Summary
Dendritic spinules are thin, membranous protrusions formed by neuronal dendritic spines that are not adequately resolved by diffraction-limited light microscopy. Hence, our understanding of spinules is inferred predominantly from fixed-tissue electron microscopy (EM). Super-resolution modalities have enabled live-cell nanoscopic imaging, but their utility for fast, time-lapse, volumetric imaging has been restricted. Herein, we utilized rapid structured illumination microscopy (SIM) and ‘enhanced resolution’ confocal microscopy to study spatiotemporal spinule dynamics in live cultured cortical pyramidal neurons. Spinules on mushroom spines typically recurred at the same topographical locations and most were short-lived, originating near simple post-synaptic densities (PSDs), while a subset was long-lived and elongated, emerging from complex PSDs. Ca2+ puncta within spinules synchronized with spine head transients and Ca2+ depletion drastically decreased spinule number. Moreover, we uncovered evidence of differential Ca2+-mediated regulation of short-lived and long-lived spinules. Thus, we identified unique spinule classes divergent in lifespan, dynamics, morphology, relationship to the PSD, and regulation. These data suggest distinct synaptic functions of spinule classes, informing future studies, while demonstrating a new application for enhanced resolution microscopy.
Introduction
Dendritic spinules, first detected in hippocampal neurons using EM (Westrum and Blackstad, 1962, Tarrant and Routtenberg, 1977), are thinner than filopodia, <1 µm in length, often detected on mushroom spines, and inducible by synaptic activity (Petralia et al., 2015). These nanoscale structures are not well-resolved by diffraction-limited light microscopy, hence our understanding of spinule structure and function has been largely inferred from fixed tissue EM (Tao-Cheng et al., 2009, Spacek and Harris, 2004) and a few limited-resolution confocal and two-photon imaging studies (Richards et al., 2005, Ueda and Hayashi, 2013). Prior work has established that spinules can project into pre-synaptic boutons or glial cell membranes and their proposed functions include post-synaptic membrane remodeling and retrograde signaling (Spacek and Harris, 2004, Petralia et al., 2015).
Numerous EM studies have reported their link to the PSD, including some showing spinule origination from PSD edges or perforations (Tao-Cheng et al., 2009, Spacek and Harris, 2004, Petralia et al., 2015), and others describing origination from nonperforated PSD edges or spine necks (Sorra et al., 1998). A study utilizing fast, multi-channel, high-resolution imaging in live neurons could provide needed clarification on the relationship between spinules, which can be highly dynamic (Tao-Cheng et al., 2009), and PSDs, which are also continually remodeling in response to activity (Okabe et al., 1999, Bourne and Harris, 2008). Moreover, Ca2+ is a key signaling ion in a multitude of synaptic functions, including synaptic transmission, which is known to induce spinule formation (Petralia et al., 2015, Tao-Cheng et al., 2009, Ueda and Hayashi, 2013). Ca2+ regulates spinule-dependent synaptic plasticity in fish horizontal cells (Country and Jonz, 2017), but the role of Ca2+ in mammalian spinule formation remains unclear, necessitating live, high-speed imaging studies.
Recently developed super-resolution imaging modalities have enabled live imaging of nanoscale structures (Cox, 2015, Godin et al., 2014), but their application for large volumetric datasets has been restricted by slow acquisition speed, photo-bleaching, and photo-toxicity. While SIM is well-suited for live three-dimensional (3D) imaging, a mechanically translated grating pattern has restricted acquisition speed, leading to missed events and motion artifacts (Fiolka et al., 2012). Here we utilized a custom-built system that employs a liquid crystal spatial light modulator to efficiently alter the grating pattern and increase speed (Kner et al., 2009, Li et al., 2015) to quantify spinule lifespan and dynamics. We additionally used enhanced resolution confocal microscopy to investigate spinule dynamics in relation to the PSD, as well as Ca2+ as a candidate regulator of mammalian spinule formation during basal activity.
Results
We first utilized a SIM system with improved optical sectioning speed compared to standard SIM to image live dissociated cortical mouse pyramidal neurons. We balanced acquisition speed against photobleaching and phototoxicity, to resolve dynamic, thin membrane protrusions that were finer than filopodia and originated from dendritic spines, i.e. spinules, at 15–20 sec intervals. Similar to a previous report (Spacek and Harris, 2004), IMARIS 3D reconstructions revealed spinules on mushroom, branched, and more rarely, on thin spines (Figure 1A). Mushroom spines displayed a substantially higher mean number of spinules over the 1000 sec imaging duration, compared to spines transitioning between classes, thin, or filopodia-shaped spines (Figure 1B). We next tracked mushroom spine volume over time (Figure 1C) and found a strong positive correlation between spinule number and mean spine head volume (Figure 1D). Spinules were often observed in non-contiguous frames at the same topographical spine head locations, and repeatedly extending and retracting along axons (Figure 1E; Video 1). Quantification of normalized spinule origination in relation to kymograph outlines of spine heads (Figure 1F) showed that the majority (76%) of spinules recurred within the 1000 sec imaging frame, and the mean number of recurring spinules per spine was substantially higher than singly occurring spinules (Figure 1G). Importantly, most spinules (70%) existed for <60 sec, while the minority (30%) existed for ≥60 sec, rarely exceeding the imaging duration (Figure 1H and Figure S1A). Short-lived spinules were accordingly more frequent per spine (Figure 1I). A representative spine displayed multiple short-lived spinules, which were often tapered or vermiform, and fewer longer-lived spinules (Figure 1J; Figure S1B-C; Video 2). Long-lived spinules also displayed diverse morphologies, including pinch-waisted and rarely filopodia-like, as well as occasional spine-neck origination and evidence of presumptive vesicular trafficking within thick spinules (Figure 1K, Figure S1D).
