|Home | About | Journals | Submit | Contact Us | Français|
Synapses continually replenish their synaptic vesicle (SV) pools while suppressing spontaneous fusion events, thus maintaining a high dynamic range in response to physiological stimuli. The presynaptic protein complexin can both promote and inhibit fusion through interactions between its alpha-helical domain and the SNARE complex. In addition, complexin’s C-terminal half is required for the inhibition of spontaneous fusion in worm, fly, and mouse, although the molecular mechanism remains unexplained. We show here that complexin’s C-terminal domain binds lipids through a novel protein motif, permitting complexin to inhibit spontaneous exocytosis in vivo by targeting complexin to SVs. We propose that the SV pool serves as a platform to sequester and position complexin where it can intercept the rapidly assembling SNAREs and control the rate of spontaneous fusion.
Regulated neurotransmitter release is essential to nervous system function. The presynaptic terminal continually replenishes a pool of synaptic vesicles (SVs) which are docked at the plasma membrane and primed to undergo fusion upon calcium entry. SVs also occasionally fuse with the plasma membrane in the absence of calcium influx, creating an unregulated basal level of secretion. This spontaneous fusion rate must be maintained at a relatively low level to prevent depletion of the SV pool and to maximize the dynamic range of the synapse by minimizing noise levels. The key proteins controlling SV fusion are the neuronal SNARE proteins syntaxin 1, SNAP-25 and VAMP2/synaptobrevin 2, as well as a core set of SNARE-binding proteins including synaptotagmin, Munc13, Munc18, tomosyn, and complexin (Ashery et al., 2009; Brose, 2008; Jahn and Fasshauer, 2012; Kasai et al., 2012; Sudhof and Rizo, 2011). Of these proteins, complexin plays a major role in preventing high rates of spontaneous fusion (Cho et al., 2010; Hobson et al., 2011; Huntwork and Littleton, 2007; Kaeser-Woo et al., 2012; Martin et al., 2011; Maximov et al., 2009), and loss of complexin reduces the SV pool by 50% in C. elegans (Hobson et al., 2011). Consistent with this inhibitory function, complexin also decreases SNARE-mediated proteoliposome fusion in vitro (Chicka and Chapman, 2009; Schaub et al., 2006). At several mammalian synapses, loss of complexin does not result in an increase in spontaneous fusion, indicating that its inhibitory function can be modulated or perhaps replaced by other synaptic proteins (Strenzke et al., 2009; Xue et al., 2008). Differences in complexin roles at distinct synapses highlight a need for better mechanistic understanding of its function.
Complexin is a cytoplasmic protein comprised of a highly charged α-helical domain subdivided into the accessory helix (residues 23–47), and the central helix (residues 48–70), flanked by N- and C-terminal unstructured regions (Bracher et al., 2002; Chen et al., 2002). The central helix binds the SNARE complex in the groove formed by synaptobrevin 2 and syntaxin 1 and is essential for all known complexin functions, whereas the C-terminal half (residues 71–134) was previously believed to be dispensable (Xue et al., 2007). In C. elegans (Martin et al., 2011), Drosophila (Cho et al., 2010; Xue et al., 2009), and mouse (Kaeser-Woo et al., 2012), complexin’s C-terminal domain (CTD) has been implicated in the inhibition of spontaneous SV fusion even though no direct mechanism has been proposed for its inhibitory function. Biochemical studies of the mammalian complexin 1 isoform found that the CTD interacts with phospholipids via a small amphipathic region near the C-terminus, suggesting a role for membrane interactions in complexin function (Seiler et al., 2009). Furthermore, some complexin isoforms associate with membranes via a prenylated CAAX motif at their C-termini, and mutating the CAAX box prevented complexin from inhibiting spontaneous fusion (Cho et al., 2010; Xue et al., 2009), indicating membrane binding may be required for this function.
Precise registration of SNARE and complexin residues is required for complexin’s inhibitory mechanism, involving conformations that may occur inefficiently in solution (Krishnakumar et al., 2011; Kummel et al., 2011). How is complexin positioned to engage the assembling SNARE complex? Is the SNARE-clamping function of complexin accelerated by its membrane interactions? Does complexin interact specifically with SV membranes or with the plasma membrane? Using a combination of biochemical, genetic, electrophysiological, and imaging approaches in C. elegans, we found that worm complexin, CPX-1, possesses a conserved amphipathic region near its C-terminus that is required for suppression of neurotransmitter secretion. This region interacts with phospholipid bilayers and localizes CPX-1 to SVs where it can rapidly cycle on and off the membrane. We identified similar amphipathic regions in all available complexin sequences across phylogeny regardless of whether a CAAX motif was present. Thus, we propose that membrane localization of complexin is a requirement for its ability to intercept the assembling SNARE complex and prevent unregulated SV fusion events.
