An active zone state switch concentrates and immobilizes voltage-gated Ca 2+ channels to promote long-term plasticity

A molecularly diverse spectrum of plasticity mechanisms orchestrates brain information processing and storage via positive (“Hebbian”) and negative (“homeostatic”) feedbacks, which, however, mechanistically converge and functionally interact in vivo. The presynaptic scaffold proteins that orchestrate active zone (AZ) function undergo plastic remodeling to regulate release potentiation. Here, voltage-gated Ca 2+ channel function and their exact AZ nanoscale distribution steer release, although how they are involved in AZ remodeling remains unknown. We here establish intravital, dynamic, single-molecule imaging of endogenously tagged Ca 2+ channel Cacophony (Cac) at Drosophila AZs triggered towards homeostatic potentiation. At potentiating AZs, Cac channel numbers increased, and their mobility decreased, while their overall distribution compacted. Mechanistically, RIM-1 and RimBP proteins and their conserved bindings sites, within the Cac channel’s C-terminus, were dispensable for Cac immobilization and compaction. Conversely, the absence of ELKS-family homolog Bruchpilot precluded Cac immobilization and compaction. We show that AZs can undergo a state switch, likely via the ELKS scaffold to concentrate and immobilize Ca 2+ channels and thus boost release. proteins were found to undergo liquid-liquid-phase separation (LLPS), with “liquid” states promoting assembly but “rigid” states promoting release function 52,54,56,57 . The relevance of such collective states changes for functional plasticity remained unexplored yet. Notably, extended regions of AZ proteins RIM-1, RimBP, ELKS/BRP, as well as the Ca 2+ channel intracellular C-terminus were shown in vitro to form LLPS-like condensates 52,54,56,57 . Given the apparently “collective nature” of this new form of AZ plasticity, relating its mechanisms to other state-switches, particularly LLPS like processes, should be a subject warranting future research.


Introduction
Synapses are key sites of information processing and storage in the brain. The synaptic transmission strength is not hardwired but adapts during synaptic plasticity to provide adequate input-output relationships, to maintain or restore transmission when compromised and to store information [1][2][3][4][5][6][7] . Mechanisms of long-term (more than seconds) postsynaptic plasticity have been extensively studied, and processes targeting postsynaptic neurotransmitter receptors have been convincingly connected to learning and memory processes 6,7 . Presynaptic long-term plasticity, however, is also prominent, not only at hippocampal mossy ber bouton synapses, but also at many other synapse types 8 . However, as we still lack a coherent understanding of how long-term presynaptic plasticity manifests itself at the molecular level, analyzing its computational and behavioral role remains challenging. Ca 2+ channels at the presynaptic membrane are the fundamental and conserved triggers of evoked synaptic vesicle (SV) release and modulate SV release probability through the precise nanoscale "coupling" between Ca 2+ channels (within 10-200 nm around the active zone (AZ) center) and release-ready SVs. Indeed, AZ scaffold proteins can cluster Ca 2+ channels with nanoscale precision to release sites within the membrane.
A remarkable and adaptive form of presynaptic plasticity is homeostatic plasticity. Homeostatic plasticity is observed from invertebrates to humans, and actively stabilizes synaptic transmission in response to neural activity perturbation. Notably, homeostatic mechanisms control both presynaptic and postsynaptic function and likely functionally intersect with Hebbian plasticity mechanisms in stable information encoding such as learning and memory 9,10 .
Meanwhile, elucidating the conserved mechanistic basis of presynaptic homeostatic plasticity at the highly-accessible neuromuscular junction (NMJ) of Drosophila larvae has provided important insights 11 . Here, presynaptic homeostatic potentiation (PHP) can be acutely triggered by blocking postsynaptic glutamate receptors with Philanthotoxin (PhTx). Within minutes, PHP precisely counterbalances the reduced postsynaptic responsiveness through enhanced neurotransmitter release, involving both increased release probability for docked SVs at existing release sites and the addition of functional release sites at AZs. During PHP, Ca 2+ in ux increases, and image analysis based on conventional confocal microscopy suggests that additional Ca 2+ channels are physically recruited to pre-existing AZs at NMJ synapses. Moreover, confocal imaging studies suggests that Unc13A and BRP levels also increase during PHP, implying that additional release sites might get recruited at pre-existing AZs. Notably, at the AZs of plastically remodeling rodent mossy ber boutons, addition of release sites and accumulation of additional Ca 2+ channels were also found to promote functional potentiation. Despite such progress, the exact molecular mechanisms by which AZ scaffolds intersect with Ca 2+ channels to mediate the complex mechanism of PHP still remain enigmatic. In this study, we explore the in vivo dynamics of individual Ca 2+ channels at AZs undergoing homeostatic remodeling over half an hour. We present evidence that the AZ scaffold, speci cally via its BRP/ELKS core component, can trigger the compaction and immobilization of Ca 2+ channels speci cally to drive stable expression of plastic potentiation.

Results
CacmEOS4b molecules in a focal plane across several boutons of a given NMJ (Fig. 1F-J, Suppl. Fig. 1A-C). Time windows of observation were both limited by the bleaching of the uorophore, which was irreversible after 3-5 minutes of continuous illumination.
To identify functional synaptic contacts, we monitored activity at NMJs by employing the Membrane-tethered Ca 2+ sensor Syn-GCamp6f in the postsynaptic muscle of CacmEOS4b animals (Suppl. Fig. 2). Syn-GCamp6f thus reports Ca 2+ entry through postsynaptic receptors, activated in response to spontaneous presynaptic transmitter release. Notably, frequencies of postsynaptic Ca 2+ signals were stable over a period of 30 minutes (Suppl. Fig. 2M). This implies that the quick and persistent loss of CacmEOS4b blinking signals did not result from drastic changes in the synaptic properties but rather re ects a limited population and a low turnover of Cac channel molecules within the AZs. For further analysis, we focused on the uorescent signals of CacmEOS4b whenever robust characteristic blinking behavior indicated the excitation of individual uorophores and the stability of the focal plane. At an acquisition frequency of 20 Hz, signal intensities were su cient to gain an average localization accuracy of ~30 nm for individual Cac molecules (Suppl. Fig. 1D).
To gain access to the local distribution and dynamics of individual Cac channels, we rst localized individual events and reconnected consecutive detections to create single channel trajectories (see materials and methods for details). We selected trajectories longer than 8 points (400ms), as suitable for direct extraction of both, Cac channel dynamic properties and radius of con nement by calculating of their mean square displacements (MSDs) (Fig. 1L-M).
