Introduction

The ability of neural circuits to effectively detect and dynamically respond to a broad range of inputs depends on a diversity of neuronal subtypes communicating through synapses with distinct properties. Neurotransmission occurs at specialized membranes called active zones (AZs) where action potentials drive the opening of voltage-gated Ca²⁺ channels (VGCCs) to trigger Ca²⁺-dependent synaptic vesicle (SV) fusion and neurotransmitter release. Neurotransmitter release properties are determined locally at individual synapses and vary considerably both between neuronal subtypes and within homogeneous populations of neurons (Ariel et al., 2012; Atwood and Karunanithi, 2002; Branco and Staras, 2009; Hatt and Smith, 1976). In fact, functional imaging studies in Drosophila demonstrate that even single neurons forming synapses with the same postsynaptic partner display heterogeneous synaptic strength among individual AZs (Guerrero et al., 2005; Melom et al., 2013; Peled and Isacoff, 2011).

Presynaptic strength is defined as the likelihood of neurotransmitter release following an action potential (probability of release, P_r). This stochastic process is determined by the number of functional SV release sites and their individual probability of vesicle release. The probability of SV release is highly dependent on transient increases in intracellular Ca²⁺ levels at vesicular sensors. Accordingly, SV release sites and VGCCs are key substrates for synaptic diversity (Akbergenova et al., 2018; Aldahabi et al., 2022; Chen et al., 2015; Fedchyshyn and Wang, 2005; Fekete et al., 2019; Gratz et al., 2019; Holderith et al., 2012; Laghaei et al., 2018; Nakamura et al., 2015; Newman et al., 2022; Rebola et al., 2019; Reddy-Alla et al., 2017; Sauvola et al., 2021; Sheng et al., 2012). Numerous studies have demonstrated that VGCC levels correlate highly with P_r (Akbergenova et al., 2018; Gratz et al., 2019; Holderith et al., 2012; Nakamura et al., 2015; Sheng et al., 2012). Paradoxically, this is not always the case. For example, a recent study investigated two cerebellar synaptic subtypes, one high-P_r formed by stellate cells and one low-P_r formed by granule cells, and found higher VGCC levels at low-P_r granule synapses (Rebola et al., 2019). This finding highlights the importance of synaptic context in understanding the role of VGCC levels in determining P_r at distinct synapses. Since VGCCs in closer proximity to release sites are expected to have a greater impact on vesicular release probability than those positioned farther away, the spatial coupling of VGCCs and SVs at AZs is a critical determinant of P_r (Chen et al., 2015; Eggermann et al., 2011; Fedchyshyn and Wang, 2005; Nakamura et al., 2015; Rebola et al., 2019). Indeed, at high-P_r stellate synapses, a ‘perimeter release’ AZ organization places VGCCs ∼40nm closer to SVs than at low-P_r granular synapses (Rebola et al., 2019). How these distinct topographies are set remains unknown.

To better understand the context-dependent contributions of VGCCs to heterogeneous neurotransmitter release properties, we sought a system where we could investigate the relationship between VGCCs and P_r at two closely related neurons that form synapses with distinct release probabilities. Drosophila muscles are innervated by two glutamatergic motor neurons, one tonic and one phasic, that form type Ib and type Is synapses, respectively. Type Ib synapses have relatively low P_r and facilitate, whereas type Is synapses have higher P_r and depress in response to high-frequency stimulation (Aponte-Santiago et al., 2020; Lnenicka and Keshishian, 2000). In this study, we investigated how VGCC number, organization, and subunit composition contribute to synaptic heterogeneity at low-P_r type Ib and high-P_r type Is inputs. We find that individual synapses formed by both low-and high-P_r inputs exhibit heterogeneous release properties that can be predicted by VGCC levels alone. However, VGCC levels do not correspond to differences in P_r between the two inputs due to underlying molecular and organizational differences that alter the relationship between VGCC levels and P_r at AZs of low-vs. high-P_r inputs. These findings reveal intersecting levels of molecular and spatial diversity that combine to generate extensive synaptic heterogeneity.