In C. elegans, CPX-1 modulates synaptic vesicle exocytosis by simultaneously inhibiting spontaneous fusion while promoting fusion triggered by a depolarizing electrical stimulus (Hobson et al., 2011; Martin et al., 2011). These two modes of exocytosis can be experimentally distinguished, as CPX-1 lacking the last 50 amino acids of its CTD can partially rescue enhancement of stimulus-evoked exocytosis, but not inhibition of spontaneous SV fusion (Martin et al., 2011). The complexin CTD regions from worm, fly and mouse contain a relatively well-conserved amphipathic region near the C-terminus (Fig. 1A&B), whereas mammalian complexins 3 and 4, along with many other complexin orthologs, also contain C-terminal CAAX boxes (Fig. S1A&B). Since the amphipathic region appears to be conserved and to confer lipid binding in mammalian complexins (Seiler et al., 2009), we asked whether the homologous region of worm CPX-1 can interact with liposomes.
To examine potential interactions of the CTD with lipids, recombinant CPX-1::GFP fusion protein was incubated with liposomes and centrifuged in a sucrose gradient. Fractions were recovered from the top of the gradient and examined by SDS PAGE and immunoblotting. Wild-type (WT) CPX-1 co-migrated with liposomes in the topmost fractions (Fig. 1C), and the interaction did not depend on calcium (data not shown). Insertion of a helix-breaking proline (K123P) into the amphipathic region modestly decreased liposome binding, while deleting the C-terminal 50 residues (ΔCT) or exchanging L117 and V121 with glutamates (LV/EE) largely eliminated the liposome interaction (Fig. 1C&D). About 13% of CPX-1::GFP co-sedimented with liposomes containing anionic phospholipids, consistent with a relatively low affinity interaction (Fig. 1E). Protein concentrations were well below the expected saturation point of liposome binding sites (Experimental Procedures). Liposomes composed entirely of phosphatidylcholine (PC) no longer co-sedimented CPX-1 suggesting that liposome association depended in part on interactions with anionic phospholipids. Thus, despite a low overall sequence homology within the CTD, worm and mammalian complexins share a common lipid-binding amphipathic motif near the C-terminus.
To further investigate the interaction between CPX-1 and phospholipids, we acquired and compared proton-nitrogen single quantum coherence (HSQC) NMR spectra of full-length CPX-1 in the absence and presence of liposomes (Figure S2A). A subset of resonances exhibited decreased intensity in the presence of liposomes, suggesting a lipid interaction. Regions outside of complexin’s central helix are largely disordered, so residues not in direct contact with lipids should retain significant mobility (and therefore high signal intensity) in the presence of liposomes. In contrast, residues that interact directly with the phospholipid bilayer will be considerably immobilized, resulting in broad resonance lines and low signal intensity. Figure 2A shows HSQC intensity ratios for well-resolved residues within positions 87–143 of the full-length CPX-1. Complexin residues preceding position 92 were largely unaffected by the presence of liposomes. In contrast, the amphipathic region and neighboring amino acids exhibited a pronounced dose-dependent decrease in signal intensity (Fig. 2B). This drop in signal intensity was attenuated when the LV/EE mutation was introduced, consistent with the liposome flotation results above. Interestingly, the final 12 residues of the CTD demonstrated the strongest lipid interaction, suggesting that although the amphipathic region is required for flotation of CPX-1, lipid binding extends beyond the amphipathic region to the final amino acids. This region includes five hydrophobic side chains and two basic residues, and a similar pattern exists in the final 12 residues of other complexins. Deletion of the last 12 amino acids (Δ12) eliminated co-flotation (Fig. S1C&D) and co-sedimentation (Fig. 1E) with liposomes and almost completely eliminated lipid binding based on HSQC signal (Fig. 2B). The amphipathic region and last 12 residues of complexin appear to comprise a novel lipid-binding motif conserved throughout the complexin protein family.
Guided by the flotation, sedimentation, and NMR results, several CPX-1 CTD variants with altered lipid binding were introduced into cpx-1 mutant animals to determine whether lipid-binding correlates with complexin function in vivo. Nematodes paralyze when exposed to the cholinesterase inhibitor aldicarb, and increased secretion of acetylcholine accelerates the rate of paralysis (Dittman and Kaplan, 2008; Miller et al., 1996; Nurrish et al., 1999). Reflecting the loss of complexin’s clamping function, cpx-1 mutants rapidly paralyze when exposed to aldicarb (Hobson et al., 2011; Martin et al., 2011). Wild-type aldicarb sensitivity was restored in cpx-1 mutants expressing full-length CPX-1, while CPX-1 failed to rescue when missing its last 50 residues (ΔCT), last 12 residues (Δ12), or with the LV/EE substitution (Fig. 3A–C). Failure to rescue complexin’s clamping function was due neither to poor expression nor trafficking of CPX since all CPX variants were tagged with GFP, and only transgenic animals expressing comparable levels of synaptic complexin were used in these studies (Fig. S3A&B). Rescue with the K123P variant of complexin was intermediate, consistent with its mild defect in lipid binding revealed by liposome flotation (Fig. 3C).
