-STORM. Imaging of pre- and postsynaptic membrane proteins allows the width of the synaptic cleft to be estimated. Figures from Dani et al. (2011). (D) STED image of dendrites and axons labeled with GFP (green) and YFP (red), respectively, inside living organotypic brain slices. (E) STED image of dendritic spines loaded with Alexa Fluor 488 in living brain slices. The plot profile of a line through the spine neck is shown in the insert. (F) Lifeact-labeled dendritic spines deep inside living brain slice imaged by STED microscopy. The arrow points to an actin dense area under the spine base. Figures from Tønnesen et al. (2011) and Urban et al. (2011). (G) Widefield image of fixed actin-labeled dendrite. (H) Same frame as (G) but imaged by PALM. Figures taken from Izeddin et al. (2011). (I) Velocity map of actin filament dynamics (“treadmilling”) inside a living dendritic spine acquired by spt-PALM imaging. Red colors represent higher velocities. (J) Same frame as (I), yellow vectors indicate velocity map of actin molecules along actin filaments superimposed on PALM image of polymerized actin density (green) and widefield image of postsynaptic density protein PSD-95 (red). Scale vector: 200 nm/s. Figures taken from Frost et al. (2010).

38

J. Tønnesen, U.V. Nägerl / Experimental Neurology 242 (2013) 33–40

In principle, this technique is still limited by diffraction, even though it improves on the Abbé limit (which is based on homogenous illumination) by a factor of two. Consistently, images of cells with a resolution of 120 nm have been achieved, about twice the normal resolution (Schermelleh et al., 2008). This limitation comes from the fact that diffraction puts the same limit on spatial frequencies for illumination as for imaging, which means that the maximal shift to lower spatial frequencies that can be achieved cannot be greater than a factor of two. Confocal microscopy can in theory also achieve this resolution, provided a tiny pinhole is used in front of the detector. However, the penalty is that most of the signal, even from inside the focal spot, would be rejected by the pinhole. By comparison, SIM is more economical in terms of signal photons as it does not require a pinhole. Interestingly, the SIM technique has been combined with saturating illumination of the fluorophores (referred to as saturating or non-linear SIM), which improves the resolution effectively beyond the diffraction barrier, i.e. beyond the factor of two, down to about 50 nm (Gustafsson, 2005). Similar to STED microscopy, saturating SIM relies on a non-linearity in the fluorescence cycle to produce a gain in resolution beyond the diffraction limit. In this approach the intensity of the patterned illumination is increased to saturating levels, which effectively constricts the size of the non-illuminated areas that lie in between (Heintzmann et al., 2002). In this way steep spatial gradients between molecules that are fluorescent (switched on) and non-fluorescent (switched off) are set up on a spatial scale beyond the diffraction limit. Because of the wide-field nature of SIM, the temporal resolution for image acquisition can be reasonably high (on the order of 10 Hz), generally limited by the need to acquire multiple images of the beat frequency or Moiré patterns (typically 10 to 20 frames to reconstruct a given image) (Kner et al., 2009). SIM poses no major constraints on imaging multiple colors, and has been used for three-color imaging (Schermelleh et al., 2008). On the downside, wide-field illumination limits depth penetration of SIM. Moreover, to get meaningful images requires offline image processing, thereby preventing the experimenter from immediately viewing the images, which is an advantage of STED microscopy. The offline picture reconstruction in SIM requires considerable computing power, to process the large data files underlying each image, and therefore the 10 Hz acquisition rate does not translate into 10 Hz image processing and online visualization. Applications in neuroscience A central challenge for neuroscience is investigating the dynamic organization of synapses in the context of development, experiencedependent plasticity and diseases of the nervous system. Superresolution fluorescence microscopy opens up tremendous experimental opportunities in this regard, by making it possible to monitor the nanoscale dynamics of synapses in living neurons, non-invasively and over considerable time periods (>hours). While it is possible to identify and count presynaptic boutons and postsynaptic spines by diffraction-limited light microscopy, important structural and functional domains of synapses, e.g. spine necks and the postsynaptic density (PSD) are too small to be properly resolved by conventional techniques. The same problem holds true for axons, which are less than 200 nm in diameter, and for the synaptic cleft, which is around 30 nm wide (Harris and Kater, 1994; Mishchenko et al., 2010), and glial processes at synapses, which are similarly small (Ventura and Harris, 1999). Likewise, the topography of intracellular organelles, such as the endoplasmatic reticulum and mitochondria within synapses, and the nanoscale distribution and dynamics of synaptic proteins has essentially remained unexplored for lack of spatial resolution. Despite the attention they have received, the new superresolution techniques have not yet been adopted by many neuroscience groups, and the number of publications based on them is still small. Also, the

