Introduction

The broad and complex functions of neural circuits depend on diverse neuronal subtypes communicating through synapses with distinct properties. Thus, understanding how synaptic diversity is established is critical for understanding neural circuit function. 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 between neuronal subtypes and within homogeneous populations of neurons (Ariel, Hoppa, and Ryan 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 probabilistic 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 generating diversity of synaptic function (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; Miki et al. 2017; 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 abundance is highly correlated with P_r across species (Akbergenova et al. 2018; Gratz et al. 2019; Holderith et al. 2012; Miki et al. 2017; 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 inhibitory stellate cells and one low-P_r formed by excitatory granule cells, and found higher VGCC levels at low-P_r granule synapses (Rebola et al. 2019). 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). Another recent study investigated two functionally distinct connections formed by CA1 pyramidal cells (Aldahabi et al. 2022). While Ca²⁺ influx was higher at the high-P_r synapse, raising Ca²⁺ influx at the low-P_r synapse to match the high-P_r synapse did not equalize P_r.

To further investigate this paradox, we sought a system where we could investigate the relationship between VGCCs and P_r both within and between 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 abundance, spatial organization, and subunit composition contribute to synaptic heterogeneity at AZs of low-P_r type Ib and high-P_r type Is inputs to the same postsynaptic targets. We find that individual synapses formed by both low- and high-P_r inputs exhibit heterogeneous release properties that can be predicted by VGCC abundance alone. However, VGCC abundance does not correspond to differences in P_r between the two inputs. We identify underlying molecular and organizational differences that may alter the relationship between VGCC abundance and P_r at low- vs. high-P_r inputs. We further find that the homeostatic potentiation of neurotransmitter release triggered by glutamate receptor inhibition involves dynamic increases in VGCCs, the α2δ-3 subunit Straightjacket (Stj), and the AZ cytomatrix protein Bruchpilot (Brp) across AZs of both inputs. These findings provide insight into how VGCC and AZ protein abundance intersects with underlying molecular and organizational differences between inputs to contribute to greater synaptic diversity.

Results

VGCC levels predict P_r within, but not between, inputs

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, Felling, and Ordway 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}) and confirmed that the tag does not impair synaptic function (Figs. 1A-F, S1; (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-Nanimals for a plus/minus readout. 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 inputs, we simultaneously imaged type Ib and 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. As has been previously reported, we found that type Is synapses exhibited significant heterogeneity and higher average P_r than type Ib synapses (Fig. 1B, C; (Lu et al. 2016; Newman et al. 2022)). Consistent with their higher P_r, type Is connections contain relatively fewer low-P_r and more high-P_r AZs (Figs. 1D). 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). We also observe a strong positive correlation between VGCC levels and P_r at type Ib inputs to the same muscles (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 Cac levels will be higher at synapses of type Is inputs than type Ib. We analyzed Cac^sfGFP-N levels at individual type Ib and Is synapses 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 AZs 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 directly compare the best-fit lines for the relationship between Cac levels and P_r at type Ib and Is inputs from our correlative functional imaging data (Fig. 1E, F), we find that the slopes are significantly different (Fig. 1J). Across type Is AZs, a similar range of VGCC levels supports a higher range of release probabilities. Thus, VGCCs can predict P_r within synaptic subtypes, but not between AZs 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

Many differences between low-P_r type Ib and high-P_r type Is AZs have been described (Aponte-Santiago and Littleton 2020; Aponte-Santiago et al. 2020; Atwood, Govind, and Wu 1993; He et al. 2023; Jetti et al. 2023; Kurdyak et al. 1994; Lu et al. 2016; Medeiros and OConnor-Giles 2023).

Perhaps most notably, type Is AZs experience ∼2-fold greater Ca²⁺ influx than type Ib (He et al. 2023; Lu et al. 2016). While this alone could explain the estimated 3-fold greater P_r at type Is AZs and is certainly a key factor, several lines of evidence argue for additional contributors. A recent study using a botulinum transgene to isolate type Ib and Is synapses for electrophysiological analysis found that increasing external [Ca²⁺] from physiological levels (1.8 mM) to 3 mM or even 6 mM does not result in a 3-fold increase in EPSCs or quantal content at type Ib synapses and type Ib synapses continue to facilitate at 3 mM external [Ca²⁺] (He et al. 2023). Using this approach, they further found that type Ib synapses are more sensitive to the slow Ca²⁺ chelator EGTA, indicating looser VGCC-SV coupling.