To examine the effects of perturbing the CTD on the rate of spontaneous fusion, we recorded synaptic activity at the neuromuscular junction (NMJ) in dissected animals in the absence of external calcium. As reported previously, the spontaneous exocytosis rate was considerably elevated in the absence of CPX-1 (Hobson et al., 2011; Martin et al., 2011), and inhibition could not be restored in the absence of the CTD, whereas full-length CPX-1 could fully rescue (Fig. 3D–F). Consistent with the aldicarb sensitivity data, the LV/EE substitution and Δ12 truncation also completely eliminated complexin’s clamping function. Thus, complexin’s ability to inhibit SV fusion in vivo mirrors its binding to phospholipids in vitro.
As complexin’s lipid binding is critical for its function at the synapse, we next explored complexin’s synaptic localization at the NMJ. CPX-1::GFP is diffusely distributed throughout motor axons with enrichment at presynaptic terminals (Martin et al., 2011). The synaptic localization of CPX-1::GFP was compared with cytoplasmic mCherry as well as with two canonical synaptic markers: the synaptic vesicle associated RAB-3 and an active zone protein ELKS-1 (Fig 4A–C). Although complexin appeared to be diffusely distributed throughout the axon, CPX-1::GFP was about 40% more enriched at synaptic boutons than mCherry, indicating that its synaptic enrichment is not merely due to the increased volume of synaptic varicosities. By quantifying the spatial correlation of CPX-1::GFP with either mCherry::RAB-3, soluble mCherry, or ELKS-1::mCherry across a collection of images taken from motor axons of intact animals, CPX-1 appeared to be more robustly colocalized with SVs than with the active zone (Fig. 4D). Although the similarity between axonal CPX-1 distribution and SV distribution could imply CPX-1 is associated with SVs, CPX-1 may be bound to the plasma membrane or to SNARE complexes within the bouton, a distinction well beyond the limits of confocal microscopy. To distinguish between these possibilities, CPX-1::GFP was imaged in axons lacking either SVs or the v-SNARE, synaptobrevin. Axonal SVs are largely eliminated in animals harboring a mutation in the anterograde motor unc-104 KIF1A as a result of impaired vesicle trafficking (Hall and Hedgecock, 1991). Synaptobrevin is mistrafficked in snb-1(md247) animals due to a mutation in the transmembrane region (Nonet et al., 1998). While CPX-1::GFP abundance decreased by more than 80% when SVs were eliminated (Fig. 4E), there was actually an increase in synaptic CPX-1::GFP when v-SNARE abundance was reduced (Fig. 4F). Finally, CPX-1 and RAB-3 were both decreased in an endocytosis mutant (unc-57 endophilin) while both proteins increased in several exocytosis mutants (unc-64 syntaxin, unc-13, and unc-18)(Fig. 4G), further supporting the correlation of complexin with vesicle number. Thus, complexin’s spatial distribution, insensitivity to SNARE abundance, and strong bidirectional dependence on vesicle number indicate an association with SVs rather than the plasma membrane.
Having determined that synaptic CPX-1 localization correlates with the SV pool at the NMJ, we next asked whether the CTD is required for sequestering CPX-1. Steady-state imaging does not report whether CPX-1 is permanently anchored within the synapse or rapidly exchanging between compartments. To measure the residence time of CPX-1 within a synapse, CPX-1 was fused to photo-activatable GFP (pGFP) and expressed in motor neurons along with mCherry::RAB-3 to visualize the SV pool. By photo-activating CPX-pGFP at single synapses in intact animals, the mobility of fluorescent CPX-1 could be monitored and quantified (Fig. 5A,B). Proteins that are bound to immobile structures within the synapse would be expected to remain within the synapse on a time scale of seconds to minutes, whereas unbound cytosolic proteins should rapidly diffuse out of the synapse and into the adjacent axonal volume. The majority of untagged pGFP diffused out of a synaptic bouton within one second while about half of the CPX-pGFP exhibited a much longer residence time at the synapse, with about 20% remaining after 30 seconds (Fig. 5C–E). Thus a significant population of CPX-pGFP does not behave like a freely-diffusing protein within the synapse, consistent with the steady-state comparisons to soluble mCherry.
Using complexin’s long synaptic dwell time as an indication of an interaction at the synapse, we investigated whether the SNARE binding central helix (CH) or the lipid binding CTD influence the kinetics of CPX-pGFP exchange. Deletion of either complexin domain increased mobility, but the CTD played a more significant role in synaptic retention with CPX(ΔCT)-pGFP indistinguishable from pGFP alone by 30 seconds after photoactivation (Fig. 5C–E). The mobility of the CPX-pGFP variants did not correlate with their synaptic abundance (Fig. S3C). Since the only known high-affinity binding partner of complexin is the ternary SNARE bundle, abolishing the ability of complexin to bind synaptic SNARE bundles might be expected to decrease the dwell time of synaptic complexin. However, these experiments show that the CTD is more important in sequestering a pool of synaptic complexin consistent with the localization results described above. The low affinity lipid binding conferred by the CTD provides a simple explanation for synaptic localization of complexin, perhaps reversibly trapping complexin on SVs where it can engage the SNARE proteins.