majority of the published work still mostly concerns methodological proof-of-principle. Nevertheless, several noteworthy discoveries in the area of synapse biology have already been made with them. For a more general overview of superresolution imaging in cell biology we refer to other recent reviews (Huang et al., 2010; Toomre and Bewersdorf, 2010). Nanoscale imaging of synaptic proteins and synaptic vesicles The organization of organelles and proteins within subsynaptic domains has been investigated by STED and single-molecule based superresolution imaging. While early superresolution studies only looked at protein distributions in fixed tissue preparations, dynamic events at synapses such as vesicle movements and protein trafficking have been investigated more recently as well. The fate of synaptic vesicles after exocytosis and fusion with the cell membrane was studied by STED microscopy of the vesicular protein synaptotagmin 1 in fixed hippocampal rat neurons. It was shown that synaptotagmin 1 remains clustered in the cell membrane after fusion of synaptic vesicles with the cell membrane, suggesting that a significant fraction of vesicles is recycled in an intact fashion as opposed to merging completely with the surrounding membrane (Opazo et al., 2010; Willig et al., 2006). Subsequently, the dynamics of single vesicles inside presynaptic boutons of living cultured neurons was studied using STED imaging at video-rate (28 Hz), revealing that vesicles move more rapidly outside of boutons than within boutons (Westphal et al., 2008). Also in living cultured rat hippocampal neurons, time-lapse STED microscopy demonstrated that newly endocytosed synaptic vesicles are initially highly mobile in the presynaptic bouton, before becoming immobilized within clusters of vesicles (Kamin et al., 2010). This report was complemented by a study using STED microscopy and electron microscopy on synaptosomes, describing how endocytosed vesicles are rapidly sorted and reorganized within the synapse (Hoopmann et al., 2010). Another early STED study resolved the spatial organization of the presynaptic protein Bruchpilot (BRP) at presynaptic active zones of Drosophila melanogaster neuromuscular junctions, revealing doughnut shaped assemblies that appear to facilitate vesicular release by clustering calcium channels near the presynaptic active zone (Kittel et al., 2006). In addition, BRP was shown to be part of the electron dense T-bars, which form part of active zones, and to be important for clustering Ca2+ channel at developing presynaptic active zones (Fouquet et al., 2009). Multi-color, three-dimensional STORM was used in fixed mouse brain sections to characterize the nanoscale distribution of several pre- and postsynaptic proteins and receptor subunits, such as Bassoon and Homer (Dani et al., 2011). This approach made it possible to resolve the relative spatial organization of these proteins in three dimensions by combining superresolved images taken at different imaging depths into a 3D image stack and using a cylindrical lens to localize single fluorophores along the z-axis (Figs. 2A–C). The study revealed an orderly alignment of the proteins along the longitudinal axis of the synapse, with receptors being localized at the periphery of the postsynaptic membrane, followed by anchoring proteins and other proteins of the postsynaptic density. Moreover, based on the superresolved maps of protein distributions, the authors estimated the size of the synaptic cleft to be around 20 nm, which is in range with the 30 nm reported by electron microscopy (Harris and Kater, 1994; Mishchenko et al., 2010). The spatial resolution was reported to be around 15 nm laterally and 35 nm axially, which is about 20-fold higher than what is achievable with diffraction-limited techniques. Together with the possibility for multi-color labeling, the 3D STORM approach represents a powerful alternative to electron microscopy for protein localization studies at synapses. The dimensions, variation and putative dynamics of the span of the synaptic cleft in live samples are still to be investigated.

