Deciphering the Nanoscale Architecture of Presynaptic Actin Using a Micropatterned Presynapse-on-Glass Model

Abstract

Chemical synapses are fundamental units for the transmission of information throughout the nervous system. The cytoskeleton allows to build, maintain, and transform both pre- and postsynaptic contacts, yet its organization and the role of its unique synaptic nanostructures are still poorly understood. Here we present a presynapse-on-glass model based on cultured neurons from rat pups of either sex. Presynaptic specializations are robustly induced along axons by micropatterned dots of neuroligin, allowing the controlled orientation and easy optical visualization of functional induced presynapses. We demonstrate the relevance and usefulness of this presynapse-on-glass model for the study of presynaptic actin architecture, showing that a majority of induced presynapses are enriched in actin, with this enrichment being correlated to higher synaptic cycling activity. We confirm our previous results on bead-induced presynapses by identifying distinct actin nanostructures within presynapses: corrals, rails, and mesh. Furthermore, we leverage the controlled orientation of the presynapse-on-glass model, visualizing the arrangement of these actin structures relative to the active zone nanoclusters using multicolor 3D single-molecule localization microscopy (SMLM) and relative to the subdiffractive localization exocytic events using a correlative live-cell and SMLM approach.

Significance Statement

The actin cytoskeleton plays important but poorly understood roles at presynapses, fundamental compartments for communication in the nervous system. We developed a presynapse-on-glass model to induce isolated, optically accessible presynaptic specializations along the axon of cultured neurons. This model recapitulates the presynaptic actin enrichment and distinct nanostructures we previously uncovered using presynapses induction by large beads. The controlled orientation of presynapses in our new model allows to go further: we visualized the nanoscale arrangement of actin and presynaptic components by multicolor nanoscopy and could link actin nanostructures to the precise location of synaptic vesicle release thanks to a correlative live-cell/super-resolution microscopy approach. This demonstrates the relevance of our model for deciphering the nanoarchitecture of presynapses and understanding their molecular functioning.

Introduction

Neuronal communication is a complex process involving coordinated interactions between cell-adhesion molecules, signaling complexes, and cytoskeletal elements, which are essential for precise signal transmission across chemical synapses (McAllister, 2007; Togashi et al., 2009). The actin cytoskeleton supports the establishment of the neuronal morphology and enables dynamic changes both during early development and maturity (Luo, 2002; Coles and Bradke, 2015). In addition to this structural role, the actin cytoskeleton is vital for synaptic transmission (Cingolani and Goda, 2008; Papandréou and Leterrier, 2018; Gentile et al., 2022), being implicated in presynaptic vesicle clustering, mobilization, and docking (Bernstein and Bamburg, 1989; Kim and Lisman, 1999; Doussau and Augustine, 2000). Actin is also important for endocytosis and the generation of synaptic vesicles at presynapses (Shupliakov et al., 2002; Watanabe et al., 2013; Ogunmowo et al., 2023). Studies using short-term application of actin-perturbing agents have led to contradictory findings of reduced (Cole et al., 2000) or enhanced (Sankaranarayanan et al., 2003) vesicular release, suggesting that actin acts as a meshwork that gathers vesicles before exocytosis or as a physical barrier within the presynaptic subdomains (P. Li et al., 2018).

Understanding the roles of actin at the presynapse requires the knowledge of its structural architecture (Papandréou and Leterrier, 2018). Early ultrastructural studies using electron microscopy have provided significant insights into the distribution and organization of actin within synaptic vesicle clusters (Landis et al., 1988; Hirokawa et al., 1989; Wen et al., 2016) and at the active zone periphery (Bloom et al., 2003). Fluorescence microscopy has expanded our knowledge about the spatial relationship between presynaptic components, but its resolution fails to resolve the presynaptic nanostructures. Fortunately, super-resolution microscopy approaches can overcome the diffraction limit and allow nanoscale imaging studies (Jacquemet et al., 2020; Werner et al., 2021). Among these, single-molecule localization microscopy (SMLM) techniques such as stochastic optical reconstruction microscopy (STORM) and DNA point accumulation in nanoscale topography (PAINT), which rely on stochastic blinking of single fluorophores followed by their precise localization, have refined our knowledge of presynaptic organization (Dani et al., 2010; Glebov et al., 2017; Carvalhais et al., 2021).

Despite these advances, dense postsynaptic actin may obscure the presynaptic signal even at nanometer spatial resolution (Reshetniak and Rizzoli, 2019; Nosov et al., 2020). To overcome this obstacle, our team has recently used a model of bead-induced isolated presynapses (Burry, 1980; Lucido et al., 2009). SMLM of actin within isolated presynapses has revealed three distinct presynaptic actin nanostructures: a faint active zone mesh, linear actin rails found between the active zone and the reserve pools, and actin corrals that surround the presynaptic compartment (Bingham et al., 2023). However, presynapses formed on ∼5 µm spherical beads result in heterogeneous orientation and sizes of induced presynapses. This limits optical accessibility and unambiguous localization of the active zone, impeding the analysis of actin structure roles in presynaptic processes.

An alternative is to induce the formation of presynapses directly on the glass, using postsynaptic adhesion proteins such as neuroligin applied on the whole coverslip (Funahashi et al., 2018) or on micropatterns of SynCAM1 separated by cytophobic regions (Czöndör et al., 2013). In this study, we designed and optimized a presynapse-on-glass approach to orient presynaptic formation toward patterned dots displaying a fragment of neuroligin-1 (nlgn) in addition to a layer of polylysine and laminin. This allowed unrestricted growth of neuronal processes as well as robust induction of presynaptic specialization near the patterned nlgn. We demonstrate how this model recapitulates actin distribution and the nanoscale organization we observed in bead-induced and endogenously tagged presynapses (Bingham et al., 2023). Furthermore, presynapses-on-glass allow to refine the 3D visualization of actin nanoarchitecture at presynapses. Finally, we used our model to combine the visualization of spontaneous synaptic vesicle exocytosis in living neurons with the nanoscopy of presynaptic components, linking actin nanoarchitecture to the precise location of vesicle exocytosis.

Material and Methods

Plasmids

The expression vector for Fc-neuroligin (nlgn, gift from T. Hirano; Funahashi et al., 2018) consists of the cDNA encoding the extracellular region (amino acids 1–675) of neuroligin 1 (−A + B) fused to the N-terminal of human immunoglobulin–Fc region in the pCAGplayII expression vector (Kawaguchi and Hirano, 2006). VAMP2-pHluorin (gift from S Gupton; Urbina et al., 2018) consists in the human VAMP2 fused to superecliptic pHluorin (Miesenböck et al., 1998).

