Complexin Membrane Interactions: Implications for Synapse Evolution and Function

Abstract

The molecules and mechanisms behind chemical synaptic transmission have been explored for decades. For several of the core proteins involved in synaptic vesicle fusion, we now have a reasonably detailed grasp of their biochemical, structural, and functional properties. Complexin is one of the key synaptic proteins for which a simple mechanistic understanding is still lacking. Living up to its name, this small protein has been associated with a variety of roles differing between synapses and between species, but little consensus has been reached on its fundamental modes of action. Much attention has been paid to its deeply conserved SNARE-binding properties, while membrane-binding features of complexin and their functional significance have yet to be explored to the same degree. In this review, we summarize the known membrane interactions of the complexin C-terminal domain and their potential relevance to its function, synaptic localization, and evolutionary history.

Introduction

Precise and rapid regulation of neurotransmitter exocytosis is a key feature of neuronal communication. More than two decades of research into the molecules underlying exocytosis have revolutionized our understanding of synaptic transmission, and a coherent picture of the core presynaptic release machinery is beginning to take shape.^1-4 Briefly, neurotransmitter-containing synaptic vesicles (SVs) become tightly associated with the synaptic bouton plasma membrane in part through interactions at the protein-dense active zone (AZ) where voltage-gated calcium (Ca²⁺) channels also localize.^5-6 Like many other forms of eukaryotic lipid bilayer fusion, SNARE proteins (Synaptobrevin/VAMP2 on the vesicle and Syntaxin 1 and SNAP25 on the plasma membrane) are required for SV fusion. At close proximity, VAMP2 and Syntaxin 1/SNAP-25 partially assemble into the trans-SNARE complex (Figure 1(A)). A subset of SVs with SNAREs in this state are considered docked and primed, meaning that they are poised for rapid (submillisecond) fusion upon local Ca²⁺ elevation supplied by nearby voltage-gated Ca²⁺ channels and subsequent conformational changes of synaptotagmin (Syt1) following Ca²⁺ binding to its two C2 domains.^4,7 Several other Ca²⁺-binding presynaptic proteins such as Munc13-1 also regulate the processes of SNARE assembly and SV fusion.^8-10 While many biochemical, structural, and functional aspects of this account remain obscure or unsettled, this core machinery for Ca²⁺-triggered exocytosis appears to be shared across all animals, implying a singular and fundamental mechanism exists for SNARE-mediated vesicle fusion.^9,11-13 In this description of the neurotransmitter release machinery, we intentionally omitted a critical player: the SNARE-binding synaptic protein complexin (Cpx), first discovered as a SNARE-binding protein more than 27 years ago.^14-17 Unlike other core fusion machinery proteins, Cpx function and its conservation across animal taxa has proven more difficult to elucidate. Most prominently, the issue of whether Cpx acts primarily to inhibit or promote SV fusion has endured for more than 15 years (Figure 1(B)). Interested readers are directed to several excellent reviews that cover these issues.^4,18-19 Here, we will focus on Cpx membrane binding – a somewhat neglected property that is shared among almost all known Cpx homologs. Notably, most experimental efforts to connect Cpx membrane biochemical and structural features to its synaptic function have been centered on the Cpx C-terminal domain (CT). What mechanisms underlie Cpx CT membrane binding and are they important for Cpx function at the synapse? Are there distinct forms of membrane interactions between the different isoforms of Cpx? Studies over the past few years across experimental systems and species have begun to shed light on these critical and still unresolved questions.

Figure 1. Cpx packs a big punch for a small protein.

(A). Prefusion model for Cpx on the trans-SNARE complex while tail-anchored to the synaptic vesicle (SV) membrane. Helices are indicated by cylinders. Cartoon adapted from Söllner and colleagues.⁴⁰ (B). Schematic of the four Cpx protein domains: NT = N-terminal domain, AH = accessory helix, CH = central helix, and CT = C-terminal domain. The secondary structure is largely disordered with several alpha-helical regions indicated with a gray helix. Published effects on synaptic vesicle fusion when each domain is perturbed are indicated by the arrows with green for a fusogenic function and red for an inhibitory function (both in vitro and in vivo).

