Complexin gains effective access to the assembling SNAREs via its membrane-binding C-terminal domain

Abstract

The conserved presynaptic SNARE-binding protein Complexin (Cpx) promotes Ca²⁺-triggered synaptic vesicle (SV) fusion and inhibits spontaneous fusion at some synapses. A membrane-binding motif in the C-terminal domain (CTD) of Cpx plays a critical role in Cpx function, but it remains unclear whether the CTD participates in Cpx regulation of synaptic transmission beyond targeting Cpx to membranes. We examined the impact of the C. elegans CPX-1 CTD in vivo and found that this domain profoundly boosted the efficiency of CPX-1 mediated inhibition of spontaneous SV fusion as a function of protein abundance at the synapse. Removing the C-terminal half of CPX-1 and substituting it with the SV protein RAB-3 could fully restore both the fusogenic and inhibitory functions of CPX-1 whereas other SV proteins failed to restore CPX-1 function with the same efficiency regardless of abundance. These results indicate that regulation of spontaneous SV fusion requires a specific interaction of CPX-1 with the SV membrane. We propose that Cpx cannot efficiently access assembling SNAREs from the cytoplasm and that interactions of its CTD with the SV membrane guide Cpx to these sites of SNARE assembly.

Introduction

Neurotransmitter secretion arises from a tightly orchestrated series of molecular events culminating with the assembly of the neuronal SNARE proteins (VAMP2/Synaptobrevin 2, SNAP25, and Syntaxin 1) and subsequent fusion of synaptic vesicles (SVs) with the plasma membrane (Sudhof, 2013; Brunger et al., 2018; Dittman & Ryan, 2019; Rizo, 2022). Complexin (Cpx) is a relatively small cytoplasmic protein (~130-140 residues) associated with SV fusion machinery and proposed to play several functions at the synapse by binding the assembling SNARE complex (Pabst et al., 2000; Tokumaru et al., 2001; Chen et al., 2002; Pabst et al., 2002; Giraudo et al., 2006; Melia, 2007; Brose, 2008). The Cpx peptide contains four functional domains: an N-terminal domain, accessory helix, central helix (CH), and an unstructured C-terminal domain (CTD) (Xue et al., 2007; Trimbuch & Rosenmund, 2016; Lottermoser & Dittman, 2023) (Fig 1A). One enduring mystery surrounding this critical presynaptic protein is its seemingly opposite roles in promoting and inhibiting release at different synapses. In particular, loss of the mammalian Cpx paralogs mCpx1 and mCpx2 decreases both stimulus-evoked and spontaneous neurotransmitter release in several types of neurons whereas loss of the fly and nematode Cpx orthologs results in massive enhancement of spontaneous SV fusion along with decreased stimulus-evoked fusion (Huntwork & Littleton, 2007; Xue et al., 2007; Xue et al., 2008; Xue et al., 2009; Hobson et al., 2011; Martin et al., 2011; López-Murcia et al., 2019). The CH domain binds the ternary SNARE complex with high affinity, and all known functions of Cpx are impaired if this domain is disrupted (Giraudo et al., 2006; Xue et al., 2007; Brose, 2008; Martin et al., 2011; Radoff et al., 2014; Trimbuch et al., 2014). The Cpx CTD is less well understood but also shown to play a critical role in Cpx function (Chicka & Chapman, 2009; Xue et al., 2009; Cho et al., 2010; Martin et al., 2011; Kaeser-Woo et al., 2012; Wragg et al., 2013; Wragg et al., 2017; Makke et al., 2018; Courtney et al., 2022; Wang et al., 2025). In vitro studies identified a conserved amphipathic helical region near the end of the CTD required for interactions with liposomes, and a model was proposed for the CTD as a vesicle membrane anchor for Cpx (Seiler et al., 2009). Subsequent work from multiple groups revealed that this amphipathic CTD region binds preferentially to highly curved membranes and targets Cpx to SVs in vivo (Wragg et al., 2013; Snead et al., 2014; Gong et al., 2016; Zdanowicz et al., 2017; Lottermoser & Dittman, 2023). Notably, a single point mutation at the end of the Cpx1 CTD is associated with severe neurological impairments in humans (Redler et al., 2017). While these studies buttress the notion that the Cpx CTD plays a critical functional role in regulating SNARE assembly and synaptic transmission, many questions remain. What is the functional significance of membrane interactions with the CTD? Is membrane binding itself part of the Cpx inhibitory or facilitatory function? Does the CTD play a mechanistic role in regulating SNARE assembly and fusion or does it exclusively recruit Cpx to its site of action?

Figure 1. Impact of CPX-1 and its CTD on synaptic transmission.

A. Schematic of CPX-1 fused to mGFP with a short flexible linker. The N-terminal domain (NTD), accessory helix (AH), central helix (CH) and C-terminal domain (CTD) are indicated along with their known binding targets. B. Cartoon of trans-SNAREs bound to full-length CPX-1::mGFP (left) or a truncated variant lacking its membrane-binding C-terminal domain (right). C. Example recordings of miniature EPSCs in 0 external Ca²⁺ (spontaneous) for wild type (gray), $cpx\text{-}1$(ok1552) null mutant (blue), $cpx\text{-}1$ null with a single-copy integrated transgene of full-length CPX-1::mGFP (black, $\text{CPX(SC)}$), and $cpx\text{-}1$ null with a multi-copy integrated transgene of truncated CPX-1 lacking its C-terminal domain (red, CPX∆). Average spontaneous EPSC rate (D) and amplitude (E) for each of the four genotypes as well as the CPX∆/– hemizygote (green) (wild type N=7, cpx-1 N=12, $\text{CPX(SC)}$ N=9, CPX∆ N=10, CPX∆/– N=7). F. Example recordings of tonic EPSCs in 1 mM external Ca²⁺ for the same four genotypes. Average tonic EPSC rate (G) and amplitude (H) for each of the four genotypes (wild type N=14, cpx-1 N=19, $CPX(SC)$ N=11, CPX∆ N=10). I. Representative stimulus-evoked EPSCs in 1 mM external Ca²⁺. J. Average EPSC peak for the five genotypes (wild type N=19, cpx-1 N=17, $\text{CPX(SC)}$ N=7, CPX∆ N=10, CPX∆/– N=8). Representative sucrose-evoked response (K) and average integrated synaptic charge transfer (L) for wild-type (gray, N=7) and $cpx\text{-}1$ (blue, N=7). 1 M sucrose was applied for 2 sec. Data are mean ± SEM. For statistical comparisons, ** and * indicate significant difference from wild type with p < 0.01 and p < 0.05, respectively. # indicates a significant difference from both wild type and $cpx\text{-}1(-)$ with p < 0.01. Not significant, n.s. Data were compared using ANOVA, and the Tukey-Kramer test for multiple comparisons was used to generate p values. Strains: N2, JSD347, JSD1258, and JSD1260.

