Neuro at the Nanoscale

Introduction

Synapses have long been recognized as fundamental units of neural computation, and we understand, in impressive detail, some of the molecular-scale processes that support and modify fast, point-to-point neuronal communication. Much of this understanding has come from light microscopy studies that have localized key molecules, tracked fine-scale morphological changes, and investigated ionic, membrane, and cytoskeletal dynamics at the scale of the single synapse. The success and ingenuity of these approaches aside, they were long thought to be fundamentally limited in resolution due to the wave-nature of light. In practical terms, observing the synapse at length scales below ~200 nm was thought to be strictly the domain of electron microscopy (EM). Pushing microscopy beyond these scales is critical for studying neuronal function, as virtually all synaptic phenomena—including the trafficking and release of vesicles—are governed by small protein complexes with highly interdependent components. Using diffraction-limited techniques, it can be difficult to even know whether a given protein is localized to the pre- vs post-synaptic side of a synapse, as the synaptic cleft is ~20 nm wide.

Whereas EM remains a gold standard for studying the ultrastructure of the synapse (especially its membranous components), its possibilities for molecule-specific labeling are limited, and it currently offers no way to image live specimens. These constraints dramatically limit the study of synapse dynamics, and additionally limit investigations of how discrete synaptic components [which may number in the hundreds of thousands at a typical synapse and span several hundred distinct protein species (Collins et al. 2006; Wilhelm et al. 2014)] are distributed and co-localized. Fluorescent labels, on the other hand, can be used to tag specific molecules of interest in both fixed and living tissue. Additionally, the broad range of available fluorophores readily allows multiple protein species to be localized in the same sample. In light of these desirable features, considerable efforts have been directed towards breaking the diffraction barrier in fluorescence microscopy.

The development of super-resolution (‘nanoscopy’) techniques in the 1990’s and 2000’s allowed researchers to image fluorescent molecules at unprecedentedly small scales (Eggeling et al. 2015), and with this improvement in resolving power came new possibilities for studying synapses and neurons more generally. Here, we review recent progress in the field of fluorescence nanoscopy, emphasizing how pushing imaging beyond the diffraction limit has contributed to our understanding of synaptic function, and neuronal function more generally. For the purposes of this review, we focus on developments in stimulated emission depletion (STED) microscopy, which is one of several techniques currently available for imaging at the nanoscale. As a light microscopy specimen, the brain presents challenges to which STED is well suited and, not surprisingly, many technological advances in STED have been developed in the context of neuroscience applications. Neurons have complex three-dimensional (3D) features that are organized on cellular scales, yet span large volumes of tissue. The inherent optical sectioning of STED makes it an attractive imaging modality for investigating thick specimens such as brain tissue, where understanding function relies critically on resolving 3D details.

The Diffraction Barrier

The diffraction of a light wave through a circular aperture (i.e., a lens) imposes a fundamental limit to the spatial resolution of any far-field imaging system. This hurdle has been well known since its discovery by Ernst Abbe well over a century ago (Abbe 1873). In more practical terms, the ability to resolve nearby objects in a (confocal) microscope is limited by the size of the focal volume produced by the objective lens. A common convention uses the full-width at half-maximum (FWHM) of the focus as a measure of resolution. The lateral (in the focal plane, perpendicular to the optical axis) FWHM, Δr, is given by Abbe’s equation:

Δ r ≈ λ 2 N A

where λ is the wavelength of light, and NA is the numerical aperture of the lens. Assuming the use of visible light and a high-NA objective lens, the lateral resolution in a conventional fluorescence microscope is >200 nm. Axial resolution, Δz, is typically 2-3 times larger than the lateral resolution and is given by:

Δ z ≈ n λ N A 2

where n is the index of refraction of the mounting medium. In the context of image formation, the adverse effect of diffraction is that the image of a point-like object (i.e., a fluorophore) will have a size given by Eqns. (1) and (2). Furthermore, identical fluorophores separated by less than these distances will appear as a single image and cannot be resolved. Correspondingly, the image of any extended object is blurred, and spatial resolution is therefore limited.

