Complexins are small soluble proteins that bind to assembling SNARE complexes during synaptic vesicle exocytosis, which in turn mediates neurotransmitter release. Complexins are required for clamping of spontaneous “mini ” release and for the priming and synaptotagmin-dependent Ca2+ triggering of evoked release. Mammalian genomes encode four complexins that are composed of an N-terminal unstructured sequence that activates synaptic exocytosis, an accessory α-helix that clamps exocytosis, an essential central α-helix that binds to assembling SNARE complexes and is required for all of its functions, and a long, apparently unstructured C-terminal sequence whose function remains unclear. Here, we used cultured mouse neurons to show that the C-terminal sequence of complexin-1 is not required for its synaptotagmin-activating function but is essential for its priming and clamping functions. Wild-type complexin-3 did not clamp exocytosis but nevertheless fully primed and activated exocytosis. Strikingly, exchanging the complexin-1 C terminus for the complexin-3 C terminus abrogated clamping, whereas exchanging the complexin-3 C terminus for the complexin-1 C terminus enabled clamping. Analysis of point mutations in the complexin-1 C terminus identified two single amino-acid substitutions that impaired clamping without altering the activation function of complexin-1. Examination of release induced by stimulus trains revealed that clamping-deficient C-terminal complexin mutants produced a modest relative increase in delayed release. Overall, our results show that the relatively large C-terminal complexin-1 sequence acts in priming and clamping synaptic exocytosis and demonstrate that the clamping function is not conserved in complexin-3, presumably because of its distinct C-terminal sequences.
Synaptic vesicle fusion, like all intracellular fusion reactions except for mitochondrial fusion, is thought to be mediated by the assembly of SNARE complexes and the action of SM proteins (Sec1/Munc18-like) on these SNARE complexes (for review, see Südhof and Rothman, 2009). SNARE and SM proteins are controlled by synaptotagmins as Ca2+ sensors that trigger fusion-pore opening when Ca2+ binds to their C2 domains and induces binding of these C2 domains to phospholipids and SNARE complexes (Geppert et al., 1994; Fernández-Chacón et al., 2001) and by complexins, which are small soluble adaptor proteins that bind to SNARE complexes (McMahon et al., 1995).
A large number of studies of complexin function with different approaches has uncovered three principal functions for complexin in fusion: activation of SNARE complexes for subsequent Ca2+ triggering by synaptotagmin (Reim et al., 2001; Tang et al., 2006; Xue et al., 2007, 2009, 2010; Maximov et al., 2009; Strenzke et al., 2009); clamping of SNARE complexes preventing fusion (Giraudo et al., 2006, 2008, 2009; Schaub et al., 2006; Tang et al., 2006; Huntwork and Littleton, 2007; Yoon et al., 2008; Cho et al., 2010; Yang et al., 2010; Hobson et al., 2011; Martin et al., 2011); and priming of vesicles for fusion (Cai et al., 2008; Xue et al., 2010; Yang et al., 2010).
Mammals express four complexins, of which complexin-1 and complexin-2 are present in all neuronal and non-neuronal cells (McMahon et al., 1995), whereas complexin-3 and complexin-4 are expressed at high levels in retina (Reim et al., 2005). Crystal structures of complexin bound to SNARE complexes revealed that complexin contains unstructured N-terminal and C-terminal sequences (residues 1–22 and 71-134, respectively, in mouse complexin-1) that flank two α-helices, namely a more N-terminal accessory α-helix (residues 23–47) and a central α-helix (residues 48–70; Bracher et al., 2002; Chen et al., 2002). The N-terminal complexin region is essential for activating fusion (Xue et al., 2007; Maximov et al., 2009) and for vesicle priming (Yang et al., 2010), the accessory α-helix is essential for clamping fusion (Maximov et al., 2009), and the central α-helix of complexin binds to the SNARE complex and is necessary for all complexin functions (Tang et al., 2006; Maximov et al., 2009; Martin et al., 2011). However, the functions of the C-terminal region of complexins that accounts for almost half of their sequences remain unclear. Initial experiments suggested that this sequence is functionally inert (Xue et al., 2007) or that it may participate in Ca2+ triggering of fusion by binding to synaptotagmin-1 (Tokumaru et al., 2008) or to phospholipids (Malsam et al., 2009; Seiler et al., 2009). Recently, moreover, it was proposed in Caenorhabditis elegans that this sequence may clamp fusion (Martin et al., 2011). Thus, to explore the functional significance of C-terminal domain of complexins, we here performed structure–function experiments using complexin knockdown (KD) neurons and tested different forms and mutations of complexins. We find that the complexin C-terminal sequences are required for both clamping spontaneous fusion and for priming evoked fusion but are dispensable for activating synaptagmin-dependent Ca2+ triggering of fusion.
Materials and Methods
Generation of KD lentiviruses and rescue constructs.