Results

VGCC levels predict P_r within, but not between, synaptic subtypes

To investigate the relationship between VGCC levels and neurotransmitter release properties at functionally distinct synapses, we took advantage of the two motor neuron subtypes with low and high release probabilities that innervate most Drosophila muscles (Aponte-Santiago and Littleton, 2020; Kurdyak et al., 1994). These glutamatergic neuromuscular junctions (NMJs) contain hundreds of individual synapses that are accessible to single AZ functional imaging using genetically encoded Ca²⁺ indicators. In Drosophila, Cacophony (Cac) is the sole Ca_v2 pore-forming subunit and is the VGCC responsible for triggering synaptic transmission (Kawasaki et al., 2000; Macleod et al., 2006; Peng and Wu, 2007; Smith et al., 1996). To simultaneously monitor neurotransmitter release and VGCC levels, we swapped the N-terminal sfGFP tag in our well-characterized cac^sfGFP-Nline for a Td-Tomato tag (cac^Td-Tomato-N). We have previously incorporated a number of protein tags at this site without disrupting function and confirm below that a larger tandem tag does not impair synaptic function (See Figs 1A-F; (Ghelani et al., 2023; Gratz et al., 2019)). We then expressed postsynaptically targeted GCaMP6f (SynapGCaMP6f; (Newman et al., 2017)), which reports Ca²⁺ influx through glutamate receptors in response to neurotransmitter release in Cac^Td-Tomato-N animals. We and others have previously shown that Cac levels are highly predictive of P_r at individual type Ib AZs (Akbergenova et al., 2018; Gratz et al., 2019). To determine if VGCC levels are similarly predictive at high-P_r type Is AZs, we measured Cac^Td-Tomato-N fluorescence intensity and monitored neurotransmitter release in response to 0.2-Hz stimulus at individual synapses. To enable direct comparisons between the two synaptic subtypes, we simultaneously imaged type Ib and type Is synapses at NMJ 6/7 (Fig. 1A). We quantified the number of times a vesicle was released over 120 stimuli to determine single-synapse P_r, and found that, as expected, type Is synapses exhibited significant heterogeneity and higher average P_r than type Ib synapses (Fig. 1B, C; (Newman et al., 2022)). Consistent with their higher P_r, type Is connections contain relatively fewer low-P_r AZs (Figs. 1D, (Jetti et al., 2023)). We next investigated the correlation between P_r and VGCC levels and found that at type Is inputs, single-AZ Cac intensity positively correlates with P_r (Fig. 1E; REFS). We also observe a strong positive correlation between VGCC levels and P_r at type Ib inputs (Fig. 1F), consistent with our and others’ prior findings (Akbergenova et al., 2018; Gratz et al., 2019; Newman et al., 2022).

Figure 1.

A simple prediction of the observation that VGCC levels correlate highly with P_r at individual AZs of both low-and high-P_r inputs is that, when comparing between them, Cac levels will be higher at type Is inputs than type Ib. We analyzed Cac^sfGFP-N levels at individual AZs of type Ib and Is motor neurons and found that average Cac levels are the same at type Ib and Is AZs (Fig. 1G, H). Cac levels are also similarly distributed across active zones of the two inputs (Fig 1I). Together, these findings indicate that the relationship between VGCC levels and P_r differs between the two inputs. Consistently, when we compare the relationship between Cac levels and P_r at the type Ib and Is inputs from our correlative functional imaging data, we find that the slopes of the best-fit lines are significantly different (Fig. 1J). Across type Is active zones, a similar range of VGCC levels supports a broader range of release probabilities. Thus, VGCCs can predict P_r within synaptic subtypes, but not between active zones of different synaptic subtypes, providing a framework for understanding seemingly contradictory findings on the role of VGCCs in determining P_r.

VGCC clusters are more compact at AZs of high-P_r type Is inputs

To understand how spatial differences might alter the relationship between VGCC levels and P_r at low-P_r type Ib and high-P_r type Is AZs, we turned to 3D dSTORM single-molecule localization microscopy (SMLM). An individual VGCC complex is estimated to be ∼10 nm in diameter with the most common immunolabeling techniques adding significantly to their size and creating a linkage error of ∼20 nm between the target molecule and fluorescent reporter (Früh et al., 2021; Liu et al., 2022; Thomas, 2000). For following VGCC dynamics using single-particle tracking via photoactivation localization microscopy (sptPALM), we recently incorporated mEOS4b (Paez-Segala et al., 2015) at the N-terminus of Cac, achieving a linkage error of less of than 5 nm (Ghelani et al., 2023). To gain more flexibility in labeling Cac without adding to the linkage error, we swapped the mEOS tag for a similarly sized HaloTag (cac^HaloTag-N). As expected, these flies are fully viable and Cac^HaloTag-N exhibits normal localization to AZs, where Brp is arranged in a ring that surrounds a puncta of VGCCs in superresolution optical reassignment images (Fig. 2A-C; (Ghelani et al., 2023; Gratz et al., 2019)). HaloTag, which covalently binds synthetic ligands, is 3.3 nm in diameter (Los et al., 2008; Yazaki et al., 2019), yielding a linkage error well under 5 nm.

Figure 2.

Cac^HaloTag-N larvae were stained with JaneliaFluor646 HaloTag ligand (Grimm et al., 2015) and horseradish peroxidase (HRP) to distinguish between type Ib and Is branches and enable simultaneous imaging of the two synaptic subtypes innervating NMJ 6/7. We then used density-based spatial clustering of applications with noise (DBSCAN) analysis to identify Cac clusters at type Ib and Is AZs (Fig. 2D,E; (Ehmann et al., 2014)). We find that the average size of Cac^HaloTag-N clusters is similar at low-and high-P_r AZs (Fig. 2F), with mean diameters of approximately 102 nm and 105 nm, respectively. This is similar to the Cac^mEOS4b-N type Ib cluster size observed by sptPALM imaging (Ghelani et al., 2023). In agreement with our confocal level data, the number of localizations per cluster was similar at low-and high-P_r AZs (Fig. 2G). We then calculated the average Cac density per AZ, and found that VGCCs are significantly more densely organized at high-P_r type Is AZs than low-P_r type Ib AZs (Fig. 2H, I). Greater AZ density at type Is AZs is consistent with recent SMLM studies using antibodies to label Cac or Brp (Mrestani et al., 2021; Newman et al., 2022) and a recent electrophysiological analysis of VGCC-SV coupling that revealed significantly tighter coupling at type Is synapses (He et al., 2022). Together, these findings suggest that more compact organization of VGCCs increases their proximity to SVs and contributes to the steeper relationship between VGCC levels and P_r at high-P_r type Is AZs.