For more than a century, this diffraction limit was considered insurmountable. However, motivated in large part by the desire to visualize biological systems on sub-diffraction length scales, efforts over the past two decades have produced a variety of imaging modalities that provide resolution far below the classical limit in fluorescence microscopy. The term ‘nanoscopy’ was recently coined (Hell 2007) to distinguish these diffraction-unlimited techniques from conventional approaches (e.g., confocal or widefield microscopy). Below, we review the fundamental principles behind one such technique, STED nanoscopy.

STED Nanoscopy

Initially proposed in 1994 (Hell and Wichmann 1994) and experimentally demonstrated in 1999 (Klar and Hell 1999), the first concept to overcome the long-standing diffraction barrier and provide diffraction-unlimited resolution in a light microscope was stimulated emission depletion (STED). Broadly speaking, a STED system is a point-scanning microscope with an effective focal volume smaller than dictated by Eqns. (1) and (2). The underlying principle of resolution enhancement in a STED microscope is that the effective focal volume can be reduced to sub-diffraction dimensions through the targeted switching of fluorophores (Hell 2009).

In a conventional confocal microscope, an excitation source (typically a laser) is focused to a diffraction-limited spot into a sample labeled with fluorescent markers. Fluorescence is then generated in the same focal region (Fig. 1A) and an image is formed pixel by pixel as the signal is detected in coordination with a raster scan (either by scanning the laser focus through the sample or by scanning the sample through a stationary focus). Recorded images will have lateral and axial resolution as determined by Eqns. (1) and (2), respectively.

OPEN IN VIEWER

Figure 1.

Comparison of confocal microscopy and STED nanoscopy. (A) Excitation and corresponding fluorescence profiles for a conventional confocal microscope. (B) The addition of a depletion focus confines fluorescence to a sub-diffraction-sized volume in STED nanoscopy. (C) Cartoon of dendritic spines and simulated (D) confocal and (E) STED images demonstrate the effect of enhanced spatial resolution in a fluorescence microscope. Scale (C–E) 500 nm.

In a STED microscope a second, so-called depletion laser is additionally focused into the sample and superimposed onto the excitation focus (Klar et al. 2000; Donnert et al. 2006) (Fig. 1B). The wavelength of the depletion laser is selected to reside in the red-shifted tail of the emission spectrum of the fluorophore being used. This choice allows fluorescence to be easily separated from excitation and depletion light using common filters and also minimizes unwanted fluorescence induced by anti-Stokes excitation (Vicidomini et al. 2012). By the process of stimulated emission, the depletion laser forces fluorophores from the excited state back to the ground state before they spontaneously emit fluorescence (Fig. 2A). Interestingly, stimulated emission, the process by which light is amplified in a laser, is here used to quench (turn off) fluorescence. This on/off transition produced by exciting and quenching fluorophores (i.e., the coordinated actions of the excitation and depletion lasers) provides the switch that allows the diffraction barrier to be surpassed (Hell 2009). Finally, the depletion focus must be appropriately shaped in order for this on/off switching to produce a sub-diffraction sized focal volume.

OPEN IN VIEWER

Figure 2.

Principles of STED nanoscopy. (A) Jablonski diagram showing fluorophore transitions between the ground and excited states for excitation, fluorescence, and depletion. Resolution enhancement through targeted switching is achieved when fluorophores are forced out of the excited state by the depletion laser. Lateral (B) and axial (C) depletion profiles generated by a helical phase ramp (inset, B). Lateral (D) and axial (E) depletion profiles generated by central half-wave phase step (inset, D).

In the most common implementation of STED (which provides lateral resolution enhancement only), the depletion beam is modulated by a helical phase ramp to form a toroid-shaped (colloquially called ‘doughnut-shaped’) focus with zero intensity at the center (Figs. 1B, 2B). In this configuration, fluorophores residing outside the center of the tandem foci are quenched by stimulated emission while fluorophores in the immediate vicinity of the center (where the depletion focus has zero intensity) emit photons that are detected to form an image. This synergy results in an effective focal volume of dimensions well below the diffraction limit (Fig. 1B, 1E).