Here, individual Cac channels were con ned but not static. From trajectories that were longer than 8 time points, we calculated a median diffusion coe cient (diff.-coeff.) of 0.0075 µm²/s and from the mean square displacement (MSD) analysis we observed that on average Cac channels were distributed within an average con nement radius of 120 nm (Fig. 1M). Our data show mobile and immobile Cac channel trajectories within the AZs. Here, the immobile fraction of Cac channels comprised about 25% of the channels while 75% of channels stayed mobile but con ned over the observation time window (Fig. 1K).
To further characterize the biophysical milieu in which Cac channels operate at AZs, we applied statistical methods to include all trajectories as short as two time points, increasing the statistical power of the analysis. As a result, we extracted additional parameters, such as the drift eld and attraction energy for Cac channels con ned within the AZs (Suppl. Fig. 3G). We here observed that the trajectories were mostly located in high-density regions ("energy wells"), which also resemble an elliptic Cac channel cluster geometry (long axis 145 nm, short axis 95 nm) (Suppl. High-and low-density zones of Ca 2+ channels in active zones We nd that the Cac channels in wild type synapses are strongly con ned within the AZs (Fig. 1). This observation allowed us to further exploit the total population of Cac channel localizations, from the sptPALM imaging, to probe whether the channels would preferentially localize to speci c "nanodomains" within AZs, as previously proposed for mammalian synapses 21,22 . To this end, we employed Dual-Delaunay Triangulation Tessellation, a segmentation method to de ne local density distributions based on cumulative localizations 23 . This method allowed identi cation of the preferred region within which the majority of Cac channel localizations in an AZ occur, de ning a boundary for each AZ. We excluded AZs with diameters larger than 600 nm (Fig. 1J, N).
Furthermore, the tessellation process also robustly retrieved a sub-region within the AZ clusters, wherein a signi cant increase in localization density was determined. Henceforth, we call these AZ-central areas "nanoclusters" (Fig. 1J, N). In addition, we employed further ltration parameters similar to the ltering and segmentation parameters of AZ area limits and localizations (for details see materials and methods), to those previously reported in super resolved localization microscopy of scaffold protein BRP at Drosophila NMJ AZs 13,14 . Employing these strict parameters, we robustly found AZ borders of the best infocus AZs and found that these AZs typically form a single nanocluster in their center.
The Cac channels localizations formed AZ clusters of about 160 nm diameter with a central nanocluster of about 60 nm diameter ( Fig. 1N-O). On average, the Cac nanoclusters harbored about 50% of all Cac localizations found within an AZ ( Fig. 1N-O). PALM-based localizations of the CacmEOS4B depicts Cac nanocluster diameters of ∼60 nm, resembling results of previous STED based quanti cations 24 . In addition, the diameter of Cac nanoclusters derived from tessellation analysis of our sptPALM data also falls within the size range of 50 to 90 nm, obtained from nanobody-based detection by 2D STED microscopy at a lateral resolution of 40 nm (Fig. 1A-E). STED based quanti cation of Cac channel cluster size thus obviously "emphasizes" the central AZ nanocluster derived from our sptPALM data. In addition, these results concur with previous STED data which determined a Cac enrichment zone size of about ∼70 nm in the AZ center, which is the nanoscopic spacing between SV release sites marked by Unc13A release clusters and Cac channels, and thus ensures the high release probability at NMJ and also central Drosophila synapses 25,26 . Notably, the Ca 2+ channel cluster sizes observed here are similar to those previously measured in rodent hippocampal synapses by immunohistochemistry, gSTED (∼70-100 nm) and electron microscopy 27−29 .
Ca 2+ channel numbers increase, mobility decreases and distribution compacts in the course of presynaptic homeostatic plasticity As mentioned above, Drosophila NMJ AZs can be triggered towards PHP within minutes by the application of the glutamate receptor antagonist Philanthotoxin (PhTx) 11,30 . We triggered PHP in CacmEOS4 larvae by application of 50 µM PhTx for 10 minutes in the bath solution, before transferring the larva into the imaging chamber containing uorescent horseradish peroxidase (HRP)-antibody in imaging buffer to outline the NMJ for better orientation during imaging. After 5 minutes, the animals were washed with plain imaging buffer to remove unbound label. We began live sptPALM imaging of Cac channel populations at NMJ I b boutons (muscle 4, 6, and 7) 15 minutes after the start of the PhTx application. Of note in this regard, PhTx treatment was shown previously to irreversibly block postsynaptic GluRIIA receptors 31,32 . Thus, we here intentionally analyzed the behavior of individual Ca 2+ channels within the time window of upcoming PHP expression, which means after 20 minutes of the pharmacological glutamate receptor blockage ( Fig. 2A-H).
In addition, we also counted the relative Cac channel molecule numbers per AZ, at control as well as at plastically remodeling AZs from the same sptPALM data. Here, we analyzed the bleaching behavior of the CacmEOS4b signals at individual AZs upon steady illumination. Discrete bleach steps were identi ed, which obviously indicate the bleaching of individual mEOS4b tagged Cac channel uorophores. Under control conditions, about nine channel molecules were measured within an AZ (9.2 ± 0.5). In contrast, after PhTx treatment, on average twelve channels per AZ were measured (12.2 ± 0.8) (Fig. 2I-K). This analysis thus provides direct single molecule derived evidence that the numbers of Ca 2+ channels per AZ increase upon PhTx treatment, consistent with previous interpretations based on confocal imaging and Ca 2+ imaging [31][32][33] . We would like to emphasize that this approach does not depend on the use of antibody labeling necessarily confounded with issues of inhomogeneous labeling or limitation in the accessibility of the relevant epitopes. Still, the Cac channel numbers retrieved might well represent an underestimation of the total channel number, as we extracted these relative channel numbers from the live imaging sptPALM data with limited control over uorophore conversion.
These sptPALM data were further used to compare the dynamics and distribution of Cac channels of animals undergoing PhTx-triggered plasticity to controls (see details in materials and method). We rst analyzed the Cac channel dynamic behavior. Notably, we found that our PhTx treatment reduced Cac channel mobility, evident in a profound decrease of the median diffusion coe cient (PhTx treated channels median diffusion coe cient: 0.0027 µm 2 /s; untreated controls: 0.0074 µm 2 /s) (Fig. 2L). In addition, the mean radius of con nement of Cac trajectories was slightly but signi cantly decreased for PhTx treated compared to the untreated animals (Control-113±0.6 nm S.E.M, PhTx treated 107±0.5 nm S.E.M) (Fig. 2M). Thus, the PhTx treatment seemingly promotes an immobilization and con nes the spread of Cac channels within the AZ membrane during PHP.