In this study, we examined the function of the C. elegans ortholog CPX-1 in vivo at the neuromuscular junction (NMJ) while focusing on the role of its CTD to address these questions. We found that loss of the CTD markedly impaired CPX-1 inhibitory function while partially affecting its facilitatory function. Furthermore, over-expression of CTD-truncated variants of CPX-1 could restore some inhibitory function, consistent with the idea that CTD perturbations lower the effective concentration of CPX-1. A truncated variant of CPX-1 lacking its CTD could be enriched on membranes using vesicle-associated proteins as artificial tethers, and CTD function could be entirely replaced by substitution with the SV-associated small GTPase RAB-3, indicating that the Cpx CTD does not play an essential mechanistic role in vivo. These results suggest that the CTD specifically acts to steer CPX-1 to the assembling SNARE complex to regulate SV fusion. We propose that assembling trans-SNARE complexes are not freely accessible by cytoplasmic Cpx and that distinctive SV membrane interactions with the CTD permit Cpx to bind and ‘clamp’ the SNAREs, thereby preventing unregulated SNARE assembly and SV fusion.

Results

Inhibitory function of CPX-1 at the neuromuscular junction

We have previously reported that deletions in the CPX-1 CTD impact complexin inhibitory function (Martin et al., 2011; Wragg et al., 2013; Wragg et al., 2017). Several of the relevant transgenic strains for those past studies contained highly over-expressed CPX-1 variants and made it difficult to compare relative expression levels between the strains. Here, we revisited the functional significance of the complexin CTD using a series of mGFP-tagged CPX-1 variants expressed at or near single-copy expression levels (Fig 1A-B). Specifically, a truncated variant of CPX-1 lacking its CTD (residues 94 – 143) was fused to mGFP and used to generate a multi-copy integrated array. Additionally, two transgenic strains expressing full-length CPX-1::mGFP were generated: a single-copy integrant and a modestly over-expressed multi-copy integrant. All three of these transgenic strains were crossed with the $cpx\text{-}1$(ok1552) complexin null mutant to eliminate endogenous CPX-1 and isolate the functional consequences of the mGFP-tagged variants by electrophysiological and behavioral assays.

To monitor the direct impact of CPX-1 on synaptic vesicle (SV) fusion rates, we recorded cholinergic synaptic activity at the worm NMJ via whole-cell voltage-clamp recordings in body-wall muscle of dissected animals as described previously (Richmond et al., 1999; Wang et al., 2001; Madison et al., 2005; Hu et al., 2013). Consistent with past studies, loss of CPX-1 strikingly enhanced spontaneous fusion in 0 mM external Ca²⁺ by more than 10-fold (3.2 ± 0.9 Hz for wild type versus 33.8 ± 1.8 Hz for the $cpx\text{-}1$ null mutant), confirming that CPX-1 is required for inhibiting spontaneous release of acetylcholine (ACh) in the absence of Ca²⁺ influx (Fig 1C-E) (Huntwork & Littleton, 2007; Hobson et al., 2011; Martin et al., 2011). Expression of the full-length single-copy CPX-1::mGFP under a pan-neuronal promoter in $cpx\text{-}1(-)$, hereon referred to as $\text{CPX(SC)}$, fully restored complexin clamping function. By contrast, expressing the truncated variant lacking its CTD in $cpx\text{-}1(-)$, hereon referred to as CPX∆, provided only marginal rescue of clamping function (Fig 1C-E). We also examined the impact of lowering the expression of CPX∆ by recording from hemizygous transgenic animals in the $cpx\text{-}1(-)$ background (CPX∆/–). These animals were indistinguishable from the null mutant, suggesting that clamping function may be sensitive to the abundance of the CPX∆ protein. In the presence of 1 mM external Ca²⁺, the tonic EPSC rate was significantly elevated in the absence of CPX-1 (55 ± 3.8 Hz for wild type versus 86.3 ± 5.2 Hz for $cpx\text{-}1$) (Fig 1F-H). Like the 0 Ca²⁺ conditions, this clamping defect was fully rescued with $\text{CPX(SC)}$ but not with CPX∆.

Fusogenic function of CPX-1 at the neuromuscular junction

In stark contrast to the inhibitory function of CPX-1, stimulus-evoked EPSCs in 1 mM external Ca²⁺ were reduced by about 90% in the absence of CPX-1 as measured by the EPSC peak (1.95 ± 0.13 nA for wild type and 0.19 ± 0.02 nA for $cpx\text{-}1$) (Fig 1I-J). Note that the small amplitude of the stimulus-evoked EPSC in $cpx\text{-}1$ was not due to a massive depletion of the readily releasable pool of SVs since the SV pool released by hypertonic sucrose was unchanged in the $cpx\text{-}1$ null mutant Fig 1K-L. These changes in synaptic transmission illustrate the two opposing functions of CPX-1 at the worm NMJ: clamping spontaneous fusion and promoting Ca²⁺-triggered fusion (Martin et al., 2011). In contrast to the clamping function, the positive role of CPX-1 during stimulus-evoked secretion was largely restored with and without the CTD. However, CPX∆/– hemizygous animals were similar to the null mutant, consistent with the notion that the fusogenic function of CPX-1 was highly sensitive to CPX∆ abundance.

Assessing the behavioral impact of CPX-1 clamping in vivo

We and others have previously shown that $cpx\text{-}1$ null mutants are hypersensitive to the cholinersterase inhibitor, aldicarb, and paralyze within minutes of exposure due to high levels of spontaneous ACh release (Hobson et al., 2011; Martin et al., 2011; Wragg et al., 2013). Thus, the time course of paralysis on aldicarb provides an in vivo metric of CPX-1 clamping function (Fig 2A-B). Consistent with CPX-1 suppression of spontaneous and tonic SV fusion, sensitivity to aldicarb was rescued to wild-type levels using $\text{CPX(SC)}$ but not by CPX∆ (Fig 2C). Aldicarb-induced paralysis is driven by both spontaneous and activity-driven ACh secretion, and stimulus-evoked ACh release was largely restored in CPX∆. To provide a quantitative measure of CPX-1 clamping function across all transgenic strains in the $cpx\text{-}1$ null background, the time to 50% paralysis was compared to $\text{CPX(SC)}$;$cpx\text{-}1$ and $cpx\text{-}1$ as indicated in Fig 2C. Thus, if expressing a CPX variant in $cpx\text{-}1$ shifted the paralysis time course closer to $\text{CPX(SC)}$;$cpx\text{-}1$, this variant would be scored as providing some degree of rescue in proportion to how far the time course shifted. By this measure, CPX∆ rescued CPX-1 inhibitory function by approximately 10%, in accord with the impact of CPX∆ on tonic release (Fig 2D). Thus, the time course of aldicarb-induced paralysis provides a quantitative in vivo measure of CPX clamping function while potentially over-estimating the degree of impairment because of a contribution by activity-dependent ACh release.

Figure 2. In vivo assay of CPX-1 clamping function.