The lateral FWHM of the STED focal volume—and, therefore, spatial resolution— is well approximated (Westphal and Hell 2005; Harke et al. 2008a) by a modified version of Abbe’s equation (Eq. 1):

Δ r S T E D ≈ λ 2 N A 1 + I S T E D / I S A T

Here, I_STED is the peak intensity of the depletion focus and I_SAT is the characteristic saturation intensity (the intensity at which 50% of fluorescence is quenched) specific to the fluorophore being used. More simply stated, the STED resolution scales according to the conventional resolution (Eq. 1) divided by the square root of the depletion intensity. Given that Δr_STED scales inversely with the square root of the depletion intensity, the resolution of a STED microscope is considered to be diffraction-unlimited in principle (Hell 2007). However, in addition to standard considerations, such as labeling specificity, background signal, optical aberrations, and signal-to-noise ratio, the resolution achieved in practice will, of course, depend on the tolerance of both the fluorophores and the specimen to laser intensity.

One virtue of STED as a tool for neuroscience comes in part from its ability to image intact and semi-intact specimens, which preserve critical 3D information about circuit organization and fine-scale connectivity. Though the STED configuration discussed above has been reported to provide lateral resolution of less than 10 nm (Rittweger et al. 2009), using a helical phase ramp to shape the depletion focus does not provide axial resolution enhancement (Fig. 2B, 2C). 3D STED imaging requires alternative configurations in order to additionally reduce the size of the focal volume along the optical axis. This requirement can be straightforwardly achieved by imposing a central half-wave phase step (in place of a helical phase ramp) onto the depletion beam (Fig. 2D, 2E) and has been demonstrated to yield ~100-nm resolution in 3D (Donnert et al. 2006). The drawback of this approach is that axial resolution enhancement is introduced at the expense of reduced lateral resolution (Donnert et al. 2006). However, two depletion beams, one masked with the helical ramp (Fig. 2B), the other with the central half-wave step (Fig. 2D), can be used simultaneously to exploit the benefits of each phase mask and yield optimal 3D resolution (Harke et al. 2008b). Alternatively, implementing STED in the 4Pi microscope geometry (Hell and Steltzer 1992) has been demonstrated to provide isotropic 3D resolution down to 30 nm (Schmidt et al. 2008; Schmidt et al. 2009). Though this approach has set the current bar for resolution in STED nanoscopy, it has so far been restricted to imaging thin samples that can be sandwiched between opposing high NA objective lenses.

Whereas EM allows clear visualization of synapse ultrastructure, it provides a limited perspective on how discrete neuronal components co-localize and interact. With fluorescence microscopy, on the other hand, one can, in-principle, image as many unique molecules as there are unique fluorophores, revealing far more biological detail about the multi-component protein complexes that govern neuronal structure, motility, and function. In practice, of course, there are practical limits to how many fluorophores one can image, and multicolor imaging with STED nanoscopy poses additional complexities beyond those encountered with conventional confocal microscopy. For example, the conventional approach of using spectrally separated “green” and “red” imaging channels requires two depletion wavelengths—one for each fluorophore—and therefore a total of four laser lines. Aside from the challenges of aligning four laser foci with nanometer precision, this scheme has required sequential imaging of each channel (red then green) to avoid bleaching of the red-shifted fluorophore by the green channel’s depletion laser (Donnert et al. 2007; Meyer et al. 2008). To alleviate some complexity, several alternative approaches have been developed for multicolor STED imaging. For instance, using a fluorophore with a long Stokes-shift (i.e., the spectral shift between excitation and emission maxima) allows dual-color imaging with a single depletion beam (Schmidt et al. 2008). In this approach, each dye shares the same detection channel and is distinguished by its respective excitation source. Though still requiring sequential imaging of each probe, this approach allows repeated imaging of each staining and has been used for quasi-simultaneous, dual-color imaging of live cells using line by line alternation of the excitation lasers (Pellett et al. 2011). Other alternative approaches have used fluorescence lifetimes (Buckers et al. 2011) or spectral detection combined with linear unmixing (Tonnesen et al. 2011) to distinguish multiple fluorophores. A more comprehensive review of multicolor STED imaging can be found in (Eggeling et al. 2015).