The imaging, biochemistry, and physiology results presented above indicate that complexin is recruited to SVs to engage the SNARE bundle and block spontaneous fusion. Is complexin active or inactive when bound to the SV? Could the SV pool serve as a reservoir of inactive complexin, or does complexin require SV interactions in order to function as a clamp? To address these questions, CPX-1 was targeted either to the plasma membrane or the SV pool using proteins known to be strongly enriched either on the synaptic PM or SVs as artificial anchors. Three anchors were selected to target CPX-1 to the plasma membrane (Fig. 6A). K-ras localizes predominantly to the plasma membrane and is anchored by a C-terminal polybasic stretch and farnesylation of its CAAX box (Hancock, 2003; Willumsen et al., 1984; Zhang and Casey, 1996). The last 23 residues of let-60 K-ras were attached to full-length CPX-1 to create a plasma-membrane targeted complexin (CPXCAAX). Despite targeting sufficient protein levels within the synapse (Fig. 6B, Fig. S4A), CPXCAAX could only partially rescue the aldicarb hypersensitivity of cpx-1 mutants (Fig. 6D). If membrane association was disrupted by replacing the terminal cysteine with serine, CPX-1 function could be largely restored (Fig. S4B&C), demonstrating that anchoring CPX-1 to the plasma membrane interferes with its inhibitory function. Similar results were found when CPX-1 was targeted to the plasma membrane using unc-64 syntaxin 1A and the PDZ/C2 fragment of unc-10 RIM1 (Fig. 6D). We cannot rule out the possibility that some of the CAAX box or TMR-anchored CPX-1 localizes to SVs as low levels of K-ras and other CAAX box small GTPases have been co-purified with SVs (Takamori et al., 2006), and a small percentage of total synaptic syntaxin is found on vesicles (Mitchell and Ryan, 2004; Walch-Solimena et al., 1995). This residual vesicle binding may account for the partial rescue we observed in the plasma-membrane anchored complexins. Taken together, these experiments suggest that tethering full-length complexin to the plasma membrane disrupts its ability to inhibit neurotransmitter release.
Full-length CPX-1::GFP was fused to full-length RAB-3 to create a chimeric complexin CPX(SV) specifically targeted to SVs based on the known SV enrichment of this Rab (Schluter et al., 2002). CPX(SV) localization was indistinguishable from RAB-3 based on confocal imaging and fully restored aldicarb sensitivity in cpx-1 mutants, in contrast to the plasma membrane anchor chimeras (Fig. 6C,D). Furthermore, CPX(SV) restored wild-type locomotory behavior whereas CPX(CAAX) was not functional (Fig. 6E). Thus in two independent behavioral assays, SV-targeted CPX-1 fully restored the function of endogenous complexin whereas plasma membrane-associated CPX-1 was not sufficient. Finally, muscle patch-clamp recordings in wild-type, cpx-1, and cpx-1 mutants expressing CPX(SV) demonstrated that SV-targeted CPX-1 could suppress spontaneous fusion with the same efficacy as endogenous CPX-1 (Fig. 6F–H). These experiments indicate that full-length complexin is more effective at clamping the SNAREs when tethered to the SV than when targeted to the plasma membrane.
To test the hypothesis that complexin’s CTD functions chiefly to anchor complexin to SVs, the ΔCT variant of CPX-1 was targeted to SVs using RAB-3 and to the plasma membrane using the let-60 K-ras C-terminus (Fig. 7A). Complexin lacking its CTD failed to suppress elevated ACh levels in the absence of endogenous complexin (Martin et al., 2011). However, when the ΔCT variant was fused to RAB-3 (ΔCT-SV), the majority of complexin’s inhibitory function was restored (Fig. 7B). Targeting the LV/EE variant to SVs (LV/EE-SV) could also restore the majority of its function in vivo (Fig. 7C). In contrast, fusion of the ΔCT variant to the K-ras CAAX domain (ΔCT-CAAX), a plasma membrane-targeted protein, could not replace complexin function (Fig. 7B&C). It is possible that the CTD participates in other processes beyond tethering complexin to SVs because neither ΔCT-SV nor LV/EE-SV could fully restore CPX-1 function. Nevertheless, much of complexin’s inhibitory function could be restored solely by targeting the N-terminal half of complexin to SVs. This highlights a critical role for the lipid-binding CTD in the fusion clamp activity of complexin: steering the SNARE-binding region of complexin onto SVs is required for efficient clamping and prevention of spontaneous fusion.