The four domains of Cpx and their proposed functions

Cpx can be parsed into four distinct domains despite its diminutive size (typically 100–160 residues in length depending on the species and Cpx isoform) (Figure 1(B)). The N-terminal domain (NT) of mammalian Cpx1 has been shown to interact with SNAP25 as well as membranes and may support Cpx fusogenic activity, although there is only a small region of sequence conservation across metazoa, and a fusogenic role for the NT domain was not observed in C. elegans synapses or in vitro.^20-25 While the accessory helix domain (AH) is poorly conserved across phylogeny, this region is generally rich in acidic residues and forms a stable alpha helix in solution.^26-29 Functionally, the AH domain is required for Cpx-mediated inhibition of fusion both in vitro and in vivo, but there is little consensus on a mechanistic basis for its role.^20,25,28-35 The highly conserved central helix domain (CH) tightly binds assembled SNARE proteins (in the groove between VAMP2 and Syntaxin 1) and is required for all known functions of Cpx.^26,36-37 This 25-residue sequence effectively defines Cpx and supplies the only high sequence conservation in the protein when comparing homologs in widely divergent animal taxa.³⁸ For example, mouse and C. elegans CH display 80% sequence similarity and 94% sequence identity along the SNARE-binding region of the helix (Figure 2(A)). Lastly, the largely disordered CT domain is poorly conserved (39% identity between worm and mouse) but has been proposed to aid in proper localization of Cpx via membrane interactions and in some cases to promote neurotransmitter release.^38-46 As described below, broad comparisons of metazoan and unicellular Cpx homologs indicate that there are two subfamilies based on the presence of a CAAX box motif at the C-terminus (Figure 2(B&C)).^38,47 In the following sections, we will illustrate the Cpx family from an evolutionary perspective and then discuss several Cpx CT features adapted for membrane interactions as well as their functional ramifications at the synapse.

Figure 2. Conserved and divergent features of Cpx across holozoa.

(A). CH alignments from 16 representatives across metazoan and choanoflagellate taxa. Positions with chemically similar residues across species are shaded in blue and the consensus sequence for these sequences is given below.⁸⁸ The seven most conserved residues are indicated in blue on a helix wheel model of the central helix with the SNARE-binding interface indicated below. The final 33 residues of CAAX Cpx (B) and final 30 residues of non-CAAX Cpx (C) are shown for 12 species spanning holozoan taxa as indicated with no attempt to align individual sequences. The CAAX box motif is indicated in green and the predicted amphipathic region (limited to 18 consecutive residues based on HELIQUEST predictions) is encircled (blue)⁸⁹. Acidic residues outside the amphipathic motif are indicated in red. The corresponding amphipathic moment is indicated after each sequence.⁸⁹ Helical wheel models of the amphipathic motifs for several bilaterian and nonbilaterian representatives of CAAX variants (D) and non-CAAX variants (E). Residue color scheme: hydrophobic/aromatic (yellow), Gly/Pro (white), all others (blue). Species abbreviations are: Homo sapiens (Hs), Mus musculus (Mm), Monodelphis domestica (Md), Gallus gallus (Gg), Xenopus laevis (Xl), Danio rerio (Dr), Branchiostoma floridae (Bf), Ciona intestinalis (Ci), Drosophila melanogaster (Dm7A & Dm7B), Caenorhabditis elegans (Ce), Hydra vulgaris (Hv), Exaiptasia pallida (Ep), Nematostella vectensis (Nv), Orbicella faveolata (Of), Mnemiopsis leidyi (Ml), Pleurobrachia bachei (Pb), Trichoplax adherens (Ta), Sycon ciliatum (Sc), Monosiga brevicollis (Mb), Salpingoeca rosetta (Sr).

What has Cpx been doing for the past 600 million years?