Generalization and Alternative Methods

The concept of enhancing resolution through the targeted switching of fluorophores has also been generalized to include switching mechanisms other than stimulated emission. Collectively, this family of techniques has been termed RESOLFT microscopy (Hell et al. 2003; Hell 2009). Recently, the use of reversibly-switching fluorescent proteins (FPs) in RESOLFT microscopy has come to fruition in biological imaging (Brakemann et al. 2011; Grotjohann et al. 2011; Grotjohann et al. 2012; Testa et al. 2012). Reversibly-switching FPs offer the combined benefits of genetically encoded labeling and up to 10⁶-fold reduction in laser intensities (Grotjohann et al. 2011). In combination with the development of faster switching FP’s (Grotjohann et al. 2012) and parallelization (Chmyrov et al. 2013), this approach holds great promise for imaging living specimens at the nanoscale.

Through continued technological and conceptual advances after its initial conception, STED had become an important tool for biological imaging by the mid-2000’s—several examples of which will be discussed below. Though it will not be treated in depth in this review, during this time, another approach at breaking the diffraction barrier was also developed. Building on earlier work in single-molecule imaging (for reviews see (Moerner 2007; Gould et al. 2012b)), in 2006, three groups published the concept of repeatedly imaging sparse subsets of single molecules, localizing their positions, and reconstructing a final super-resolution image from the molecular positions (Betzig et al. 2006; Hess et al. 2006; Rust et al. 2006). These approaches relied on the stochastic switching of photoactivatable or photoswitchable dyes to image large numbers of single fluorophores within a diffraction limited volume—an essential ingredient, as the resolution of the resulting reconstructed images depend on both the localization precision and the density of localized molecules (Betzig et al. 2006; Shroff et al. 2008). Termed FPALM (Hess et al. 2006), PALM (Betzig et al. 2006), and STORM (Rust et al. 2006) in the original reports, numerous acronyms have subsequently been reported as alternative switching schemes have been realized [see recent review (Nelson and Hess 2014)]. However, the fundamental principle of switching, localization, and image reconstruction underlies this entire class of approaches, which is often collectively termed localization microscopy (LM) (Gould et al. 2012b).

Both STED microscopy and LM are becoming increasingly important imaging tools in the biomedical sciences, now allowing direct visualization of structure and processes that have previously required indirect verification. Each of these nanoscopy techniques of course has their respective pros and cons. Therefore, choosing a particular imaging modality will generally depend on the needs of the specific application. One notable benefit of STED microscopy over LM is that the resolution enhancement in STED is a purely physical phenomenon and does not require post-acquisition image analysis. As a result, acquired images can be observed in real time (as with conventional laser-scanning or widefield microscopy).

Early Neuroimaging Successes and Developments Using STED Nanoscopy

The first fixed-tissue applications of STED nanoscopy in neurons were published in 2006, and the five subsequent years saw rapid technological development that extended the technique’s reach to include applications in live, in-tact preparations, with the ability to resolve multiple fluorophores, and to reveal temporal dynamics.

The earliest neurobiological STED studies shed light on two fundamental and long-investigated issues in synaptic physiology: 1) how the organization of presynaptic proteins contributes to the efficacy of transmitter release, and 2) how vesicular proteins are recycled after exocytosis. The first question was studied at the Drosophila neuromuscular junction (NMJ) by Kittel et al. (2006), who investigated the role of bruchpilot—a protein with homology to mammalian active zone components—in synaptic transmission. Flies lacking bruchpilot had motor abnormalities (Wagh et al. 2006) as well as diminished excitatory post-synaptic current EPSCs recorded at the NMJ. In addition, subunits of calcium channels critical for release were significantly less clustered in knockouts, pointing to a key organizational role for bruchpilot. Whereas confocal imaging of bruchpilot simply showed diffraction-limited spots, STED images revealed that the protein was organized into donut-shaped structures (Fig. 3). These were subsequently shown to be a core, central component of the electron-dense ‘T-Bars’ characterizing the invertebrate presynaptic density (Fouquet et al. 2009).

OPEN IN VIEWER

Figure 3.