We next employed tessellation analysis to calculate the spread and densities of Cac channels within the broader AZ cluster and the central nanocluster from Cac channel PALM localization data obtained from PhTx treated or untreated animals ( Fig. 2C-H & N-Q). Upon analysis of these AZs, we found that within AZ cluster boundaries, overall diameters signi cantly decrease (10%) and their channel localization densities signi cantly increased (21%) in PhTx treated animals compared to controls (Fig. 2N-O). Similarly, analysis of the localization within nanocluster boundaries of PhTx treated animals also displayed a signi cant increase (16%) in Cac channel density, while the diameter of nanocluster clusters signi cant decreased by about 8% (Fig. 2P-Q), when compared to controls. This suggest that the high-density, central nanocluster of Cac channels gets further compacted during PHP.
Thus, taken together the PhTx treatment seemingly promotes an immobilization and "compaction" of Cac channel distributions within minutes at the level of the individual AZ.

RIM-1/RimBP binding is dispensable for Ca 2+ channel immobilization and compaction
We asked how this in-AZ distribution and compaction of Cac channels could be molecularly orchestrated, and whether AZ scaffold proteins might contribute to this process. Ca v 2 class voltage-gated Ca 2+ channels (including Cac channels) bind to the AZ scaffold via their conserved binding motifs within their intracellular C-termini bind 24,34−39 , and form contacts to RIM-1 and RimBP. These two highly conserved multidomain proteins are responsible for orchestrating SV release and organize the functional SV release site architectures across evolution 40 . To our surprise, in both mutants, Cac channels showed signi cantly lower diffusion coe cients compared to controls, indicating that Cac channel mobility at these AZs is indeed reduced, with effects being more pronounced for Thus, instead of increasing Cac motility, as might be expected, lack of RimBP and RIM-1 surprisingly provoked a "pre-compaction" of Cac distributions "already" in the absence of PhTx application.
Other mechanistic means, rather than the individual binding of RimBP or RIM-1 to Cac channels, might be at play here, particularly when considering the severe release de cits characterizing both mutants. Thus, in the background of CacmEOS4b, we speci cally disrupted, in the background of CacmEOS4b, the conserved RIM-1 and RimBP binding sites within the C-terminal region of the Cac channels, allowing us to directly study the role of both the RimBP and RIM-1 binding sites. To do so, we eliminated the C-terminal last six amino acids (to interfere with binding of the RIM-PDZ domain) and introduced a double point mutation (PPTP to APTA at the 2nd PxxP motif within the Cac C-terminus at amino acid positions 1690-1693) to disturb Cac/RimBP interaction 39 ). Following suit, the dynamic localization and distribution parameters of individual Cac channels derived from the live sptPALM data obtained at these mutant AZs were essentially identical to controls ( Fig. 3O-T). In detail, neither the in-AZ mobility, nor the nanodomain organization were different for the on-locus engineered Cac-APTA-Δlast6aa mutant channels. In short, consistent with our results analyzing RimBP/RIM-1 mutant AZs ( Fig. 3 and Suppl. Fig. 4), the discrete "classical" binding motifs connecting Ca v 2 class voltage-gated Ca 2+ channels with these scaffold proteins are obviously dispensable for AZ tethering and immobilization in this context.

The BRP/ELKS scaffold mediates Ca 2+ channel immobilization in homeostatic plasticity
We went on by genetically analyzing the compaction process by taking a Cac channel perspective. Recent studies demonstrate that the distal C-terminus interaction surface of Ca 2+ channel harbors a signi cant number of channel-scaffold interactions, and some have been shown to mediate Ca 2+ channel localization in cultured rodent neurons 42 . Thus, we generated an on-locus Cac channel C-terminal deletion mutant that removes the last 160 amino acids of the C-terminus (Cac-Δlast160aa) ( Fig. 4 and Suppl Fig. 5A). To evaluate the functional consequences of this deletion, we performed TEVC recordings in mutant and control animals. The Cac-Δlast160aa mutants, different from cognate controls, suffered from a severe reduction of evoked response, and consequently had signi cantly lower quantal content levels compared to controls (Suppl Fig. 5B-M; 5E). We then performed live sptPALM imaging on Cac-Δlast160aa mutant NMJs, and analyzed their dynamics, distribution, and relative channel numbers ( Fig. 4A-H, I-N and O respectively). Our subsequent analysis of channel dynamics from live sptPALM imaging revealed a signi cantly higher Cac mobility ( Fig. 4I: median diffusion coe cient of 0.0163 µm 2 /s for Cac-Δlast160a.a. compared to 0.0062 µm 2 /s for controls) and a signi cantly larger radius of con nement ( Fig. 4J: Control:111.5±0.9 nm; Cac-Δlast160a.a: 155.8±1.7 nm). Tessellation analysis of Cac localization distribution within the AZ cluster and nanocluster boundaries showed that the Cac-Δlast160aa channels appear dispersed over a larger area as nanocluster diameters were also signi cantly larger (Fig. 4K,M), while the localization densities within AZ-and nanoclusters reduced concomitantly (Fig. 4L,N). In addition, bleach curves analysis of Cac-Δlast160aa PALM data uncovered a signi cant reduction in channel numbers in Cac-Δlast160aa mutants compared to controls (Fig. 4O).
Collectively, the distal 160 amino acids of the Cac C-terminus are thus obviously critical to effectively tether Cac channels at the AZ presynaptic membrane, and densely pack them into the AZ central nanocluster. The remaining Cac channels at Cac-Δlast160aa AZs were fewer in number at mutant AZs and rendered atypically mobile.
At Drosophila AZs, BRP is a member of the AZ evolutionarily generic Cast/ELKS protein family and is a major building block of the NMJ AZ ultrastructural scaffold (also known as "T-bar"). Combinations of electron and super-resolution light microscopy have previously shown that a concentrated density of Cac channels clusters beneath the AZ center, i.e., the BRP-based T-bar, and that, additionally, the in-AZ Cac channels levels were reduced in the absence of BRP 24,43 . Similarly, the reduction of ELKS proteins decreased the presynaptic Ca 2+ in ux at hippocampal synapses 44 and reduced the clustering of Ca v 2.1 channels at the Calyx of Held.