A. Cartoon of a worm neuromuscular junction indicating synaptic cleft acetylcholinesterase as the site of action for the cholinesterase inhibitor aldicarb. B. Time course of animal paralysis induced by 1 mM aldicarb illustrating that loss of inhibition by CPX-1 produces a left-shift in the paralysis time course (blue). C. Average paralysis time course on 1 mM aldicarb for wild type (gray), $cpx\text{-}1$(ok1552) null mutant (blue), $cpx\text{-}1$ null with a single-copy integrated transgene of full-length CPX-1::mGFP (black, $\text{CPX(SC)}$), and $cpx\text{-}1$ null with a multi-copy integrated transgene of truncated CPX-1 lacking its C-terminal domain (red, CPX∆). Rescue of CPX-1 clamping function was defined as the time to 50% paralysis relative to the null mutant and $\text{CPX(SC)}$;$cpx\text{-}1$ as illustrated. D. Mean time to 50% paralysis (top) and percent rescue (bottom) are shown for the four genotypes (wild type N=43, cpx-1 N=17, $CPX(SC)$ N=12, CPX∆ N=16). Data are mean ± SEM. For statistical comparisons, ** indicates significant difference from wild type with p < 0.01. Data were compared using ANOVA, and the Tukey-Kramer test for multiple comparisons was used to generate p values. Strains: N2, JSD347, JSD1258, and JSD1260.

Impact of the CPX-1 C-terminal domain and synaptic abundance in vivo

The failure of CPX∆ to restore clamping function by both electrophysiological and behavioral measures may have been due to poor expression or shortened half-life of the truncated protein rather than a direct result of losing the CTD and its membrane-binding properties. To address this issue, we quantified the axonal abundance of $\text{CPX(SC)}$ and CPX∆ homozygotes and hemizygotes in the dorsal nerve cords of living intact animals based on mGFP fluorescence using confocal microscopy. In animals expressing mGFP-tagged CPX variants in all neurons, dorsal cord fluorescence predominantly arises from the neurites of motor neurons synapsing onto dorsal body-wall muscles (Fig 3A). We have previously shown that CPX-1 is somewhat enriched at en passant presynaptic boutons in the dorsal nerve cord (Dittman & Kaplan, 2006; Martin et al., 2011; Wragg et al., 2013; Wragg et al., 2015). Notably, the relatively diffuse distribution of CPX and the superposition of fluorescence from multiple axons within the nerve bundle generated a complex fluorescence pattern (Fig 3B). To analyze CPX abundance, we utilized custom-written automated analysis routines in Igor Pro (Wavemetrics). Confocal image stacks were corrected for background fluorescence, and several image parameters including fluorescence peak intensity, width, and density were identified and quantified for each genotype (Fig 3C-D). We chose absolute peak fluorescence (normalized to a fluorescent slide standard for each imaging session) as a quantitative measure of axonal protein abundance and normalized all genotypes to $\text{CPX(SC)}$. Relative to $\text{CPX(SC)}$, CPX∆ was expressed at almost a 2-fold higher abundance (183 ± 6.9%) (Fig 3E) and therefore the functional failure of CPX∆ was not due to poor expression of the truncated protein. Alternatively, perhaps the 2-fold higher expression of CPX∆ caused some impairment in CPX-1 function at the synapse. To address this possibility, we generated hemizygous transgenic animals in the $cpx\text{-}1$ background as a means of decreasing CPX∆ abundance as well as $\text{CPX(SC)}$ levels. In addition, we generated a multi-copy integrant of full-length CPX-1 tagged with mGFP (hereon referred to as CPX(MC)) to examine the impact of CPX-1 over-expression. All of these CPX-1 variants were analyzed by plotting their rescue of aldicarb sensitivity as a function of abundance (both normalized to the single-copy integrant) to reveal the impact of expression levels on clamping as shown in Figure 3F. Notably, hemizygous $\text{CPX(SC)}$ animals displayed an intermediate clamping defect (65 ± 2.7% of wild-type clamping), suggesting that this abundance (58 ± 2% of the homozygous single-copy integrant) was insufficient to maintain normal CPX-1 inhibitory function. A 2-fold decrease in the abundance of CPX∆ further impaired clamping consistent with the electrophysiological recordings described in Figure 1, whereas a nearly 4-fold increase in full-length CPX-1 did not interfere with clamping function. These observations strengthen the argument that loss of the CTD weakened CPX-1 clamping function.

Figure 3. Quantifying axonal CPX-1 abundance in vivo.

A. Cartoon of a worm indicating a region of the dorsal nerve cord imaged using confocal microscopy. B. Example confocal maximum-intensity projection image of a dorsal nerve cord expressing CPX-1::mGFP together with a fluorescence line scan used to identify synaptic puncta (arrowheads, see Methods). C. Average representative images are shown for single-copy CPX-1::mGFP hemizygous and homozygous animals ($\text{CPX(SC)}$/– and $\text{CPX(SC)}$, respectively) as well as CPX-1::mGFP lacking its CTD (CPX∆) and an integrated multi-copy array of CPX-1::mGFP (CPX(MC)). Note that all genotypes are in the $cpx\text{-}1$(ok1552) null background. Scale bar is 5 μm. D. Summary of average peak fluorescence, peak relative to the axon (∆F/F), full-width half-max, and puncta density for $\text{CPX(SC)}$;$cpx\text{-}1$ hemizygotes (gray, N=20), homozygotes (black, N=61), CPX∆;$cpx\text{-}1$ (red, N=22), and CPX(MC);$cpx\text{-}1$ (blue, N=22). Each genotype was compared to $\text{CPX(SC)}$ (arrowhead). E. Average peak fluorescence normalized to $\text{CPX(SC)}$ is shown for the $\text{CPX(SC)}$ hemizygotes (gray, N=20), homozygotes (black, N=61), CPX∆ hemizygotes (pink, N=20), CPX∆ homozygotes (red, N=22), and CPX(MC) (blue, N=22), with all transgenes expressed in the $cpx\text{-}1$(ok1552) null background. F. Average percent rescue of CPX clamping based on aldicarb sensitivity is plotted versus the average axonal abundance for each of the genotypes as indicated. Solid lines are fits to a simple equilibrium binding model with a single fit parameter for the dissociation constant (${\mathrm{K}}_{\text{fit}}$) for the full-length (blue) and CTD-truncated (red) variants (see Methods). G. Cartoon of cytoplasmic Cpx (magenta) binding to the trans-SNARE complex (SNAREpin) with ${\mathrm{K}}_{\text{fit}}$ indicated. Data are mean ± SEM. For statistical comparisons, ** indicates significant difference from $\text{CPX(SC)}$ with p < 0.01. Data were compared using ANOVA, and the Tukey-Kramer test for multiple comparisons was used to generate p values. Strains: N2, JSD347, JSD1259, and JSD1260.

The synaptic availability of Cpx is likely a crucial factor in setting the efficacy of its clamping because this inhibitory function requires direct binding via the Cpx CH domain to assembling trans-SNARE complexes (hereon referred to as SNAREpins). The relationship between CPX inhibitory function and CPX protein abundance was fit by a highly simplified equilibrium binding model with a single free parameter (a normalized dissociation constant ${\mathrm{K}}_{\text{fit}}$ – see Methodsfor model details) to characterize the efficiency of CPX clamping based on its binding to SNAREpins (Fig 3G). Using this fit to quantify the difference in full-length versus truncated CPX-1, loss of the CTD decreased CPX-1 clamping efficiency by about 70-fold as a function of protein abundance. This large decrement in sensitivity to CPX-1 abundance implies that the expression of the truncated CPX variant would need to be increased by a factor of ~70 to fully restore wild-type levels of inhibition. Thus, CPX-mediated inhibition of secretion was highly sensitive to axonal CPX-1 protein abundance and required its C-terminal membrane-binding domain for efficient clamping.