We investigated the spatial distribution of VGCCs at type Ib and Is AZs using 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, Hoess, and Ries 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 same N-terminal site we previously used to endogenously tag Cac, achieving a linkage error of less of than 5 nm (Ghelani et al. 2023; Gratz et al. 2019). 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). cac^HaloTag-N flies are fully viable, do not display significant defects in synaptic function, and exhibit normal Cac localization at AZs as observed in super-resolution optical reassignment images, where Brp is arranged in rings surrounding puncta of VGCCs (Fig. S1; 2A-C, (Ghelani et al. 2023)). 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.

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 inputs at a single NMJ. 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 VGCC AZ density at type Is AZs is consistent with a recent SMLM study using antibodies to label Cac, indicating these results are robust to different labeling approaches (Newman et al. 2022). Together, these findings suggest that more compact organization of VGCCs increases their coupling to SVs and contributes to the steeper relationship between VGCC levels and P_r at high-P_r type Is AZs.

Figure 2:

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 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, Liu, and Kaeser 2016; Kittel et al. 2006; Liu et al. 2014; McDonald, Fetter, and Shen 2020; Radulovic et al. 2020). Like Cac, Brp is more densely arranged at type Is AZs as measured through SMLM and stimulated emission depletion (STED) imaging studies (He et al. 2023; Jetti et al. 2023; Mrestani et al. 2021). 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 AZs of the two inputs, 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 Ib inputs (Gratz et al. 2019; Muhammad et al. 2015; Newman et al. 2017; Peled, Newman, and Isacoff 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 inputs, likely due to differences in AZ organization between the two synaptic subtypes.

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 inputs (Fig. 3G). 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 in Brp dependence at 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, Zwiefelhofer, and Frank 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 type Ib AZs (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 minutes, 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 the 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 becomes more densely organized during PHP (Ghelani et al. 2023), consistent with its denser organization 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 both inputs (Fig. 4G). In contrast to no change at type Ib AZs, Cac^sfGFP-N levels are significantly decreased at type Is AZs (Fig. 4G), revealing an input-specific role for Brp in maintaining VGCC levels during the dynamic reorganization of 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 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. Beyond 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; Risher et al. 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, Kurshan, and Schwarz 2008; Hoover et al. 2019; Kurshan, Oztan, and Schwarz 2009; Ly et al. 2008; Schöpf et al. 2021; Wang et al. 2016). While low sequence homology between α2δ subunits within and across species makes 1:1 mapping difficult, the remaining two Drosophila α2δs map more closely to mammalian α2δ-3 and -4 than α2δ-1 and -2. 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 in sequence common to all isoforms of stj, stolid, and Ca-β (Bruckner et al. 2017; Gratz et al. 2014). We inserted V5 after the N-terminal signal peptides of Stj and Stolid and near the C-terminus of Ca-β (see Materials and Methods for details), and confirmed that the incorporation of the peptide tags did not impair neurotransmission (Fig. 5B-E). 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 punctum at each AZ and found that only Ca-β^V5-C and Stj^V5-N are present (Fig. 5F-H). This aligns with the recent finding 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 Drosophila 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 Ca-β^V5-C and Stj^V5-N localization at type Ib and Is AZs, we used super-resolution 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 Ca-β^V5-C (Fig. 6A). The tight localization of both subunits to central AZ puncta suggests they are associated with α subunits and predicts that Ca-β^V5-C and Stj^V5-N levels, like Cac, will be similar at the low- and high-P_r synapses. To test this, we simultaneously imaged Ca-β^V5-C and Stj^V5-N levels at both 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 levels are significantly lower at high-P_r type Is AZs (Fig. 6F). Thus, while Cac and Ca-β are present in similar ratios at AZs of both inputs, surprisingly, the same is not true of Stj/α2δ-3 with high-P_r type Is AZs exhibiting lower levels of Stj. This unexpected finding indicates that α:α2δ-3 stoichiometry is not always 1:1 in vivo and differs at low- and high-P_r synapses. This 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 indicate this pool may be 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, inputs.