Many of the core proteins required for chemical synaptic transmission – including Cpx – predate the evolution of neurons and even animals that lack neurons such as sponges and Trichoplax.^11-12,48 The most distantly related Cpx homologs are found (together with neuronal SNARE homologs and other SNARE binding proteins) in unicellular organisms such as the slime mold Fonticula alba, as well as in choanoflagellates, which are the closest living unicellular relatives of all animals.^12,38 Analysis of the Cpx CH domain across taxa (Figure 2(A)) suggests that ancestral Cpx-SNARE interactions likely played a role in regulating some form of membrane fusion prior to the emergence of neurons and nervous systems. Moreover, available Cpx sequences in unicellular organisms typically end in a CAAX box motif, implying that ancestral Cpx was tethered to membranes via prenylation (Figure 2(B)).⁴⁷ Interestingly, these distant Cpx sequences do not harbor amphipathic motifs at their C-termini, supporting the notion that this CT feature arose later in the evolution of synapses (Figure 2(C)). Cnidaria (jelly fish, coral, hydra, and anemone for example), which are thought to have arisen between 540 and 580 million years ago, possess neurons and are more closely related to bilaterian animals (such as vertebrates, nematodes, insects, and molluscs).^49-50 A variety of Cpx homologs with and without CAAX motifs and C-terminal amphipathic regions can be found within cnidaria (Figure 2(B-E)). Notably, Yang and colleagues demonstrated that a cnidarian Cpx homolog (from the starlet sea anemone Nematostella vectensis) could restore the fusogenic function of mouse Cpx (mCpx) in cultured neurons from the mCpx1/2 double knock out.³⁸ Finally, bilaterian phyla generally possess both CAAX and non-CAAX variants of Cpx, all containing amphipathic CTs (Figure 2(D&E)), thus spawning new localization rules for Cpx in complex nervous systems.⁵¹ Bilaterian Cpx CTs also contain long acidic stretches of residues upstream from the amphipathic motif, which have yet to be assigned any functional significance, although one study concluded that this acidic region mediates an electrostatic interaction with a polybasic patch of residues on Syt1.⁵² Together with the highly acidic accessory helix adjacent to the SNARE-binding region of Cpx, this CT feature accounts for the low isoelectric point of bilaterian complexins (net charge in −7 to −10 range for most bilaterian Cpx sequences).²⁹ While the SNARE-binding region of Cpx displays far less sequence divergence than the CT, the selection pressures operating on the two regions may also stem from non-neuronal roles of Cpx. In vertebrates, Cpx is involved in both endocrine and exocrine functions as well as sperm acrosome fusion.^53-56 Thus, some of the evolutionary divergence described here for the Cpx CT may have originated from non-synaptic membrane fusion.

Some of the controversies regarding Cpx function arise from comparisons between mammalian and invertebrate synapses.^4,18-19 Numerous functions have been uncovered and quantified through careful assessments of multiple vertebrate and invertebrate synapses, and by consensus, vertebrate Cpx1/2 homologs generally support facilitatory rather than inhibitory functions at the synapse, whereas vertebrate Cpx3/4 homologs and invertebrate Cpx support both facilitatory and inhibitory functions.^{17,21-,22,57-63} For instance, in the absence of Cpx, stimulus-evoked release decreased by 50–90% in mouse, fly, and worm neurons whereas spontaneous fusion (SV fusion in the absence of Ca²⁺) increased by 2-fold to 20-fold in worm synapses and by more than 40-fold in fly synapses.^{17,20-22,57,64-65} In the most thorough study yet conducted in mouse cultured hippocampal neurons, spontaneous fusion decreased by more than 2-fold in the absence of mCpx1, 2, and 3.⁶⁵ While some of this functional divergence will be accounted for by differences in membrane interactions of the Cpx CT, it is probable that coevolution of other core fusion proteins will also be important to consider when comparing Cpx manipulations across widely divergent species. Notwithstanding ongoing debates associated with Cpx function, its evolutionary history suggests that Cpx has been part of the core synaptic machinery since the emergence of synaptic transmission as far back as 600 million years ago.