Replacing the CPX-1 CTD with foreign vesicle anchors

The poor clamping performance of CPX∆ suggests that the membrane-binding domain serves some key role in clamping SNAREpins. Indeed, previous studies have suggested that the CTD could be involved in direct SNARE interactions (Bowen et al., 2005; Makke et al., 2018). Alternatively, the CTD could serve a localizing or positioning function by directing the SNARE-binding CH of CPX-1 to a specific SV location where assembling trans-SNARE complexes could be efficiently and rapidly engaged (Wragg et al., 2013). We wondered if it was sufficient to target CPX∆ to SVs by fusing the CPX∆ peptide to a known SV protein thereby achieving a high membrane-proximal abundance. To explore this idea, we screened through a series of SV proteins to serve as artificial anchors and monitored CPX∆ inhibitory function versus axonal protein abundance as described above. CPX∆::mGFP was fused to the N-terminus of a variety of known SV-associated proteins such as Synaptogyrin (SNG-1), RAB-3, RAB-27, and RAB-35 and expressed as multi-copy arrays in all neurons in the $cpx\text{-}1(-)$ background (Schlüter et al., 2002; Mahoney et al., 2006; Yu et al., 2008; Pavlos et al., 2010; Sheehan et al., 2016) (Fig 4A). Paralysis kinetics on aldicarb revealed a range of rescue across the artificial SV tethers (Fig 4B). This variable restoration of clamping function may have arisen from poor expression or trafficking of these chimeric constructs, so all variants were imaged to assess axonal abundance (Fig 4D-E). In all cases, the multi-copy arrays expressed at significantly higher levels than $\text{CPX(SC)}$, ranging from 5-fold to 13-fold over-expression. Thus, poor clamping function was not due to low abundance at the synapse. Plotting the restoration of CPX inhibition versus axonal abundance revealed a broad range of clamping efficiencies for the artificial SV anchors. (Fig 4F). For each transgenic strain, the rescue versus abundance data were fit to the same binding model used in Figure 3F and the clamping efficiency was quantified using 1/$1∕{\mathrm{K}}_{\text{fit}}$ normalized to the ${\mathrm{K}}_{\text{fit}}$ for full-length CPX (Fig 4G). By this measure, clamping efficiency spanned almost two orders of magnitude across the artificial Cpx anchors.

Figure 4. Replacing the CPX-1 CTD with artificial vesicle anchors.

A. (left) Schematic of a chimeric truncated CPX-1 variant using a synaptic vesicle-associated protein (blue) fused to mGFP (green) and the N-terminal half of CPX-1 (CPX∆, magenta) with small flexible linkers in between. (right) A cartoon of several anchor types tethering CPX∆ to a synaptic vesicle (SV). B. Average paralysis time course on 1 mM aldicarb for the full-length single copy integrant $\text{CPX(SC)}$ (black), $cpx\text{-}1$(ok1552) null mutant (blue), truncated CPX-1 lacking its C-terminal domain (red, CPX∆), CPX∆ fused to RAB-35 (cyan), and CPX∆ fused to Synaptogyrin 1 (SNG-1, brown). Note that all transgenes are expressed in the $cpx\text{-}1$(ok1552) null background. C. Average percent rescue of CPX-1 clamping function is shown for the full-length single-copy (black, N=12) and multi-copy (gray, N=11) integrants, CPX∆ (red, N=17), CPX∆ fused to RAB-35 (cyan, N=10), CPX∆ fused to RAB-27 (purple, N=10), CPX∆ fused to SNG-1 (brown, N=14), and CPX∆ fused to RAB-3 as a hemizygous array (light green, N=10), and CPX∆ fused to RAB-3 as a homozygous array (dark green, N=16). D. Average Representative images are shown for CPX∆ fused to RAB-35, RAB-27, SNG-1, and RAB-3 as well as full-length CPX-1 fused to RAB-3 as indicated. Scale bar is 5 μm. E. Summary of average peak fluorescence normalized to $\text{CPX(SC)}$ is shown for $\text{CPX(SC)}$ (black, N=61), CPX∆ fused to RAB-35 (cyan, N=51), RAB-27(purple, N=47), SNG-1 (brown, N=33), RAB-3 expressed as a hemizygous array (light green, N=23) and RAB-3 expressed as a homozygous array (dark green, N=83) as well as full-length CPX-1 fused to RAB-3 (orange, N=24). All genotypes are in the $cpx\text{-}1$ null background. F. Average percent rescue of CPX clamping based on aldicarb sensitivity is plotted versus the average axonal abundance for each of the genotypes as indicated. Solid lines are fits to an equilibrium clamping model (see Methods). G. Clamping efficiency normalized to CPX for each of the anchors and CPX∆ plotted on a log scale. Data are mean ± SEM. For statistical comparisons, ** and * indicate significant difference from $\text{CPX(SC)}$ with p < 0.01 and p < 0.05, respectively. # indicates a significant difference from both $\text{CPX(SC)}$ and $cpx\text{-}1$ with p < 0.01. Data were compared using ANOVA, and the Tukey-Kramer test for multiple comparisons was used to generate p values. Strains: N2, JSD347, JSD1105, JSD1258, JSD1259, JSD1260, JSD1281, JSD1344, and JSD1345.

Among the three RAB anchors, RAB-3 provided the highest clamping efficiency with nearly full restoration of CPX inhibition albeit at a ~3-fold over-expression level when assaying heterozygous transgenic animals (Fig 4F). Previous studies suggest that all three Rab proteins are associated with SVs, so the superior performance of RAB-3 may be due in part to its effector interactions with proteins comprising the pre-fusion complex such as UNC-13/Munc13-1 and UNC-10/RIM1 (Betz et al., 2001; Schluter et al., 2002). Alternatively, CPX-1 may normally clamp a specific subset of SVs at the synapse, and RAB-3 could preferentially target these SVs (Guzikowski & Kavalali, 2021). If CPX-1 normally reaches SNAREpins while bound to the vesicle membrane, we would anticipate that SNG-1-anchored CPX∆ would inhibit secretion with high efficiency. The SNG-1 anchor partially restored inhibitory function despite permanently tethering CPX∆ to the SV membrane via its four transmembrane domains. Thus, permanently enriching CPX∆ on vesicle membranes failed to reconstitute the high clamping efficiency observed with full-length CPX-1.