Stj/α2δ-3 levels are modulated at AZs of both low- and high-P_r inputs during presynaptic homeostatic potentiation

α2δ subunits are critical regulators of α subunit forward trafficking. In flies and mammals, overexpression of α2δ subunits increases α subunit abundance, whereas overexpression of the α subunit alone does not (Cao et al. 2004; Cunningham et al. 2022; Hoppa et al. 2012). These findings suggest that α2δ may be dynamically regulated together with Cac during PHP, a prediction we can now test with our endogenously tagged line. Following PhTx exposure, we find that Stj^V5-N is recruited on a rapid timescale to both low- and high-P_r AZs, increasing by a similar percentage at both type Ib and Is AZs (27% and 26%, respectively) as predicted (Fig. 7A-C). Cac levels are similarly increased (33% at type Ib and 30% at type Is; See Fig. 4D), suggesting coordinated regulation. We have previously shown that Cac abundance is also increased in chronic PHP, which is induced by genetic loss of the GluRIIA receptor subunit (Gratz et al. 2019; Li et al. 2018; Petersen et al. 1997). We investigated Stj dynamics in GluRIIA^-/- null animals and observed elevated Stj^V5-N levels at AZs of both type Ib and Is inputs (18% and 17%, respectively; Fig. 7D-F), indicating similar dynamics during chronic PHP. These data are consistent with the recent finding that in addition to its previously uncovered role in promoting an increase in the readily releasable pool of SVs (Wang et al. 2016), Stj is required for the accumulation of Cac at type Ib AZs during both acute and chronic PHP (Zhang et al. 2023). Thus, the abundance of multiple key AZ proteins distinguishes low- and high-P_r synapses within, but not between, inputs at baseline and during homeostatic plasticity.

Figure 7:

Discussion

Complex nervous system function depends on communication at synapses with heterogeneous and plastic properties. Paradoxical findings in the field have raised questions about the role of VGCCs in establishing neurotransmitter release properties. Our findings suggest a model in which two broad intersecting mechanisms contribute to synaptic diversity in the nervous system: (1) nanoscale spatial organization and relative molecular content establish distinct average basal release probabilities that differ between inputs and (2) coordinated modulation of VGCC and active zone protein abundance independently tunes P_r among individual synapses of distinct inputs. This model provides a framework for integrating diverse findings in the field and understanding how multiple levels of molecular and organizational diversity can 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; Miki et al. 2017; Nakamura et al. 2015; Newman et al. 2022; Sheng et al. 2012). This holds true among mature synapses in the hippocampus or immature synapses of the calyx of Held (Holderith et al. 2012; Sheng et al. 2012) and at the developing Drosophila NMJ where the differences in P_r and channel number between AZs of a single input also correlate with synapse maturity (Akbergenova et al. 2018; Gratz et al. 2019; Newman et al. 2022). While this conclusion corresponds neatly with the dependence of neurotransmitter release on Ca²⁺ influx, counterintuitively, a disconnect between VGCC and P_r is observed in some studies. Although the number of functional VGCCs positively correlates with P_r among synapses in the immature calyx of Held, the mature calyx has higher P_r yet recruits fewer Ca²⁺ channels (Fedchyshyn and Wang 2005; Sheng et al. 2012; Wang and Augustine 2014). Similarly, in the cerebellum, inhibitory stellate cells form high-P_r synapses with lower VGCC levels than low-P_r synapses formed by excitatory granular cells (Rebola et al. 2019). Adding to these paradoxical examples, we find that 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 abundance is similar at the two inputs (Figs. 1, 2) despite an ∼3-fold difference in P_r (Lu et al. 2016; Newman et al. 2017). Accordingly, our correlative functional imaging confirms that the same number of channels can support greater release at these high-P_r AZs (Fig. 1). While two-fold greater Ca²⁺ influx at high-P_r type Is AZs can explain much of this difference (He et al. 2023; Lu et al. 2016), mounting evidence suggests that here and in other systems differences in P_r are separable from Ca²⁺ influx. At distinct inputs formed by CA1 pyramidal neurons, Ca²⁺ influx is greater at high-Pr synapses, but doesnt explain differences in synaptic strength as raising Ca²⁺ entry at low-Pr synapses to high-Pr synapse levels was not sufficient to increase synaptic strength to high-Pr input levels (Aldahabi et al., 2022). Similar findings have been reported at tonic and phasic synapses of the Crayfish NMJ (Msghina, 1999).