Function and localization of prenylated Cpx isoforms

Prenylation is the post-translational covalent attachment of a lipid (either a farnesyl or geranylgeranyl isoprenoid group) to a cysteine residue at a protein’s C-terminus, thereby anchoring its C-terminal region to the cytoplasmic leaflet of a lipid bilayer. As mentioned above, in many Cpx isoforms, the CT ends in a four-residue CAAX motif (CAAX), thereby enabling prenylation of the CT in situ (Figure 3(A)). Vertebrates express four Cpx paralogs: Cpx1 and Cpx2 are non-CAAX proteins widely expressed within and outside of the nervous system whereas Cpx3 and Cpx4 are CAAX proteins that display much more restricted expression patterns in the brain – high retinal expression in zebrafish and mice.^18,47,66 Reim and colleagues demonstrated that mCpx3 and mCpx4 are localized to ribbon synapses in photoreceptors and retinal bipolar cells (mCpx3 also was also found at conventional synapses in the retina), and that farnesylation of their CTs is required for this localization.⁴⁷ Disrupting Cpx3/4 function in ribbon synapses of both fish and mouse enhanced spontaneous and tonic neurotransmitter release while decreasing stimulus-evoked release akin to Cpx mutants in Drosophila and C. elegans.^{21-22,57-58,60,62-63,67} Kaeser-Woo and colleagues demonstrated that the N-terminal half of mCpx1 fused to the C-terminal half of mCpx3 (thereby gaining a lipid modification) failed to inhibit spontaneous fusion in cultured mouse neurons whereas the reverse chimera inhibited to a similar degree as full-length mCpx1, indicative of distinct functions for CAAX and non-CAAX variants in some synapses.⁴¹

Figure 3. Cpx CAAX variants are also found at some synapses.

(A). The CT of CAAX Cpx variants is farnesylated to create a membrane-anchored protein. (B). Two Drosophila melanogaster mutants disrupting DmCpx farnesylation. C. Postsynaptic current recordings of spontaneous release at fly muscle 6 NMJs in control (black), cpx⁵⁷² (brown), and cpx^−/− (cpx^SH1 – red) in 2.0 mM external Ca²⁺ saline. From⁶⁴ Figure 2A. (D). Maximal projections of confocal stacks show colocalization of Cpx (green, α-Cpx) with SVs (magenta, α-Csp) in WT motoneurons. Cpx staining intensity is strongly reduced in cpx¹²⁵⁷boutons (center) and no longer matches the SV distribution (right– laser power increased 6-fold). From⁶⁹ Figure 2E. €. Head homogenates from cpx/Df hemizygotes and cpx¹²⁵⁷/Df were subjected to phase partitioning in Triton X-114. Aqueous (left) and detergent (right) phases were then analyzed by immunoblotting with an anti-CPX antibody. From⁶⁸ Figure 6. (F). Neurons transfected with expression vectors encoding either WT EGFP-Cpx III or farnesylation-defective EGFP-Cpx III-C156S and immunostained for synaptophysin (red) to mark presynaptic terminals were analyzed by laser-scanning confocal microscopy. From⁴⁷ Figure 7.