RAB-3 can replace the CPX-1 CTD to restore clamping and fusogenic functions at the synapse

Based on the high efficiency of RAB-3-anchored CPX∆, we further explored the functional consequences of this variant as well as full-length CPX-1 fused to RAB-3 using a broad range of expression levels (Fig 5A). In principle, the decrease in ACh secretion indicated by the slower paralysis kinetics may have been due to over-expression of RAB-3 rather than a true rescue of CPX-1 clamping function. If the decrease in secretion was specifically driven by CPX-1 binding to SNAREpins, then disruption of the SNARE interaction while maintaining high expression of RAB-3 would eliminate the apparent rescue. To test this scenario, we over-expressed a CPX∆::RAB-3 variant harboring a double point mutation in the SNARE-binding central helix domain of CPX-1 (K71A Y72A – hereon referred to as CPX∆^KY/AA::RAB-3) (Fig 5B). In $cpx\text{-}1$ null mutant animals, CPX∆^KY/AA::RAB-3 was over-expressed by about 14-fold (14.3 ± 1.8) relative to $\text{CPX(SC)}$. Despite this high over-expression, almost no rescue of CPX inhibitory function was detected based on aldicarb sensitivity (5.23 ± 1.9% rescue) (Fig 5C). Hence, the inhibition of ACh secretion observed in CPX∆::RAB-3 and full-length CPX::RAB-3 (FL CPX::RAB-3) resulted from CPX-SNARE interactions. We next examined the impact of CPX∆::RAB-3 on suppression of spontaneous SV fusion as well as the fusogenic effects of these SV-anchored complexins at the NMJ. CPX∆ fully restored clamping of spontaneous release when attached to RAB-3 in the $cpx\text{-}1(-)$ background (Fig 5D-F). Moreover, CPX∆::RAB-3 also fully restored the fusogenic function of CPX-1 as measured by peak EPSC and by total synaptic charge transfer (Fig 5G-I). These data confirm that the membrane-binding CTD does not play an essential role in either the clamping or facilitatory function of CPX-1. Furthermore, the relatively high efficiency of RAB-3 in positioning CPX∆ to clamp assembling SNAREpins supports the hypothesis that CPX-1 normally engages the SNAREs from the SV membrane and not from the cytoplasm. Loss of the CTD weakens this efficient access rather than disrupting some fundamental mechanistic aspect of CPX-1 function.

Figure 5. Anchoring CPX∆ with RAB-3 fully restores all CPX-1 functions.

A. Summary of average peak fluorescence normalized to $\text{CPX(SC)}$ is shown for $\text{CPX(SC)}$ (black, N=61), CPX∆ (red, N=22), CPX∆ fused to RAB-3 as a single-copy integrant (RAB-3(SC), dark green, N=21), CPX∆-RAB-3 multi-copy integrants: hemizygous (medium green, N=23) and homozygous (light green, N=83), and full-length CPX-1 fused to RAB-3 (orange, N=24), and CPX∆ containing the central helix mutation K71A/Y72A (KY/AA) fused to RAB-3 (maroon, N=16). All genotypes are in the $cpx\text{-}1$ null background. B. CPX central helix alignment of worm CPX-1 and mouse mCpx1 highlighting the conserved KY motif that aids in high-affinity binding to the SNARE complex (yellow). C. Average percent rescue of CPX clamping based on aldicarb sensitivity is plotted versus the average axonal abundance for full-length CPX-1 variants (black), RAB-3 anchored CPX∆ variants (green), and CPX∆ with KY/AA fused to RAB-3 (maroon). Solid lines are fits to an equilibrium clamping model (see Methods). D. Representative spontaneous EPSC recordings in 0 Ca²⁺ are shown for wild type (black), $cpx\text{-}1$(ok1552) (blue), CPX∆ in $cpx\text{-}1(-)$ (red), and CPX∆ fused to RAB-3 in $cpx\text{-}1(-)$ (green). Average spontaneous EPSC rate (E) and amplitude (F) for each of the four genotypes (wild type N=7, $cpx\text{-}1(-)$ N=12, CPX∆ N=10, CPX∆::RAB-3 N=10). G. Representative stimulus-evoked EPSCs in 1 mM external Ca²⁺. Average evoked EPSC peak (H) and integrated synaptic charge (I) for each of the four genotypes (wild type N=19, $cpx\text{-}1(-)$ N=17, CPX∆ N=10, CPX∆-RAB-3 N=10). Data are mean ± SEM. For statistical comparisons in panel A, ** indicates significant difference from $\text{CPX(SC)}$ with p < 0.01. For statistical comparisons in panels D – I, ** and * indicate significant difference from wild type with p < 0.01 and p < 0.05, respectively. # indicates a significant difference from $cpx\text{-}1(-)$ but not wild type with p < 0.01. Not significant, n.s. Data were compared using ANOVA, and the Tukey-Kramer test for multiple comparisons was used to generate p values. Strains: N2, JSD347, JSD374, JSD390, JSD1258, JSD1260, JSD1281, and JSD1292.

A SNARE access provider model of the Cpx CTD

We have previously shown that sequestering full-length CPX-1 to the SV by fusing it to RAB-3 did not impair its inhibitory function (Wragg et al., 2013). By contrast, localizing CPX-1 to the plasma membrane disrupted clamping. In addition, the membrane-binding CTD of complexin is highly sensitive to membrane curvature and preferentially binds to small vesicles with a diameter of 100 nm or smaller (Snead et al., 2014; Gong et al., 2016; Zdanowicz et al., 2017). In vitro studies previously demonstrated complexin lacking its CTD can still function to inhibit SNARE-mediated fusion, but only when supplying high concentrations of the truncated CPX (Chicka & Chapman, 2009; Bera et al., 2022). Based on these observations and the large shift in clamping efficiency in the absence of the membrane-binding CTD reported here, we hypothesize that CPX-1 cannot efficiently access and clamp the assembling SNAREpins from the cytoplasm unless expressed at a very high abundance. We propose that CPX-1 localizes to the SV membrane (perhaps at a specific location) and gains direct unobstructed access to assembling SNAREpins from this location (Fig 6 – mode 1 route). In this model, the assembling SNAREpins are shielded from cytoplasmic access by chaperone proteins such as Munc13 and Munc18 as has previously been observed in the context of protection from SNARE disassembly by N-ethylmaleimide sensitive factor (NSF) (Ma et al., 2013; He et al., 2017; Rizo, 2018). If CPX-1 abundance is elevated to sufficiently high levels, CPX-1 can engage the assembling SNAREpins with a much lower effective affinity due to shielding by other SNARE-binding proteins (Fig 6 – mode 2 route). This model can account for a broad range of observations regarding the Cpx CTD and its functional impact both in vitro and at the synapse.

Figure 6. Dual-clamp equilibrium binding model of CPX-1.

Cpx (magenta) binds to the assembling SNAREpins either from the synaptic vesicle (mode 1) or directly from the cytoplasm (mode 2). Mode 2 is intrinsically less efficient than mode 1 due to obstruction by chaperones bound to the assembling SNAREpins (privileged compartment – blue). Mode 1 is highly efficient but requires that Cpx bind to the synaptic vesicle membrane prior to accessing the SNAREpins.

