Flies were raised on standard cornmeal and molasses medium at 25^°C. Drosophila melanogaster male 3rd instar larvae of the following strains were used for experiments:

Wildtype: w¹¹¹⁸ [Bloomington Drosophila Stock Center (BDSC), Bloomington, IN, United States]. unc-13^GFSTF: y[1] w[*]; Mi{PT-GFSTF.0}unc-13[MI00468-GFSTF.0] (stock sent out to BDSC). Original MiMIC-strain (MI00468): y[1] w[*]; Mi{y[+ mDint2] = MIC}unc-13[MI00468]/In(4)ci[D], ci[D] pan[ciD] (BDSC #31015).

The Minos-mediated integration cassette (MiMIC) strain MI00468 was used to generate the GFP-tagged Unc-13 line (Venken et al., 2011; Nagarkar-Jaiswal et al., 2015a). Microinjection of the plasmid pBS-KS-attB1-2-PT-SA-SD-0-EGFP-FlAsH-StrepII-TEV-3xFlag (Drosophila Genomics Resource Center #1298, Venken et al., 2011) and all transgenesis steps were performed at Bestgene Inc. (Chino Hills, CA, United States). Larvae for injection were obtained from crosses of MI00468 males to ΦC31 integrase expressing female virgins. Removal of the genomic ΦC31 integrase source and PCR confirmation of the attP sites in MI00468 and of the correct orientation of the recombinase-mediated cassette exchang (RMCE) event were performed by the company. y[1] w[*]; Mi{PT-GFSTF.0}unc-13[MI00468-GFSTF.0] (unc-13^GFSTF) was shipped as an unbalanced, homozygous viable stock. Molecular confirmation of precise incorporation of the EGFP-FlAsH-StrepII-TEV-3xFlag (GFSTF, sfGFP used in recent versions of the multi-tag, Venken et al., 2011) tag was performed by genomic PCR with primer pairs fg_30f (AATGATAAAGGACAGGGACAAGGT) + fg_31r (CTGCTTCATGTGATCGGGGT), fg50_f (TGGATGGCGA CGTGAAC) + fg51_r (GGTTCCATGCAGCATCC) and fg_58f (CACAACGTGTACATCACCGC) + fg_59r (CTTGA GAACCTGCCGTCCAT). The PCR products were sequenced with fg_31r, fg50_f, and fg_58f.

For immunofluorescence imaging of larval neuromuscular junctions and ventral nerve cords, larvae were dissected in ice-cold hemolymph-like solution (HL-3, Stewart et al., 1994), fixed with 4% paraformaldehyde (PFA) in phosphate buffered saline (PBS) for 10 min and blocked for 30 min with PBT (i.e., phosphate buffered saline including 0.05% Triton X-100, Sigma, St. Louis, MI, United States) including 5% natural goat serum (NGS, Dianova, Hamburg, Germany). Primary antibodies were added for overnight staining at 4^°C. After two short and three long washing steps with PBT (60 min each for data shown in Figure 2C, 20 min each for all other imaging data), preparations were incubated with secondary antibodies for 3 h at room temperature, followed by two short and three 20 min long washing steps with PBT. Preparations were kept in PBS at 4^°C until imaging. All data were obtained from NMJs formed on abdominal muscles 6/7 in segments A2 and A3. Directly compared data were obtained from larvae stained in the same vial and measured in one imaging session. Primary antibodies were used in the following concentrations: mouse α-Brp (Brp^Nc82, 1:100; AB_2314866, Developmental Studies Hybridoma Bank, Iowa City, IA, United States), rabbit-α-GFP (1:1,000 for dSTORM or 1:3,000 for Zeiss Axiovert and structured illumination microscopy; A11122, ThermoFisher, Waltham, MA, United States), mouse α-GluRIIA (1:100; AB_528269, Developmental Studies Hybridoma Bank, Iowa City, IA, United States), and Cy5-conjugated α-horseradish-peroxidase (α-HRP, 1:250, AB_2338714, Jackson ImmunoResearch, West Grove, IA, United States).

For immunofluorescence imaging of adult brains, fly heads were dissected and brains removed in ice-cold PBS, fixed in 4% PFA in PBS for 30 min and blocked for 30 min with 0.6% PBT including 5% NGS. Incubation in primary antibodies, washing in 0.6% PBT, incubation in secondary antibodies and again washing followed the protocol explained in detail above. The same primary antibodies were used in the following concentrations: mouse α-Brp (Brp^Nc82, 1:100) and rabbit-α-GFP (1:1,000).