J. Tønnesen, U.V. Nägerl / Experimental Neurology 242 (2013) 33–40

STED imaging of synapse morphology Because STED microscopy is an ensemble imaging technique, which means that many fluorophore molecules are excited at the same time within a diffraction-limited volume, it is well suited for imaging cellular structures that are volume-labeled with a diffusible dye. We have used STED microscopy to faithfully capture the morphology of dendritic spines in living brain slices using YFP as volume label (Nägerl et al., 2008). Recently, we extended this approach to imaging the dynamic distribution of actin inside dendritic spines with around 60 nm spatial resolution using lifeact to label actin (Fig. 2F) (Urban et al., 2011). Because lifeact is diffusible and binds actin with low affinity, the fluorescence signal is much less prone to bleaching as compared with approaches based on constructs of fluorescent proteins fused to actin monomers, where the label is essentially immobile and cannot replenish itself by diffusion. Using this approach, we observed that spine necks are dynamic structures that undergo size changes after the induction of chemical LTP (Urban et al., 2011). While increases in spine volume after the induction of LTP have been well documented by two-photon imaging (Lang et al., 2004; Matsuzaki et al., 2004), changes at the level of the spine neck have not been reported to date for lack of adequate spatial resolution. In combination with other opto-physiological approaches such as two-photon uncaging of glutamate and patch-clamp electrophysiology STED microscopy is bound to greatly facilitate the study of spines and their role for electrical and chemical compartmentalization of synaptic signals, which has been a long-standing controversy (Yuste, 2011). PALM imaging of actin dynamics inside synapses Combining PALM with single particle tracking (spt-PALM) makes it possible to acquire data with high spatial and temporal resolution. spt-PALM is basically a massively parallelized version of single particle tracking, where the trajectories of hundreds of single molecules are visualized and localized at the same time beyond the diffraction limit by mathematical fits of their centers of mass. This approach was used in several recent studies to image actin dynamics in spines of cultured neurons, either by tagging actin molecules with photo-activatable proteins (EosFP Tatavarty et al., 2009, or PA-GFP Frost et al., 2010, Figs. 2I–J) or using an actin-binding sequence similar to lifeact fused to a photo-activatable protein (tdEosFP Izeddin et al., 2011) (Figs. 2G–H). Beyond the single molecule information, this approach makes it also possible to obtain superresolved images of synapse morphology because the single molecule trajectories also outline the accessible volume of the synapse (Izeddin et al., 2011). While electron microscopy can provide beautiful images of filamentous actin structures inside fixed tissue (Korobova and Svitkina, 2010), these single molecule based imaging studies demonstrate that it is possible to resolve the directed flow of actin molecules along actin filaments in a dynamic setting, which is going to be necessary for unraveling the dynamic organization of actin structures and understanding their role for synapse function. Perspectives Importantly, the new methods should reproduce and validate previous studies based on traditional approaches such as electron microscopy and biochemistry. As described above, previous electron microscopy studies on the synaptic cleft dimensions agree well with reports on fixed tissue using the single-molecule based superresolution imaging techniques, and estimates of neck widths averaging 150 nm both by electron microscopy and live cell STED seem to validate the methods, setting the ground for future studies on the dynamics of these structures in live preparations.