Previous yeast-two hybrid and pulldown assays revealed that a N-terminal BRP fragment (amino acids 1-320), which is conserved between ELKS family members, interacts with the Cac C-terminus. The BRP N-terminus in STED microscopy localizes considerably closer to the Cac channel accumulation when probed via a C-terminal fused GFP. We thus tested whether the distal 160 amino acids of Cac, which are critical for its AZ immobilization, would be critical for binding the BRP N-terminal region. Thus, we mapped the interaction site using yeast two hybrid (Suppl. Fig. 6). Binding is attenuated by deleting the last 60 amino acids from the Cac C-terminus (Δ81aa) and is further weakened upon removal of the last 130 amino acids from the Cac channel C-terminus (Suppl. Fig. 6). Thus, it appeared plausible that binding of the AZ membrane proximal N-terminal of BRP to the C-terminal of Cac is crucial for Cac channel immobilization. To further elaborate on this idea, we analyzed the dynamics of CacmEOS4b channels at BRP null mutant AZs by sptPALM (Fig. 5). Notably, individual Cac molecules were atypically mobile at BRP null mutant AZs and less con ned than at control AZs ( Fig. 5A-F, G-H). Cac channel mobility and radius of con nement were higher than in controls and, importantly, remained high even after PhTx treatment ( We also attempted to use the localization data from the sptPALM recording for Tessellation analysis of the BRP null mutant AZs. Here however, far fewer AZ and nanoclusters clusters could be retrieved per NMJ terminal area by the standard tessellation parameters (AZ clusters: minus 62% in AZs of BRP null mutant and BRP null mutant after PhTx treatment; NC clusters: minus 54% in AZs of BRP null mutant and 69% BRP null mutant after PhTx treatment compared to controls) (Suppl. Fig. 7). Given the exceptionally low numbers of remaining clusters that were quanti able with the tessellation method, we refrained from this type of analysis in this situation.
We also analyzed CacmEOS4b bleach curves from the sptPALM imaging data to extract a measure for the average numbers of Cac molecules per AZ in BRP mutant AZs with and without PhTx treatment. Consistent with our previous work using cDNA derived overexpressed GFP-tagged Cac constructs 24,35 , the endogenously tagged CacmEOS4B line displayed reduced numbers of Cac channels in BRP null mutants compared to controls (Fig. 5I). Moreover, there was no signi cant increase in Cac channel number upon PhTx treatment (Fig. 5I). Important in this regard, we recently found that BRP is crucial to sustain the expression of homeostatic potentiation beyond the initial induction period, while induction persists undisturbed. Concretely, BRP was fully dispensable for PHP at a 10-minute time interval, while its absence eliminated PHP measured after 30 minutes of PhTx application 45 . Collectively, our data thus imply that the immobilization of Cac channels via BRP might be a critical process to drive the stable expression of PHP. This is interesting considering that BRP is essential for the AZ structural remodeling which sustains the PHP 30,45 and that BRP "itself" has been shown to undergo PhTx-triggered compaction.
Taken together, our data suggest that within about 20 minutes, PhTx triggered plasticity can still trigger a certain accumulation of Cac channel even at AZs lacking BRP (Fig. 5I,J also see discussion), likely derived from extra-AZ sources. However, these additional channels then fail to stably anchor at the BRPlacking AZs and were rendered atypically mobile, a state incompatible with PhTx plasticity beyond the initial induction period (Fig. 5G-H). This is interesting considering that BRP is essential for the AZ structural remodeling which sustains the 45 PHP and that BRP "itself" has been shown to undergo PhTx-triggered compaction.
Our analysis therefore suggests that the BRP scaffold is critical for mediating the Cac channel immobilization and compaction (Fig. 2L-Q), thus likely directly contributing to sustaining release increases to allow for stable PHP expression.

Discussion
Exact numbers, density, and nanoscale positioning of voltage-gated Ca 2+ channels (VGCCs), within the AZ at the presynaptic membrane, are central for sensitively tuning AZ release function. Ca 2+ channels seemingly get con ned to speci c positions within the presynaptic membrane 28,29,46 , and AZ scaffold proteins have been shown to dynamically change on the minute timescale 29 . To better understand presynaptic plasticity processes at the level of VGCC dynamics, we here exploited the Drosophila NMJs as a particularly well-suited model to establish quasi intravital imaging of individual Ca 2+ channels at remodeling AZs. A fundamental nding here is that the lateral dynamics of individual Ca 2+ channels declines, in the context of homeostatic potentiation, accompanied by a compaction of the Ca 2+ channels within the AZs, which coincides and most likely directly contributes to the sustained increase in presynaptic SV release. We also suggest BRP/ELKS as the critical scaffold protein mediating these changes in Ca 2+ channel number, immobilization, and compaction, to promote these sustained release increases (stable PHP). We suggest this based on the following ndings: 1. BRP is essential for Cac channel immobilization and compaction, regardless of PhTx application (Fig. 5); 2. BRP is essential for stable expression but not induction of PHP (Fig. 5); 3. the distal Cac C-terminus including the BRP binding region is essential for immobilization and compaction ( Fig. 4 and 5 In detail, we found Ca 2+ channels to signi cantly slow their mobility at AZs when undergoing homeostatic remodeling (Fig. 2L). At the same time, their con nement area became "compacted", evident in a decreased size of both AZ cluster and the AZ-central nanocluster (Fig. 2M, N, P). Consequently, the density of individual Ca 2+ channels per unit area increased strongly, particularly in the center of the AZ (Fig. 2N-Q). In addition, here, the (relative) numbers of Ca 2+ channels per AZ, estimated via bleach curves from in vivo sptPALM movies increased signi cantly (Fig. 2I-K). As mentioned above, during PHP, Ca 2+ in ux is increased, and previous confocal microscopic analyses have shown Cac channel increases during PHP. Biophysical analysis has suggested that SV release dominantly operates in a distance less than 100 nm from the AZ center, i.e., the nanocluster center 12,47 . It thus appears most plausible that the increased Ca 2+ channel density accounts for the PhTx-triggered increase of Ca 2+ in ux and consequently directly contributes to PHP expression. An increase in the total number of channels has been proposed as a rst order mechanism 11,33 assuming that Ca 2+ channels will behave independently in respect to their voltage dependent activation. However, it has also been shown that Ca 2+ channels can act cooperatively, depending on their proximity to each other and on their activation history 48, 49 . It is thus conceivable that the AZ and nanocluster compaction of Ca 2+ channels within the AZ probably serves to promotes channel-scaffold interactions in favor of more reliable Ca 2+ signaling, particularly within the identi ed nanocluster of each AZ. The condensation of Cac channels within just a single nanocluster per AZ supports the idea that beyond the increased channel number, a change in channel density entails a cooperative modulation of Ca 2+ channel activity with an impact on the PHP related increase of SV release. The compaction of Cac channels may further add to the isolation of Ca 2+ in ux dedicated to SV release from peri-AZ Ca 2+ in ux related to SV endocytosis 50 . Due to the supra-linear dependence of the SV release on intracellular calcium concentration as shown for many synapses, even already the nanoscale change in Cac-channel organization can be very effective and work as fast feedback for acute activity changes. Several mechanisms can be envisioned to support and maintain the compaction of the central An interesting result of our analysis is also that the absence of RimBP and RIM-1 rendered the AZs in a pre-compacted state of high Cac channel density and immobility. This is likely explained by the fact that these mutant AZs operate under conditions of chronically ine cient SV release and reduced Cac levels that likely renders the channel distribution into a compact state despite the absence of a PhTx treatment (Fig. 3). This result is now to be kept in mind when interpreting mutants interfering with PHP expression but also from evoked base line release de cits. BRP mutants, despite suffering from reduced evoked release as well (although not as severely as in RimBP mutants), are obviously unable to pre-compact given the role of BRP/ELKS as a "master scaffold for compaction".