The synapse must maintain tight control over the rate of spontaneous exocytosis to sustain a pool of SVs while minimizing basal noise levels, with complexin functioning as the predominant inhibitor at many synapses. The question of how complexin is directed to the assembling SNARE bundle has not previously been addressed. Several results drawn from our in vitro and in vivo experiments revealed that a region near the end of complexin’s CTD interacts with SVs to regulate spontaneous fusion. First, a conserved amphipathic region near the C-terminus bound phospholipids as measured by liposome flotation, sedimentation, and NMR spectroscopy. Second, a novel region of complexin’s C-terminus consisting of the final 12 residues interacted with liposomes. Third, mutations that disrupt lipid binding also prevented complexin from inhibiting neurotransmitter secretion in vivo. Fourth, complexin was sequestered at the synapse primarily through its CTD and colocalized with SVs. And lastly, tethering complexin to SVs bypassed the requirement for the CTD whereas plasma membrane targeting prevented complexin from fully functioning. Since other SNARE-binding proteins such as synaptotagmin and tomosyn are thought to engage the SNARE complex from the SV (McEwen et al., 2006; Takamori et al., 2006; Wang et al., 2011), this dependence on localization and enrichment may be a general feature of proteins tasked with binding a highly dynamic SNARE complex in the act of assembly.
The SV pool has been proposed to serve as a scaffold capable of buffering synaptic proteins, and complexin was identified as a highly abundant protein enriched within the bouton (Denker et al., 2011). A comprehensive SV protein study found no co-purification of complexin with SVs, consistent with a transient, low affinity SV interaction (Takamori et al., 2006). Our binding and dynamic imaging data suggest that the complexin has a low affinity interaction with liposomes in vitro and SVs in vivo. Since complexin is one of the more abundant proteins at the synapse (Denker et al., 2011), complexin likely saturates the available vesicle binding sites at the synapse and is rapidly exchanging between vesicle surface and the cytoplasm. Even if, as our biochemical and imaging results suggest, complexin’s affinity for SVs is low, a cluster of hundreds of closely packed SVs at the synapse would provide the avidity to trap and buffer complexin, maintaining the total abundance at a higher concentration than axonal complexin. Our data do not exclude the possibility that complexin also interacts with other membranes at the synapse, but the artificial tether experiments indicate that complexin’s inhibitory function originates from the SV. Further studies will determine whether a specific interaction such as preference for highly curved membrane, particular phospholipids, or binding to SV proteins directs complexin to the SV pool. Thus, the data presented here provide direct mechanistic evidence for the vesicle pool as a platform for the capture and local organization of a key synaptic protein.
In both worm and fly, the spontaneous fusion rate was ten times higher in NMJs lacking complexin, (Hobson et al., 2011; Huntwork and Littleton, 2007; Martin et al., 2011), highlighting complexin’s major role in suppressing exocytosis in the absence of stimulation. In addition, over-expression of the ΔCT variant, which does not bind SVs, failed to rescue complexin’s inhibitory function in worm, fly, and mouse, suggesting that simply providing a high cytoplasmic concentration may not be sufficient to trap the assembling SNARE bundle effectively (Cho et al., 2010; Kaeser-Woo et al., 2012; Martin et al., 2011; Xue et al., 2009). Why might vesicle-anchored complexin be more effective at stopping SNARE assembly? Complexin is thought to inhibit SV fusion by binding to the partially-zippered SNARE complex through its central helix and accessory domain (Giraudo et al., 2009; Kummel et al., 2011; Maximov et al., 2009; Xue et al., 2007). The central helix sits in the groove formed by syntaxin 1 and synaptobrevin 2 while the accessory domain can interact with a neighboring partially-assembled SNARE bundle situated anti-parallel to the first SNARE bundle, in a ‘zig-zag’ array conformation (Krishnakumar et al., 2011; Kummel et al., 2011). Thus, an otherwise rapidly assembling trans-SNARE complex requires complexin to engage several sites on multiple polypeptide chains simultaneously. This clamping mechanism will not occur efficiently if complexin is limited by its diffusion from the cytoplasm directly onto the SNARE complex. However, if complexin levels are locally elevated near the SV membrane through low affinity vesicle binding, the protein could be positioned to catalyze a significantly more rapid and reliable blockade of SNARE assembly. This sort of proximity-accelerated reaction has been observed in other systems where interfacial catalysis occurs, such as enzymes which act on phospholipids (Ramirez and Jain, 1991).
Since complexin binding to the trans-SNARE blocks spontaneous fusion, both the abundance and local position of complexin have a direct effect on spontaneous fusion rates. A highly simplified conceptual model of this relationship is depicted in Figure S5. In this model, complexin binds to the pre-fusion SNARE complex prior to full SNARE assembly to reduce spontaneous fusion. Locally elevating and orienting complexin through interactions with the SV will increase the effective rate of binding to the SNARE complex thereby reducing the rate of spontaneous fusion. Complexin function was impaired when tethered to the plasma membrane, so we propose that intercepting the assembling SNARE complex requires complexin to act from a specific location on the vesicle. Simply creating a high local concentration of complexin on the plasma membrane does not restore blockade of spontaneous fusion. Further exploration of complexin interactions with the SV will be needed to describe the positional requirements of complexin in the act of binding the assembling SNARE complex.