Preparation, fixation, and staining was performed as described above. Secondary antibodies were used in the following concentrations for larval neuromuscular junctions, ventral nerve cords and adult fly brains: goat α-rabbit conjugated Cy3 (1:500, AB_2338006, Jackson ImmunoResearch, West Grove, IA, United States) and goat α-mouse conjugated Alexa Fluor488 (1:500, A32723, Invitrogen, Waltham, MA, United States). Larval preparations were mounted in Vectashield (Vector Laboratories, Burlingame, CA, United States) for Axiovert imaging or Prolong Glass (ThermoFisher, Waltham, MA, United States) for structured illumination microscopy. Images were acquired at room temperature from NMJs on muscles 6/7 in segments A2 and A3. An Apotome System (Zeiss, Jena, Germany, Axiovert 200M Zeiss, objective 63x, NA 1.4, oil) was used for low-resolution characterization of larval NMJs, larval ventral nerve cords and adult brains (Figures 2A–C). For SIM we used a Zeiss Elyra S.1 structured illumination microscope equipped with a sCMOS camera (pco.edge 5.5 m) and an oil-immersion objective (Plan-Apochromat 63x, 1.4 NA). Lasers with 488, 531, and 641 nm wavelength were used. Z step size was set to 0.1 μm, and imaging was performed using five rotations of the grating at five different phase steps. Fourier transformation of structured illumination images was performed using ZEN software (ZEISS Efficient Navigation, Carl Zeiss, Jena, Germany), and subsequent analysis was done with ImageJ.

direct stochastic optical reconstruction microscopy imaging of the specimen was performed essentially as previously reported (Ehmann et al., 2014; Paul et al., 2015, 2022; Mrestani et al., 2021). Preparations were incubated with respective primary antibodies as described above. The following secondary antibodies were used: goat α-rabbit F(ab’)₂ fragments labeled with Alexa Fluor647 (1:500; A21246, Thermofisher, Waltham, MA, United States) and goat α-mouse IgGs labeled with Alexa Fluor532 (1:500; A11002, ThermoFisher, Waltham, MA, United States). After staining, larval preparations were incubated in 100 mM mercaptoethylamine (MEA, Sigma, St. Louis, MI, United States) in a 0.2 M sodium phosphate buffer, pH ∼7.9, to allow reversible switching of single fluorophores during data acquisition (van de Linde et al., 2008). Images were acquired using an inverted microscope (Olympus, Tokio, Japan, IX-71, 60x, NA 1.49, oil immersion) equipped with a nosepiece-stage (IX2-NPS, Olympus, Tokio, Japan). 647 nm (F-04306-113, MBP Communications Inc., Quebec, Kanada) and 532 nm (gem 532, Laser Quantum, Stockport, United Kingdom) lasers were used for excitation of Alexa Fluor647 and Alexa Fluor532, respectively. Laser beams were passed through clean-up filters (BrightLine HC 642/10 and Semrock, ZET 532/10, respectively), combined by two dichroic mirrors (LaserMUX BS 514-543 and LaserMUX BS 473-491R, 1064R, F38-M03, AHF Analysentechnik, Tübingen, Germany), and directed onto the probe by an excitation dichroic mirror (HC Quadband BS R405/488/532/635, F73-832, AHF Analysentechnik, Tübingen, Germany). The emitted fluorescence was filtered with a quadband-filter (HC-quadband 446/523/600/677, Semrock, West Henrietta, NY, United States) and a long pass- (Edge Basic 635, Semrock, West Henrietta, NY, United States) or bandpass-filter (Brightline HC 582/75, Semrock, West Henrietta, NY, United States) for the red and green channel, respectively, and divided onto two cameras (iXon Ultra DU-897-U, Andor, Oxford, United Kingdom) using a dichroic mirror (HC-BS 640 imaging, Semrock, West Henrietta, NY, United States). For the red channel, image resolution was 127 × 127 nm per pixel to obtain super-resolution of Unc-13^GFSTF. For the green channel, image resolution was 130 x 130 nm per pixel. Single fluorophores were localized and high resolution-images were reconstructed with rapidSTORM (Heilemann et al., 2008; Wolter et al., 2010, 2012; van de Linde et al., 2011)¹. Only fluorescence spots with an A/D count over 12,000 were analyzed and a subpixel binning of 10 nm px^–1 was applied.