Our analysis showed that RIM-1 and RimBP with their structurally well de ned, "punctate" Ca 2+ channel C-term interaction motifs were dispensable for PhTx triggered Ca 2+ channel immobilization measured via live sptPALM imaging. In contrast, the ELKS homologue BRP clearly is essential to serve as an AZ anchor and immobilize the Ca 2+ channels, while additional Ca 2+ channels are seemingly still tra cked to the presynaptic plasma membrane upon PhTx (Fig. 5I).
However, PhTx treatment failed to trigger an in-AZ immobilization of Ca 2+ channels when BRP was missing ( Fig. 5A-H, J). Given that, BRP is essential for stable PHP expression (but not induction), these results further suggest that Ca 2+ channels need to be stably anchored, at least to display their full impact in triggering SV release. The Ca 2+ channel association with RIM-1 and RimBP in our intravital system is not crucial for stable in vivo Ca 2+ channel anchorage at AZs, which is obviously essential for proper SV priming as shown for rodent synapses 40,42,55 .
How then might BRP mediate this Cac channel compaction? Notably, a recent study using STORM analysis of xed NMJs showed that the BRP AZ scaffold seemingly also undergoes a compaction-like process in response to PhTx, as suggested by measuring an increased nanoscale density of the C-terminal BRP epitope evaluated in this study 14 . The biophysical nature of BRP scaffold and its ability to undergo compaction, like Ca 2+ channels, might contribute to a consequent connection to Ca 2+ channels, which likely drives their distribution and compaction. Here, liquid-liquid phase separation (LLPS)-like processes are prime candidates in our eyes for the interaction of Ca 2+ channels with the BRP/ELKS C-terminus. Recently, ELKS-type AZ scaffold proteins were found to undergo liquid-liquid-phase separation (LLPS), with "liquid" states promoting assembly but "rigid" states promoting release function 52,54,56,57 . The relevance of such collective states changes for functional plasticity remained unexplored yet. Notably, extended regions of AZ proteins RIM-1, RimBP, ELKS/BRP, as well as the Ca 2+ channel intracellular C-terminus were shown in vitro to form LLPS-like condensates 52,54,56,57 . Given the apparently "collective nature" of this new form of AZ plasticity, relating its mechanisms to other state-switches, particularly LLPS like processes, should be a subject warranting future research.

Page 7/19
Fly husbandry, stocks, and handling Fly strains were reared under standard laboratory conditions and raised at 25 °C on semi-de ned medium (Bloomington recipe). For RNAi experiments ies and larvae were kept at 29 °C. For experiments male third instar larvae were used. The following genotypes generated in the wild-type background of CacmEOS4b were created as follows: Drosophila genetics and genome engineering: Endogenous expression of calcium channels and on-locus deletion mutants: Fly strains were reared under standard laboratory conditions at 25°C and 65-70% humidity in incubators. Wandering third instar male larvae were used for analysis in all experiments, except when indicated differently. The endogenously tagged y lines CacsfGFP and CacmEOS4b were generated in the O'Connor Giles laboratory as previously described in Gratz et al., 2019 33 . Brie y, N-terminally tagged Cac was generated using a scarless CRISPR/piggyBac-based approach ( yCRISPR.molbio.wisc.edu).
The following generated transgene animals based on CRISPR mediated mutagenesis were performed by Well Genetics (Taipei City, Taiwan) Inc. using modi ed methods of Kondo and Ueda (2013). The following transgene animals have been generated using the CRISPR/Cas9 mediated genome editing approach by homology dependent repair (HDR) using 1 guide RNA and a dsDNA plasmid donor: CacmEOS4b-Δlast160AA: We generated a deletion allele of the Cac gene by truncating its distal C-terminal sequence of the last 160 AA.
In brief, the gRNA sequence GACGGTTTGCTGGGAGTCGG[AGG] (same gRNA used for CacmEOAS4b-APTA) and AGGAGGATTGGTGCTAGCAA[AGG] were cloned into U6 promoter plasmids. Cassette PBacDsRed containing two PBac terminals 3xP3 DsRed and two homology arms were cloned into pUC57 Kan as a donor template for repair. Cac/CG43368 targeting gRNAs and hs-Cas9 were supplied in DNA plasmids, together with a donor plasmid for microinjection into embryos of control strain CacmEOS4B/FM7a. F1 ies carrying the selection marker of 3xP3 DsRed were further validated by genomic PCR and sequencing.
CRISPR generates a 734 bp deletion allele of Cac/CG43368, deleting the last 160 AA of Cac/CG43368 and is replaced by cassette PBacDsRed. CacmEOS4b-APTA-ΔLast6AA: In brief, the gRNA sequence AGGAGGATTGGTGCTAGCAA[AGG] w a s cloned into U6 promoter plasmids. Cassette PBacDsRed containing two PBac terminals and 3xP3 DsRed and two homology arms were cloned into pUC5 7 Kan as donor template for repair. cac/CG43368 targeting gRNAs and hs-Cas9 were supplied in DNA plasmids, together with donor plasmid for microinjection into embryos of control strain Cac mEos APTA CRISPR / FM7a. F1 ies carrying selection marker of 3xP3 DsRed were further validated by genomic PCR and sequencing. CRISPR generates a 21 bp deletion allele of cac/CG43368, deleting the last AA of cac/CG43368 and is replaced by cassette PBacDsRed.