An early result of STED microscopy, imaging the Drosophila active zone component bruchpilot [from (Kittel et al. 2006)]. Left: Confocal images of bruchpilot fluorescently labeled with the selective antibody Nc82. Whereas discrete protein complexes can be resolved, labeled domains are diffraction limited, and exhibit no discernible substructure. Middle: higher resolution STED images of the same field of view, revealing the same labeled domains to be toroid-shaped. Right: higher magnification of bruchpilot, showing single and multi-ring clusters. White arrows and arrowheads indicate single rings and multi-ring clusters, respectively. Red arrow indicates a protein cluster viewed parallel to the plane spanned by the synaptic cleft. Scale, 1 μm.

In parallel with this early work, Willig et al. (2006) used STED nanoscopy to investigate the degree to which individual vesicles retain their identity after exocytosis and subsequent endocytosis. Broadly speaking, the two prominent models in the field are that vesicular proteins disperse rapidly upon fusion with the presynaptic membrane and, alternatively, that proteins remain clustered post-fusion to facilitate vesicle re-use [reviewed in (Sudhof 2004) and (Rizzoli and Betz 2005)]. Before the advent of nanoscopy, evidence for either proposal was only indirect, as the critical test—monitoring protein distribution post-fusion at the single vesicle level—was beyond the resolution of classical microscopy. Using STED imaging, Willig et al. (2006) showed that fluorescently labeled antibodies for the lumenal portion of synaptotagmin, a vesicular protein, remained clustered following synaptic stimulation and subsequent endocytosis. This result was taken as evidence that vesicle integrity is maintained after fusion. Other work using STED during this early period provided critical data on the size and distribution of presynaptic clusters of syntaxin, another prominent presynaptic protein; this work informed theories about the assembly and organization of membrane protein complexes (Sieber et al. 2007).

This early investigation of synaptic vesicles using STED was rapidly extended to include studies of vesicle dynamics. With the development of video-rate STED in 2008, Westphal et al. monitored the tracking of single antibody-labeled vesicles in cultured hippocampal neurons—a feat that was again beyond the resolution of traditional light microscopes. This work provided a more dynamic portrait of the vesicle population, revealing static “hotspots” of vesicles that remained relatively stationary on the timescale of minutes, in addition to highly mobile vesicles that diffused rapidly between different boutons. Investigations of single-vesicle organization and dynamics have remained an important application of STED (Wilhelm et al. 2010; Hua et al. 2011; Jorgacevski et al. 2011), and improvements in the technique’s temporal resolution will almost certainly open new possibilities for future investigations.

Another line of early STED work investigated post-synaptic phenomena, in particular, the organization and dynamics of dendritic spines. These small (~ 1 μm) and numerous protuberances, particularly well-studied in hippocampal and cortical pyramidal neurons, are discrete sites of synaptic input. First hypothesized by Cajal in 1888 to allow for the establishment of “more intimate contacts with the axonal arborizations,” (quoted from (Garcia-Lopez et al. 2007)) spines have been intensely investigated for several decades, and interest in their function has consistently spurred the development of new microscopy and photoactivation techniques. Two major lines of work in the spine field have involved studying 1) the degree to which these structures are electrically and chemically compartmentalized, and 2) the degree to which the spine population is dynamic and plastic, particularly in the context of experience and activity-dependent change. Although a great number of confocal and two-photon studies have investigated spine morphology, STED allows for better quantification of spine size, particularly of the thin, tapering necks that join spines to dendrites.

The first spine study using STED was performed by Nagerl et al. (2008) in which they demonstrated that spines could be monitored longitudinally over minutes in cultured hippocampal neurons using STED nanoscopy. The authors induced long-term potentiation (LTP) by applying a high-potassium and high-calcium perfusate, and monitored resultant changes in spine morphology. Whereas this initial study was a proof-of-principle, the same group later extended these techniques (Tonnesen et al. 2014) to perform a technical tour-de-force that challenged longstanding ideas about spine function (see below).

One of the major microscopy developments that revolutionized the study of dendritic spines prior to STED was two-photon microscopy. Developed and refined in the 1990s and 2000s (Denk et al. 1990; Yuste and Denk 1995; Svoboda et al. 1996) [reviewed in (Denk and Svoboda 1997)], this technique was especially revolutionary for neurobiologists studying brain tissue in vivo and in semi-intact preparations (i.e., brain slices), as its use of long-wavelength illumination allowed deeper penetration into heavily light-scattering brain tissue. Recently, Ding et al. (2009) have demonstrated the feasibility of two-photon STED nanoscopy, which couples the superior depth penetration of two photon imaging with the superior resolution of STED. Using two-photon excitation for the first time in combination with a standard STED depletion laser, these investigators imaged spines at depths of ~120 microns in acute brain slices. These initial pioneering studies promise to revolutionize future work in imaging spines in in-tact tissues, and the recent success of in-vivo STED imaging of mouse somatosensory cortex (Berning et al. 2012) is an exciting initial step in this direction.