Discussion

We previously identified the CTD of CPX-1 as playing a critical role in the regulation of synaptic transmission (Martin et al., 2011; Wragg et al., 2013; Snead et al., 2014; Wragg et al., 2015; Wragg et al., 2017). Here, we uncovered several aspects of complexin CTD function in the context of synaptic transmission in a living nervous system. First, loss of the CTD preferentially disrupts CPX-1 clamping function at higher abundance while both the clamping and fusogenic functions of CPX-1 are impaired at lower abundance. Second, increasing the synaptic abundance of mutated CTD variants could overcome functional impairment of CPX-1, suggesting that CTD mutations impair Cpx availability. Third, artificially increasing the vesicle membrane-proximal abundance of CPX∆ did not generally restore CPX-1 function. Moreover, most of the SV anchors failed to boost the efficiency of CPX-1 function. And fourth, RAB-3-anchored CPX∆ fully restored both the clamping of spontaneous fusion and the facilitation of stimulus-evoked fusion with moderately high efficiency. Together, these observations suggest that the CTD is not essential for any function of CPX-1 beyond localizing CPX-1 to the SV membrane. Moreover, this SV localization is specific (either to a subtype of SV or to a particular location on the vesicle) since several SV anchors failed to position CPX-1 efficiently despite providing high ambient levels of membrane-proximal CPX-1. We hypothesize that the curvature sensor of the CPX-1 CTD directs the CH domain to assembling SNAREs at specific locations on docked SVs rather than CPX-1 diffused onto SNAREs directly from the cytoplasm.

Molecular sequence of events during CPX-mediated inhibition

Numerous investigators have documented that Cpx can interact with both membranes and ternary SNARE complexes via its C-terminus and CH domains, respectively (Pabst et al., 2000; Chen et al., 2002; Giraudo et al., 2006; Chicka & Chapman, 2009; Seiler et al., 2009; Snead et al., 2014; Gong et al., 2016; Snead et al., 2017; Zdanowicz et al., 2017; Bera et al., 2022; Lottermoser & Dittman, 2023). However, it is not clear whether membrane binding precedes or follows SNARE binding during the process of clamping SNAREpins. Several lines of evidence lead us to propose that Cpx first binds to the SV membrane and subsequently interacts with SNAREpins as they assemble. First, in vitro reconstituted fusion assays demonstrate that Cpx-mediated inhibition of fusion require an intact CH domain but Cpx lacking its CTD can still inhibit fusion if used at sufficiently high concentrations (Giraudo et al., 2006; Chicka & Chapman, 2009; Bera et al., 2022). Thus, CH-SNARE interactions are strictly required whereas CTD-membrane interactions can be bypassed in vitro. Second, we have previously reported that Cpx is localized and sequestered within synaptic boutons via its CTD rather than its SNARE-binding CH in vivo, indicating that the primary mechanism for confining Cpx at the synapse is through its membrane interactions (Wragg et al., 2013; Wragg et al., 2015). And third, current models of SV priming and fusion assert that only a small number of SNAREpins are formed and these SNAREpins may be in dynamic states of assembly/disassembly (Shi et al., 2012; Bao et al., 2018; Rothman et al., 2023). Compared to the effective number of binding sites on the membrane of an SV, there are likely fewer available binding sites for assembling SNAREpins. Thus, on average, Cpx would be more likely to occupy an available membrane binding site than a rare SNAREpin binding site. Overall, these observations suggest that a pool of membrane-bound Cpx could supply the inhibition of assembling SNAREpins. It remains to be determined whether the SNARE-bound Cpx could remain simultaneously bound to the SV membrane via its CTD.

Targeting Cpx to SNAREs using SV proteins as artificial anchors

We employed several SV proteins as artificial anchors for CPX∆ to better understand the role of the CTD in synaptic transmission. These anchors serve distinct functions on SVs, and some are more dynamically associated with SVs than others. For instance, tetraspanins such as Synaptophysin and Synaptogyrin represent a major class of integral membrane protein enriched on SVs with an estimated ~ 35 copies per vesicle (Takamori et al., 2006). Some of these tetraspanins are proposed to play a role in determining the high curvature of SVs (Park et al., 2024). And like Cpx, there is evidence that some tetraspanins clamp SV fusion in mammals (Raja et al., 2019) and promote Ca²⁺-regulated exocytosis (Sugita et al., 1999). Moreover, SNG-1 was shown to modulate nervous system function in worms (Abraham et al., 2011). Despite these superficial similarities, CPX∆ anchored to SNG-1 poorly restored CPX-1 function relative to other anchors. Perhaps the permanent tethering of CPX∆ to the SV membrane restricted SNARE binding due to steric constraints. Alternatively, SNG-1 may be organized on the SV in a location that is distal to the assembling SNAREpins.

We also utilized SV-associated Rab proteins as anchors since these are membrane-associated but not permanently attached to the SV membrane. RAB-3 and RAB-27 were chosen because both have been associated with SVs in worm and mouse and known to regulate SV exocytosis (Schluter et al., 2004; Mahoney et al., 2006; Takamori et al., 2006; Yu et al., 2008; Pavlos et al., 2010). To broaden the diversity of the presynaptic Rab anchors tested here, we also examined RAB-35, a presynaptic Rab that is proposed to play a role in SV protein degradation (Uytterhoeven et al., 2011; Sheehan et al., 2016). If driving a high abundance of membrane-proximal CPX-1 was the only function of the CPX-1 CTD, we would anticipate that all three Rab anchors could restore clamping function. However, only RAB-3 was able to restore full Cpx function at minimal overexpression. The selective positioning of CPX∆ by RAB-3 suggests that the CTD normally provides specificity beyond enrichment at the SV membrane. Because Rab3 binds to AZ proteins such as RIM1 and Munc13, this Rab may selectively position CPX∆ nearby assembling SNAREpins via its protein interactions. Alternatively, we cannot exclude the possibility that CPX-1 clamping normally occurs on a specific subset of SVs that are selectively targeted by RAB-3 (Schluter et al., 2006; Guzikowski & Kavalali, 2021). We propose that the CPX-1 CTD positions the SNARE-binding CH domain on a specific location of a docked SV where CPX-1 gains access to assembling SNAREpins that are not accessible from the cytoplasm. In addition, we speculate that RAB-3 localizes nearby and provides some degree of efficient access to the assembling SNAREpins. Despite the relatively high clamping efficiency achieved by the RAB-3 anchor, this strategy was markedly less efficient that the endogenous CTD of CPX-1. Protein interactions mediated by RAB-3 may be a poor substitute for the specific membrane interactions of the CPX-1 CTD; but with sufficiently high abundance and proximity to the assembling SNAREs, the CPX∆::RAB-3 variant could successfully engage and clamp these SNAREs. The C-terminal geranylgeranyl modification of RAB-3 may also contribute to the localization of RAB-3. Moreover, some Cpx variants contain a CAAX-box motif at the end of the CTD, and farnesylation is critical for proper function of this Cpx subfamily (Reim et al., 2005b; Buhl et al., 2013). For instance, of the four Cpx paralogs in mammals, only mCpx3/4 are CAAX proteins (Reim et al., 2005a), while in fly, CAAX and non-CAAX variants of Cpx are produced by alternative splicing of a single gene transcript (Buhl et al., 2013). CPX-1 is a non-CAAX variant and thus relies entirely on its curvature sensitive amphipathic CTD helix for membrane interactions. The potential for lipid modifications to steer proteins to specific locations on a vesicle remains an open question.