Yeast-2 Hybrid Assay
In principle all experiments were conducted according to the yeast two-hybrid protocols of Clontech using the strain AH109 with minor adjustments meeting the requirements of the interaction-domain mapping experiments. Co-transformation of the yeast strain AH109 with both prey and bait constructs (Clontech) was conducted using the following protocol: The yeast was plated on an YPDA (yeast, peptone, dextrose, adenine) medium agar plate and cultured for two days at 30° C. An appropriate amount of YPDA medium was inoculated with a single clone and cultured at 30° C overnight while shaking at approximately 200 rpm. After dilution to OD600 0.2 (10 mL culture volume per transformation) the culture was grown until OD600 reached 0.6. The cells were pelleted, washed in 1xTE (Tris-EDTA)-medium at pH 8.0, re-pelleted and resuspended in TE/LiAc (100 μl per 10 mL culture volume; LiAc: Lithium acetate). 100 μl resuspended yeast were mixed with the transformation mix which was composed as follows, 500 ng of each of the two plasmids and 10 μl (= 100 μg) Herring Testes carrier DNA (Clontech). After adding 600 μl PEG (Polyethylene glycol)/LiAc, the mix was vortexed and left to incubate at 30°C for 30 minutes in a shaker. Then 70 μl DMSO (10 % nal concentration) were added, and the cells were heat shocked for 15 minutes at 42°C. After chilling on ice, the yeast was carefully sedimented, the pellet suspended in 100 μl 1xTE-medium pH 8.0, and the transformation plated on minimal SD (synthetic de ned) /-Leu/-Trp medium plates, to ensure the presence of both plasmids. After growing for 2 -3 days at least 10 clones each were analyzed on SD/ -Ade/ -His/ -Leu/ -Trp/ X-αgal medium plates to select for positive interaction. If > 70% of the clones plated on SD/ -Ade/ -His/ -Leu/ -Trp/ X-α-gal medium plates grew, this was regarded as weak positive interaction (+), > 80% as Intermediate interaction (++) and > 90% and blue clones as strong interaction (+++). Negative controls consisted both of laminin as bait and the prey to be tested and the corresponding bait and the empty prey vector. In the positive control pGBKT7 p53 was transformed with pGADT7 containing the SV40 large T antigen.
The BRP D1 and Cac fragments depicted in Suppl 4 were cloned into pGADT7 using the following primer pairs: Dissections were performed following standard protocols 60 and are described here: Third instar larvae were opened in HL3 by opening dorsally along the midline and removing the innards. Filets were xated with ice-cold MeOH for 10 min for all experiments to quench auto uorescence of sfGFP/mEOS4b to subsequently boost them with an antibody or nanobody booster, after testing all antibodies function optimally in MeOH. After xation, the lets were washed with PBS plus 0.05% Triton X-100 and blocked for 60 min in 5% normal goat serum. The larvae were incubated with primary antibodies at 4°C overnight and subsequently washed in a PBS plus 0.05% Triton X-100 solution for 2 h at room temperature for immunostaining. Larvae were then incubated for 4h with secondary antibodies at room temperature followed by the same washing procedures. Immunocytochemistry was done the same way for both conventional confocal and STED microscopy except for the mounting medium. Larvae were nally mounted either in Mowiol (Sigma-Aldrich) for confocal microscopy or Prolong Gold (ThermoFisher-Scientifc) for STED. The following antibodies were used at the indicated concentrations: mouse anti-Bruchpilot (Brp) Nc82/ BRP Cterm (DSHB, catalog #nc82; RRID:AB_2314866) and BRP N-term (1:500 61 ), rabbit RIMBP C-term (1:500 62 ); mEOS4B Nanobody coupled to atto580 or GFP nanobody coupled to STAR635P at 1:300 (Nanotag, Catalog N0304 or N3102), Rabbit or Guineapig anti-GluRIIA (generated in house), and anti-HRP conjugated to AlexaFluor 405 at 1:200-1:500 (Jackson ImmunoResearch). Species-speci c Alexa Fluor 488, atto490LS and Alexa Fluor 594 secondary antibodies (Invitrogen, Jackson ImmunoResearch) were used at 1:300.
Acute PHP/PhTx assay: Third instar control and mutant larvae in CacsfGFP/CacmEOS4b backgrounds were treated with a 10-minute PHP assay using Philantotoxin-433 drug (PhTx) (AOBIOS). Larvae were carefully pinned down at mouth and tail while avoiding any stretching of body wall muscles and minimally dissected in 50μl solution containing 50 μm PhTx in Ca 2+ and Mg 2+ -free HL3 or No PhTx-HL3. Animal's heamolyph was mixed with the treatment solution by regularly mixing of the solutions during the 10-minute treatment window undertaken with care to not stretch or pull the larvae muscles. For live-sptPALM assays treated animals were moved into the imaging solution containing HL3 with 4mM Mg 2+ (to reduce muscle contractions) and basal 1.5mM Ca 2+ . STED imaging and analysis: gSTED microscopy was performed using an Abberior Instruments Expert Line STED setup equipped with an inverted IX83 microscope (Olympus), two pulsed STED lasers for depletion at 775 nm (0.98-ns pulse duration, up to 80-MHz repetition rate) and at 595 nm (0.52-ns pulse duration, 40-MHz repetition rate) and pulsed excitation lasers (at 488 nm, 561 nm, and 640 nm).
Multi-channel 2D confocal and gSTED images were acquired with a 100× oil-immersion objective lens (UPLSAPO100XO, Olympus, NA = 1.4), with a pixel dwell time of 2 μsec, with 10x and 30x line accumulation, respectively, at 16-bit sampling and a eld of view of 10 μm x 10 μm. Lateral pixel size was set to 20 nm.
The dyes STAR 635P, Alexa Fluor 594 and ATTO490LS were depleted at 775 nm. Alexa Fluor 488 was depleted at 595 nm. Time gating was set at 750 ps with a width of 8 ns. Fluorescence signals were detected sequentially by line by avalanche photodiode detectors at appropriate spectral regions (STAR 635P and ATTO490LS: 680 nm-765 nm, Alexa Fluor 594: 584 nm-630 nm, Alexa Fluor 488: 500 nm-551 nm). Alexa Fluor 488 confocal and gSTED images were acquired following acquisition of the other channels. These procedures were operated by the software Imspector (version 16.3.13367, Abberior Instruments, Germany). Raw gSTED images were processed for Richardson-Lucy deconvolution with default settings using the Imspector software (version 16.3.13367, Abberior Instruments, Germany). The point spread function was automatically computed with a 2D Lorentzian function having a full-width half-maximum of 40 nm, based on measurements with 40 nm crimson beads. gSTED analysis to determine CacsfGFP spot size (Feret's diameter) was performed by particle segmentation above the threshold signal with the function "Find maxima" in ImageJ (ImageJ-1.52g, NIH).