Several complexin sequences contain CAAX box motifs and the biological significance of prenylation has been demonstrated in both Drosophila DmCpx1 and mouse mCpx3 (Cho et al., 2010; Reim et al., 2005), suggesting complexin is specifically targeted to a membrane-associated compartment in some synapses. In fly, at least one splice variant of DmCpx1 must be farnesylated to block spontaneous SV fusion at the larval NMJ (Cho et al., 2010; Xue et al., 2009). Interestingly, another splice variant of DmCpx1 eliminates the CAAX box but preserves the amphipathic region (Fig. 1A), suggesting the possibility that fly complexin splice variants may be differentially targeted, perhaps within distinct neuronal subtypes. Mouse complexin isoforms mCpx3 and 4 are farnesylated and targeted to ribbon synapses of the retina (Landgraf et al., 2012; Reim et al., 2009; Reim et al., 2005). However, the specific effects on SV fusion at these synapses when farnesylated complexins are removed or altered are not yet known. Interestingly, forcing either worm CPX-1 or mCpx1 to be prenylated impairs its inhibitory function in the worm NMJ and mouse hippocampal cultures, respectively (Kaeser-Woo et al., 2012); Fig. 6). Perhaps the C-terminal sequences of CAAX-containing complexins target them to synaptic specializations such as the ribbon in photoreceptor terminals, or the T-bar in the fly larval NMJ (Prescott and Zenisek, 2005; Wichmann and Sigrist, 2010), while conventional synapse complexins target SVs directly.
In both worm and mouse, the facilitatory role of complexin during stimulus-evoked synaptic transmission can be partially restored without its CTD (Kaeser-Woo et al., 2012; Martin et al., 2011). Moreover, complexin promotes stimulus-evoked fusion even when tight SNARE binding is disrupted, despite complete elimination of its clamping function (Martin et al., 2011). If neither SV interactions nor tight SNARE binding are required to promote calcium-triggered exocytosis, then how does complexin facilitate this process? Perhaps interactions with other critical synaptic proteins such as synaptotagmin in the context of elevated cytoplasmic calcium can account for this separate functionality of complexin. Further studies will be required to determine the details of these interactions, but the results presented here offer a key mechanistic insight into complexin’s inhibitory function, and establish a general mechanism for controlling the abundance and positioning of synaptic proteins guided by low affinity interactions with the SV pool.
To measure aldicarb sensitivity, 20–25 young adult animals were placed on agar plates containing 1 mM aldicarb (Chem Services). Worms were scored for paralysis at ten minute intervals for two hours. Each genotype was tested at least 10 times and paralysis curves were generated by averaging paralysis time courses for each plate as described previously (Dittman and Kaplan, 2008).
Young adult animals were picked to 6 cm agar plates with no bacterial lawn (40 worms per plate). Imaging began 1 hour after worms were transferred. At least 200 animals were assayed for each genotype. One minute digital videos of individual animals were captured on an ORCA-05G CCD camera (Hamamatsu) mounted on an Olympus SZX16 stereomicroscope at 5 Hz using 10X magnification. Plates were lightly tapped before tracking capture. The center of mass was recorded for each animal on each video frame using custom object tracking software in MatLab (Ramot et al., 2008). Fewer than 10% of worms were excluded due to lack of movement for 30 seconds or longer.
Animals were immobilized using a 30 mg/mL solution of 2,3-butanedione monoxime (J.T Baker) mounted on 2% agarose pads, and imaged on an inverted Olympus microscope (IX81), using a laser scanning confocal imaging system (Olympus Fluoview FV1000 with dual confocal scan heads) and an Olympus PlanApo 60X 1.42 NA objective. Rescuing complexin variants were C-terminally tagged with GFP separated by a 12 residue linker (GGSGGSGGSAAA), and synaptic protein levels were estimated by measuring background-subtracted fluorescence within dorsal cord varicosities. Data were analyzed with custom software in IGOR Pro (WaveMetrics, Lake Oswego, OR) (Burbea et al., 2002; Dittman and Kaplan, 2006). A fluorescent slide was imaged daily to monitor the laser stability and the dorsal cord fluorescence was normalized to the slide value. For the spatial correlation experiments shown in Figure 4, individual line scans where computed from maximal intensity projections of image stacks in each color channel. Green (CPX-1) and red (RAB-3 or ELKS-1) line scans from the same animal were cross-correlated to compute the Pearson correlation coefficient in the paired measurement. Since some of the overlapping fluorescence intensity could have arisen due to chance, we also shuffled the green and red line scans so that green-red correlations between animals could be determined (shuffled correlation). In all cases, the shuffled correlation coefficient was nearly zero, confirming that the measured correlation is not due to chance. ΔF/F calculations in Figure 4, were made as described previously (Dittman and Kaplan, 2006).