Nlgn production

Nlgn production followed published procedures (Tanaka et al., 2014). HEK 293T cells cultured at 70% confluence in 100-mm-diameter dishes were transfected with a mixture of Lipofectamine 3000 (Thermo Fisher Scientific; 16 µl/dish) and nlgn plasmid (8 µg/dish) for 4 h in DMEM. Protein expression was carried out for 2 d in DMEM, 10% ultralow-IgG fetal bovine serum (Thermo Fisher Scientific), and 100 UI/ml penicillin/streptomycin. The culture medium was then recovered and centrifuged for 10 min at 3,000 × g to remove cell debris. Two hundred microliters of protein A Sepharose beads (Sigma-Aldrich) was added to the 50 ml of supernatant in the presence of antiprotease (Sigma-Aldrich) and incubated overnight at 4°C with stirring. After centrifugation for 5 min at 1,200 rpm, the beads were washed three times with phosphate-buffered saline (PBS), pH 7.4, nlgn was eluted with 600 µl of 0.1 M glycine buffer, pH 2.6, and immediately buffered to pH 7 with 1 M Tris, pH 9. Purified nlgn concentration determined by dot blot was consistently ∼250 µg/ml.

Micropatterned substrate onto glass coverslips

Eighteen-millimeter-diameter round, #1.5H thickness coverslips were washed in acetone and ethanol and then affixed with dots made from two-component orthodontics paste (Rotec) as spacers. They were then treated for 4 min in an oxygen plasma cleaner (Femto, Diener Electronic). Microcontact printing was performed using PDMS stamps with 1.8-µm-diameter pillar arrays with 4 µm center-to-center distance (gift from Jianping Fu, University of Michigan; Weng and Fu, 2011) according to a previously published protocol (Théry and Piel, 2009). Stamps were coated and incubated for 40 min at 37°C with a mixture of bovine serum albumin (BSA)–Alexa Fluor 488 (80 µg/ml; Thermo Fisher Scientific) and nlgn (25 µg/ml) or with BSA-AF488 (80 µg/ml) mixed with the Fc fragment of human IgG (25 µg/ml; Jackson ImmunoResearch Laboratories) as a control. PDMS stamps were then inverted onto ionized coverslips and incubated at 37°C for 20 min and then treated overnight with a mixture of laminin (Thermo Fisher Scientific; 10 µg/ml) and poly-l-lysine (Sigma-Aldrich; 38 µg/ml), prior to neuron seeding.

Animals and cell culture

The use of Wistar rats followed the guidelines established by the European Animal Care and Use Committee (86/609/CEE) and was approved by the local ethics committee (agreement G13055). Rat hippocampal neurons were cultured following the Banker method, above a feeder glia layer (Kaech and Banker, 2006). Hippocampi from E18 rat pups of either sex were dissected and homogenized by trypsin treatment followed by mechanical trituration and seeded on the patterned coverslips at a density of 4,000–6,000 cells/cm² for 3 h in serum-containing plating medium (MEM with 10% FBS, 0.6% glucose, 0.08 mg/ml sodium pyruvate, and 100 UI/ml penicillin/streptomycin). Coverslips were then transferred, cells down, to Petri dishes containing confluent glia cultures conditioned in neurobasal medium (NB; Thermo Fisher Scientific) supplemented with 2% B27 (Thermo Fisher Scientific), 100 UI/ml penicillin/streptomycin, and 2.5 µg/ml amphotericin and cultured in these dishes at 37°C, 5% CO₂. For this work, neurons were fixed at 9–16 d in vitro (DIV).

Anti-synaptotagmin vesicular cycling assays

Vesicular cycling assays were performed using an antibody directed to an extracellular epitope of synaptotagmin 1 (syt; clone 604.2; Synaptic Systems; Kraszewski et al., 1995) that was fed to neurons before fixation. Neurons were incubated for 1 h at 37°C, 5% CO₂ with the syt antibody at 1.6 µg/μl in HBS medium (in mM: 1,356 NaCl, 10 HEPES, 10 d-glucose, 2 CaCl₂, 1.3 MgCl₂). Cells were then washed three times with PBS before fixation.

Antibodies

Primary antibodies used are chicken anti-map2 (1:1,000, ab5392, RRID:AB_2138153; Abcam), mouse anti-βII-spectrin (1:100, clone 42/aa 2,101–2,189, #612563, RRID:AB_399854; BD Biosciences), guinea pig anti-synaptophysin (1:400, 101 004, RRID:AB_1210382; Synaptic Systems), mouse anti-synaptotagmin1 (1:600, clone 604.2/aa 1–12, 105 311, RRID:AB_993036; Synaptic Systems), rabbit anti-adducin alpha (1:100, ab40760, RRID:AB_722627; Abcam), mouse anti-bassoon (1:200, clone SAP7F407, ab82958, RRID:AB_1860018; Abcam), rabbit anti-GFP (1:750, ab290, RRID:AB_303395; Abcam), mouse anti-GFP (1:750, clone 9F9.F9, ab1218, RRID:AB_298911; Abcam), and rabbit anti-Munc13-1 (1:500, 126 103, RRID:AB_887733; Synaptic Systems). Secondary antibodies used were conjugated to Alexa Fluor 488, 555, or 647 (Thermo Fisher Scientific), or DyLight 405 (Rockland Immunochemicals) and diluted 1:200–1:400. Actin was labeled with phalloidin Alexa Fluor 647 Plus (1:400, Thermo Fisher Scientific).

Fluorescence immunocytochemistry

Fluorescence immunocytochemistry of neuronal cultures for wide-field microscopy, SMLM, and live-cell imaging was performed as in published protocols (Jimenez et al., 2020) with minor modifications. Cells were fixed using 4% PFA in PEM buffer (80 mM PIPES, 5 mM EGTA, 2 mM MgCl₂), pH 6.8, for 10 min at room temperature (RT). After rinses in 0.1 M phosphate buffer (PB), neurons were blocked for 1–2 h at RT in immunocytochemistry buffer (ICC, 0.22% gelatin; 0.1% Triton X-100 in PB) and incubated with primary antibodies diluted in ICC for either 1.5 h at RT or overnight at 4°C. After rinses in ICC, neurons were incubated with secondary antibodies diluted in ICC for 1 h at RT and rinsed. Actin staining was then performed by incubating in fluorescent phalloidin at 0.5 μM for either 1 h at RT or overnight at 4°C. Stained coverslips were kept in PB + 0.02% sodium azide at 4°C until imaging by SMLM. For wide-field microscopy, coverslips were mounted in ProLong Glass (Thermo Fisher Scientific).

Epifluorescence microscopy

Diffraction-limited epifluorescence images were obtained using an Axio-Observer upright microscope (Zeiss) equipped with a 40× NA 1.4 or 63× NA 1.4 objective and an ORCA Flash 4.0 camera (Hamamatsu Photonics). Appropriate hard-coated filters and dichroic mirrors were used for each fluorophore. A thin Z-stack of 3–10 slices spaced by 0.2 µm was acquired to include the signal from all neuronal processes within the whole field of view. For illustration images, image editing was performed using Fiji (Schindelin et al., 2012) with linear contrast adjustment to highlight the faint actin labeling along axons.

Epifluorescence image analysis

Percentage of patterned axon quantifications were performed by defining linear regions of interest (ROIs) and measuring their length using the Fiji software. Patterned axon ROIs were identified along isolated axons (devoid of dendritic contact) as presenting at least three successive presynaptic aggregates next to nlgn dots; other isolated axon segments were defined as nonpatterned. The percentage of patterned axons was calculated over the total length (patterned + nonpatterned ROIs).