Electrophysiology
Two-electrode voltage clamp (TEVC) recordings were performed at room temperature on muscle 6 of 3rd instar larval NMJs in the abdominal segments A2 and A3. Male third instar larvae were dissected in modi ed Ca 2+ -free hemolymph-like saline (HL3; in mM: NaCl 70, KCl 5, NaHCO3 10, MgCl2 20, Sucrose 115, Trehalose 5, HEPES 5) Recordings were obtained with a bath solution of HL3 with 1.5 mM CaCl 2 . Recordings were made from cells with an initial Vm between -50 and -80 mV, and input resistances of ≥ 4 MΩ, using intracellular electrodes with resistances of 30-50MΩ, lled with 3 M KCl. Glass electrodes were pulled using a Flaming Brown Model P-97 micropipette puller (Sutter Instrument, CA, USA). Recordings were made using an Axoclamp 2 B ampli er with HS-2A x0.1 head stage (Molecular Devices, CA, USA) on a BX51WI Olympus microscope with a 40X LUMPlanFL/IR water immersion objective (Olympus Corporation, Shinjuku, Tokyo, Japan). mEJCs were recorded for 90 seconds with the voltage clamped at -80 mV, all other recordings were performed while clamping the voltage at -60 mV. eEJCs were recorded after stimulating the appropriate motor neuron bundle with 5 V, 300 µs at 0.2 Hz using an S48 Stimulator (Grass Instruments, Astro-Med, Inc., RI, USA). Signals were digitized at 10 kHz using an Axon Digidata 1322 A digitizer (Molecular Devices, CA, USA) and low pass ltered at 1 kHz using an LPBF-48DG output lter (NPI Electronic, Tamm, Germany). The recordings were analyzed with pClamp 10 (Molecular Devices, Sunnyvale, CA, USA), GraphPad Prism 6 (GraphPad Software, Inc., San Diego, CA, USA). Stimulation artifacts of eEJCs were removed for clarity. mEJCs were further ltered with a 500 Hz Gaussian low-pass lter. Using a single template for all cells, mEJCs were identi ed and averaged, generating a mean mEJC trace for each cell. An average trace was generated from 20 eEJC traces per cell for 0.2 Hz stimulation and 10 ms ISI paired pulse recordings and from 10 traces for total amplitude before the peak. Decay constant τ was calculated by tting a rst order decay function to the average trace of the 0.2 Hz stimulation recording starting from 60% of the total amplitude after the peak until the baseline was reached. The amplitude of the average eEJC trace from the 0.2 Hz stimulation recording was divided by the amplitude of the averaged mEJC, for each respective cell, to determine the quantal content. 10 ms and 30 ms ISI paired pulse ratios were calculated by dividing the amplitude after the second pulse by the amplitude after the rst pulse. The baseline for the second amplitude was set at the last point before stimulation artifact onset.

Single-particle tracking PALM
All live sptPALM experiments were conducted on male third instar larvae. Larval body walls designated for single particle tracking were prepared according to Ramachandran andBudnik 2010 andMarter et al., 2019 20,63 and imaging experiments were performed using an inverted total internal re ection uorescence (TIRF) setup. The microscope (Nikon Eclipse Ti) was equipped with a 100x NA 1.49 Apo TIRF oil objective (Nikon). Up to 10.000 images were captured using a EMCCD camera (iXon+ 897, Andor Technology) controlled by NIS-Elements (Nikon) at a frame rate of 20 Hz. The TIRF set-up is based upon an inverted microscope within a whole microscope incubator at 25°C. (eclipse Ti microscope,Nikon GmbH) was equipped with a 100 x Apo TIRF oil objective (1.49 NA; Nikon). CacmEOS4b containing larvae were imaged at the live HRP-488 stained Z-plane. Brie y, we used laser diodes to photoconvert and excite the mEOS4b uorophore by continuous illumination of the probe with a 405 nm laser (2-5 % of 100 mW) and a 561 nm laser (25% to 40% of 100 mW), and to further improve the separation of the mEOS signal from auto uorescence and background signals, an emission bandpass lter (ET620/60 nm; AHF analysentechnik) was used. Fluorescence was excited by oblique illumination of the probe with a combined laser system (Coherent; MPB communications Inc.) and image sequences (up to 10000frames) were captured by an EMCCD camera (iXon+ 897, Andor Technology) controlled by NIS-Elements Advanced Research acquisition software (Nikon). Images were recorded at a frame rate of 20 Hz, controlled by MetaMorph imaging software (Universal Imaging). We used a 1.6 magni cation lens to reduce the pixel size to 107×107nm. The N-terminal expressed mEOS4b on the Cacophony channel was excited by continuous illumination with 405/561 nm light using 2-5% of the initial laser power of the 405 nm laser line (100mW) and 25-40 % of the 561 nm laser line (100 mW) for upto 10000 frames at 50ms acquisition rate. (1:1000), serving as a surface marker to visualize the outline of NMJs during subsequent imaging, for 5 minutes. The larvae were brie y washed in the HRPfree imaging buffer to remove excess HRP label. Motor nerves were cut close to the ventral nerve cord and the brains were removed to ensure reduced muscle contractions during live imaging. Clamped at the tip, body walls were turned around such that the muscles were facing the cover slip of the imaging chamber.
Samples were straightened carefully using six clips and subjected to TIRF illuminated PALM imaging. An optimal imaging focal plane was chosen, guided by the HRP signal, within which the highest number of boutons of one type 1b NMJ (from segments A2-A4 on Muscle 4 or 6/7) could be captured within the focal plane.
Male Mutant and control larvae were subjected to 10-minute PhTx or HL3 incubation, followed by a live HRP-488 stain were imaged at the HRP Z-plane of type 1b NMJs (from segments A2-A4 on Muscle 4 or 6/7). Localization maps were generated from the acquired data using PalmTracer (MetaMorph, provided by J.-B. Siberita, Bordeaux). Prior to localization detection the movies were drift corrected and cropped to exclude movement artefact by using the ImageJ plugin NanoJ 64 or Thunderstorm 65 . The subsequently generated trajectories (PalmTracer) were further analyzed by calculating the mean square displacement (MSD). Analysis of the local channel density within con ned regions were performed by cluster analysis based on Voronoï tessellation constructed from of localized channels using the software package SR-Tessler 23 .
Localization and trajectory reconnection of mEOS signals was carried out using wavelet-based algorithm and a simulated annealing algorithm as previously described, which considers molecule localization and total intensity. It has been reported that mEOS4b molecules can show blinking-like behavior 66 . To avoid false reconnections between trajectories, all sub-trajectories were analyzed as individual trajectories. Diffusion coe cients (D) were calculated by linear tting of the rst four points of the MSD plots. MSD plots of immobilized molecules (on xed samples) revealed that, under our imaging conditions, molecules with D ≥ 0.002 μm 2 /s can be considered as mobile. The offset of MSD curves from immobile molecules was taken as empirical measure for the localization accuracy under the live imaging conditions (31.3 nm IQR 25.9/36.6 nm, 247 trajectories, suppl. Fig. 1 C). Calculation of the localization accuracy and number of collected photons per individual uorophores (suppl. Fig.1 A,B) were calculated by the use of the Thunderstorm plugin in ImageJ, matching the value obtained from the MSD curves of immobile molecules.