Differences in Bruchpilot levels and function at low-and high-P_r inputs

To understand how these nanoscale differences in VGCC organization might be established, we investigated the AZ scaffolding protein Bruchpilot (Brp). Brp/CAST/ELKs family proteins function as central organizers of both VGCCs and SV release sites at developing synapses (Dai et al., 2006; Dong et al., 2018; Hallermann et al., 2010; Held et al., 2016; Kittel et al., 2006; Liu et al., 2014; McDonald et al., 2020; Radulovic et al., 2020).

We simultaneously imaged type Ib and Is inputs and found lower Brp levels at type Is AZs (Fig. 3A, B). Since Cac levels are similar at the two synaptic subtypes, lower Brp levels result in a significantly higher Cac:Brp ratio at type Is synapses, which we hypothesize promotes compact organization of VGCCs (Fig. 3C). In contrast, we and others have previously shown that Brp levels positively correlate with P_r among AZs of low-P_r type Is inputs (Gratz et al., 2019; Muhammad et al., 2015; Newman et al., 2017; Peled et al., 2014; Reddy-Alla et al., 2017). Consistently, Brp and Cac levels strongly correlate at type Ib AZs (Gratz et al., 2019) and we observe a similarly strong correlation across individual type Is AZs (Fig. 3D). Thus, like VGCCs, Brp levels contribute in distinct ways to synaptic heterogeneity within vs. between low-and high-P_r synaptic subtypes depending on input-specific AZ organization.

Figure 3.

We next investigated the requirement for Brp in promoting VGCC accumulation at low-and high-P_r inputs by analyzing Cac^sfGFP-N levels in brp null mutants (brp^-/-; Fig. 3E, F). Cac^sfGFP-N levels are diminished at both type Ib and Is AZs, demonstrating a conserved role for Brp in promoting Cac accumulation at both synaptic subtypes (Fig. 3G). Strikingly, the relative decrease in Cac levels at type Ib AZs is significantly greater than at type Is AZs, indicating a greater requirement for Brp in regulating VGCC levels at low-P_r type Ib synapses (Fig. 3H). This suggests that an additional factor or factors function with or upstream of Brp to establish differences between low and high-P_r AZs.

Brp differentially regulates VGCC dynamics at low-and high-P_r synapses during presynaptic homeostatic potentiation

In response to acute or chronic inhibition of glutamate receptors at NMJs, Drosophila motor neurons homeostatically increase neurotransmitter release to maintain synaptic communication (Davis and Muller, 2015; Frank, 2014; James et al., 2019). Pharmacological inhibition of glutamate receptors with the wasp toxin Philanthotoxin-433 (PhTx) induces acute presynaptic homeostatic potentiation of release (PHP) within minutes (Frank et al., 2006). We and others have demonstrated that acute PHP involves rapid changes in VGCCs and other AZ protein levels at AZs of type Ib inputs (Bohme et al., 2019; Gratz et al., 2019; Weyhersmuller et al., 2011). Recent studies have revealed significant differences in the induction of PHP at low-and high-P_r synaptic inputs under different conditions (Genc and Davis, 2019; Newman et al., 2017; Sauvola et al., 2021). PhTx induces acute PHP at both type Ib and Is synapses (Genc and Davis, 2019), but the molecular changes underlying PHP at high-P_r type Is AZs remain unknown. To compare the dynamic modulation of VGCCs at low-and high-P_r AZs, we treated cac^sfGFP-N larvae with non-saturating concentrations of PhTx for 10 min, then quantified Cac and Brp levels at type Ib and Is AZs (Fig. 4A, B). We observe a significant PhTx-induced increase in Brp and Cac^sfGFP-N levels at type Is AZs similar to type Ib (Fig. 4C, D; (Gratz et al., 2019)). Thus, despite their distinct baseline transmission and organizational properties, PhTx-induced potentiation of neurotransmitter release involves rapid accumulation of VGCCs at both low-and high-P_r AZs.

Figure 4.

At low-P_r type Ib AZs, Brp is a critical regulator of PHP-induced accumulation of proteins associated with SV priming and release, specifically Unc13A and Syntaxin-1A (Bohme et al., 2019). At type Ib AZs, PhTx also induces a Brp-dependent increase in Cac density and decrease in channel mobility (Ghelani et al., 2023). Notably, Brp itself is more densely organized during PHP (Ghelani et al., 2023) and at high-P_r type Is AZs (Mrestani et al., 2021). Since baseline accumulation of VGCCs depends less on Brp at high-P_r type Is AZs, we investigated the role of Brp in promoting dynamic increases in VGCC levels at type Ib and Is AZs by treating cac^sfGFP-N; brp^-/- larvae with PhTx followed by quantification of Cac^sfGFP-N levels (Fig. 4E, F). We find that PhTx failed to induce accumulation of Cac at either type Ib or Is AZs in brp^-/- mutants, demonstrating a shared requirement for Brp in regulating VGCC dynamics at low-and high-P_r AZs (Fig 4G). Notably, in contrast to no change at type Ib AZs, Cac^sfGFP-N levels are significantly decreased at type Is AZs (Fig 4G), revealing subtype-specific roles for Brp during the dynamic reorganization of VGCCs at low-and high-P_r AZs. Consistently, Ghelani et al. (2023) found that whereas PhTx induces a decrease in the Cac mobility at wild-type type Ib AZs, in brp^-/- mutants Cac mobility increases (Ghelani et al., 2023). Together, these findings suggest potentiating synapses must coordinate the accumulation of new VGCCs with the stabilization of existing channels, and that meeting this challenge is more dependent upon Brp at high-P_r AZs.