The separability of VGCC abundance, Ca²⁺ influx, and P_r appears to be due to molecular and spatial differences between synaptic subtypes. In CA1 pyramidal neurons, differences in Munc-13-dependent SV priming are proposed to establish synapse-specific release properties, possibly due to the presence of distinct isoforms at low- vs. high-P_r connections. In the cerebellum, fewer VGCC are more tightly coupled to SVs at high-P_r stellate synapses (Rebola et al. 2019). 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). We find that Cac clusters are denser at high-P_r AZs formed by Drosophila phasic motor neurons (Fig. 2). Brp, which organizes both VGCCs and SVs, is also more densely organized at type Is synapses (He et al. 2023; Jetti et al. 2023; 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. 2023). Together, these findings suggest that more compact AZs may be a common organizing principle of high-P_r synapses. Consistent with this model, Brp and Unc-13 density increase during PHP when P_r is increased (Dannhauser et al. 2022; Ghelani et al. 2023; Mrestani et al. 2021).

We also observe molecular differences between Drosophila motor inputs that may contribute to distinct P_r. Somewhat counterintuitively, high-P_r type Is AZs have lower levels of both Brp and Stj/α2δ-3 (Figs. 3, 6). 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, Liu, and Kaeser 2016; Kittel et al. 2006; Liu et al. 2014; McDonald, Fetter, and Shen 2020; Radulovic et al. 2020). Brp interacts broadly with AZ proteins to promote Cac clustering, 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). While Brp clearly plays a central role in AZ organization and reorganization, we also find that type Ib and Is synapses have distinct requirements for Brp during synapse formation and homeostatic potentiation (Figs. 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, Liu, and Kaeser 2016), and suggest additional factors establish neuron-specific differences in Brp dependence. A recent single-cell transcriptomic study of type Ib and Is motor neurons provides an unbiased starting point for identifying candidate regulators of molecular differences at low- and high-P_r AZs (Jetti et al. 2023). Differentially expressed genes encode 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 between type Ib and Is AZs.

The VGCC complex itself provides an additional potential mechanism for diversifying synaptic function. Both β and α2δ subunits can influence the membrane localization and function of VGCCs (Campiglio and Flucher 2015; Dolphin and Lee 2020; Weiss and Zamponi 2017). 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 additional functional diversity (Cingolani et al. 2023; Lipscombe, Andrade, and Allen 2013; Lipscombe and Lopez Soto 2019) – an area of great interest for further investigation at the Drosophila NMJ. Because both β and α2δ subunits are generally considered positive regulators of channel trafficking and function, we were surprised to find upon endogenously tagging Stj/α2δ-3 that its levels are lower at high-P_r AZs (Figs. 5, 6). Since AZ levels of α subunit Cac are similar at the two inputs, lower levels of Stj at type Is synapses indicates a difference in α:α2δ-3 stoichiometry between low- and high-P_r synaptic subtypes. While some previous studies have observed a tight association between the two subunits, a single-molecule tracking and modeling study of mammalian VGCCs found that α2δ subunits have a relatively low affinity for α subunits and predicted 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 may be a pool of VGCCs lacking an α2δ subunit at endogenous synapses, and further suggest that this pool is specific to or enriched at high-P_r type Is AZs. Stj is both required (Dickman, Kurshan, and Schwarz 2008; Kurshan, Oztan, and Schwarz 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), so complexes may not need to be maintained. Tools for following Stj dynamics in developing neurons will help clarify its precise role in Cac delivery at type Ib and Is AZs.