Intensity quantifications were performed on maximum projections of the raw epifluorescence data with no further adjustment. Short linear ROIs were defined onto the presynaptic cluster aggregates at nlgn dot-induced presynapses (S+), at natural (axon–dendrite) synapses (NS), and at the axonal shaft flanking the presynapses as control (S−). Those tracings were translated into Fiji ROIs and refined to the presynaptic cluster (S+ and NS) or axon shaft (S−) using the ProFitFeat script (available at https://github.com/cleterrier/Measure_ROIs/blob/master/Pro_Feat_Fit.js), limiting it to the segment with an intensity above 40% of the maximum intensity along the linear ROI. The background-corrected mean intensity along these ROIs was then measured for each labeled channel. Induced presynapses were visually categorized as “actin enriched” (A+ on figures) or “low-actin” (A− on figures) depending on the relative intensity of actin at the presynapse compared with the neighboring axon shaft (Bingham et al., 2023).

SMLM: STORM and PAINT

Both STORM and DNA-PAINT acquisitions were performed on an N-STORM microscope (Nikon Instruments). The N-STORM system uses an Agilent MLC-400B laser launch with 405 nm (50 mW maximum fiber output power), 488 nm (80 mW), 561 nm (80 mW), and 647 nm (125 mW) solid-state lasers, a 100× NA 1.49 objective, and an iXon DU-897 camera (Andor). After locating a patterned axon using low-intensity epifluorescence illumination, a STORM or DNA-PAINT acquisition was performed using laser illumination in HiLo (grazing angle) configuration. An astigmatic lens was added to the light path to achieve 3D imaging. For STORM, stained coverslips were mounted in a silicone chamber filled with STORM buffer (Smart Buffer kit; Abbelight), and 30,000–60,000 images (256 × 256 pixels, 15 ms exposure time) were acquired at 100% 647 nm laser power. Reactivation of fluorophores was performed during acquisition by increasing illumination with the 405 nm laser. For DNA-PAINT, stained coverslips were mounted in a Ludin chamber filled with an imaging buffer (Massive Photonics). Imaging strands conjugated to Atto 643 (Massive Photonics) were added from a starting concentration of 0.1 nM each and then adjusted to optimize blinking density. The 30,000–45,000 images (256 × 256, 40 ms exposure time) were acquired at 50–60% laser power for 647 nm laser excitation.

To obtain optimal STORM images of presynaptic actin, we carefully selected induced presynapses with unambiguous proximity with a nlgn dot along an isolated, continuous axon. We avoided situations where the nlgn dot was partially printed or erased, as well as regions of axons crossing or bundling.

For STORM and DNA-PAINT images, acquired stacks were processed using DECODE (Speiser et al., 2021). Briefly, PSFs were modeled using spline fitting in SMAP (Y. Li et al., 2018) and used to simulate sequences of blinking events using characteristics (photon number range and lifetime distribution) inferred from real acquisition data from the N-STORM microscope. A PyTorch model was trained to infer the 3D coordinates and uncertainty of the simulated blinking events and then applied to the experimental acquired sequence. The resulting localizations (fitted blinking events) were filtered based on uncertainty, and drift during acquisition was corrected in 3D using a redundant cross-correlation algorithm (Wang et al., 2014) implemented as an independent module of SMAP (Ries, 2020). After translation of the coordinate files, image reconstructions were performed using the ThunderSTORM ImageJ plugin (Ovesný et al., 2014) in the Fiji software. Custom scripts and macros were used to translate coordinate files, as well as automate image reconstruction for whole images at 16 nm/pixel for visualization and at 8 nm/pixel for cluster analysis (https://github.com/cleterrier/ChriSTORM).

SMLM: spectral-demixing STORM

Neurons were fixed and stained for actin using phalloidin–Alexa Fluor 647 Plus and for Munc13-1 using a secondary antibody coupled to CF680 (Biotium). Coverslips were secured to a silicone perfusion chamber (Electron Microscopy Sciences) filled with reducing imaging buffer (Smart Kit; Abbelight) fixed to a glass slide (Jimenez et al., 2020). Prepared samples were mounted on a Nikon ECLIPSE Ti2 inverted microscope (Nikon), equipped with a 100×, 1.49 NA oil-immersion objective lens. HiLO illumination was used to restrict illumination to ∼1 μm above the glass interface. Illumination was provided by two 640 nm continuous wave diode lasers (Oxxius) with a combined power between 300 and 400 mW measured at the back aperture. Fluorescence emission was detected in a split light path optimized for three-dimensional spectral demixing (Friedl et al., 2023), and 60,000 frames were recorded at an exposure of 20 ms. Paired single-molecule localizations were detected using the globLoc fitting algorithm (Li et al., 2022), incorporated in the SMAP software (Ries, 2020). Localization coordinates were further processed and reconstructed using ThunderSTORM and ChriSTORM.

Identification of actin nanostructures on STORM images

A high-magnification image (4 nm/pixel) was generated for each induced presynapse image and manually registered by translation with the corresponding epifluorescence images in other channels, including presynaptic marker. Actin structures present at the induced presynapse were categorized into three types using the following criteria, developed after independent assessment by two experimenters (Bingham et al., 2023):

1. Actin mesh: a small cluster of actin next to the nlgn dot contact and within the presynaptic active zone marker cluster

2. Actin rails: linear structures within the presynaptic marker cluster

3. Actin corrals: large actin clusters at the periphery or just surrounding the presynaptic marker cluster

To assess intraobserver reproducibility and minimize classification bias, selected datasets were independently reanalyzed by the same experimenter after a minimum interval of 2 weeks, over a period of 6 months. Consistent region selection and category assignment were observed across analyses.

The size of the actin nanostructures was determined using manual outlines on 2D projection of actin STORM images in Fiji, aided by visualization of 3D renderings using the ChimeraX software (Goddard et al., 2018).

Spectral-demixing STORM image analysis

For the determination of the axial distances of the actin structures and Munc13-1 clusters to the glass in spectral-demixing STORM experiments, 4 nm per pixel isotropic stacks of induced presynapses were reconstructed using ThunderSTORM. The contour of the presynaptic ROI was drawn based on an epifluorescence image of synaptophysin labeling, and then the axial intensity profile was generated across the Z-stack within this ROI for the actin and Munc13-1 channels. A ROI devoid of axons next to the presynapse was also drawn, and axial intensity profiles generated which presented a sharp peak at the glass level in the Munc13-1 channel due to the presence of background staining by the secondary antibody. Gaussian fits of the intensity profiles were used to determine the average Z level of the glass, as well as the actin and Munc13-1 labeling within the presynapse, from which actin to glass and Munc13-1 to glass distances were derived.