Localization data were analyzed essentially as described previously (Mrestani et al., 2021) with an extension for two-channel localization data. Analysis was performed with custom written Python code (language version 3.6)² and the web-based Python interface Jupyter³. Localization tables from rapidSTORM were directly loaded and analyzed. Prior to the Python-based analysis the regions of interest (ROI) were masked in the reconstructed, binned images from rapidSTORM using FIJI (Schindelin et al., 2012). These ROIs corresponded to the terminal six boutons. For cluster analysis we used the Python implementation of HDBSCAN (McInnes et al., 2017)⁴ which takes “minimum cluster size” and “minimum samples” as the main free parameters. In the first step we extracted Brp clusters in the Alexa Fluor532 channel, corresponding to individual AZs, with the combination 100 and 25 for minimum cluster size and minimum samples (Figure 4Aiii; compare Mrestani et al., 2021). All unclustered localizations were discarded from further analysis. Brp clusters were used to remove noise from the Unc-13^GFSTF Alexa Fluor647 channel such that all Unc-13^GFSTF localizations with an euclidean distance >20 nm to Brp localizations were discarded. The H function (Figure 4B) as derivative of Ripley’s K function was computed using Python package Astropy (Robitaille et al., 2013) according to our previous algorithm (Mrestani et al., 2021) for the denoised Unc-13^GFSTF localizations and for the random Poisson distribution. Curves for display were averaged (mean ± SD). The function was evaluated in nm steps for radii from 0 to 120 nm and without correction for edge effects. A second HDBSCAN to extract the individual Unc-13^GFSTF subclusters (SC) was performed with minimum cluster size seven and minimum samples two to get SCs with a radius that matches the maximum of the H function. Unc-13^GFSTF SCs were assigned to individual Brp clusters by computing the euclidian distance between the center of mass (c.o.m.) of each SC and the c.o.m. of each Brp cluster and selecting the lowest distance. In that way, the number of SCs per AZ could be quantified. The distance between the c.o.m.s is referred to as radial distance and was computed for each individual AZ as mean of the assigned Unc-13^GFSTF SCs. To quantify cluster areas, we computed 2D alpha shapes using CGAL (Computational Geometry Algorithms Library)⁵ in Python. To get the alpha shapes of Brp clusters and Unc-13^GFSTF SCs we choose α-values of 800 and 300 nm², respectively. Exclusion criteria for Brp clusters were area <0.03 and >0.3 μm² (Mrestani et al., 2021). Unc-13^GFSTF SCs that were assigned to those Brp clusters were also excluded from further analysis, as well as SCs where alpha shape determination failed due to sparse signal that yielded SC areas of 0 μm². The Unc-13 area per AZ (Figure 4E) was computed as the sum of all Unc-13 SC areas belonging to an individual AZ. Brp cluster circularity was computed as described previously (Mrestani et al., 2021). For the analysis of Unc-13^GFSTF superclusters (SpC, Figures 5B,D,E and Supplementary Figures 2B,C) and distances between SC c.o.m.s (Figure 5C and Supplementary Figure 2A) only AZs with a circularity ≥0.6 were selected. The Python module scikit-learn (Pedregosa et al., 2011) was used to compute distances between SC c.o.m.s. To extract SpCs, HDBSCAN was performed taking the SC c.o.m.s of an individual AZ as input (minimum cluster size and minimum samples two, respectively). The cluster selection method was changed to “leaf” clustering, which comprises a tendency to more homogeneous clusters by extracting those that lie on leaf nodes of the cluster tree rather than the most stable clusters. The default setting “excess of mass,” that was used for all other cluster analyses throughout this study, delivered similar statistical results and median values (data not shown) but less intuitive clustering. For the statistical comparison of SpC numbers per AZ between experimental groups (Figure 5D) only AZs where SpCs could be detected were included. The SpC c.o.m. was defined as the c.o.m. of its respective SC c.o.m.s and the euclidean distance between these SC c.o.m.s and the SpC c.o.m. was computed as mean per SpC.