Highlights of Recent Breakthroughs with STED

STED and other super-resolution techniques are continuing to revolutionize the study of neuronal ultrastructure, and these breakthroughs are expected to accelerate as STED systems are more widely adopted and commercialized. The field of nanoscopy has paved the way for unique possibilities for localizing and colocalizing specific molecules in live tissue, and has helped corroborate several long-held ideas about neuronal organization that previously had lacked direct verification. At the same time, STED has also provided a number of genuine surprises, some of which run contrary to longstanding intuitions and assumptions in the field

Recent work by Tonnesen et al. (2014) provides a representative case study (Fig. 4). In a technically stunning set of experiments, these investigators used STED nanoscopy in conjunction with FRAP and photo-uncaging to quantify heterogeneity among dendritic spines, as well as to monitor changes in spine geometry concomitant with changes in synapse strength. The first important finding of this work was the direct observation that all basic shape parameters, including spine neck width and head width, were smoothly and unimodally distributed over the spine population, with no evidence of clustering or pairwise correlation between parameters. This result strongly argues against the general validity of the long-used and widely-adopted classification scheme that groups spines into ‘stubby’, ‘thin’, and ‘mushroom’ types (Peters and Kaiserman-Abramof 1970). The observation that spine shape spans a continuum is also consistent with detailed EM reconstructions of spines and their post-synaptic densities (Arellano et al. 2007). The near-complete absence of so-called ‘stubby’ spines (those lacking a visible neck) in the Tonnesen et al. (2014) study was also of note, given that stubby spines are a sizeable fraction of those observed in conventional (diffraction-limited) light microscopy studies. The authors suggest, provocatively, that this morphological class—long studied and remarked on—may largely be an artifact of insufficient spatial resolution.

OPEN IN VIEWER

Figure 4.

Nanoscale changes in dendritic spine morphology evoked by focal uncaging of glutamate [after Fig. 6 of Tonnesen et al. (2014)]. (A) A single spine before (left) and after (right) induction of spine-specific long-term potentiation (LTP) by two-photon uncaging of glutamate (uLTP). Note the marked and highly resolved changes in spine head volume and spine neck length and width. (B) Enlargements of the boxed areas in (A). (C) Group data showing large, input-specific increases in spine volume following uLTP. Red trace shows head-volume changes in the stimulated spine. Black shows relative absence of volume changes in neighbor (unstimulated) spines, illustrating synapse specificity. Blue shows relative absence of uLTP in high magnesium, illustrating NMDA receptor dependence. (D) Group data showing decreases in spine-neck length following uLTP. Trace colors and corresponding conditions as in (C). (E) Group data showing increases in spine-neck width following uLTP. Trace colors and corresponding conditions as in (C). Scale bars, 500 nm.

The other major findings of Tonnesen et al. (Tonnesen et al. 2014) concerned changes in spine geometry that accompany changes in synapse strength (Fig. 4). Repetitive photo-uncaging of glutamate at single spines potentiated unitary, light-evoked EPSCs selectively for stimulated spines. Additionally, these physiological changes were paralleled by dramatic, spine-specific increases in head volume and neck width (~200% and ~30%, respectively), as well decreases in neck length (~20% to 30%) (Fig. 2C–2E). The effects of activity on neck width were of special interest, given that neck width is predicted to be one of the principal determinants of electrical and chemical compartmentalization in spines, and has, historically, been beyond the resolving power of conventional light microscopy. Intriguingly, the coordinated effects of these shape changes had qualitatively different effects on electrical vs chemical compartmentalization in spines. Whereas spine diffusion (measured by FRAP) was essentially unaffected, owing to the opposing influences of the neck and volume changes, electrical compartmentalization (modeled as a passive compartment with geometric parameters taken from STED images) was predicted to increase. Although this prediction awaits more direct verification, as conductance changes at the spine might amplify or counteract this effect, this study sets and impressive benchmark for investigating morphological correlates of plasticity at the level of single spines.