Complexin variants were C-terminally tagged with photoactivatable GFP (pGFP) separated by a 12 residue linker (GGSGGSGGSAAA) and co-expressed with mCherry::RAB-3 in dorsal cord motor neurons using a modified unc-129 promoter. pGFP was photo-activated with a one millisecond pulse from a stationary 405-nm laser beam centered on the synapse of interest based on mCherry::RAB-3 fluorescence. Photo-activated fluorescence was initially distributed over the synapse with a full-width half-maximum ranging from 500 to 800 nm. Line scans of length five to ten microns centered on the target synapse were collected every 1.5 – 2 milliseconds, depending on the line scan length. Kymographs of sequential line scans were assembled and analyzed to generate time courses of fluorescence decay averaged over a one micron region centered on the target synapse. For the CPX-pGFP experiments shown in Figure 5C, small regions of interest measuring 10 by 0.2 microns (400 by 8 pixels) centered on the target synapse were collected every 30 to 40 milliseconds, and each image was collapsed to a line scan using maximal intensity projection. Consecutive projected line scans were then assembled into a kymograph and analyzed as described above. Experiments were rejected if mCherry::RAB-3 fluorescence revealed large drift or movement artifacts. All analysis was performed in Igor using custom written software. The decay time courses reflect a nontrivial combination of diffusion and binding/unbinding kinetics. In the absence of a detailed reaction-diffusion model to account for the details of synapse geometry, diffusion, and membrane binding, we chose to fit the data empirically using a double exponential decay function. For each synapse, weighted decay times constants were obtained by fitting the normalized, background subtracted data with a double exponential function and computing τw = (Afast τfast + Aslow τslow)/(Afast + Aslow) where Afast and Aslow are the fast and slow amplitudes, while τfast and τslow are the fast and slow time constants, respectively. These parameters allow for a quantitative comparison of decay kinetics but are not intended to correspond directly to a measure of either diffusion or unbinding kinetics. For untagged pGFP, diffusion out of the synapse could be estimated by fitting each line scan with a Gaussian centered on the target synapse and analyzing the increase in variance of the Gaussian over time using the formula:
where σo2 is the initial variance of fluorescence. For simple diffusion, this rate is constant and thus variance will increase linearly with time (see Fig. 5B). Using this approach, a moderate range of diffusion constants was observed ranging between 5 and 15 µm2/sec at the worm NMJ, consistent with small globular proteins diffusing within a restricted compartment.
CPX-1(WT)::pGFP (JP249), CPX-1(K/P)::pGFP (JP252), CPX-1(ΔCT)::pGFP (JP253), CPX-1(LV/EE)::pGFP (JP254), and CPX-1(Δ12)::pGFP (JP318) constructs were cloned into the pET28a vector containing a His6 tag to facilitate purification. These constructs also contain a 34 amino acid sequence at the N-terminus consisting of a T7 tag and thrombin cleavage site. BL21-DE3 E. coli were transformed and grown in rich media containing 50 µg/mL kanamycin to an optical density of 0.6. Cells were then induced with 400 µg/ml isopropyl thiogalactopyranoside (IPTG), grown for three hours at 37 °C, pelleted, resuspended in buffer (35 mM NaCl, 20 mM imidazole, 20 mM Tris-HCl, 1.5 mM BME, 2 mM DTT), lysed by sonication, and pelleted at 40,000 rpm for 40 minutes. For NMR samples, cells grown on rich media were spun down and resuspended in a minimal M9 media containing 15N-ammonium chloride with or without 13C-glucose prior to induction. The supernatant was purified on a Ni-NTA column (Qiagen, Hilden, Germany). Protein was eluted in elution buffer (350 mM NaCl, 250mM Imidazole, 20 mM Tris-HCl, 1.5 mM BME, 2 mM DTT). For assays in Figure 1, concentrated fractions underwent buffer exchange (150 mM NaCl, 50 mM Tris-HCl, 1 mM CaCl, 1 mM DTT) using Sephadex G-25 Fine beads (Sigma). For NMR, pure complexin was dialyzed into 60 mM phosphate, 2 mM DTT, 0.5 mM EDTA, pH 6.1.
Lipids were obtained from Avanti Polar Lipids and stored at −20 °C. For anionic liposomes, a lipid mixture composed of 70% 1-palmitoyl-2-oleoyl-phosphatidylcholine (POPC), and 30% 1-palmitoyl-2-oleoyl-phosphatidylserine (POPS) was dried under a stream of N2 gas, and residual solvent was removed under vacuum for 1 hour. The lipid film was then rehydrated in buffer (150 mM NaCl, 50 mM Tris-HCl, 1 mM CaCl2, 1 mM DTT for flotation, or 60 mM phosphate, 2 mM DTT, 0.5 mM EDTA, pH 6.1 for NMR) to obtain a final lipid concentration of either 4 mM (flotation and co-sedimentation) or 20 mM (NMR). The resulting liposomes underwent 10 cycles of freezing in liquid nitrogen and thawing (50 °C water bath), and were then extruded 21 times through either 400 or 200 nm then 100 nm pore-size polycarbonate films using a 10 mL Lipex Extruder from Northern Lipids (Canada) to form unilamellar vesicles.