Endogenous tagging of VGCC auxiliary subunits reveals distinct synaptic expression patterns

In addition to spatial organization, functional differences in the VGCCs localized at low-and high-P_r AZs may contribute to differences in the relationship between channel levels and P_r. Consistent with this possibility, Ca²⁺ influx is reported to be ∼2x higher at type Is AZs and neurotransmitter release saturates at lower external Ca²⁺ concentrations than type Ib AZs (He et al., 2022; Lu et al., 2016). In addition to the pore-forming α subunits, VGCCs comprise auxiliary α2δ and β subunits that regulate forward channel trafficking, membrane insertion, and function (Fig. 5A; (Campiglio and Flucher, 2015; Dolphin and Lee, 2020; Weiss and Zamponi, 2017)). β subunits interact with pore-forming α subunits intracellularly, whereas GPI-anchored α2δ subunits are largely extracellular. In addition to their interaction with α subunits, α2δs have been shown to interact with a growing number of extracellular proteins to promote synaptogenesis (Bauer et al., 2010; Dolphin, 2018). The Drosophila genome encodes one synaptic Ca_v2 α subunit (Cac), one β subunit, and three α2δ subunits (Littleton and Ganetzky, 2000). Auxiliary subunits are both spatially and temporally regulated, and broadly able to interact with α subunits. Thus, the subunit composition of channel complexes is a potential source of significant diversity in both the spatial and functional regulation of VGCCs.

Figure 5.

ca-β encodes the sole Drosophila β subunit and has been shown to enhance Ca²⁺ transients in sensory neurons (Kanamori et al., 2013). Drosophila α2δ-3, also known as Straightjacket (Stj), has well-characterized roles at the NMJ in promoting Ca²⁺ channel clustering, homeostatic plasticity, and, independently of Cac, synapse formation and organization (Dickman et al., 2008; Hoover et al., 2019; Kurshan et al., 2009; Ly et al., 2008; Wang et al., 2016). The remaining two α2δ subunits don’t map to a specific mammalian counterpart. CG4587/Stolid was recently shown to promote dendritic Cac expression in motor neurons, whereas Ma2d is known to function in muscle where it is broadly expressed (Heinrich and Ryglewski, 2020; Reuveny et al., 2018). The synaptic localization of endogenous auxiliary subunits with VGCCs remains unknown in Drosophila. To explore potential differences in VGCC subunit composition at type Ib and Is synapses, we used CRISPR gene editing to incorporate endogenous V5 tags at the N-termini of Stj and Stolid and C-terminus of Ca-β, and confirmed that the incorporation of the peptide tag did not impair neurotransmission (Fig. 5B-E; (Bruckner et al., 2017; Gratz et al., 2014). We investigated the expression of each endogenously tagged subunit in the larval ventral ganglion and found that all subunits are expressed in the synaptic neuropil in a pattern similar to the α subunit Cac (Fig. 5F-H; (Gratz et al., 2019)). Similar to Cac, Ca-β^V5-C is highly enriched in the mushroom bodies of the larval brain. We next investigated expression at the larval NMJ where Cac localizes in a single puncta at each AZ and found that only Ca-β^V5-C and Stj^V5-N are present (Fig. 5F-H). This aligns with a recent study indicating that Stolid does not play a role in regulating Ca²⁺ transients at the larval NMJ (Heinrich and Ryglewski, 2020). We also observe Ca-β^V5-C expression in muscle as expected for the sole β subunit (Fig. 5F and see Fig. 6C).

Figure 6.

Stj/α2δ-3 levels are lower at AZs of high-P_r type Is inputs

To investigate Caβ^V5-C and Stj^V5-N localization at type Ib and Is AZs, we used superresolution optical reassignment microscopy. Both subunits localize to AZs labeled with Cac or the CAST/ELKS AZ cytomatrix protein Brp (Fig 6A, B). We observe Brp rings surrounding puncta of VGCCs, including the Ca-β^V5-C and, based on its colocalization with Cac, Stj^V5-N subunits (Fig. 6A, B; (Fouquet et al., 2009b; Kittel et al., 2006)). Their tight colocalization at AZs suggest that the three subunits associate as a complex and predict that Ca-β^V5-C and Stj^V5-N levels, like Cac, will be similar at the two synapses. To compare subunit levels at low-and high-P_r AZs, we simultaneously imaged type Ib and Is inputs using confocal microscopy and measured fluorescence intensity (Fig. 6C, D). As predicted, we found that Ca-β^V5-C levels are similar at type Ib and Is AZs (Fig. 6E). In contrast, Stj^V5-N fluorescence levels are significantly lower at higher-P_r type Is AZs (Fig. 6F). Thus, while Cac and Ca-β are present in similar ratios at both synaptic subtypes, surprisingly, the same is not true of Stj/α2δ-3 with high-P_r type Is AZs exhibiting a greater Cac:Stj ratio. This unexpected finding is consistent with studies of mammalian subunits indicating that in contrast to β subunits, α2δ interactions with α subunits may be transient, leading to a pool of VGCCs lacking α2δ (Muller et al., 2010; Voigt et al., 2016). Our results suggest this pool is present in vivo and larger at high-P_r type-Is inputs.