39

Certainly, the resolving capability of these new techniques is wellestablished by now, and they have come of age in many ways. Nevertheless, to successfully transfer these technologies into neurobiology labs still represents a challenge. Also, it must be said that despite all the fancy technology, working with superresolution techniques sometimes feels like a step back in time as key advantages of conventional fluorescence microscopy (e.g. as life-cell compatibility, depth penetration, choice of fluorophores etc.) can be significantly compromised. But these drawbacks are fast getting overcome as the pressure from the user community grows and more technical innovations and workarounds are developed. Progress in superresolution methods hinges a lot on the development of new fluorophores and labeling strategies for them, concerning brightness, stability, genetic targeting, photo-activation etc. The classic trade-offs for conventional microscopy between temporal versus spatial resolution and speed versus photodamage will continue to hold for the new superresolution techniques, and define the frontier of method development. Nanoscopy techniques are likely to transform cell-biological research, providing new insights into dynamic processes inside living cells at the nanoscale, which was deemed a pipe dream just a few years ago. As many cellular functions are regulated on a scale that is just a little bit too small to be resolved reliably with conventional microscopy, the impact of these new techniques is expected to be broad and long-lasting. Acknowledgments We thank A. Panatier and E. Avignone for comments on the manuscript. The authors acknowledge funding from the Lundbeck Foundation, the Marie Curie Program (JT), and Inserm, ANR and HFSP (UVN). References Abbe, E., 1873. Beiträge zur Theorie des Mikroskops und der Mikroskopischen Wahrnehmung. Arch. Mikrosk. Anat. 9, 53. Bates, M., Huang, B., Dempsey, G.T., Zhuang, X., 2007. Multicolor super-resolution imaging with photo-switchable fluorescent probes. Science 317, 1749–1753. Berning, S., Willig, K.I., Steffens, H., Dibaj, P., Hell, S.W., 2012. Nanoscopy in a living mouse brain. Science 335, 551. Betzig, E., Patterson, G.H., Sougrat, R., Lindwasser, O.W., Olenych, S., Bonifacino, J.S., Davidson, M.W., Lippincott-Schwartz, J., Hess, H.F., 2006. Imaging intracellular fluorescent proteins at nanometer resolution. Science 313, 1642–1645. Born, M., Wolf, E., 1999. Principles of Optics: Electromagnetic Theory of Propagation, Interference and Diffraction of Light. Cambridge University Press, Cambridge; New York. Bückers, J., Wildanger, D., Vicidomini, G., Kastrup, L., Hell, S.W., 2011. Simultaneous multi-lifetime multi-color STED imaging for colocalization analyses. Opt. Express 19, 3130–3143. Dani, A., Huang, B., Bergan, J., Dulac, C., Zhuang, X., 2011. Superresolution imaging of chemical synapses in the brain. Neuron 68, 843–856. Dempsey, G.T., Vaughan, J.C., Chen, K.H., Bates, M., Zhuang, X., 2011. Evaluation of fluorophores for optimal performance in localization-based super-resolution imaging. Nat. Methods 8, 1027–1036. Ding, J.B., Takasaki, K.T., Sabatini, B.L., 2009. Supraresolution imaging in brain slices using stimulated-emission depletion two-photon laser scanning microscopy. Neuron 63, 429–437. Donnert, G., Keller, J., Wurm, C.A., Rizzoli, S.O., Westphal, V., Schonle, A., Jahn, R., Jakobs, S., Eggeling, C., Hell, S.W., 2007. Two-color far-field fluorescence nanoscopy. Biophys. J. 92, L67–L69. Fouquet, W., Owald, D., Wichmann, C., Mertel, S., Depner, H., Dyba, M., Hallermann, S., Kittel, R.J., Eimer, S., Sigrist, S.J., 2009. Maturation of active zone assembly by Drosophila Bruchpilot. J. Cell Biol. 186, 129–145. Frost, N.A., Shroff, H., Kong, H., Betzig, E., Blanpied, T.A., 2010. Single-molecule discrimination of discrete perisynaptic and distributed sites of actin filament assembly within dendritic spines. Neuron 67, 86–99. Grotjohann, T., Testa, I., Leutenegger, M., Bock, H., Urban, N.T., Lavoie-Cardinal, F., Willig, K.I., Eggeling, C., Jakobs, S., Hell, S.W., 2011. Diffraction-unlimited all-optical imaging and writing with a photochromic GFP. Nature 478 (7368), 204–208 (Sep 11). Gustafsson, M.G., 2000. Surpassing the lateral resolution limit by a factor of two using structured illumination microscopy. J. Microsc. 198, 82–87. Gustafsson, M.G., 2005. Nonlinear structured-illumination microscopy: wide-field fluorescence imaging with theoretically unlimited resolution. Proc. Natl. Acad. Sci. U. S. A. 102, 13081–13086. Harris, K.M., Kater, S.B., 1994. Dendritic spines: cellular specializations imparting both stability and flexibility to synaptic function. Annu. Rev. Neurosci. 17, 341–371. Heintzmann, F., Cremer, C., 1999. Lateral modulated excitation microscopy: improvement of resolution by using a diffraction grating. Proc. SPIE 3568, 11.