Protein and lipids were incubated at room temperature for 1 hour. A sucrose gradient (10 mM HEPES, 150 mM NaCl, 1 mM CaCl) containing 500 µL of the sample mixed with 500 µL 60% sucrose (Amresco), 1mL 25% sucrose, 1mL 20% sucrose and 500 µL buffer was constructed. 100 µL fractions were obtained after ultracentrifugation in an MLS-50 rotor (Beckman) at 35,000 rpm for 2 hours. Fractions were run on a 10% SDS-PAGE gel followed by Western blotting. Protein was detected with a polyclonal rabbit GFP antibody and a horseradish peroxidase conjugated goat anti-rabbit secondary antibody (Genescript) and visualized using chemiluminescence (Thermo Scientific). For the flotation assays, the total amount of protein added was estimated using SDS-PAGE (WT 0.92 µg; K123P 1.2 µg; ΔCT 0.89 µg; LV/EE 0.74 µg for assay shown in Figure 1C, and WT 9.6 µg; Δ12 40.8 µg for the assay shown in Figure S1C) with bands quantified using ImageJ. The protein:liposome ratio for these experiments ranged from 4:1 to 200:1 which is more than an order of magnitude below the expected ratio of binding sites:liposome (~4,400:1, estimated using the surface area occupied by a 30-residue helix freely rotating about its center on the surface of a 150 nm diameter liposome) (Georgieva et al., 2010).
Full-length and Δ12 CPX-1 proteins were spun at 75,000 rpm for 30 min to eliminate potential protein aggregates and then incubated with lipids at room temperature for 1 hour. 20 µL fractions were taken from the ~70 µL supernatant and the resuspended pellet after ultracentrifugation in a TLA100.1 rotor (Beckman) at 75,000 rpm for 30 min. The fractions were run on a 10% SDS-PAGE gel and examined using ImageJ. Percent co-sedimentation was calculated as 100×P/(P + 3.5×S) where P is the pellet band intensity and S is the supernatant band intensity. The co-sedimentation was performed six times and the calculated co-sedimentation percentage was averaged. Protein:liposome ratios were about 20:1 for these assays.
Proton-nitrogen correlation (HSQC) spectra as well as a standard set of heteronuclear triple resonance three-dimensional spectra (HNCACB/CBCACONH, HNCACO/HNCO, and HNCANH) were collected on either a Varian Unity Inova 600 MHz (Weill Cornell NMR Facility) or a Bruker AVANCE 700 MHz (New York Structural Biology Center) spectrometer equipped with a cryoprobe. Backbone resonance assignments for free wild type complexin were obtained at a completeness level of 91% for amide group, Cα, and Cβ, and 87% for CO resonances, except for the N-terminal tag region and residues 40–66, the latter of which were not assigned due to reduced signal intensity and severe overlap, likely originating from the known helical conformation of this region (Chen et al., 2002). Liposome titrations were monitored using HSQC spectra using 1024 and 2564 complex data points in the direct and indirect dimensions, respectively and spectral widths of 12,000 Hz (proton) and 1600 Hz (nitrogen). All spectra were collected at 20 °C with subsaturating protein:liposome ratios. Data were processed using NMRpipe and analyzed using NMRview (Delaglio et al., 1995; Johnson and Blevins, 1994). Spectra were referenced indirectly to water. Protein concentrations were between 50 and 75 micromolar, as measured using absorbance at 280 nm and an experimentally determined extinction coefficient of 1882 cm−1M−1.
Whole-cell patch-clamp recordings were performed on dissected animals as described previously (Madison et al., 2005; McEwen et al., 2006). Dissected worms were superfused in an extracellular solution containing 127 mM NaCl, 5 mM KCl, 26 mM NaHCO3, 1.25 mM NaH2PO4, 20 mM glucose, 5 mM MgCl2 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 cholinergic EPSCs.
For data sets requiring a single pairwise comparison, we used Student's t test to compute significance. For all other comparisons, we used the Tukey-Kramer method for multiple comparisons. This test assumes independent, normal distributions with equal variance. In cases in which sample variance was significantly different, we employed the Newman-Keuls method to test significance. Significance was defined by the criterion p < 0.01.
We thank Tim Ryan, Josh Kaplan, Daniel Radoff, Richard Komuniecki, and Mike Hoppa for advice and for critically reading the manuscript, as well as the anonymous reviewers for their helpful suggestions. This work was supported by National Institutes of Health grants R01-GM095674 (JSD), R00-MH085039 (JB), R37-AG019391 (DE), 5R01GM71041(IM), MSTP GM07739 (DS), a gift from Herbert and Ann Siegel (DE) and the Rita Allen Foundation Award 188423 (JSD).
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.