How might a higher α:α2δ-3 ratio result in higher P_r? One possibility involves α2δ-3 interactions with cell adhesion molecules. In mammals, α-Neurexin specifically inhibits Ca²⁺ currents in Ca_v2.2 channels containing α2δ-3 (Tong et al. 2017), which, if similar at the Drosophila NMJ, could result in greater inhibition of channel function at type Ib AZs and contribute to the observed difference in Ca²⁺ influx. Loss of Drosophila Neurexin reduces neurotransmitter release at the NMJ, but it also leads to reduced synaptogenesis so whether this is a direct effect remains unknown (Li et al. 2007). Another possibility is that Stj isoforms with different functions are differentially expressed at type Ib and Is AZs. A recent transcriptomic study observed no significant differences in mRNA splice isoforms, but transcript levels do not always reflect protein levels (Jetti et al. 2023). Notably, we find that among synapses of either low- or high-P_r AZs, Stj levels correlate with P_r and Stj levels are increased across AZs of both inputs during acute and chronic PHP (Fig. 7). Cac and Stj levels increase by similar amounts at the two inputs, suggesting the difference in stoichiometry is maintained following homeostatic potentiation. α2δ-3 is an important drug target for treating epilepsy, neuropathic pain, and anxiety, so understanding its cell-specific roles and how α:α2δ-3 stoichiometry impacts channel function and contributes to synaptic diversity is of great interest.

Materials and methods

Drosophila genetics and gene editing

The following fly lines used in this study were obtained from the Bloomington Drosophila Stock Center (BDSC, NIH P40OD018537): 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). brp⁶⁹ and GluRIIA^sp16 (GluRIIA^-/-) alleles were generously provided by Stephan Sigrist (Freie Universität Berlin; (Fouquet et al. 2009; Kittel et al. 2006)) and Aaron DiAntonio (Petersen et al. 1997) respectively. brp loss-of-function experiments were performed in brp⁶⁹/Df(2R)brp^6.1. 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)). We used FlyBase (release FB2024_01) to obtain genomic sequences (doi.org/10.1093/genetics/iyad211). Td-Tomato and Halo tags were incorporated at the N-terminus of Cac, where we have previously incorporated fluorescent tags (Ghelani et al. 2023; Gratz et al. 2019). V5 tags were incorporated at the N-terminus of Stj and Stolid after their signal peptide sequences, immediately after Asparagine 34 of Stj and after Isoleucine 34 of Stolid. For Ca-β, V5 was inserted immediately after Proline 446 of isoform Ca-β-PN. All endogenously tagged 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 Bouins 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 super-resolution 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 Is inputs. 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 the Bright Spots detection module in the Nikon GA3 software to identify the location of each postsynaptic event. 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 punctum 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. Pearsons 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 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 of larval NMJs, dissections 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 Is branches. To analyze individual AZs, Nikon Elements GA3 Software was used to process images with Gaussian and rolling ball filters and measure fluorescence intensity levels at individual puncta identified by the Bright Spots module. 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).

Super-resolution 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 et al. 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 (Fig. 4: PhTx; Santa Cruz, sc-255421, Lot B1417 and Fig. 7: PhTx; Sigma Aldrich, P207-2, Lot MKCK7405) diluted in HL3 containing 0.4 mM Ca²⁺ for 10 minutes 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 (cac^sfGFP-N) or 100% ice-cold methanol on ice for 5 minutes (stj^V5-N), and stained 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 DAgostino–Pearson omnibus test. Comparisons of normally distributed data were conducted by Students t test (with Welchs correction in the case of unequal variance) for single comparisons and ANOVA followed by Tukeys test for multiple comparisons. For non-normally distributed data, the Mann–Whitney U test and Kruskal-Wallis test followed by Dunns multiple comparisons tests were used for single and multiple comparisons, respectively. Paired analysis of non-normally distributed data was conducted using Wilcoxons 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 (NIH P40OD018537), Flybase (Öztürk-Çolak et al. 2024), Ehud Isacoff (UC Berkeley), and Stephan Sigrist (Freie Universität Berlin) for providing antibodies and fly stocks. The Nikon SoRa/STORM microscope was generously provided by The Neurobiology of Cells and Circuits/Center for Translational Neuroscience Microscopy Committee, Carney Institute for Brain Science. 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, Matthew Knoeppel for help generating CRISPR alleles, and Liana Lewis for her assistance with image analysis. We thank Rajan Thakur and the members of the OConnor-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.