Neuronal transfection and live-cell imaging

DIV13 cultured neurons, at a density of 6,000 cells/cm², were transfected using Lipofectamine 2000 (Thermo Fisher Scientific) with 0.15 μg of VAMP2-pHluorin. Following 30 min incubation in NB (Thermo Fisher Scientific) at 37°C, 5% CO₂, coverslips were returned to their original culture dishes. After 16 h, live-cell imaging of neurons expressing VAMP2-pHluorin was performed on an inverted microscope ECLIPSE Ti2-E (Nikon Instruments) equipped with an ORCA-Fusion sCMOS camera (Hamamatsu Photonics) and a CFI SR HP Apochromat TIRF 100× oil (NA 1.49) objective. The system was equipped with a Nikon Perfect Focus System, and images were acquired with the NIS-Elements software. Coverslips with neurons were mounted in a Ludin metal chamber in Tyrode's solution (119 mM NaCl, 25 mM HEPES, 2.5 mM KCl, 2 mM CaCl₂, 2 mM MgCl₂, 30 mM glucose), pH 7.35. Neurons were maintained in a humid chamber at 35.5–37°C for the duration of the experiments using a stage-top incubator (Okolab). To image spontaneous exocytosis occurring in the absence of stimulation, VAMP2-pHluorin present initially in the plasma membrane was photobleached by exposing it to high power 488 nm laser light for 15 s before time-lapse imaging to remove the basal membrane signal and highlight exocytic signal (Yudowski et al., 2007); 100 ms exposure frames were then continuously acquired for 5 min using 488 nm laser light at lower power (1–10%).

Time-lapse data processing

Acquired videos were preprocessed using the Filter Timelapse script (available at https://sites.imagej.net/Christopheleterrier/plugins/NeuroCyto%20Lab/Kymographs/), which performed image stabilization using the Image Stabilizer plugin (https://imagej.net/plugins/image-stabilizer) after 2× downscaling, and bleach correction using average intensity compensation for the foreground identified as the top 12% of pixel intensities within each image of the sequence. Exocytosis events along isolated axons across a field of nlgn dots were then identified with the pHusion plugin in the ImageTank software (O’Shaughnessy et al., 2024). Exocytotic events were detected by using the difference of Gaussian (DoG) method using a sigma of 3 and a scaling factor of 1.25. Using the ROI from the DoG image, a small image series was cropped from the raw image stack taking five frames before and 30 after the identified spot. Each image in the series was fitted with a Gaussian model. Only events with a goodness of fit for the Gaussian model >0.7 were considered. Events that drifted >4 pixels were removed. Identified exocytosis events were subsequently verified by visual inspection of the original time-lapse movies in Fiji.

Correlative live-cell/STORM

Live-cell time–lapse data and corresponding STORM images were registered with a two-step procedure. First, an RGB merge of the maximum-projected VAMP2-pHluorin movie and nlgn pattern from live-cell data was aligned to an RGB merge of the fixed VAMP2-pHluorin image from the STORM acquisition using a manually tuned affine transform using the BigDataViewer Fiji plugin (Pietzsch et al., 2015). The resulting affine transform was then applied to the time-lapse data and live-cell data using the TransformJ Fiji plugin to align it to STORM reconstructions and corresponding upsampled epifluorescence channels (at 16 or 4 nm per pixel depending on the zoom level). Vamp2-pHluorin time lapse was upsampled using a quintic B-spline approach (Meijering et al., 2001), and the exocytic localization was determined on the first frame where the exocytosis signal appeared, as the highest intensity pixel on the image upsampled to 16 nm per pixel. For the determination of the distance between the exocytic site and bassoon/Munc13-1 clusters, the center of mass of the first, second, and third closest clusters to the exocytic site was determined using the MTrackJ ImageJ plugin from Meijering et al. (2012), and the distance to the determined exocytic location was calculated. For the determination of the distance with the presynaptic actin nanostructures, the structures were manually segmented as ROIs and their center of mass calculated before deriving the distance to the exocytic location.

Statistics

Individual measurement points (n) from independent experiments (N) were pooled. Intensity profiles, graphs, and statistical analyses were generated using Prism. On bar graphs, dots (if present) are averages of each independent experiment, bars, or horizontal lines represent the mean, and vertical lines are the SEM unless otherwise specified. Significances were tested using one-way, nonparametric ANOVA with Šídák post hoc significance testing between selected conditions (Fig. 1; Figs. S1, S3) or an unpaired nonparametric Mann–Whitney test (Figs. 4, 5). On figures, the results of the post hoc significance are indicated as follows: ns, nonsignificant; *p < 0.05; **p < 0.01; and ***p < 0.001.

Figure 1.

Micropatterned neuroligin induces axonal presynaptic marker clustering and actin aggregation. A, Wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (green) fixed at 15 DIV, labeled for map2 (blue), adducin (gray), and bassoon (orange). B, Zooms corresponding to the areas highlighted in A. Top row, natural synapses at axo-dendritic contacts (NS); middle row, induced presynapses at isolated axon–micropattern dots (S+); bottom row, isolated axon with no induced presynapse (S−). C, Wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (green) fixed at 15 DIV, labeled for map2 (blue), βII-spectrin (gray), and synaptophysin (syp, orange). D, Zooms corresponding to the NS, S+, and S− highlighted in C. E, Wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (green) fed at 15 DIV with anti-synaptotagmin antibody (syt) during constitutive vesicular cycling (orange), fixed and labeled for map2 (blue) and actin (gray). F, Zooms corresponding to the NS, S+, and S− highlighted in E. Scale bars for A, C, and E, 20 µm. Scale bars for B, D, and F, 10 µm. G, Proportion of the length of isolated axons showing presynaptic specializations following the nlgn micropattern (patterned axon) over the total length of isolated axons selected within the experimental conditions (bassoon, synaptophysin, synaptotagmin, N = 3–5 independent experiments) compared with controls (pooled labels; Fig. S1). H, Mean intensity for bassoon (dark orange, N = 2; n > 100), synaptophysin (orange, N = 4; n > 200), synaptotagmin (light orange; N = 5; n > 200), and actin (gray) at natural synapses, induced presynapses (S+), and isolated axon devoid of presynapse (S−), normalized to NS.

Figure 4.