In contrast to the vertebrate Cpx family, many invertebrates appear to have a single Cpx gene. Interestingly, the CT of the Drosophila homolog is alternatively spliced, creating both CAAX and non-CAAX Cpx variants from its single Cpx gene.⁶⁴ Buhl and colleagues identified two major splice variants expressed at the fly NMJ termed DmCpx7A (CAAX) and DmCpx7B (non-CAAX), with distinct amphipathic motifs near the C-terminus.⁶⁴ The CT sequence differences are shown in Figure 2(B&C). Both variants localized to presynaptic terminals while displaying differential effects on SV fusion: DmCpx7A was more effective at clamping spontaneous fusion while DmCpx7B strongly facilitated stimulus-evoked release relative to DmCpx7A. Mutations that disrupted prenylation of DmCpx7A eliminated its clamping function at the fly NMJ and switched biochemically isolated adult fly DmCpx from a detergent-soluble phase to an aqueous phase (Figure 3(B,C,E)).^58,64,68 Notably, the specific subsynaptic location of CAAX variants such as DmCpx7A and mCpx3/4 is not yet entirely clear (e.g. SVs, AZ plasma membrane, ribbon, endosomes), although synaptic enrichment is sensitive to CT perturbations (Figure 3(D&F)).^47,64,68-69 Loss of mCpx3/4 in retinal photoreceptor and bipolar cell ribbons or disruption of prenylation in fly Cpx led to subsynaptic alterations in SV tethering to ribbons/T-bars, supporting the notion that these CAAX variants target ribbons or T-bars via their lipid-modified C-termini.^69-70 Interestingly, DmCpx7B associates with liposomes in vitro even without its CT, implying that this non-CAAX variant harbors other membrane interacting regions beyond its CT amphipathic helix.⁷¹ Perhaps the NT of DmCpx7B mediates additional membrane interactions as described with other Cpx isoforms.^23-24,72 Given the ubiquity of the CAAX Cpx subfamily across metazoan nervous systems and the obvious connection between prenylation and membrane interactions, there are likely critical roles for this version of the Cpx CT yet to be discovered.

The Cpx CT harbors a membrane curvature sensor

Given that the CT CAAX motif is only present in some Cpx isoforms, what is the role of CTs lacking CAAX motifs? Are there other shared CT structural or chemical features that could indicate a function? Early Cpx structural studies were performed on truncated versions of non-CAAX variants lacking most of their CT. For example, the Cpx CT was not included in the peptide used for the original Cpx-SNARE crystal structure likely because its intrinsically disordered structure precluded efficient crystallization and only its middle alpha-helical region corresponding to the AH-CH domain was required for stoichiometric binding to the SNARE complex.^26-27,36 The highly variable protein sequence across phyla and lack of a three-dimensional structure made it difficult to propose a mechanistic role for the CT in either promotion or inhibition of neurotransmitter release. Notwithstanding the lack of conservation, several studies concluded that the Cpx CT is important for Cpx function including in vitro SNARE-mediated fusion assays, acute dissected animal preps, and in cell culture.^{22,41-42,51,58-59,68,73-74} One in vitro study based on single-molecule FRET with full-length Cpx bound to the cis-SNARE complex observed that the Cpx CT could fold back onto the N-terminal half of the peptide, bringing the CT into proximity with the ternary SNARE bundle bound to its CH.³⁷ Another study utilizing cell–cell SNARE-mediated fusion assay concluded that the Cpx CT did not matter functionally for inhibition of fusion when Cpx was anchored to membranes via a GPI linkage, consistent with a membrane-tethering function for the Cpx CT.⁷⁵ The first study to address potential membrane interactions of Cpx was performed by Seiler and colleagues using human Cpx1, which lacks a CAAX motif.⁴⁰ Based on in vitro liposome binding, the authors identified a conserved amphipathic helical region near the end of Cpx required for interactions with small (~50 nm) liposomes, and a model was proposed for Cpx anchored to the synaptic vesicle (SV) membrane via its CT (Figure 1(A)). This picture of a tail-anchored Cpx attached to SVs has remained largely unaltered.