To further investigate the contribution of Stj to synaptic heterogeneity, we analyzed the relationship between Cac and Stj levels at individual AZs of type Is inputs. Stj^V5-N and Cac^sfGFP-N levels are highly positively correlated at type Is AZs (Fig. 6G). We observe the same relationship between Stj^V5-N and Cac^sfGFP-N levels at type Ib AZs (Fig. 6H). Because P_r is highly positively correlated with Cac levels within synaptic subtypes, this indicates that Stj levels are also positively correlated with P_r within, but not between, synaptic subtypes. Together, these findings reveal context-specific contributions of VGCC levels and composition to synaptic heterogeneity.

Discussion

Complex nervous system function depends on communication at synapses with heterogeneous properties. We have investigated the contributions of VGCC levels, organization, and subunit composition to synaptic heterogeneity at two closely related excitatory neurons that form synapses with distinct neurotransmitter release properties. This approach revealed that while VGCC levels alone are predictive of the strength of individual synapses within neuronal subtypes, underlying differences in spatial and molecular organization alter the relationship between channel levels and release probability at synapses formed by distinct neuronal subtypes. This provides a framework for understanding how multiple levels of molecular and organizational diversity intersect to generate extensive synaptic heterogeneity.

Investigations at diverse synapses using approaches ranging from cell-attached patch recordings to freeze-fracture immuno-electron microscopy to correlative functional imaging have revealed a strong positive correlation between VGCC number and P_r (Akbergenova et al., 2018; Gratz et al., 2019; Holderith et al., 2012; Nakamura et al., 2015; Sheng et al., 2012). While this conclusion corresponds neatly with the dependence of neurotransmitter release on Ca²⁺ influx, counterintuitively, there is a disconnect between VGCC levels and P_r at some synapses. Although VGCC number positively correlates with P_r at the immature calyx of Held, the mature calyx contains fewer VGCCs despite higher P_r (Fedchyshyn and Wang, 2005; Sheng et al., 2012; Wang and Augustine, 2014). Similarly, in the cerebellum, inhibitory stellate neurons form high-P_r synapses with lower VGCC levels than low-P_r synapses formed by excitatory granular neurons (Rebola et al., 2019). Adding to these paradoxical examples, VGCC levels positively correlate with P_r among the heterogeneous synapses formed by either low-P_r type Ib or high-P_r type Is motor neurons, but overall VGCC levels are are similar at type Ib and Is inputs despite a 2-3-fold difference in P_r (Fig. 1, 2; (Lu et al., 2016; Newman et al., 2017). Accordingly, correlative functional imaging confirms that the same number of channels can support greater release at these high-P_r AZs (Fig. 1). This difference is likely due, at least in part, to distinct spatial organization at type Is AZs. Cac clusters are denser at high-P_r AZs (Fig. 2). Brp, which organizes both VGCCs and SVs, is also more densely organized at type Is synapses (Mrestani et al., 2021), consistent with an overall more compact organization of high-P_r AZs. A straightforward prediction is that a more compact AZ organization will decrease the distance between VGCCs and SVs. Indeed, a recent electrophysiological study using new tools for genetically isolating type Ib and Is inputs demonstrated that neurotransmitter release at denser type Is synapses is less impacted by the slow Ca²⁺ chelator EGTA than type Ib synapses, indicating tighter VGCC-SV coupling (He et al., 2022). More densely organized VGCCs at the mature vs. developing calyx of Held also exhibit greater coupling with SVs (Chen et al., 2015; Fedchyshyn and Wang, 2005; Fekete et al., 2019; Nakamura et al., 2015; Sheng et al., 2012). Similarly, high-P_r stellate synapses have greater functional coupling attributable to their more compact topology (Rebola et al., 2019). Together, these findings suggest that more compact AZs may be a general organizing principle of high-P_r synapses. We propose that underlying organizational differences between low-and high-P_r synaptic subtypes combine with molecular diversity within each subtype to generate still greater synaptic heterogeneity. This explains our observation that VGCC levels, as well as Brp and Stj levels, predict P_r within, but not between, synaptic subtypes, as well as the seemingly paradoxical findings that VGCC levels do not always correlate with P_r.

Brp/CAST/ELKS AZ cytomatrix proteins are central regulators of synapse organization across species (Dai et al., 2006; Dong et al., 2018; Hallermann et al., 2010; Held et al., 2016; Kittel et al., 2006; Liu et al., 2014; McDonald et al., 2020; Radulovic et al., 2020). Brp interacts with a large number of AZ proteins to promote Cac clustering at AZs, organize SVs, and recruit Unc13A, which defines SV release sites (Bohme et al., 2016; Fulterer et al., 2018; Ghelani et al., 2023; Liu et al., 2011). Consistently, type Is AZs, which in addition to increased Brp density also have lower Brp levels (Fig. 3) and fewer release sites (He et al., 2022; Jetti et al., 2023; Mrestani et al., 2021). Given Brp’s myriad interactions, it is not difficult to envision its compaction bringing VGCCs and SVs in closer proximity and leading to higher P_r independent of the overall number of release sites and Cac levels. Consistent with this model, Brp density increases during PHP, extending the positive relationship between AZ density and P_r to dynamic changes in release properties (Mrestani et al., 2021). This denser Brp scaffold also promotes PhTx-induced increases in Cac levels and density along with the addition and reorganization of release sites (Fig. 4; (Bohme et al., 2019; Dannhauser et al., 2022; Ghelani et al., 2023). While Brp clearly plays a central role in AZ organization and reorganization, we find that type Ib and Is synapses have distinct requirements for Brp during synapse formation and homeostatic potentiation (Fig. 3, 4). Synapse-specific roles for Brp are supported by a recent functional imaging study in Drosophila (Jetti et al., 2023) and studies of ELKS at mammalian inhibitory and excitatory synapses (Held et al., 2016), and suggest that Brp acts in parallel with or downstream of other factors that establish neuron-specific differences. A recent single-cell transcriptomic study of type Ib and Is motor neurons provides an unbiased starting point for identifying candidate regulators of organizational differences at low-and high-P_r AZs (Jetti et al., 2023). A number of cytoskeletal and motor-related proteins, regulators of proteostasis, and post-translational modifying enzymes/pathway components – all of which could potentially contribute to establishing the observed molecular and/or spatial differences – are differentially expressed in either type Ib or Is motor neurons