Calcium imaging
Spontaneous synaptic activity was monitored by imaging of postsynaptic calcium transients re ecting the stochastic release of synaptic vesicles. We used the genetically encoded membrane tethered GCaMP6f reporter. Spontaneous calcium transients were recorded in imaging buffer containing 4mM Mg 2+ and 1.5mM Ca 2+ for all control situations or 0.2 mM Ca 2+ to decrease the frequency of spontaneous release events. Images were acquired with a frame rate of 20 Hz. SynapGCaMP6f was excited with a 488 nm laser and transient uorescent emission changes were detected at emission wavelengths between 520-550 nm (Suppl. Fig. 2). The analysis of calcium transient maxima (ΔF/F 0 ) was performed on background-corrected maximum projections using a custom written routine in ImageJ software (NIH) and involved detection of synaptic puncta with SynapGCaMP6f uorescence signal exceeding a threshold of 2x the standard deviation (calculated 30 frames prior to stimulus) in response to the stimulus.
Cac channel number analysis by Bleach curves: The blinking of mEOS4b labelled cacophony channels was used as an approximate measure for the number of channels within individual active zones (AZ), due to the one-to-one stoichiometry between the pore forming subunit and the genetic encoded uorophore. Only image stacks were analyzed where individual blinking events could be identi ed. The localization of the channel population within individual AZs was de ned within average projections of image stacks.
Before further analysis, images were processed by background subtraction and drift correction using the Fiji plugin NanoJ 64 . The con ned localization of the channels within the AZ allowed for the counting of uorescent units within regions of 5x5 pixels, which was su cient to cover the localization of the uorescent signal. Individual symmetric AZ within single boutons were selected manually for analysis of the bleach curve. The uorescent signal was completely bleached within less than 5000 frames (20 Hz acquisition rate), which allowed identi cation of single uorophores as one step photobleaching events. The quotient of the maximal uorescent value and the averaged single uorophore response was taken as a rough estimate for labelled channel molecules within the AZ. Due to the acquisition rate, as well as the shortcomings of live imaging parameters, and the varying positions of AZs within individual boutons there are likely many missed bleaching steps recorded and thus the calculated channel numbers are most likely are an underestimate of the actual number of molecules. Comparative Cac channel numbers under control conditions and after the PHP assay as well as for the BRP and Cacmutants (CacmEOS4B-∆160AA mutants) were analyzed (Fig 2, Fig. 4 and 5).

Well analysis
A. We scanned through all possible local grid bin sizes between 50 nm and 100 nm with a step size of 5 nm, for each bin size dx. For each value of dx, we performed the following iterative procedure starting at iteration : i. First, we restricted the analysis to the trajectory points falling in a square region of size centered at the high-density bin center (if , the square is centered at the center of the ellipse found in the previous iteration instead). We obtained the well boundary by tting the 95% con dence-ellipse to the spatial distribution of these points.
ii. Inside this ellipse, we computed the depth of the well based on a circular estimator computed on the drift map based on a local grid centered at the well center with bin sizes dx while the diffusion coe cient was obtained by considering the well as a single bin.
iii. The iteration was scored using the well parabolic score. iv. We go back to step i. increasing by 1 and stopped iterating when nm and kept the iteration with the minimum score where the well possessed more than 20 trajectory points and 5 drift bins with at least 7 trajectory displacements per bin.
B. The selected well for a given high-density region was chosen as the well with the minimum parabolic score over all local bin grid sizes.
Finally, to obtain the statistics, only the wells with a parabolic score < 0.5 and an energy < 10 kT were kept.
The ring ellipses in Supplemental Figure 3D,E,F were obtained by manually selecting wells forming an elliptic region and tting an ellipse around their centers using a minimum volume ellipsoid algorithm with a tolerance of 0.01.
The residence time of a molecule inside a well reported in Supplemental Figure 3I,J was computed for each well as: . 43,89 Tessellation Analysis of Cac localizations within AZ cluster and Nanocluster distributions: SR-Tessler software, ThunderSTORM, and the ImageJ NanoJ-SRRF were used to process and segment each Cac localization recorded by Live sptPALM of all mutant and control images. Tessellation was performed on drift corrected image stacks. The rst 300 frames were rejected to avoid the detection of multiple molecules uorescent at the same time. Individual molecules could be detected within the following frames and used to localize individual molecules.
Tessellation settings: AZ boundary settings: Voronoi object: 30 density factor, min Area: 2000, Min Localization: 5/10 NC boundary settings: Voronoi Nanocluster object: 3 density factor, min Area: 50, Min Localization: 5/10. List of AZ clusters and their nanoclusters: size, area, number of localizations, density, and number of clusters were exported and ltered to remove all AZ clusters lacking a nanocluster and were also within 30-600nm AZ and Nanoclusters limits. In addition, we applied further limitations to Cac localization number=AZ boundary: 3000 NC boundary: 500, diameter size= AZ boundary: 600-30nm. NC boundary: 300-30nm, Area: AZ boundary: 50000 nm 2 . NC boundary: 7000 nm 2 and density=AZ & NC boundary: 0.072 localizations/nm 2 to de ne well the nanodomain of Cac channels at the Drosophila NMJ at the presynaptic membrane. We took into consideration the parameters set for BRP localizations from recent PALM and QPAINT imaging studies and our average diameter size and imaging of average Cac spot size (Fig 1A-H) 13,14 . Of note here, the controls data were not independent for RIM-1 -/and RimBP -/mutant experiments (Suppl. Fig. 4 and Fig 3A-N). The data shown in Fig. 5 for BRP -/-, BRP -/-+PhTx, and controls was done in one concomitant experiment.

Quanti cation and Statistical Analysis
The statistical analysis was carried out using Prism software (GraphPad). Normality of distribution was veri ed by D'Agostino & Pearson omnibus normality test. To test the statistical signi cance of differences between two conditions Dunn's test and a Kolmogorov-Smirnov test was used for diffusion coe cient, radius of con nement data and tessellation data. For the comparison of multiple datasets, non-parametric Kruskal-Wallis one-way ANOVA was used as indicated in the corresponding gure legend. All statistical tests were performed as two-tailed tests. Effects and differences were considered as signi cant at p < 0.05. Data are presented throughout the text as mean ± standard error of the mean (SEM) or the median with the inter-quartile range (IQR) as indicated.
sptPALM data analysis for mobility, tesslation, well analysis and channel numbers are shown either as median and an interquartile range, or as mean±SEM.
Data and Software Availability