We next assessed spinule dimensions over time using rapid SIM. The current consensus is that these structures are typically < 1 µm in length and finer than filopodia, while reports on their diameter and shape have varied (Petralia et al., 2015). We found that the mean length of all spinules over their lifespans was 0.49 µm, while the mean width and volume was 210 nm and 38 nm3, respectively (Figure S2A-D). Furthermore, mean spinule length positively correlated with lifespan (Figure 2A), but not width or volume (DNS). Long-lived spinules were significantly greater in mean and maximum length (Figure 2B), with no difference in mean width between groups (Figure S2E). We tracked spinule volumes using IMARIS to generate temporal spinule volume plots (Figure 2C) and illustrate the differential size and stability of short-versus long-lived spinules (Figure 2D-E; Video 3). Overall, mean and maximum volume was greater in long-lived than short-lived spinules (Figure 2F). Strikingly, the mean spinule length change was negatively correlated with spinule lifespan (Figure 2G). While the cumulative length and volume change was greatest in long-lived spinules, the change in length (nm/sec) and volume (nm3/sec) was dramatically higher in short-lived spinules compared to long-lived spinules (Figure 2H-I; Figure S2F-G). These data suggest the existence of distinct spinule classes with divergent lifespans, morphologies, and dynamics.
Some studies have reported spinules originating from PSD edges or perforations (Tao-Cheng et al., 2009, Spacek and Harris, 2004, Petralia et al., 2015), versus nonperforated PSD edges or spine necks (Sorra et al., 1998). Considering that PSDs too are continually remodeling in response to activity (Okabe et al., 1999, Bourne and Harris, 2008), fast, multi-channel, high-resolution imaging in live neurons is needed to clarify the link between spinule origin and the PSD. Hence, we utilized fluorescently-tagged ‘intrabodies’ to label endogenous PSD95 (Gross et al., 2013) and resonance scanning enhanced resolution confocal microscopy (Figure 3A), to mitigate intrabody photo-bleaching observed with SIM, which requires a strong fluorescence signal for time-lapse imaging. The majority (91%) of mushroom spines detected by this alternate method were spinule-positive. Rare, elongated, long-lived spinules and frequent, short-lived spinules were again detected over 600 sec, and the shorter frame interval enabled the detection of an extremely short-lived subgroup (<20 sec) (Figure S3A-B). In accordance with rapid SIM, a significant positive correlation was found between mean spinule length and lifespan, and both the mean and maximum spinule length was substantially higher in long-as opposed to short-lived spinules (Figure S3C-D). Short-lived spinules typically originated near to the PSD edge, while long-lived spinules originated substantially further from the edge of more complex PSDs, at a mean distance of 0.28 and 0.43 µm, respectively (Figure 3B-D; Videos 4 and 5). Surprisingly, PSD volume and dynamics, including the change in PSD volume per sec and the net +/-volume change, did not correlate with spinule number per spine (Figure 3E-F; Figure S3E-F). Complementing prior EM findings, long-lived spinules occasionally contained mobile PSD fragments that could traverse their length (Figure 3G; Video 6). Strikingly, 78% of spines displaying long-lived spinules also contained complex, partitioned PSDs compared to 23% of long-lived spinule-negative spines (Figure 3H), and 67% of long-lived spinule-positive spines displayed fragmented PSDs versus 28% of long-lived spinule-negative spines (Figure 3I). Mean and maximum PSD fragment number per frame was significantly greater in long-lived spinule-positive compared to long-lived spinule-negative spines (Figure 3J). These data suggest a link between complex PSD remodeling and spinule stability, while simple PSDs are associated with dynamic, transient spinules.