The VGCC complex itself provides an additional potential mechanism for diversifying synapses. Both β and α2δ subunits can influence the membrane localization and function of VGCCs. In addition to the ability to mix and match subunits, many of the genes encoding VGCC subunits across species are extensively alternatively spliced to generate functional diversity (Lipscombe et al., 2013; Lipscombe and Lopez Soto, 2019) – an area of great interest for further investigation. Because both β and α2δ subunits are generally considered positive regulators of channel trafficking and function, we were surprised to find that Stj/α2δ-3 levels are lower at high-P_r AZs (Figs. 5, 6). Since AZ levels of α subunit Cac are similar at the two synaptic subtypes, lower levels of Stj at type Is synapses indicates a difference in α:α2δ-3 stoichiometry. The stoichiometry of α:α2δ appears to vary. While some studies observe a tight association between the two subunits, a single-molecule tracking study of mammalian VGCCs found that α2δ subunits have a relatively low affinity for α subunits, resulting in a population of α subunits not associated with an α2δ subunit (Cassidy et al., 2014; Voigt et al., 2016). Consistently, whereas α and β subunits were isolated at near equimolar ratios following affinity purification of Ca_v2 channels, molar levels of α2δ were a surprising 90% lower (Muller et al., 2010). Our findings suggest there is a pool of VGCCs lacking an α2δ subunit at endogenous synapses, and further suggest that this pool is greater at high-P_r type Is AZs. Stj is both required (Dickman et al., 2008; Kurshan et al., 2009; Ly et al., 2008) and rate limiting (Cunningham et al., 2022) for Cac accumulation at AZs. Stj does not appear to function in the stabilization of channels at the AZ membrane, but rather at an upstream step in the progression from ER to plasma membrane (Cunningham et al., 2022). An upstream role may explain how Cac levels are similar at type Is AZs despite lower Stj levels. Tools for following Stj dynamics in developing neurons will help clarify its precise role in Cac delivery. How might a higher α:α2δ-3 ratio result in higher P_r? A number of recent studies raise intriguing possibilities. While the correspondence of Drosophila α2δ subunits to specific mammalian counterparts is somewhat murky, Stj appears to be most closely related to α2δ-3. Strikingly, α2δ-3 inhibits activity at excitatory synapses of mature, but not immature, rat hippocampal cultures (Bikbaev et al., 2020), raising the possibility that Stj’s role may shift during synapse maturation. Further, binding of the cell adhesion molecule Neurexin-1α specifically inhibits Ca²⁺ currents in Ca_v2.2 channels containing α2δ-3 (Tong et al., 2017), which, if similar at the Drosophila NMJ, would result in greater inhibition of channel function at type Ib AZs and contribute to the observed difference in Ca²⁺ influx. Elucidating the many intersecting mechanisms underlying synaptic heterogeneity and how these differences are maintained or modulated during different forms of plasticity is a challenging and important goal for the field going forward.

Materials and methods

Drosophila genetics and gene editing

The following fly lines used in this study are available at the Bloomington Drosophila Stock Center (BDSC): w¹¹¹⁸ (RRID:BDSC_5905), vasa-Cas9 (RRID:BDSC_51324), piggyBac transposase (RRID:BDSC_8283), and Df(2R)brp^6.1 (Gratz et al., 2014; Horn et al., 2003). brp alleles were generously provided by Stephan Sigrist (Freie Universität Berlin; (Fouquet et al., 2009a; Kittel et al., 2006). brp loss-of-function experiments were performed in brp⁶⁹/Df(2R)brp6.1. Drosophila melanogaster stocks were raised on molasses food (Lab Express, Type R) in a 25°C incubator with controlled humidity and 12h light/dark cycle. Endogenously tagged cac, ca-β, stolid, and straightjacket (stj) alleles were generated using our piggyBac-based CRISPR approach as previously detailed (flyCRISPR.molbio.wisc.edu; Bruckner et al., 2017, Gratz et al., 2019). Endogenous tags were incorporated at the N-terminus of Cac, which has been shown to support a variety of tags, to make Cac^Td-Tomato-N and Cac^HaloTag-N (Gratz et al., 2019, Cunningham, Ghelani et al., 2023). V5 tags were incorporated at the N-terminus of Stj and Stolid after their respective signal peptide sequences and in the last common exon of Ca-β near the C terminus of shorter isoforms. All CRISPR-generated lines are fully viable in homozygous males and females and were molecularly confirmed by Sanger sequencing.