Multicolor localization microscopy of nlgn dot-induced presynapses reveals the arrangement of actin nanostructures relative to the active zone. A, First to third images, wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (green) fixed at 15 DIV, labeled for map2 (purple), synaptophysin (orange), and actin (gray). Scale bar, 5 µm (first image), 1 µm (second and third images, zooms of the area highlighted on the first image). Fourth image shows the nlgn pattern (green) superimposed with the PAINT image of the bassoon staining (orange) and the STORM image of the actin staining (gray). Scale bar, 1 µm. Fifth image is a zoom of the area highlighted on the fourth image showing the micropattern (green shape), bassoon clusters from the PAINT image (orange contours), and colorized presynaptic actin nanostructures from the STORM image (gray, corral in blue, mesh in red, rails in green). Scale bar, 500 nm. B, Zooms of the presynapse shown in A showing the STORM image of actin (gray) and PAINT image of bassoon (orange) in XY (left column) and transverse (XZ with a vertical Z-axis, right column) views. The transverse view is taken along the bracket shown on the XY view. Scale bars, 500 nm (XY views), 200 nm (transverse XZ views). C, ChimeraX renderings corresponding to the views shown in B with XY (left column) and transverse (right column) views of actin nanostructures (gray) and bassoon nanoclusters (orange). The transverse XZ views show how actin is present above and in-between bassoon nanoclusters. D, Top left, two-color spectral–demixing STORM image of an axon on top of the nlgn pattern (wide-field epifluorescence, green) fixed at 15 DIV and labeled for actin (gray) and Munc13-1 (orange). Scale bar, 1 µm. Top right is a zoom on the induced presynapse showing the nlgn dot contour (green circle) and actin (gray) with colorized presynaptic actin nanostructures (gray, corral in blue, mesh in red, rails in green). Scale bar, 500 nm. Bottom left is the same zoom (XY view) on the induced presynapse showing the nlgn dot contour (green circle), actin (gray), and Munc13-1 (orange). Scale bar, 500 nm. Bottom right is a transverse (XZ) view taken along the bracket shown on the XY view showing actin (gray) and Munc13-1 (orange) located below the actin structures. Scale bar, 200 nm. E, Another example of a two-color spectral–demixing STORM image of an induced presynapse with the same labels and scales as in D. F, Two-color spectral–demixing STORM image of a natural synapse (axon-dendritic spine contact) with the same labels and scales as in D. The transverse view (bottom right) shows the Munc13-1 clusters encased in presynaptic and postsynaptic actin. G, Axial intensity profiles and Gaussian fits for Munc13-1 (orange) and actin (purple) over the presynaptic area for the induced presynapses shown in D and E (top and middle profiles) and the natural synapse shown in F (bottom profile). The axial profile for the detection of the glass surface (gray) corresponds to an empty area outside of the presynapse. H, Average actin to glass (purple) and Munc13-1 to glass (orange) distances obtained from the peaks of the Gaussian fits for the axial intensity profiles in induced presynapses (n = 21 from 8 coverslips).

Figure 5.

Correlative imaging of the relationship between exocytic sites and presynaptic nanostructures. A, B, First image, live-cell imaging of the axon of a cultured neuron at 15 DIV on top of the nlgn pattern (green), transfected with VAMP2-pHluorin (VAMP2-pHl, orange, maximum projection of the 40 frames around the exocytic event). Scale bar, 10 µm. Second and third images, progressive zooms on an induced presynapse labeled for Munc13-1 (blue, epifluorescence on the second image, STORM on the third image) after fixation and aligned with the live-cell image. Scale bars, 5 µm (second image), 1 µm (third image). The right panel shows the first frame of the spontaneous exocytic event from the VAMP2-pHluorin movie (orange, top left), the corresponding STORM image of the Munc13-1 labeling (blue, bottom left) and the overlay of the exocytosis event (orange contour) with its fitted center (orange cross) with the Munc13-1 STORM image (blue) and nlgn micropattern (green shapes, right). Distances are indicated between the exocytic site and the center of the three closest Munc13-1 clusters. Scale bars, 500 nm. C, D, First image, live-cell imaging of the axon of a cultured neuron at 15 DIV on top of the nlgn pattern (green), transfected with VAMP2-pHluorin (orange, maximum projection of the 40 frames around the exocytic event). Scale bar, 10 µm. Second and third images, progressive zooms on an induced presynapse labeled for bassoon (blue, epifluorescence on the second image, STORM on the third image) after fixation and aligned with the live-cell image. Scale bars, 5 µm (second image), 1 µm (third image). The right panel shows the first frame of the spontaneous exocytic event from the VAMP2-pHluorin movie (orange, top left), the corresponding STORM image of the bassoon labeling (blue, bottom left), and the overlay of the exocytosis event (orange contour) with its fitted center (orange cross) with the bassoon STORM image (blue) and nlgn micropattern (green shapes, right). Distances are indicated between the exocytic site and the center of the three closest bassoon clusters. Scale bars, 500 nm. E, F, First image, live-cell imaging of the axon of a cultured neuron at 15 DIV on top of the nlgn pattern (green), transfected with VAMP2-pHluorin (orange, maximum projection of the 40 frames around the exocytic event). Scale bar, 10 µm. Second and third images, progressive zooms on an induced presynapse labeled for actin (gray, epifluorescence on the second image, STORM on the third image) after fixation and aligned with the live-cell image. Scale bars, 5 µm (second image), 1 µm (third image). The right panel shows the first frame of the spontaneous exocytic event from the VAMP2-pHluorin movie (orange, top left), the corresponding STORM image of the actin labeling (gray, bottom left), and the overlay of the exocytosis event (orange contour) with its fitted center (orange cross) with the actin STORM image (gray) with colored nanostructures (blue, corral; red, mesh; green, rails) and nlgn micropattern (green shapes, right). Distances are indicated between the exocytic site and the center of the two closest of each type of actin nanostructures. Scale bars, 500 nm.

Results

Micropatterned neuroligin induces axonal presynaptic marker clustering and actin aggregation

To improve optical accessibility and homogeneity of presynaptic formation, we adapted a validated method of presynaptic induction, where coating the coverslip with the extracellular domain of a postsynaptic cell-adhesion molecule, neuroligin-1 fused to the N-terminal of the human immunoglobulin–Fc region (nlgn), induces the formation of excitatory presynapses onto the glass (Funahashi et al., 2018). We used microcontact printing (Théry, 2010) of a PDMS stamp with 1.8-µm-diameter pillars to generate an array of ∼2 µm dots on the coverslip. To induce presynapses, the pillars were coated with nlgn together with fluorescent BSA for pattern visualization (nlgn dots). After printing, the coverslips were coated with a uniform layer of polylysine and laminin, so that axons and dendrites could grow freely and react when crossing printed dots. Rat hippocampal neurons were then cultured onto printed coverslips and subsequently fixed and immunolabeled for presynaptic markers after 14–15 DIV to assess presynaptic formation. Isolated axons containing at least three successive induced presynapses on neuroligin dots were classified as “patterned axons.” This avoided counting spurious isolated clusters of presynaptic components close to a nlgn dot as induced presynapses. The formation of patterned axons was measured (as the percentage of total length of isolated axons over the field of view) for two presynaptic proteins: the active zone scaffolding protein bassoon (Dieck et al., 1998) and the synaptic vesicle membrane protein synaptophysin (Jahn et al., 1985; Wiedenmann and Franke, 1985). In addition, active vesicular cycling was evaluated using feeding with a synaptotagmin antibody against an extracellular epitope that accumulates in presynapses through constitutive vesicular cycling (Kraszewski et al., 1995). Representative fluorescence images demonstrate aggregation of bassoon, synaptophysin, and synaptotagmin uptake into axonal clusters near nlgn dots (Fig. 1A–F). All three markers yielded comparable occurrences of patterned axons along nlgn dots patterns (23 ± 3% with bassoon, 16 ± 3% with synaptophysin, and 23.3 ± 2% with synaptotagmin; Fig. 1G), while neurons cultured on patterns of dots containing a human Fc without fused neuroligin or outside of the micropatterned areas resulted in almost no patterned axons (1.5 ± 0.4%, Fig. 1G with more detailed controls on Fig. S1).