40

J. Tønnesen, U.V. Nägerl / Experimental Neurology 242 (2013) 33–40

Heintzmann, R., Jovin, T.M., Cremer, C., 2002. Saturated patterned excitation microscopy — a concept for optical resolution improvement. J. Opt. Soc. Am. A Opt. Image Sci. Vis. 19, 1599–1609. Hell, S.W., 2007. Far-field optical nanoscopy. Science 316, 1153–1158. Hell, S.W., Wichmann, J., 1994. Breaking the diffraction resolution limit by stimulated emission: stimulated-emission-depletion fluorescence microscopy. Opt. Lett. 19, 780–782. Hess, S.T., Girirajan, T.P., Mason, M.D., 2006. Ultra-high resolution imaging by fluorescence photoactivation localization microscopy. Biophys. J. 91, 4258–4272. Hofmann, M., Eggeling, C., Jakobs, S., Hell, S.W., 2005. Breaking the diffraction barrier in fluorescence microscopy at low light intensities by using reversibly photoswitchable proteins. Proc. Natl. Acad. Sci. U. S. A. 102, 17565–17569. Hoopmann, P., Punge, A., Barysch, S.V., Westphal, V., Buckers, J., Opazo, F., Bethani, I., Lauterbach, M.A., Hell, S.W., Rizzoli, S.O., 2010. Endosomal sorting of readily releasable synaptic vesicles. Proc. Natl. Acad. Sci. U. S. A. 107, 19055–19060. Huang, B., Wang, W., Bates, M., Zhuang, X., 2008. Three-dimensional super-resolution imaging by stochastic optical reconstruction microscopy. Science 319, 810–813. Huang, B., Babcock, H., Zhuang, X., 2010. Breaking the diffraction barrier: superresolution imaging of cells. Cell 143, 1047–1058. Izeddin, I., Specht, C.G., Lelek, M., Darzacq, X., Triller, A., Zimmer, C., Dahan, M., 2011. Super-resolution dynamic imaging of dendritic spines using a low-affinity photoconvertible actin probe. PLoS One 6, e15611. Jones, S.A., Shim, S.H., He, J., Zhuang, X., 2011. Fast, three-dimensional super-resolution imaging of live cells. Nat. Methods 8, 499–508. Kamin, D., Lauterbach, M.A., Westphal, V., Keller, J., Schonle, A., Hell, S.W., Rizzoli, S.O., 2010. High- and low-mobility stages in the synaptic vesicle cycle. Biophys. J. 99, 675–684. Kittel, R.J., Wichmann, C., Rasse, T.M., Fouquet, W., Schmidt, M., Schmid, A., Wagh, D.A., Pawlu, C., Kellner, R.R., Willig, K.I., Hell, S.W., Buchner, E., Heckmann, M., Sigrist, S.J., 2006. Bruchpilot promotes active zone assembly, Ca2+ channel clustering, and vesicle release. Science 312, 1051–1054. Klar, T.A., Jakobs, S., Dyba, M., Egner, A., Hell, S.W., 2000. Fluorescence microscopy with diffraction resolution barrier broken by stimulated emission. Proc. Natl. Acad. Sci. U. S. A. 97, 8206–8210. Kner, P., Chhun, B.B., Griffis, E.R., Winoto, L., Gustafsson, M.G., 2009. Super-resolution video microscopy of live cells by structured illumination. Nat. Methods 6, 339–342. Korobova, F., Svitkina, T., 2010. Molecular architecture of synaptic actin cytoskeleton in hippocampal neurons reveals a mechanism of dendritic spine morphogenesis. Mol. Biol. Cell 21, 165–176. Lang, C., Barco, A., Zablow, L., Kandel, E.R., Siegelbaum, S.A., Zakharenko, S.S., 2004. Transient expansion of synaptically connected dendritic spines upon induction of hippocampal long-term potentiation. Proc. Natl. Acad. Sci. U. S. A. 101, 16665–16670. Li, Q., Wu, S.S., Chou, K.C., 2009. Subdiffraction-limit two-photon fluorescence microscopy for GFP-tagged cell imaging. Biophys. J. 97, 3224–3228. Manley, S., Gillette, J.M., Patterson, G.H., Shroff, H., Hess, H.F., Betzig, E., LippincottSchwartz, J., 2008. High-density mapping of single-molecule trajectories with photoactivated localization microscopy. Nat. Methods 5, 155–157. Matsuzaki, M., Honkura, N., Ellis-Davies, G.C., Kasai, H., 2004. Structural basis of long-term potentiation in single dendritic spines. Nature 429, 761–766. Meyer, L., Wildanger, D., Medda, R., Punge, A., Rizzoli, S.O., Donnert, G., Hell, S.W., 2008. Dual-color STED microscopy at 30-nm focal-plane resolution. Small 4, 1095–1100.