If the above study is an example of unexpectedly complex phenomena revealed by STED nanoscopy, then a recent report by D’Este et al. (2015) is an example of how STED is revealing unexpected order at the nanoscale. In brief, these authors used the fluorogenic label SirActin (Lukinavicius et al. 2014) to show that neuronal processes—both axons and dendrites—have a highly periodic and ‘ring-like’ arrangement of actin filaments that extend for the length of the neurite. This work followed closely from the pioneering study of Xu et al. (2013), who used STORM (i.e., localization microscopy; see above) to investigate the cytoskeletal organization of the axon, and garnered impressive evidence for a ‘beam and trusses’-like arrangement of actin and its associated proteins spectrin, ankyrin, and adducing (Fig. 5A). Although the functional role of this arrangement remains to be shown, Rasband (2013) has noted that a system of beams and trusses in general allows local mechanical stresses to be more effectively dissipated, providing the structural integrity that axons would almost certainly need when extending over long distances.

OPEN IN VIEWER

Figure 5.

Highly periodic arrangement of cytoskeletal components in neurites, observed at the nanoscale. (A, B) Taken from Xu et al. (2013); (C, D) taken from D’Este et al. (2015). (A) Ring-like arrangement of fluorescently labeled spectrin along the axon of a cultured hippocampal neuron. Inset shows a y-z cross-section of the boxed area. (B) Histogram of distances between spectrin rings, consistent with a highly repeated structural motif with ~180-nm spatial periodicity. (C, D) Images of sirActin-labeled axons (C) and dendrites (D) of cultured hippocampal neurons. Scale (C, D) 1 μm.

Notably, the rungs of the actin ladder [in both (Xu et al. 2013) and (D’Este et al. 2015); (Fig. 5)] were spaced at ~180 nm, just beyond the resolution of diffraction-limited microscopy, and the only ultrastructural actin motifs shown in dendrites—previously by EM—were diffuse agglomerations of actin observed in spines (Korobova and Svitkina 2010) and growth cones (Lewis and Bridgman 1992). In other words, nanoscopy techniques revealed a completely novel and unanticipated structure at sub-diffraction length scales. The observed actin rings seem to be a quite general feature of neurites, as they were observed in both axons and dendrites (Fig. 5C, 5D), as well as in myelinated and unmyelinated axons of central and peripheral neurons. Moreover, the rings become progressively more ordered and punctate in the first few days of postnatal life, and the relative proportions of actin rings vs bundles vs long filaments change dramatically during early development. Taken together, these results point to an unexpected, elegant, and developmentally regulated potential solution to the problem of how neurites maintain their structural integrity.

Outlook

Given the current (spatial and temporal) resolving capabilities and commercial availability of diffraction-unlimited imaging technology, the potential for future breakthroughs in neuroscience (and other fields) is indeed high. The rapidly ongoing development of novel and improved fluorophores and labeling strategies tailored towards super-resolution imaging will undoubtedly also drive new biological discoveries. In particular, biological applications making use of reversibly-switching fluorescent proteins for low-intensity RESOLFT nanoscopy are still relatively new and, in combination with adaptive optics (Gould et al. 2012a) for deep-tissue 3D imaging, have the potential to revolutionize super-resolution imaging in vivo. It will soon be possible, in principle, to image at the single synapse scale with molecular specificity, and to monitor changes in cytoskeletal dynamics, local mRNA abundance, and protein composition while relating these phenomena to ongoing experience.

Though only a subset of the many impressive biological applications of nanoscopy to date, the studies reviewed above certainly illustrate the importance of imaging resolution in hypothesis driven research. Visualizing structure and processes at the nanoscale not only enhances the researcher’s ability to ask essential questions but can also yield surprising discoveries that open new avenues of inquiry. In other words, the nanoscale remains a frontier for applying light microscopy in biological fields and, as such, this frontier will likely be best explored through descriptive research programs that embrace the unexpected (Saka and Rizzoli 2012).