Work from several groups has confirmed and extended the notion that non-CAAX Cpx CT interacts with membranes through either one or two motifs including an amphipathic helix.^{24,42-43,45,51,71,76} For example, Snead and colleagues observed a pronounced curvature sensitivity in the binding of the major C. elegans Cpx homolog (CPX-1) to liposomes in vitro and utilized a combination of solution-state NMR and CD spectroscopy to propose that the Cpx CT exists in a largely unstructured state in solution but becomes partially helical upon binding to highly curved membranes.⁴³ This property is reminiscent of amphipathic lipid packing sensor (ALPS) motifs found in other membrane-interacting proteins involved in vesicle trafficking pathways.^77-81 ALPS motifs comprise bulky hydrophobic residues together with a combination of charged side chains and hydroxyamino acids that selectively bind to highly curved membranes (Figure 4(C-D)). Like worm CPX-1, mammalian Cpx1 also displays a preference for high curvature membranes. Zdanowicz and colleagues utilized spectroscopic approaches to estimate that decreasing vesicle size by 4-fold (from ~ 100 nm down to ~ 25 nm) enhanced rat Cpx1 CT membrane binding by almost 80-fold (Figure 4(A)).²⁴ Gong and colleagues monitored mCpx1 binding to single labeled liposomes via TIRF microscopy and observed a striking selectivity for smaller liposomes (Figure 4(B)).⁴⁵ In addition, recent work from Grasso and colleagues (see Grasso et al this issue of JMB) described an extended bent alpha helical conformation of the last ~ 20 residues of mouse Cpx1 bound to small diameter liposomes [Grasso 2023 this issue]. Given the high sequence conservation of this motif across the vertebrate non-CAAX Cpx family (Figure 4(C)) together with similar findings in the nematode Cpx CT, the curvature sensing ALPS-like character of the Cpx CT is likely to be a general feature of all bilaterian non-CAAX Cpx proteins.

Figure 4. Cpx prefers high curvature.

(A). Electron paramagnetic resonance (EPR) spectra of Cpx labeled at T119 in the presence (red) and absence (black) of liposomes of various sizes shows dramatic changes in line shape upon CT membrane binding, and these changes were used to determine Cpx membrane affinity as a function of liposome diameter as indicated. Adapted from²⁴ Figure 4. (B). Single molecule Cpx membrane binding assay. Single Cpx-A647 molecules are monitored as they bind and unbind surface-tethered vesicles. Data points are quantification of one or more labeled Cpx molecules binding to single DiI-labeled vesicles plotted versus vesicle label intensity (log–log). The linear fit provides a measure of curvature sensitivity (larger negative slope = higher sensitivity). Adapted from⁴⁵ Figure 4. (C). Helical wheel representations of the amphipathic motif at the end of mCpx1 compared to three curvature-sensitive domains proposed to act as amphipathic lipid packing sensors (ALPS). Residue code: hydrophobic (yellow), acidic (red), basic (blue), helix-destabilizing Gly/Pro (green), and all others (gray). (D). Conceptual scheme for ALPS domains transitioning from a disordered state in solution to a folded state on highly curved lipid bilayers containing a high density of lipid packing defects as indicated.

What is the significance of Cpx preference for highly curved membranes? SVs are among the smallest bilayer-bound organelles (vesicle diameters ranging from 30 to 45 nm depending on the species), and thus an attractive hypothesis is that Cpx is recruited selectively to SV membranes via its CT where its AH-CH helix can engage SNAREs during the assembly process. Indeed, Wragg and colleagues demonstrated in C. elegans that CPX-1 colocalized with SVs in vivo and functioned efficiently when driven to SVs using foreign tethers.⁴² In contrast, forcing CPX-1 to the presynaptic plasma membrane strongly disrupted its function. Overexpression of CPX-1 lacking its entire CT (ΔCT) failed to restore CPX-1 inhibitory function whereas fusion of the ΔCT variant to the SV protein RAB-3 largely restored function in vivo.⁴² In contrast, ΔCT overexpression rescued ~ 50% of CPX-1 facilitatory function. Loss of the CPX-1 CT strongly enhanced its mobility within presynaptic terminals (where SVs are abundant) and eliminated sensitivity of CPX-1 mobility to changes in SV cycling dynamics.⁴⁴ Importantly, destabilization of the amphipathic helical conformation impaired CPX-1 inhibition even though its CT could still interact with membranes – an indication that the conformational switch itself is required for CPX-1 inhibitory function.^42-43 Intriguingly, replacement of the mCpx2 amphipathic helix with a similarly amphipathic helix derived from SNAP25 largely preserved mCpx2 inhibitory function in chromaffin cell granule secretion, indicating some form of amphipathic helical character is generally important for Cpx CT inhibitory function.⁷⁴ The role of this amphipathic region in Cpx facilitatory function is not clear, but a recent study hypothesized a direct role in stabilization of the nascent fusion pore as described in the next section.⁴⁶