Two-electrode voltage clamp recordings (Axoclamp 2B amplifier, Digidata 1440A; Molecular Devices, San José, CA, United States) were obtained from abdominal muscle 6 in segments A2 and A3 as previously described (Paul et al., 2022). All measurements were obtained at room temperature in HL-3 (Stewart et al., 1994) with the following composition (in mM): NaCl 70, KCl 5, MgCl₂ 20, NaHCO₃ 10, trehalose 5, sucrose 115, Hepes 5, and CaCl₂ 1, pH adjusted to 7.2. Intracellular electrodes had resistances of 10–30 MΩ and were filled with 3 M KCl. For analysis, only cells with an initial membrane potential of at least –50 mV and a membrane resistance of ≥4 MΩ were included. During recordings, cells were clamped at a holding potential of –80 mV (minis) or –60 mV (evoked EPSCs). To evoke synaptic currents, nerves were stimulated via a suction electrode with pulses of 300 μs length and typically at 12 V (Grass S48 stimulator and isolation unit SIU5; Astro-Med, West Warwick, United States). Signals were low-pass filtered at 10 kHz and analyzed in Clampfit 11.1 (Molecular Devices, San José, CA, United States). Paired-pulse recordings were performed with interstimulus intervals of (in ms: 10, 30, 100, 300, and 1,000). Between recordings, cells were given a 10 s rest. For analysis, 5–10 traces per interval were averaged. The amplitude of the second response in 10 ms interpulse recordings was measured from the peak to the point of interception with the extrapolated first evoked excitatory postsynaptic current (eEPSC) as described previously (Hallermann et al., 2010; Weyhersmüller et al., 2011). To assess basal synaptic transmission 10 EPSCs evoked at 0.2 Hz were averaged per cell. The quantal content was estimated by dividing the mean eEPSC amplitude by the mean evoked excitatory postsynaptic current (mEPSC) amplitude measured in the same cell. mEPSC amplitudes were corrected for the more hyperpolarized holding potential (Hallermann et al., 2010).

Philanthotoxin 433 tris (trifluoroacetate) salt (PhTx, CAS 276684-27-6, Santa Cruz Biotechnology, United States) was dissolved in dimethyl sulfoxide (DMSO) to obtain a stock solution of 4 mM and stored at –20^°C. For each experiment, the respective volume was further diluted with freshly prepared HL-3 to a final PhTx concentration of 20 μM in 0.5% DMSO. Control experiments were performed with the same DMSO concentration in HL-3. PhTx treatment of semi-intact preparations was performed essentially as described previously (Frank et al., 2006; Mrestani et al., 2021). In brief, larvae were pinned down in calcium-free, ice-cold HL-3 at the anterior and posterior endings, followed by a dorsal incision along the longitudinal axis. Larvae were incubated in 10 μl of 20 μM PhTx in DMSO for 10 min at room temperature. Following this incubation time, PhTx was replaced by HL-3 and larval preparations were completed, followed by electrophysiological measurements or dSTORM imaging.

Statistical analyses were performed with Sigma Plot 13 (Systat Software, Düsseldorf, Germany) or GraphPad Prism 9 (San Diego, United States). Shapiro-Wilk was used to test normality. If data were not normally distributed, we used the non-parametric Mann-Whitney rank sum test for statistical analysis and reported data as median (25th–75th percentile). If data were normally distributed, they were reported as mean ± SD unless indicated otherwise. In box plots, horizontal lines represent median, boxes quartiles and whiskers 10th and 90th percentiles. Scatter plots show individual data points unless indicated otherwise. Bin counts in histograms were normalized to the total number of observed events which was set to one. All plots were produced with Sigma Plot. Figures were assembled using Adobe Illustrator (Adobe, 2015.1.1 release, San José, CA, United States). Supplementary Tables 1–4 contain all numerical values not stated in text and figure legends including p-values and samples sizes.

The authors declare that custom written Python code and all data sets supporting the findings of this work are available from the corresponding authors.