On zoomed views, the morphology of bassoon (Fig. 1B), synaptophysin (Fig. 1D), and synaptotagmin uptake (Fig. 1F) clusters are similar between nlgn dot-induced presynapses (noted S+) and natural synapses (noted NS) between axon and dendrites of the same culture. We quantified the signal intensity of each presynaptic marker at ngln dot-induced presynapses and compared it with natural synapses and neighboring axonal segments devoid of presynaptic clusters (noted S−; Fig. 1H). Induced presynapses (S+) had a lower intensity compared with natural synapses for bassoon (0.89 ± 0.02) and synaptophysin (0.77 ± 0.02), and the synaptotagmin intensity was not significantly different (0.88 ± 0.04). Control axonal segment (S−) showed very low values of presynaptic component intensity (Fig. 1H). We then compared the actin content of induced presynapses with natural synapses: as expected, the presence of concentrated postsynaptic actin resulted in an actin intensity nearly three times higher in natural synapses (normalized to 1) compared with induced presynapses (S+, 0.38 ± 0.02). However, these induced presynapses had a two times higher actin content than nearby nonsynaptic axonal segments (S−, 0.19 ± 0.01; Fig. 1H), confirming the relative concentration of axonal actin at presynapses (Morales et al., 2000). Altogether, these data show that we can reliably induce the formation of presynapses on micropatterned dots of neuroligin in a minimally perturbed neuronal culture, validating our presynapse-on-glass model.

Actin enrichment correlates with increased presynaptic component concentration at nlgn dot-induced presynapses

Previous work by our group on bead-induced presynpases revealed that two categories exist: presynapses showing actin enrichment compared with the surrounding axon (actin-enriched) and presynapses lacking actin enrichment (low-actin). Furthermore, actin enrichment was associated with a higher concentration of presynaptic components and higher cycling activity (Bingham et al., 2023). Consistent with these findings, on micropatterned nlgn dots, induced presynapses could be categorized into actin-enriched and low-actin from diffraction-limited images (Fig. 2A,B): actin-enriched induced presynapses representing 63% of nlgn dot-induced presynapses and low-actin presynapses representing 37% (Fig. 2C). We then quantified the enrichment of actin and presynaptic components at nlgn dot-induced presynapses in each category, measuring the fluorescence intensity in actin-enriched and low-actin–induced presynapses as well as the adjacent axonal shaft as a control. Normalized to its intensity in actin-enriched–induced presynapses, actin intensity was reduced in low-actin presynapses and in axon shafts (0.38 ± 0.02 and 0.38 ± 0.1, respectively; Fig. 2D). Moreover, low-actin presynapses demonstrated a significantly reduced concentration of bassoon and synaptotagmin (0.73 ± 0.05 and 0.66 ± 0.05, respectively).

Figure 2.

Actin enrichment correlates with increased presynaptic component concentration at nlgn dot-induced presynapses. A, Wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (gray) fixed at 15 DIV, labeled for actin (gray), bassoon (orange), nlgn (green), and map2 (purple). Scale bars, 20 µm. B, Zooms corresponding to the areas highlighted in A displaying (from left to right) overlayed channels; actin (gray); actin with nlgn dots (green circles) near actin-enriched–induced presynapses (A+, arrowheads), and low-actin presynapses (A−, empty arrowheads); presynaptic marker bassoon (orange) and nlgn dots (green circle). Scale bars, 10 µm. C, Percentage of actin-enriched (dark gray) and low-actin–induced (light gray) presynapses along axons on nlgn dots (N = 4). D, Mean intensity of actin (gray; N = 3; n > 300), bassoon (dark orange; N = 2; n > 150), and synaptotagmin (syt, light orange, N = 3; N > 130) at actin-enriched presynapses (A+), low-actin–induced presynapses (A−), and isolated axons with no induced presynapses (S−), normalized to actin-enriched presynapses. E, Wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (gray) fixed at 9 DIV, labeled for actin (gray), synaptophysin (orange), nlgn (green), and map2 (purple). Scale bars, 20 µm. F, Zooms corresponding to the areas highlighted in E displaying (from left to right) overlayed channels; actin (gray); actin with nlgn dots (green circles) near actin-enriched–induced presynapses (A+, arrowheads) and low-actin presynapses (A−, empty arrowheads); presynaptic marker synaptophysin (syp, orange); and nlgn dots (green circle). Scale bars, 10 µm. G, Wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (gray) fixed at 14 DIV, labeled for actin (gray), synaptophysin (orange), nlgn (green), and map2 (purple). Scale bars, 20 µm. H, Zooms corresponding to the areas highlighted in G displaying (from left to right) overlayed channels; actin (gray); actin with nlgn dots (green circles) near actin-enriched–induced presynapses (A+, arrowheads) and low-actin presynapses (A−, empty arrowheads); presynaptic marker synaptophysin (syp, orange); and nlgn dots (green circle). Scale bars, 10 µm. I, Percentage of actin-enriched (dark gray) and low-actin–induced (light gray) presynapses along axons on nlgn dots at 9 DIV (left) and 14 DIV (right, N = 2). J, Mean intensity of actin (gray; N = 2; n = 28–109) and synaptophysin (syp, light orange, N = 3; n = 28–111) at actin-enriched presynapses (A+), low-actin–induced presynapses (A−), and isolated axons with no induced presynapses (S−), normalized to actin-enriched presynapses.

The coexistence of both actin-enriched and low-actin presynapse populations, for which different presynaptic component concentrations and vesicle cycling rate behaviors are associated, could be the result of an ongoing maturation for a subset of presynapses at 15 DIV. To test this hypothesis, we performed parallel experiments with fixation of micropatterned cultures at 9 and 14 DIV (Fig. 2E–H), measuring the proportion of actin-enriched and low-actin presynapses as well as the concentration of presynaptic component (Fig. 2I,J). Neurons at 9 and 14 DIV show a similar majority population of actin-enriched–induced presynapses (64% of nlgn dot-induced presynapses at both 9 and 14 DIV; Fig. 2I), with actin concentration dropping to 0.37 ± 0.04 and 0.53 ± 0.04 at 9 and 14 DIV, respectively, in low-actin presynapses (Fig. 2J). The concentration of synaptophysin was similarly distributed, being higher in actin-enriched presynapses and decreasing to 0.72 ± 0.06 and 0.74 ± 0.04 in low-actin presynapses at 9 and 14 DIV, respectively (Fig. 2J). These stable results show that progressive maturation is unlikely to explain the ∼35% population of low-actin presynapses. The two populations of presynapse coexist throughout culture stages, similarly to what we observed in bead-induced presynapses (Bingham et al., 2023). Overall, nlgn dot-induced presynapses recapitulate the presence of a majority of actin-enriched presynapses that contain more presynaptic components than their low-actin counterpart, further validating our presynapse-on-glass model.

STORM of nlgn dot-induced presynapses reveal discrete actin nanostructures

We further interrogated actin structural organization to assess if the distinct presynaptic actin nanostructures we previously identified in bead-induced presynapses (Bingham et al., 2023) were present on nlgn dot-induced presynapses. We performed STORM of actin at induced presynapses, targeting actin-enriched and low-actin presynapses distinguished at the diffraction-limited resolution and sometimes appearing side-by-side along an axon (Fig. 3A). STORM reconstruction first confirmed that the periodic actin rings are disrupted at nlgn dot-induced presynapses, as shown previously for the periodic spectrin scaffold at presynapses (He et al., 2016; Sidenstein et al., 2016; Fig. 3B).

Figure 3.