Another potentially relevant aspect of the ALPS-like motif in Cpx is the characteristic presence of hydroxyamino acids on the aqueous face of the amphipathic helix (of both CAAX and non-CAAX Cpx variants). Several examples of phosphorylation-dephosphorylation of the ALPS domain leading to changes in either membrane binding or protein–protein interactions have been described.^82-83 Of particular interest is a study from Littleton and colleagues at the fly neuromuscular junction (NMJ) where fly Cpx (DmCpx) participates in several aspects of synaptic transmission including a form of short-term plasticity induced by high-frequency stimulation.⁷¹ The authors examined the phosphorylation of a serine in the amphipathic helix of the non-CAAX DmCpx7B variant and found that this site was likely phosphorylated by protein kinase A in response to synaptic activity. Moreover, a phospho-mimetic S126D mutation impaired DmCpx7B clamping function, thereby increasing the miniature excitatory postsynaptic potential (mini) frequency and occluding the induction of activity-dependent enhancement of mini frequency. Indeed, the serine leading into the amphipathic region of mammalian Cpx (S115) has been shown to be phosphorylated,⁸⁴ and S115 mutations impaired CT fusogenic function in vitro.³⁹ And lastly, a S115,T119D double substitution abrogated CT membrane-disrupting activity in vitro (see next section for further details of this membrane activity).⁴⁶ Thus, the Cpx membrane-interacting region may be tuned by synaptic kinases and phosphatases to regulate synaptic transmission akin to other membrane trafficking proteins harboring ALPS domains.

Evidence for membrane remodeling by the Cpx CT

While numerous efforts over the past decade have compellingly developed the hypothesis that Cpx is recruited to high curvature membranes in presynaptic terminals through its amphipathic CT, a recent study reported that this region can also remodel, tubulate, vesiculate, and perforate membranes in vitro. Courtney and colleagues described a striking series of experiments where the last 21 residues of mCpx2 (a non-CAAX Cpx isoform) massively remodeled giant unilamellar vesicles (GUVs) at micromolar concentrations, converting flat membranes into highly curved and contorted structures with a complex topology (Figure 5(A&B)).⁴⁶ In addition, the mCpx2 CT perforated membranes akin to many antimicrobial peptides (AMPs). In line with these observations, the authors utilized a machine-learning approach that served to predict membrane activity in peptide sequences, and the last 21 residues of both mCpx1 and mCpx2 generated high membrane activity scores.⁸⁵ Molecular dynamics simulations indicated that a small number of CT peptides could penetrate the lipid bilayer and oligomerize into a pore-forming structure, possibly accounting for the observed membrane perforating activity of mCpx2. Presumably, higher order assemblies of the CT peptide could drive the observed membrane tubulation and vesiculation. What are the functional consequences of this profound membrane remodeling activity at the synapse? While the question remains unanswered in vivo, the authors carefully examined the impact of mCpx2 and its AMP-like C-terminus on fusion pore lifetime in a high-resolution measurement of SNARE-mediated fusion pores in black lipid membranes.^46,86 These measurements demonstrated that the mCpx2 CT could stabilize nascent fusion pores by slowing the pore closure rate and in some cases, accelerating the pore formation rate (Figure 5(C)). Interestingly, a chimeric variant of mCpx2 with its CT amphipathic motif replaced with the well-known AMP melittin exhibited a similar degree of fusion pore stabilization. Given the extremely high membrane curvature associated with prefusion structures such as membrane stalks and the post-fusion pore neck prior to pore expansion, perhaps ALPS-like or AMP-like motifs at the Cpx C-terminus could interact with and influence lipid arrangements during the co-assembly of Cpx-bound SNARE domains. Loss of SV fusion pore stability should correlate with a decrease in the fusogenic properties of Cpx, but many studies discussed above report a net loss of Cpx inhibition when its CT is disrupted or deleted.^{22,41,59-60,62,68,73-74,87} The combined loss of SNARE clamping function (due to mislocalized Cpx when the CT is disrupted) and pore stabilization function (due to loss of CT membrane remodeling) could result in distinct effects on synaptic transmission depending on the relative impact of each mode of action. Indeed, much of the controversy surrounding Cpx function originates from comparisons across distantly related species where the importance of various Cpx functional modes may have diverged as well. These mysteries highlight a need for more thorough experimental comparisons of Cpx action on SNARE assembly and membrane fusion across phyla to pinpoint the fundamental aspects of Cpx function at the synapse.