Ca2+ is a key signaling ion in synaptic transmission, which induces spinule formation (Petralia et al., 2015, Tao-Cheng et al., 2009, Ueda and Hayashi, 2013), and regulates spinule-dependent synaptic plasticity in the retina of teleosts (Country and Jonz, 2017). To investigate Ca2+ as a candidate regulator of mammalian spinules during basal activity, we again utilized fast (2.2 sec/frame) enhanced resolution confocal imaging of neurons expressing the Ca2+ sensor GCaMP6s (Chen et al., 2013). We again observed a preponderance of short-lived and infrequent long-lived spinules in negative control spines, comparable to prior enhanced resolution confocal imaging data (Figure S4A). Interestingly, we visualized isolated Ca2+ puncta appearing in long-lived spinules repeatedly during Ca2+ transients in control spines (Figure S4A; Video 7). Plots of GFP mean fluorescence intensity (MFI) in the dendritic shaft, spine head, and spinules of controls also revealed synchronization of Ca2+ transients in spinules with that of spine heads and dendritic shafts (Figure 4B; Figure S4B). To evaluate the requirement for intracellular Ca2+ in spinule formation, we compared vehicle-treated control neurons to those treated with the cell-permeant Ca2+ chelator, BAPTA-AM. The total number of short- and long-lived spinules and cumulative spinule lifespan were significantly higher in controls than in BAPTA-AM-treated spines (Figure 4C). To assess the relationship between patterns of Ca2+ transients and spinule dynamics, we first determined the change in Ca2+ fluorescence (ΔF/F0) in dendritic shafts and spine heads, and then generated corresponding spatio-temporal maps to track spinule number and recurrence over time. Mushroom spines with frequent, high-amplitude Ca2+ transients displayed many short-lived and fewer long-lived spinules (Figure 4D-E; Video 8), while spines with sporadic peaks formed only short-lived spinules (Figure S4C-D). In contrast, spines from BAPTA-AM-treated neurons displayed greatly dampened Ca2+ transients and substantially fewer spinules (Figure 4F-G; Video 9). Quantitative assessment of Ca2+ peaks in spines, i.e. 50% increase in ΔF/F0 above baseline, revealed a positive correlation between total spinule number and peak amplitude (Figure S4E), but no correlation with peak frequency or duration. Interestingly, the number of short-lived, but not long-lived, spinules per spine positively correlated with Ca2+ peak mean and maxima (Figure 4H). Conversely, long-lived spinule number positively correlated, albeit less strongly, with Ca2+ peak frequency and duration. These findings suggest distinct mechanisms of spinule regulation by local Ca2+ transients, informing future functional studies of spinule subgroups.
Discussion
Herein, we demonstrate the utility of two complementary enhanced resolution live imaging techniques for studying dynamic nanoscale post-synaptic membrane projections in volumetric samples. While rapid SIM was ideal for distinguishing fine structural details at moderate acquisition speeds, enhanced resolution was the optimal choice for faster, multi-channel imaging, particularly when the fluorescence signal is weak, while still providing the resolving power necessary for spinule visualization. We observed by rapid SIM diverse spinule morphologies, which mirrored those seen in slice cultures by EM (Tao-Cheng et al., 2009). Our findings also provide the first evidence of spinule recurrence, suggesting localized molecular machinery at spinule origination sites, and of distinctive classes of short-lived, dynamic spinules and fewer long-lived, stable spinules.
While spinules have been proposed to participate in post-synaptic membrane remodeling and PSD assembly (Spacek and Harris, 2004), their dynamic relationship with the PSD has not been investigated in living neurons. Our findings that short-lived spinules typically originate near to simple PSDs, while long-lived spinules originate farther from complex PSDs, often extending from perforations, imply a relationship predominantly between long-lived spinules and PSD remodeling. The distinctive morphologies and dynamics of spinule classes together suggest divergent modes of regulation and functions. Long-term potentiation is known to induce spinule formation (Ueda and Hayashi, 2013, Petralia et al., 2015), but the role of Ca2+ in mammalian spinule formation had not been investigated. Our results strongly indicate that Ca2+ is involved in their regulation. We also observed Ca2+ puncta within spinules, isolated from, but in synchrony with spine heads, raising the question of whether Ca2+ enters spinules from the spine head or via channels on the spinule surface. Interestingly, the nuanced regulation of short-lived and long-lived spinules by Ca2+ transients in spines also suggests distinct synaptic functions. In summary, we demonstrate a new application of enhanced resolution microscopy to study dynamic sub-spine membrane projections, informing ongoing mechanistic and functional studies.
Author contributions
PP conceived and supervised the project, and CRZ designed and conducted experiments with the assistance of KM and MDM.
Competing interests
The authors declare no competing interests. Correspondence and requests for materials should be addressed to p-penzes{at}northwestern.edu
Acknowledgments
SIM was performed at the Advanced Imaging Center of the Janelia Research Campus, funded by the Howard Hughes Medical Institute and Gordon & Betty Moore Foundation. Confocal microscopy was performed at Northwestern’s Nikon Imaging Center with the support of Drs. Constadina Arvanitis and David Kirchenbuechler. This work was supported by NIH grants, R01MH107182 and R01MH071316.