Access to a privileged compartment and release site quality control

Cytoplasmic NSF operates together with soluble NSF attachment protein (α-SNAP) to disassemble the ternary SNARE complex (Choi et al., 2018; White et al., 2018; Brunger et al., 2019). Several studies concluded that assembling trans-SNARE complexes are inaccessible to NSF/ α-SNAP while bound to Munc13 and Munc18 (Ma et al., 2013; He et al., 2017; Prinslow et al., 2019; Rizo, 2022; Xu et al., 2022). Inappropriate SNARE assembly outside of this protected compartment is antagonized by constitutive NSF disassembly activity and consequently has been described as a quality-control mechanism for the SNARE assembly pathway in vivo (Leitz et al., 2024). We hypothesized that, by analogy with these other soluble SNARE-binding proteins, cytoplasmic CPX-1 cannot readily access partially assembled trans-SNARE complexes that are bound to UNC-13 and UNC-18 under a docked vesicle. Furthermore, we speculated that these SNARE assemblies are accessible to CPX-1 bound to the SV membrane via its CTD. Cpx has been proposed to play a role in quality control of release sites (López-Murcia et al., 2024), so we conjecture that maturation of the SNARE assembly machinery under a docked vesicle utilizes Cpx as a check point within the privileged SNARE assembly compartment while NSF inhibits promiscuous SNARE assembly outside this compartment. We also speculate that the fusogenic function of Cpx is not restricted to the privileged compartment as stimulus-triggered fusion is less sensitive to perturbations in CTD function. Future studies may shed light on the mechanism by which the membrane curvature sensor in the CPX-1 CTD targets assembling SNAREs.

Materials and Methods

Strains and constructs

Strains were maintained and genetically manipulated as previously described (Brenner, 1974). Animals were raised at 20°C on nematode growth media seeded with OP50. A full list of the strains used in this study is provided in Table 1. Plasmids used for generating transgenic strains are provided in Table 2. The truncated CPX-1 construct CPX∆ protein sequence lacks residues 94-143 of the full-length protein.

Table 1. Strains.

Strain Name	Strain Description	Transgene Description	MC or SC?
JSD0347	tauIs102; $cpx\text{-}1$	Psnb-1::CPX∆-Linker::mGFP on chr V	MC
JSD0374	tauIs114; $cpx\text{-}1$	Psnb-1::CPX-Linker::mGFP-Linker::RAB-3 on IV	MC
JSD0390	tauIs124; $cpx\text{-}1$	Psnb-1::CPX∆-Linker::mGFP-Linker::RAB-3 on IV	MC
JSD1105	tauIs62; $cpx\text{-}1$	Psnb-1::CPX∆-Linker::mGFP-Linker::SNG-1 on V	MC
JSD1255	N2	Wild type	n/a
JSD1258	$cpx\text{-}1$ (ok1552)	5x outcrossed with N2	n/a
JSD1259	tauIs90; $cpx\text{-}1$	Psnb-1::CPX-Linker::mGFP on chr V	MC
JSD1260	tauSi1; $cpx\text{-}1$	Psnb-1::CPX-L::mGFP on chr II	SC
JSD1281	tauSi25; $cpx\text{-}1$	Psnb-1::CPX∆CTD-Linker::mGFP-Linker::rab-3 on IV	SC
JSD1292	tauEx508; $cpx\text{-}1$	Psnb-1::$cpx\text{-}1$(KY/AA,∆CTD)-Linker::mGFP-Linker::rab-3	MC
JSD1344	tauEx546;$\text{cpx-}1$	Psnb-1:: CPX∆-Linker::mGFP::rab-35	MC
JSD1345	tauEX547;$\text{cpx-}1$	Psnb-1:: CPX∆-Linker::mGFP::rab-27	MC

Table 2. Plasmids.

Plasmid(s)	Plasmid Descriptions	injected conc.(ng/uL)	Strains generated
JP268	Psnb-1::$cpx\text{-}1$::Linker-mGFP in pCFJ151	-	tauSi1
JP1231	Psnb-1::$cpx\text{-}1$(deltaCTD)::linker-mGFP-linker::rab-3	-	tauSi25
JP302	Psnb-1::$cpx\text{-}1$(deltaCTD)::Linker-mGFP::Linker-sng-1	-	tauIs62
JP171	Psnb-1::$cpx\text{-}1$::Linker-mGFP	-	tauIs62 tauIs90
JP104	Psnb-1::$cpx\text{-}1$ (deltaCTD)::Linker-mGFP	-
JP236	Psnb-1::$cpx\text{-}1$::Linker-mGFP::Linker-rab-3	-	tauIs102
JP233	Psnb-1::$cpx\text{-}1$ (deltaCTD)::Linker-mGFP::Linker-rab-3	-	tauIs114
KP1484	Pmyo2::NLS::maxFPGreen::ß-galactosidase	2	tauIs124
JP1244	Psnb-1::$cpx\text{-}1$(KY/AA+deltaCTD)::Linker-mGFP::Linker-rab-3 in $cpx\text{-}1(-)$	6	used in all extrachromosomal arrays
JP1257	Psnb-1::CPX(fly CH+deltaCTD)::Linker-mGFP::Linker-rab-3 in $cpx\text{-}1(-)$	6	tauEx508;$cpx\text{-}1$
JP1261	Psnb-1::$cpx\text{-}1$ (deltaCTD)::Linker-mGFP::Linker-rab-35 in $cpx\text{-}1(-)$	6	tauEx523;$cpx\text{-}1$ and tauEx524;$cpx\text{-}1$
JP1297	Psnb-1::cpx-1(deltaCTD)::GFP::rab-27 in $cpx\text{-}1(-)$	6	tauEx546;$cpx\text{-}1$
JP81 + JP1277	Punc-129::mCherry::rab-3 (no linkers) and Punc-129::mGFP::Linker-rab-27	10 + 10	tauEx547;$cpx\text{-}1$
JP81 + JP1311	Punc-129::mCherry::rab-3 (no linkers) and Punc-129::mGFP::Linker-rab-35	10 + 6	tauEx558
JP1351	Psnb-1::CPX1(delta6+mouse7) ::L-mGFP::L-rab-3 (worm) in $cpx\text{-}1(-)$	10	tauEx559
JP1351	Psnb-1::CPX-1(delta6+mouse7) ::L-mGFP::L-rab-3 (worm) in $cpx\text{-}1(-)$	10	tauEx566; $cpx\text{-}1$ (iJL44.3)
JP584	Psnb-1::CTD (LV/EE delta12) untagged in tauIs124;$cpx\text{-}1$	100	tauEx567;$cpx\text{-}1$ (iJL44.18)
JP1383	Psnb-1::$cpx\text{-}1$ (deltaCTD)::Linker-mGFP::Linker-unc-57 H0 domain (1ng/uL) in $cpx\text{-}1(-)$	2	tauEx568;tauIs124;$cpx\text{-}1$ (iJL46.4)
JP1383	Psnb-1::$cpx\text{-}1$ (deltaCTD)::Linker-mGFP::Linker-unc-57 H0 domain (10ng/uL) in $cpx\text{-}1(-)$	10	tauEx583;$cpx\text{-}1$
JP835	Psnb-1::CPX1(delta6+mouse7)::Linker-mGFP in $cpx\text{-}1(-)$	1	tauEx584;$cpx\text{-}1$
JP448	Psnb-1::mouse CplxI::Linker-mGFP in $cpx\text{-}1(-)$	3	tauEx355;$cpx\text{-}1$

All transgenes were outcrossed at least 4x before generating double mutants. Single-copy transgenes were outcrossed a minimum of 5 times and multi-copy integrated transgenes 8 times. Single-copy integrants were generated using MosSCI (Frokjaer-Jensen et al., 2008). Multi-copy integrants were generated by UV integration using extrachromosomal arrays (Stinchcomb et al., 1985). Flexible linkers comprise a triple repeat of GGS.