Immunostaining

All antibodies used, associated fixation methods, and incubation times can be found in Table S2. Male wandering third-instar larvae were dissected in ice-cold saline and fixed either for 6 minutes at room temperature with Bouin’s fixative, 5 minutes on ice with 100% methanol, or 30 minutes at room temperature (RT) in 4% PFA. Dissections were permeabilized with PTX (PBS with 0.1% Triton-X 100) and blocked for 1 hour at RT using 5% goat serum and 1% bovine serum albumin. Stained larvae were mounted in Vectashield (Vector Laboratories, #H-1000) under Fisherbrand coverglass (Fisher Scientific, #12541B) for confocal microscopy, with Prolong glass mounting medium (ThermoFisher Scientific, #P36980) under Zeiss High Performance Coverglass (Zeiss, #474030-9000-000) for superresolution optical reassignment microscopy, or buffer (see STORM imaging and analysis section) under Zeiss coverglass with edges sealed using vacuum grease for STORM microscopy.

Ca²⁺ imaging and analysis

Functional imaging was performed on a Nikon A1R resonant scanning confocal mounted on a FN1 microscope using a Nikon Apo LWD 25x 1.1 NA objective and a Mad City Labs piezo drive nosepiece. Dissections and data collection were performed as previously described in Gratz et al., 2019. Briefly, Cac^Td-Tomato-N; Mhc-GCaMP6f male 3rd instar larvae were dissected in HL3 containing 0.2mM Ca²⁺ and 25mM Mg²⁺ with motor axons severed and the larval brain removed. Larval filets were placed in HL3 containing 1.5mM Ca²⁺ and 25mM Mg²⁺ for recording. Nerves were suctioned into 1.5mm pipettes and stimulus amplitude was adjusted to recruit both type Ib and type Is input. Motor terminals from segments A2-4 at NMJ 6/7 were imaged for Cac^Td-Tomato-N levels first using a galvanometer scanner, then a resonant scanner to collect GCaMP6f events in a single focal plane continuously for 120 stimulations at a 0.2 Hz stimulation frequency.

Z-stacks and movies were loaded into Nikon Elements Software (NIS) where movies were motion corrected, background subtracted, and denoised. Change in fluorescence (ΔF) movies were then created by subtracting the average of the previous 10 frames from each frame. A substack of only stimulation frames was further processed using a gaussian filter followed by BrightSpots detection to identify the location of each postsynaptic event using the Nikon GA3 module. Cac^Td-Tomato-N fluorescence intensity levels and coordinate locations were measured for 531 AZs for type Ib and 365 AZs for type Is terminals across 6 animals. X-Y coordinate positions of fluorescent signals from GCaMP6f postsynaptic events were aligned to Cac^Td-Tomato-N puncta locations and each post synaptic event assigned to a Cac puncta using nearest neighbor analysis. Postsynaptic events that did not map within 960 nm of a Cac^{Td- Tomato-N} punctum were discarded from the analysis. Pearson’s correlation was used to determine the correlation between P_r and Cac levels normalized to average to account for variability between imaging sessions. Cac intensity-P_r heat maps were generated using Python matplotlib and seaborn plotting packages.

STORM imaging and analysis

STORM imaging was performed on a Nikon Eclipse Ti2 3D NSTORM with an Andor iXon Ultra camera, Nikon LUN-F 405/488/640 nm lasers, and a Nikon 100x 1.49 NA objective. STORM buffer (10mM MEA (pH 8.0), 3 U/mL pyranose oxidase, and 90 U/mL catalase, 10% (w/v) glucose, 10 mM sodium chloride, and 50 mM Tris hydrochloride) was made fresh each imaging day and pH adjusted to between 7.0-8.0 using acetic acid. Cac^HaloTag-N NMJs were labeled as detailed in Table S2 and immediately imaged for HRP in 488 channel to identify type Ib and type Is terminals. Cac^HaloTag-N was then imaged using the 640 nm laser line at 33 Hz for 5000 frames. 405 nm laser power was gradually increased over the course of imaging to compensate for run-down of blinking rates. A back aperture camera was used to ensure beam focus and position for each imaging session to ensure high signal to noise. Data were binned with a CCD minimum threshold of 100 and drift correction was applied using the NIS Software STORM package. ROIs of single boutons were drawn in NIS using HRP in the 488 nm channel followed by a DBSCAN analysis with criteria of 10 molecules within 50 nm to determine clusters. Positional coordinates of localizations within clusters from DBSCAN were exported from NIS and run through a Python script published with this manuscript. Using the implementation developed in Mrestani et al., 2021 as a starting point, we wrote custom code to use the alpha shapes component of the CGAL package (https://www.cgal.org), via a python wrapper (https://anaconda.org/conda-forge/cgal), to measure the area of Ca²⁺ channel clusters, the number of localizations, and calculate cluster density. To achieve an average lateral localization accuracy of ∼30 nm, all localizations with >50 nm localization accuracy were removed prior to analysis. Using this custom code, Cac^HaloTag-N area was analyzed using an alpha value of 0.015, which controls the complexity of cluster boundaries (not restricted to be convex).