Mishchenko, Y., Hu, T., Spacek, J., Mendenhall, J., Harris, K.M., Chklovskii, D.B., 2010. Ultrastructural analysis of hippocampal neuropil from the connectomics perspective. Neuron 67, 1009–1020. Moneron, G., Hell, S.W., 2009. Two-photon excitation STED microscopy. Opt. Express 17, 14567–14573. Nägerl, U.V., Willig, K.I., Hein, B., Hell, S.W., Bonhoeffer, T., 2008. Live-cell imaging of dendritic spines by STED microscopy. Proc. Natl. Acad. Sci. U. S. A. 105, 18982–18987. Opazo, F., Punge, A., Buckers, J., Hoopmann, P., Kastrup, L., Hell, S.W., Rizzoli, S.O., 2010. Limited intermixing of synaptic vesicle components upon vesicle recycling. Traffic 11, 800–812. Patterson, G.H., Lippincott-Schwartz, J., 2002. A photoactivatable GFP for selective photolabeling of proteins and cells. Science 297, 1873–1877. Rittweger, E., Han, K.Y., Irvine, S.E., Eggeling, C., Hell, S.W., 2009. STED microscopy reveals crystal colour xcentres with nanometric resolution. Nat. Photonics 3. Rust, M.J., Bates, M., Zhuang, X., 2006. Sub-diffraction-limit imaging by stochastic optical reconstruction microscopy (STORM). Nat. Methods 3, 793–795. Schermelleh, L., Carlton, P.M., Haase, S., Shao, L., Winoto, L., Kner, P., Burke, B., Cardoso, M.C., Agard, D.A., Gustafsson, M.G., Leonhardt, H., Sedat, J.W., 2008. Subdiffraction multicolor imaging of the nuclear periphery with 3D structured illumination microscopy. Science 320, 1332–1336. Schwentker, M.A., Bock, H., Hofmann, M., Jakobs, S., Bewersdorf, J., Eggeling, C., Hell, S.W., 2007. Wide-field subdiffraction RESOLFT microscopy using fluorescent protein photoswitching. Microsc. Res. Tech. 70, 269–280. Tatavarty, V., Kim, E.J., Rodionov, V., Yu, J., 2009. Investigating sub-spine actin dynamics in rat hippocampal neurons with super-resolution optical imaging. PLoS One 4, e7724. Tønnesen, J., Nadrigny, F., Willig, K.I., Wedlich-Soldner, R., Nägerl, U.V., 2011. Two-color STED microscopy of living synapses using a single laser-beam pair. Biophys. J. 101, 2545–2552. Toomre, D., Bewersdorf, J., 2010. A new wave of cellular imaging. Annu. Rev. Cell Dev. Biol. 26, 285–314. Urban, N.T., Willig, K.I., Hell, S.W., Nägerl, U.V., 2011. STED nanoscopy of actin dynamics in synapses deep inside living brain slices. Biophys. J. 101, 1277–1284. Vaziri, A., Tang, J., Shroff, H., Shank, C.V., 2008. Multilayer three-dimensional super resolution imaging of thick biological samples. Proc. Natl. Acad. Sci. U. S. A. 105, 20221–20226. Ventura, R., Harris, K.M., 1999. Three-dimensional relationships between hippocampal synapses and astrocytes. J. Neurosci. 19, 6897–6906. Westphal, V., Rizzoli, S.O., Lauterbach, M.A., Kamin, D., Jahn, R., Hell, S.W., 2008. Videorate far-field optical nanoscopy dissects synaptic vesicle movement. Science 320, 246–249. Wildanger, D., Medda, R., Kastrup, L., Hell, S.W., 2009. A compact STED microscope providing 3D nanoscale resolution. J. Microsc. 236, 35–43. Willig, K.I., Rizzoli, S.O., Westphal, V., Jahn, R., Hell, S.W., 2006. STED microscopy reveals that synaptotagmin remains clustered after synaptic vesicle exocytosis. Nature 440, 935–939. Willig, K.I., Stiel, A.C., Brakemann, T., Jakobs, S., Hell, S.W., 2011. Dual-label STED nanoscopy of living cells using photochromism. Nano Lett. 11 (9), 3970–3973 (Sep 14). Yuste, R., 2011. Dendritic spines and distributed circuits. Neuron 71, 772–781.