Electrophysiology

Electrophysiology was performed on dissected C. elegans as described previously (Madison et al., 2005; McEwen et al., 2006). Worms were superfused in an extracellular solution containing 127 mM NaCl, 5 mM KCl, 26 mM NaHCO₃, 1.25 mM NaH₂PO₄, 20mM glucose, 1 mM CaCl₂ and 4 mM MgCl₂, bubbled with 5% CO₂, 95% O₂ at 20°C. Whole cell recordings were carried out at −60 mV using an internal solution containing 105 mM CH₃O₃SCs, 10 mM CsCl, 15 mM CsF, 4 mM MgCl₂, 5 mM EGTA, 0.25 mM CaCl₂, 10 mM HEPES and 4 mM Na₂ATP, adjusted to pH 7.2 using CsOH. Under these conditions we only observed acetylcholine EPSCs. For low Ca²⁺ experiments, 1 mM CaCl₂ was replaced with additional MgCl₂ for a total of 5 mM.

Aldicarb Sensitivity Assay

To measure aldicarb sensitivity, 20–25 young adult animals were placed on agar plates containing 1 mM aldicarb (CarboSynth) and scored for paralysis at 10 minute intervals for 2 hours. 100 mM Aldicarb stock solution was dissolved in 70% ethanol and stored at −20 °C. Each genotype was tested 10 times and paralysis curves were generated by averaging paralysis time courses for each plate as described previously (Dittman & Kaplan, 2008; Martin et al., 2011; Wragg et al., 2013; Wragg et al., 2017). Percent rescue based on ${\mathrm{t}}_{0.5}$ was calculated by first interpolating the time at which 50% of the worms paralyzed for each trial, averaging the single-trial ${\mathrm{t}}_{0.5}$ values together, and normalizing to $\text{CPX(SC)}$ (100%) ${\mathrm{t}}_{0.5}$ and $cpx\text{-}1$(ok1552) (0%) ${\mathrm{t}}_{0.5}$ values according to the equation below.

$$\%\text{Rescue [strain]}=100\cdot \frac{{\mathrm{t}}_{0.5}[\text{Strain}]-{\mathrm{t}}_{0.5}[cpx\text{-}1(\text{-})]}{{\mathrm{t}}_{0.5}[\text{CPX}(\text{SC})]}$$

Steady-State Confocal Imaging and Quantification

To measure protein expression levels, animals were immobilized using 2,3-butanedione monoxime (Alfa Aesar) (30 mg/mL) mounted on 2% agarose pads. The dorsal nerve cords of animals expressing GFP-tagged variants of CPX-1 were imaged by laser-scanning confocal microscopy (Olympus Fluoview FV1000) mounted on an Olympus IX81 inverted microscope with a PlanApo 60X 1.42 NA objective. Microscopy conditions were uniform across all images. Complexin constructs were C-terminally tagged with monomeric GFP separated by a 12-residue linker (GGSGGSGGSAAA). Example images in figures are maximal intensity projections of 12-bit images mapped to 8-bit image files with minimum and maximum pixel values set to 30 and 1200, respectively. All display images were generated using a gamma factor of 0.75. Synaptic protein levels were quantified in the raw 12-bit image stacks by measuring background-subtracted fluorescence within dorsal cord varicosities. A reference fluorescent slide was imaged to monitor laser stability over time and the dorsal cord axonal fluorescence was normalized to the slide value for all measurements. For all imaging data plotted, the normalized axonal fluorescence was normalized to the worm $\text{CPX(SC)}$ strain. Data were analyzed with custom software in IGOR Pro (WaveMetrics, Lake Oswego, OR, United States) (Burbea et al., 2002; Dittman & Kaplan, 2006).

Equilibrium binding model for CPX-1 clamping

A simple physical model of $\text{Cpx}$ binding to assembling SNAREs was generated by assuming that there is an excess of free $\text{Cpx}$ relative to the number of SNARE binding sites and assuming that $\text{Cpx}$ is in equilibrium with the SNARE sites. With these assumptions, the percentage of assembling SNAREs bound to normalized $\text{Cpx}$ ($\mathrm{C}$) at equilibrium is given by

$$\%bound=100\cdot \frac{\frac{[Cpx]}{[CPX(SC)]}}{\frac{[Cpx]}{[CPX(SC)]}+\frac{K_D^{\prime }}{[CPX(SC)]}}=100\cdot \frac{C}{C+K_D}$$

where ${\mathrm{K}}_{\mathrm{D}}$ is the normalized dissociation constant for the SNARE binding site. We used this percentage as a metric for clamping mediated by CPX-1 and fit the rescue versus abundance data in Figures 3-5 to this function with ${\mathrm{K}}_{\text{fit}}$ as the dissociation constant fit parameter. Thus, a high ${\mathrm{K}}_{\text{fit}}$ value would correspond to low affinity Cpx binding and inefficient clamping. The normalized efficiency in Figure 4G is defined as

$$E(genoX)=K_{fit}(CPX(SC))∕K_{fit}(genoX)$$

Statistics

Data sets were compared by Tukey-Kramer to allow for multiple comparisons after one-way ANOVA with 95% confidence and significance set to p < 0.01. For all aldicarb data, each independent trial contained at least 20 animals and at least 10 trials were included to generate the average time course curves. The investigator was blind to the genotype during the assay in all cases. For confocal imaging data sets, at least 20 independent line scans were collected and each line scan typically revealed at least 8 local enrichments of CPX-1 along at 42 μm segment of the dorsal cord. The same CPX∆ aldicarb data is used in Figures 2-4. Data sets for $cpx\text{-}1$ and $\text{CPX(SC)}$ are used in Figures 2-5. CPX(MC) data is used in Figures 3-5. Figures 4-5 contain the same aldicarb data points for CPX∆::RAB3(SC), hemizygous CPX∆::RAB3(MC), and CPX∆::RAB3(MC). The example spontaneous EPSC traces for wild type, $cpx\text{-}1$, and CPX∆ in Figure 1 are repeated in Figure 5 for comparison with CPX∆::RAB-3.

Key Points.

• Complexin (Cpx) regulates presynaptic SNARE assembly to control synaptic transmission.

• A membrane curvature-sensing motif within the Cpx C-terminal domain (CTD) recruits Cpx to vesicles.

• Replacement of the CTD with the synaptic vesicle protein Rab3 can restore full Cpx function whereas other vesicle proteins fail to substitute regardless of abundance.

• The efficiency of Cpx-mediated inhibition of synaptic vesicle fusion is profoundly enhanced by the specific localization supplied by its CTD.

• These results suggest that Cpx reaches the assembling SNARE complexes via its specific CTD-membrane interactions and these SNAREs are inaccessible from the cytoplasmic compartment.

Data Availability Statement

We will provide the raw data file for all data presented in this manuscript