Confocal imaging and analysis

For quantitative AZ analysis, larvae stained in the same dish were imaged on a Nikon Eclipse Ni A1R+ confocal microscope using an Apo TIRF 60x 1.49 NA oil-immersion objective for larval NMJs. NMJs containing both type Is and type Ib branches from muscles 6/7 in segments A2-4 were collected. ROIs were drawn using HRP staining to differentiate between type Ib and type Is branches. To analyze individual AZs, Nikon Elements Software was used to process images using Gaussian and rolling ball filters and analyze fluorescence intensity levels for individual puncta. When experimental design allowed, Brp fluorescence signal was used to create a binary mask to aid in the identification of AZ ROIs for analysis. Otherwise, binary masks were created based on the fluorescence signal of the channel analyzed. Confocal fluorescence intensity level data are reported as the sum fluorescence intensity per AZ averaged over individual NMJs. For Fig. 5, larvae were stained separately and imaged using a Nikon Plan-Apo 20x 0.75 NA objective (ventral ganglia) or Apo TIRF 60x 1.49 NA oil-immersion objective (NMJs).

Superresolution optical reassignment images were obtained on a Nikon CSU-W1 SoRa (Spinning Disk Super Resolution by Optical Pixel Reassignment) with a Photometrics Prime BSI sCMOS camera and a 60x 1.49 NA oil-immersion objective. Images were acquired using Nikon NIS and deconvolved using Richardson-Lucy deconvolution with 15-20 iterations.

Electrophysiology

Current-clamp recordings were performed as previously described (Bruckner, 2017). Male third-instar larvae were dissected in HL3 (70 mM NaCl, 5 mM KCl, 15 mM MgCl2, 10 mM NaHCO3, 115 mM sucrose, 5 mM trehalose, 5 mM HEPES, pH 7.2) with 0.25 mM Ca²⁺. Recordings were performed in HL3 at the external Ca²⁺ concentration indicated. Sharp borosilicate electrodes filled with 3 M KCl were used to record from muscle 6 of abdominal segments A3 and A4. Recordings were conducted on a Nikon FN1 microscope using a 40x 0.80 NA water-dipping objective and acquired using an Axoclamp 900A amplifier, Digidata 1550B acquisition system, and pClamp 11.0.3 software (Molecular Devices).

For each cell with an initial resting potential between −60 and −80 mV and input resistance ≥5 MΩ, mean miniature excitatory junctional potentials (mEJPs) were collected for 1 minute in the absence of stimulation and analyzed using Mini Analysis (Synaptosoft). EJPs were generated by applying a stimulus to severed segmental nerves at a frequency of 0.2 Hz using an isolated pulse stimulator 2100 (A-M Systems). Stimulus amplitude was adjusted to consistently elicit compound responses from both type Ib and Is motor neurons. At least 25 consecutive EJPs were recorded for each cell and analyzed in pClamp to obtain mean amplitude. Quantal content was calculated for each recording as mean EJP amplitude divided by mean mEJP amplitude.

Acute homeostatic challenge

Acute PHP was induced by incubating semi-intact preparations in 20 µM Philanthotoxin-433 (PhTx; Santa Cruz, sc-255421, Lot B1417) diluted in HL3 containing 0.4 mM Ca²⁺ for 10 min at room temperature (Frank et al., 2006). Control preparations were given a mock treatment. Following control and experimental treatment, dissections were completed, fixed in 4% PFA for 30 minutes, and stained for Cac^sfGFP-N and Brp in the same dish. Analyses of fluorescent intensity levels were performed as previously described in the Confocal imaging and analysis section.

Experimental Design and statistical analysis

Statistical analyses were conducted in GraphPad Prism 9. Normality was determined by the D’Agostino–Pearson omnibus test. Comparisons of normally distributed data were conducted by Student’s t test (with Welch’s correction in the case of unequal variance) for single comparisons and ANOVA followed by Tukey’s test for multiple comparisons. For non-normally distributed data, the Mann–Whitney U test and Kruskal-Wallis test followed by Dunn’s multiple comparisons tests were used for single and multiple comparisons, respectively. Paired analysis of non-normally distributed data was conducted using Wilcoxon’s matched-pairs signed rank test. One-dimensional Pearson correlation coefficients (r) were used to compare intensity levels and neurotransmitter release probability. ANCOVA test was performed on all regression lines to determine if slopes were significantly different. Reported values are mean ± SEM. Sample size, statistical test, and p values for each comparison are reported in Table S1.

Acknowledgements

We thank the Developmental Studies Hybridoma Bank, the Bloomington Drosophila Stock Center, Ehd Isacoff (UC Berkeley), and Stephan Sigrist (Freie Universität Berlin) for providing antibodies and fly stocks. We are grateful to Joel Hirsch (Tel Aviv University) for consultations on tagging Ca-β, Nicholas Deakin (Nikon) for guidance on STORM imaging, the Heckmann lab (University of Würzburg) for guidance on STORM image analysis pipelines, and Liana Lewis for her assistance with image analysis. We thank Rajan Thakur and the members of the O’Connor-Giles lab for thoughtful discussions and comments on the manuscript. This work was supported by grants from the National Institute of Neurological Disorders and Stroke, National Institutes of Health to K.M.O.G (R01NS078179) and Audrey Medeiros (F31NS122424), Brown Neuroscience Graduate Program training grant T32 MH020068, and funds from the Brown University Carney Institute for Brain Science.