STORM of nlgn dot-induced presynapses reveals discrete actin nanostructures. A, Wide-field fluorescence image of cultured neurons on top of the nlgn micropattern (green) fixed at 15 DIV, labeled for actin (gray) and bassoon (orange) with an axon showing an actin-enriched (A+) and a low-actin–induced (A−) presynapse. Scale bar, 4 µm. B, Zooms corresponding to two distinct induced presynapses highlighted in A. Left column shows the epifluorescence image of actin (gray) and the position of nlgn dots (green circle). Second column displays the STORM image of actin (gray), epifluorescence bassoon (orange), and the position of nlgn dots (green circle). Third column shows the STORM image of actin alone (gray). Fourth column shows colorized presynaptic actin nanostructures: corrals (blue), mesh (red), and rails (green). Scale bar, 1 µm. C, Additional examples of axon segments with an actin-enriched (A+, first row) and a low-actin–induced (A−, second row) presynapse. Left column shows the epifluorescence image of actin (gray) and the position of nlgn dots (green circle). Second column displays the STORM image of actin (gray), epifluorescence bassoon (orange), and the position of nlgn dots (green circle). Third column shows the STORM image of actin alone (gray). Fourth column shows colorized presynaptic actin nanostructures: corrals (blue), mesh (red), and rails (green). Scale bar, 1 µm. D, Proportion of presynapses where each nanostructure is detected for actin-enriched and low-actin–induced presynapses (N = 3): actin corrals (blue), actin mesh (red), and actin rails (green). E, Areas of presynaptic actin nanostructures as measured on STORM images, comparing actin-enriched and low-actin–induced presynapses. Each point represents an individual measurement (N = 3; n = 41 presynapses for actin-enriched and 16 for low-actin presynapses).

Furthermore, we were able to identify the three distinct types of actin nanostructures previously described on bead-induced presynapses (Bingham et al., 2023) within nlgn dot-induced presynapses (Fig. 3B,C). Firstly, a large branched and dense structure surrounding the periphery of the presynapse is identified as the actin corral (blue). Secondly, the small linear structures within the bouton often connecting wider and branched structures of the presynapse are classified as actin rails (green). Thirdly, the faint cluster of actin nanostructures colocalizing with the active zone bassoon clusters is recognized as the actin mesh (red). Quantitative analysis of these discrete actin nanostructures showed that all actin-enriched and low-actin presynapses exhibited a peripheral corral. The presence of the fainter rails was detectable in 71 and 69% of actin-enriched and low-actin presynapses, respectively, while the active zone mesh was detected in 83 and 81% of these (Fig. 3D). We refined our analysis by measuring the area of each nanostructure on STORM images: only the area of corrals showed a significant difference between actin-enriched and low-actin–induced presynapses, with actin-enriched presynapses being larger (0.23 ± 0.01 vs 0.17 ± 0.03 µm² for actin-enriched and lows actin, respectively; Fig. 3E). The presence of all three nanostructures in nlgn dot-induced presynapses and the main difference between the two populations of induced presynapses being the relative size of the actin corral surrounding them are in line with our previous results on bead-induced presynapses (Bingham et al., 2023).

Multicolor localization microscopy of nlgn dot-induced presynapses reveals the arrangement of actin nanostructures relative to the active zone

At this point we confirmed that our presynapse-on-glass model recapitulates the results obtained on the bead-induced presynapse model both in terms of actin content and nanostructures. To demonstrate the added value of our presynapse-on-glass model, we leveraged the reproducible and controlled orientation of the induced presynapses by mapping the relationship between actin nanostructures and presynaptic components at the nanoscale. We first performed two-color localization microscopy at nlgn dot-induced presynapses via sequential STORM of actin and DNA-PAINT of the active zone scaffold protein bassoon, using somatodendritic map2 and axonal βII-spectrin costaining to identify presynapses along isolated axons apposed to nlgn dots (Fig. 4A). We linked the position of the identified actin nanostructures to the bassoon nanoclusters by examining XY and transverse XZ reconstructions (Fig. 4B) as well as ChimeraX renderings (Fig. 4C). Bassoon labeling was made of distinct nanoclusters, as previously described (Glebov et al., 2017). These clusters were located at the bottom of the axon, consistent with the formation of the presynapse toward the glass. The actin corral outlines the periphery of the presynapse, in continuity with rails branching inwardly toward the central part of the presynapse. The structure identified as the mesh based on diffraction-limited views of presynaptic markers could often be seen above the bassoon clusters rather than below them, interdigitating between the clusters to separate them.

We then took advantage of the controlled orientation of the nlgn dot-induced presynapses to obtain a more quantitative view of the axial distribution of presynaptic actin, using spectral-demixing 3D STORM (Friedl et al., 2023) to image induced presynapses stained for actin and Munc13-1, a component of the presynaptic release machinery (Augustin et al., 1999). Spectral-demixing STORM could visualize the nanoscale actin structures (corral, mesh, and rails) as well as the clusters of Munc13-1 on zoomed XY views (Fig. 4D,E). Transverse XZ views (Fig. 4D,E) as well as intensity profiles within the presynapse along the Z-axis (Fig. 4G, top and middle profiles) consistently showed actin structures located above the Munc13-1 clusters, in contrast to natural synapses such as those found on dendritic spines (Fig. 4F) where presynaptic and postsynaptic actin are overlapping and Munc13-1 clusters are not closer to the glass (Fig. 4G, bottom profile). Quantification of axial intensity profiles from 21 presynapses revealed the average distances to the glass at 62 ± 21 and 116 ± 18 nm for Munc13-1 and actin, respectively (Fig. 4H). These experiments demonstrate how our presynapse-on-glass model can help map the nanoscale organization of presynaptic components thanks to its controlled orientation of presynapse induction.

Correlative imaging of the relationship between exocytic sites and presynaptic nanostructures

To further demonstrate the usefulness of our presynapse-on-glass approach, we took advantage of its controlled en face orientation to set up a live-cell/STORM correlative workflow and resolve the nanoscale environment of synaptic vesicle release events. Neurons were cultured on nlgn-patterned coverslips and transfected with VAMP2-pHluorin, a fusion of the VAMP2/synaptobrevin membrane protein (Baumert et al., 1989) with a pH-sensitive GFP that fluoresces upon exocytosis of the tagged synaptic vesicles (Miesenböck et al., 1998; Burrone et al., 2006), followed by live-cell total internal reflection fluorescence imaging to image spontaneous synaptic vesicle exocytosis at nlgn dots (Fig. 5, first column). The neurons were then fixed and immunolabeled for epifluorescence and single-color STORM imaging of either active zone proteins or presynaptic actin, and the precise location of the exocytic event (from fitting the appearing spot on time-lapse images of VAMP2-pHluorin) was correlated to the nanoscale arrangement from the aligned STORM image (Fig. 5).