To generate an endogenously tagged unc-13 locus, we screened the previously established MiMIC library (Venken et al., 2011; Nagarkar-Jaiswal et al., 2015a,b) for an appropriate line. The MiMIC transposon harbors selection markers (yellow⁺ and enhanced GFP [EGFP]) and two inverted attP sites for recombinase-mediated cassette exchange (RMCE), as well as a gene-trap cassette (Venken et al., 2011). The insertion MI00468 carries a MiMIC element in a coding intron and, via RMCE with a protein-trap plasmid, allows the generation of a new exon that gets spliced in all annotated unc-13 transcripts (Figures 1A,B). This strategy was applied to insert a multi-tag, consisting of superfolder (sf) GFP (sfGFP), a modified Fluorescein arsenical helix binder (FlAsH) binding tetracysteine motif, the StrepII peptide tag, a TEV protease cleavage site and the Flag peptide tag [EGFP-FlAsH-StrepII-TEV-3xFlag (GFSTF) tag], using a previously published protein-trap plasmid (see Materials and methods; Venken et al., 2011; Nagarkar-Jaiswal et al., 2015b). The multi-tag combines the advantages of different protein and peptide tags including several possibilities for antibody staining, live imaging and RNAi knockdown. GFSTF is inserted at the same amino acid positions in the two major isoforms Unc-13A (Figure 1A) and Unc-13B (Figure 1B) but is C-terminally followed by alternative amino acids in Unc-13C, E, and F, resulting from the additional exon in the respective transcripts. The multi-tag is located at the C-terminal side of the large unstructured N-Terminus and is followed by all annotated domains of the proteins (Figure 1C). The animals are homozygous viable and appear healthy (e.g., emerging and effective flight). In the following, we focus on the detection of sfGFP using a combination of primary IgG antibodies and secondary F(ab’)₂ fragments to establish a protocol comparable to earlier applications of dSTORM imaging at the Drosophila NMJ (Ehmann et al., 2014; Paul et al., 2015, 2022; Mrestani et al., 2021). We refer to the fusion protein and the genotype as Unc-13^GFSTF and unc-13^GFSTF, respectively.

To assess expression of the endogenously tagged Unc-13 in the Drosophila nervous system, we performed immunostaining using a polyclonal antibody against GFP (see Materials and methods) for detection of the sfGFP within the GFSTF-tag and a well-characterized, highly specific monoclonal antibody Brp^Nc82 mapping to the C-terminal region of Bruchpilot (Brp, Kittel et al., 2006; Fouquet et al., 2009; Figure 2). We found strong and specific expression of Unc-13^GFSTF in the adult and larval central nervous system (Figures 2A,B). In addition, we detected reliable Unc-13^GFSTF expression at larval NMJs in co-expression with Brp (Figure 2C). Since Brp is an abundant presynaptic epitope and covers the spatial extent of an individual presynaptic AZ (Mrestani et al., 2021) we set out to analyze the co-localization of Brp and Unc-13^GFSTF. We employed SIM at individual type Ib boutons that show structural adaptation during presynaptic homeostasis (Weyhersmüller et al., 2011). Our data revealed co-localization of Unc-13^GFSTF and Brp at presynaptic AZs and of Unc-13^GFSTF and postsynaptic glutamate receptors using a monoclonal antibody against the GluRIIA subunit (Figure 2D). We conclude that incorporation of the fluorescent GFSTF-tag at the endogenous locus of unc-13 in Drosophila reports reliable and reasonable localization of the protein at presynaptic AZs allowing further investigations at the nanoscale level (compare Figures 4, 5).

To evaluate spontaneous and evoked synaptic release at larval NMJs of unc-13^GFSTF we performed two-electrode voltage clamp recordings (TEVC, Figure 3). First, mEPSCs were recorded to examine spontaneous vesicle release (Figure 3A). mEPSC amplitudes and the frequency of spontaneous fusion events were unchanged in unc-13^GFSTF compared to wildtype controls (Figure 3B). In addition, eEPSCs in response to nerve stimulation were unaltered in unc-13^GFSTF compared to controls (Figures 3C,D). Furthermore, quantal content was unchanged in unc-13^GFSTF (Figure 3D). Next, we tested whether insertion of the tag alters synaptic short-term plasticity. We found paired pulse ratios (PPR) unchanged in unc-13^GFSTF (Figure 3D). Taken together, these data reveal no differences between unc-13^GFSTF and control larvae in basal transmission properties of spontaneous and evoked synaptic release, making the endogenously tagged Unc-13 variant a valuable tool for the assessment of AZ function and structure.

In addition to its role in docking and priming of synaptic vesicles Unc-13 is involved in diverse presynaptic plasticity processes including Ca²⁺, DAG, or RIM-dependent short-term plasticity (Dulubova et al., 2005; Shin et al., 2010; Xu et al., 2017) and presynaptic homeostatic potentiation (PHP, Böhme et al., 2019). In Drosophila, upregulated Unc-13A levels and increased numbers of Unc-13A nanomodules at presynaptic AZs in acute and chronic PHP were observed (Böhme et al., 2019). Furthermore, functional PHP is abolished in unc-13A null mutants and the functional dependence can be attributed to the Unc-13A N-terminus. We decided to probe if the GFSTF-tagged unc-13 Drosophila strain is suitable for PHP analyses. To this end we measured the electrophysiological response to an acute homeostatic challenge using Philanthotoxin (PhTx) in unc-13^GFSTF animals compared to control larvae (Figure 3E). We found that upon PhTx treatment unc-13^GFSTF larvae showed the same increase in quantal content and thus evoked EPSC restoration as control larvae, thus, indicating that unc-13^GFSTF animals can still exhibit functional PHP to the full extent. We conclude that incorporation of the GFSTF-tag into the endogenous unc-13 locus does not disrupt this form of presynaptic plasticity in vivo.