Figure 5. Cpx remodels membranes.

(A). Illustration of the GUV-pore formation assay. Recombinant HaloTag protein is encapsulated in GUVs and JF635i ligand is in the media. GUV lipids are labeled with 0.1% rhodamine-PE (white). Pore formation allows ligands to enter the GUVs and bind to HaloTag to produce JF635i fluorescence (magenta). Addition of mellitin or Cpx allows JF635i access into GUVs. Cpx additionally drives overt vesiculation. (B). Cryo-ET of 100 nm LUVs with 10 μM Cpx. 2-nm thick virtual cross-section of the reconstructed volume (left) and surface representations of segmented features (right). (C). Illustration of the transmembrane domain Cpx (TMD-Cpx) fusion protein with each domain labeled and embedded in a 13 nm nanodisc (ND) with syb2. (D). Representative open-pore traces from recordings with several types of Cpx NDs as indicated. Fusion pore current (E₁), average pore open probability (E₂), average pore closing rate (E₃), and average pore opening rate (E₄) for each Cpx ND type. Figure panels reproduced from⁴⁶ Figures 3 - 5.

Conclusion

The previous sections have outlined numerous lines of evidence pointing to interactions between the Cpx CT and membranes via both amphipathic motifs and lipid modifications. These interactions are important both for the synaptic localization of Cpx as well as its function in regulating SV exocytosis. Functionally, the non-CAAX CT may play an indirect role in Cpx1/2 inhibitory activity by properly localizing Cpx on SVs during docking/priming to prevent spontaneous fusion while promoting evoked fusion through its membrane remodeling properties.^39,42,46 While a similar scenario may also apply to CAAX variants such as those found in retinal ribbon synapses, more research is required to expand our understanding of their subsynaptic localization and membrane shaping properties. Perhaps a larger question is whether differences observed between distinct synapse types and different species indicate divergent mechanistic roles for Cpx. Why does spontaneous SV fusion increase by a factor of 40–80-fold in the fly NMJ when DmCpx is removed whereas spontaneous SV fusion decreases by more than 2-fold in cultured mouse neurons when mCpx1,2, and 3 are removed^57-58,64-65? What is different about the docked/primed state of a mammalian SV compared to an invertebrate SV that could account for such a remarkable discrepancy? And given the stunning observations of membrane vesiculation and perforation by the mCpx2 CT, what prevents this lipid remodeling in vivo where mCpx1/2 are expressed at high levels in neurons and other cell types? The answers to these and many other unresolved questions will be eagerly awaited by those interested in the molecular underpinnings of synaptic transmission.

Abbreviations:

AH

accessory helix

ALPS

amphipathic lipid packing sensor

AMP

antimicrobial peptide

Ca²⁺

calcium ion

CAAX

C = cysteine, A = aliphatic, X = any residue

CD

circular dichroism

CH

central helix

Cpx

complexin

CPX-1

worm complexin

CT

C-terminal domain

DmCpx

fly complexin

EPR

electron paramagnetic resonance

FRET

fluorescence resonance energy transfer

GPI

glycosylphosphatidylinositol

GUV

giant unilamellar vesicle

mCpx

mouse Cpx

ND

nano disc

NMJ

neuromuscular junction

NMR

nuclear magnetic resonance

NSF

NEM sensitive factor

NT

N-terminal domain

SNARE

soluble NSF Attachment protein Receptor

SV

synaptic vesicle

Syt1

synaptotagmin 1

TMD-Cpx

transmembrane domain Cpx fusion protein

TIRF

total internal reflection

WT

wild type

DATA AVAILABILITY

Data will be made available on request.