As a proof of concept, we first contextualized the exocytic event by localizing presynaptic components with STORM images of Munc13-1 and the active zone scaffold protein bassoon (Südhof, 2012). Subresolution fitting of VAMP2-pHluorin flashes at nlgn-induced presynapses resulted in exocytic localizations located 75–200 nm from Munc13-1 clusters (Fig. 5A,B) and from bassoon clusters (Fig. 5C,D). We then performed a similar experiment, this time combining live-cell imaging of VAMP2-pHluorin with STORM of actin, counterstaining for bassoon to confirm presynapse induction at individual nlgn dots (Fig. 5E,F). Comparing the localization of single spontaneous exocytic events to labeled actin nanostructures revealed that at the nanoscale, exocytosis occurs in presynaptic regions devoid of actin, with the actin mesh and rails being closer to the exocytic site than the corrals (Fig. 5E,F). These results are qualitative due to the low throughput of the correlative approach, and together with the two-color nanoscale imaging of bassoon and actin (Fig. 4), they support an emerging model where actin is interspaced with bassoon clusters to structure the active zone in different nanoscale areas, with further segregation within these areas between bassoon scaffold clusters and Munc13-1/RIM-enriched exocytic zones (Glebov et al., 2017; Miki et al., 2024).

Discussion

In this study, we designed and optimized a presynapse-on-glass model to more precisely investigate actin nanoarchitecture in induced presynapses. By printing the extracellular domain of neuroligin-1 before coating the coverslips with polylysine and laminin, we developed a micropatterned substrate that promotes natural development of neurons over the whole coverslip. This contrasts with previous methods that coaxed presynapse induction by overexpressing presynaptic neurexin in addition to coating the coverslip with neuroligin (Funahashi et al., 2018; Tanaka et al., 2023) or restrict axon attachment to small spots separated by cell-repellent substrates (Czöndör et al., 2013). Compared with our previous model of presynapses induced by ∼5 µm beads (Bingham et al., 2023), our new model drives presynaptic orientation toward the glass, improving visual accessibility for nanostructure visualization and unbiased exocytosis analysis.

We demonstrate that nlgn dots can induce presynaptic specializations along isolated axons, with concentration of presynaptic components (bassoon, synaptophysin) and active vesicular cycling demonstrated by synaptotagmin antibody uptake. Induced presynapses show a similar morphology, fluorescence intensity of presynaptic components, and vesicular cycling activity compared with natural synapses found at axon–dendrite contacts. The quantification of the fraction of axons exhibiting induced presynapses over nlgn dots might appear low (∼20%), but it is a consequence of our stringent criteria for identifying them, excluding segments with less than three successive induced presynapses to avoid overestimating induction by including spurious clusters. It should be noted that our micropatterning strategy differs from previously published ones, where the presynapse-inducing patterns (1.5-µm-diameter dots) were complementary with a cytophobic substrate (polyethyleneglycol; Czöndör et al., 2013). In our model, an adhesion-promoting coating of polylysine and laminin is present uniformly, superimposed with the nlgn dots to allow natural growth of neurites over the coverslip with induction of presynapses when axons cross the nlgn dots. The concentration of neuroligin is likely to be higher on the dots than in natural postsynapses, and this could allow axons to react to their presence through the adhesive layer but may impact presynapse formation and maintenance. It would be interesting to test other postsynaptic proteins known to instruct presynaptic assembly such as SynCAM (Biederer et al., 2002; Czöndör et al., 2013) or LRRTM2 (Linhoff et al., 2009).

Besides their similarity to natural synapses, the absence of the more intense postsynaptic actin at induced presynapses allowed us to distinguish two populations of presynapses: 63% that are actin-enriched and 37% that are low-actin, a proportion close to the results obtained on bead-induced presynapses (67 and 33%; Bingham et al., 2023). Furthermore, actin-enriched–induced presynapses similarly accumulated ∼50% more presynaptic components compared with their low-actin counterparts, validating our nlgn dot-induced presynapse model. The presence of the low-actin population is unlikely to be a consequence of partial maturation (Yao et al., 2006), as these populations are similarly present after 9 and 14 d in culture. Whether the actin enrichment state of presynapses is stable or can interconvert during development or due to synaptic activity and more generally the nanoscale dynamics of actin rearrangement at presynapses will be a promising goal for future studies.

In addition to closely recapitulating the properties of bead-induced presynapses, the presynapse-on-glass model allowed nanoscale visualization with improved optical accessibility due to the formation of an active zone with a controlled orientation and proximity to the substrate. We could thus refine the characterization of the presynaptic actin nanostructures we previously identified (Bingham et al., 2023). Actin corrals are consistently found distributed at the periphery of presynapses, with the corrals of actin-enriched presynapses being significantly larger than those of low-actin presynapses. These variations in shape and size and the difference in presynaptic content between populations of presynapses suggest that corrals are the primary remodeled actin structure during presynaptic plasticity (O’Neil et al., 2021). Interestingly, we observed a greater percentage of both actin rails and actin mesh in nlgn dot-induced presynapses compared with the bead-induced presynapses (Bingham et al., 2023). This indicates that nlgn dot-induced presynapses indeed exhibit enhanced visibility and allow to more effectively distinguish between the corrals, rails, and mesh nanostructures.

One of the key advantages of the presynapse-on-glass model is the controlled orientation of the induced presynapses that form en face to the glass substrate. This allows unambiguous interpretation of the 3D nanoscale architecture, without the need for a postsynaptic marker that is often used to determine synapse orientation (Dani et al., 2010; Tang et al., 2016). We could verify the orientation of the induced presynapses by imaging the nanoscale arrangement of bassoon/Munc13-1 together with presynaptic actin architecture: bassoon and Munc13-1 nanoclusters were consistently found at the bottom of the axon, even when the presynapse formed on one side of the shaft, and this allowed easier identification of actin corrals, rails, and mesh within presynapses. Oriented 3D, multicolor SMLM images allowed us to refine our previous results by demonstrating that the actin mesh, previously coated at the active zone (Bingham et al., 2023), is found above and between bassoon clusters rather than below them. This suggests that rather than a submembrane assembly directly shaping the vesicle exocytosis area, the mesh could have a role in organizing nanoclusters within the active zone (Glebov et al., 2017), forming nanocompartments that could help guide synaptic vesicles to the release site (Miki et al., 2024).

Being able to visualize synaptic vesicle exocytosis in living cells, including rare spontaneous events, and to correlate them with the nanoarchitecture of presynaptic component is another application we demonstrate of our presynapse-on-glass model. We visualized individual exocytic events that likely correspond to single-vesicle release (Funahashi et al., 2018) and fit their subdiffraction localization. By correlating with fixed-cell STORM images of components within the same presynapse, we confirmed that exocytosis occurs next to bassoon and Munc13-1 clusters (Sakamoto et al., 2018), in line with their proposed complementary segregation in distinct nanoclusters within the active zone (Tang et al., 2016; Glebov et al., 2017). Furthermore, we used this correlative approach to further understand the role of presynaptic actin. We consistently visualized exocytic events away from actin nanostructures, including the actin mesh, confirming that actin seems to be mostly absent from the exocytic points and further suggesting that absence of actin is a predictor of vesicular release in the active zone (Morales et al., 2000). Our results on the nanoscale relationship between actin structures and presynaptic components and vesicular release are primarily qualitative at this point, due to the low throughput of our current correlative approach, and will undoubtedly guide future work by helping to refine functional hypotheses about the roles of actin in presynapses.