Next, we performed localization microscopy in terms of two-color dSTORM (Figure 4; Heilemann et al., 2008; van de Linde et al., 2011; Löschberger et al., 2012; Ehmann et al., 2014; Paul et al., 2015; 2022; Mrestani et al., 2021; Pauli et al., 2021). We used Brp^Nc82 and a polyclonal antibody against GFP for detection of Unc-13^GFSTF in type Ib boutons and found Unc-13^GFSTF in discrete co-localization with Brp (Figure 4A). The HDBSCAN analysis extracted individual Unc-13^GFSTF subclusters (SCs) matching the maximum of the H function with diameters of ∼26 nm (Figures 4B,C). We aimed to investigate if induction of acute PHP leads to reorganization of Unc-13^GFSTF within the presynaptic AZ. Thus, we compared Brp and Unc-13^GFSTF localization data in PhTx treated preparations (phtx) and DMSO controls (ctrl, Figures 4C–E and Supplementary Figure 1). Using Alexa Fluor532 for detection of Brp^Nc82 we found decreased Brp cluster areas with similar localization numbers in phtx, as described before using Alexa Fluor647 (Supplementary Figure 1; compare Mrestani et al., 2021). Analysis of Unc-13^GFSTF SCs revealed no changes in localization numbers per SC or SC area in phtx (Figure 4D) with equal numbers of SCs per AZ (Figure 4E). Yet, the overall Unc-13^GFSTF area per AZ and the radial distance between the AZ c.o.m. and individual Unc-13^GFSTF SC c.o.m.s was decreased in phtx (Figure 4E). Our data reveal that Unc-13^GFSTF forms distinct subclusters within the AZ in co-localization with Brp which are reorganized during acute PHP.

Our super-resolution analysis so far showed ∼11 Unc-13^GFSTF SCs per AZ (Figure 4E). However, the number of docked vesicles at Drosophila type Ib boutons is known to be lower (Böhme et al., 2016; Reddy-Alla et al., 2017). Furthermore, the diameter of individual SCs is only ∼26 nm and thus smaller than the diameter of individual synaptic vesicles (Karunanithi et al., 2002). Therefore, we wondered whether Unc-13^GFSTF SCs might be spatially organized at a higher level into release sites. We employed a second round of HDBSCAN analysis focusing only on AZs in planar view (indicated by circularity values ≥0.6, Figure 5A). As illustrated in Figure 5B this evaluation revealed what we termed Unc-13^GFSTF superclusters (SpC), each containing several Unc-13^GFSTF SCs (Figure 5B). Distances between nearest-neighbors of all Unc-13^GFSTF SCs within an AZ increased in phtx compared to ctrl (Figure 5C and Supplementary Figure 2A). However, whereas median, 10th, 25^th, and 75^th percentiles were increased, the 90th percentile was decreased in phtx suggesting narrower distance distributions (Figure 5C). This indicates a more even allocation of Unc-13^GFSTF SCs at the available space in PHP. Next, we asked whether Unc-13^GFSTF superclustering occurs at all AZs at the NMJ and whether it is influenced by PHP. We found that most AZs contained 2–3 SpCs in both phtx and ctrl, however, a large fraction of AZs did not show superclustering at all (Figure 5D). Overall, about 50% of the Unc-13^GFSTF SCs of an individual AZ were clustered and one individual SpC consisted of ∼3 Unc-13^GFSTF SCs in both phtx and ctrl (Supplementary Figures 2B,C). Finally, we analyzed the distance between individual Unc-13^GFSTF SC c.o.m.s and their corresponding SpC c.o.m. in PHP (Figure 5E). Remarkably, this distance was increased in phtx compared to ctrl indicating expansion of Unc-13^GFSTF SpCs following the homeostatic challenge. In summary, PHP induces a more homogenous distribution of Unc-13^GFSTF SCs forming enlarged SpCs that may correlate with increased vesicle traffic.