Complexin KD, wild-type complexin (CpxWT) rescue, and C-terminally truncated complexin-1 (Cpx11–86) rescue constructs were described previously (Maximov et al., 2009; Yang et al., 2010). CpxI101W, CpxL117W, CpxL117K, and CpxV120P mutations were generated by gene synthesis and were cloned into the downstream of human ubiquitin promoter of L309 lentiviral vector (Pang et al., 2010).
Cortical neurons were cultured from newborn wild-type mice of either sex as described previously (Maximov et al., 2009). Briefly, mouse cortices were dissected from postnatal day 1 mice, dissociated by papain digestion (10 U/ml, with 1 μm Ca2+ and 0.5 μm EDTA) for 20 min at 37°C, and plated on Matrigel-coated circular glass coverslips (11 mm diameter). Neurons were maintained for 14–16 d in MEM (Invitrogen) supplemented with B27 (Invitrogen), glucose, transferrin, fetal bovine serum, and AraC (Sigma).
Lentiviral production and infection of neuronal culture.
Cpx KD lentiviruses were prepared as described previously (Maximov et al., 2009; Yang et al., 2010). The lentiviral expression vector (complexin-1/2 KD vectors and/or the complexin rescue vectors) and three helper plasmids (pRSV–REV, pMDLg/pRRE, and VSV-G) were cotransfected with the lentiviral vectors into human embryonic kidney 293T (HEK293T) cells (American Type Culture Collection) at 4, 2, 2, and 2 μg of DNA, respectively, per 25 cm2 culture area using FUGENE6 transfection reagent (Roche). At 48 h after transfection, HEK cell medium was collected and clarified by centrifugation (500 × g for 5 min), and the supernatant was added directly to the medium of cultured cortical neurons maintained in 24-well plates (200 μl of supernatant per well). Cultured cortical neurons were infected at 4–5 d in vitro (DIV4–DIV5) and used for physiological analysis on DIV14–DIV16. In all experiments, the infection efficiency of neurons was monitored by the GFP signal driven by the ubiquitin promoter in the lentiviral vectors. All steps were performed under biosafety level 2 conditions. Quantitative RT-PCR indicated that the KD efficiency was 70–80% (Maximov et al., 2009).
Cultured neurons infected with the control or complexin KD lentiviruses were washed twice with PBS at DIV14, lysed in SDS–lysate buffer (25 mm HEPES-NaOH, pH 7.2, 25 mm NaCl, and 4 mm EDTA; 25 μl of sample buffer per well for 24-well plates), and boiled for 20 min. Equivalent amounts of protein (20 μl) were analyzed by SDS-PAGE and immunoblotting using the following antibodies: rabbit polyclonal antibody against the C terminus of Cpx1 (P942), rabbit polyclonal antibody against full-length GST-Cpx1 (L669), complexin-3 (122 301; Synaptic Systems), syntaxin-1 (78.2; Synaptic Systems), SNAP-25 (71.1; Synaptic Systems), and synaptobrevin-2 (69.1; Synaptic Systems).
Whole-cell patch-clamp recordings were made with neurons at DIV14–DIV16 with a pipette solution containing the following (in mm): 135 CsCl, 5 NaCl, 1 MgCl2 10 HEPES-NaOH, pH 7.4, 10 EGTA, 4 Mg-ATP, 0.4 Na-GTP, and 5 QX-314 (lidocaine N-ethyl bromide). The bath solution contained the following (in mm): 140 NaCl, 5 KCl, 2 CaCl2, 2 MgCl2, 10 HEPES-NaOH, pH 7.4, and 10 glucose. AMPA-receptor-mediated EPSCs (recorded at a holding potential of −70 mV) were isolated pharmacologically with 50 μm d-APV and 100 μm picrotoxin in the bath solution, and NMDA-receptor-mediated EPSCs (recorded at a holding potential of +40 mV) were isolated with 20 μm CNQX, 100 μm picrotoxin, and 15 μm glycine and by raising the MgCl2 concentration to 4 mm. mEPSCs was recorded in 1 μm tetrodotoxin, identified using the pClamp template program, and individually proofread (event threshold of 5 pA). Synaptic responses were triggered by a 1 ms current pulse (900 μA) through a local extracellular electrode (FHC) and recorded in whole-cell voltage-clamp mode using a Multiclamp 700A amplifier (Molecular Devices). Data were digitized at 10 kHz with a 2 kHz low-pass filter and analyzed using Clampfit 9.02 (Molecular Devices) or Igor 5.01 (Wavemetrics).
All data shown are means ± SEMs. Statistical significance was determined by Student's t test comparing various conditions to controls examined in the same experiments, using the number of cells as the statistical n with each experiment performed in at least three independent cultures, and with similar numbers of cells recorded for each condition and in each batch of culture (typically five independent cells per culture and condition). All data acquisition and analyses were performed with anonymized samples, such that the experimenter was unaware of the type of manipulation being analyzed in a sample.
Complexin C-terminal sequence is required for clamping spontaneous exocytosis and for priming of vesicles but not for activation of Ca2+ triggering of exocytosis
We generated complexin-deficient neurons using shRNA-dependent KD of complexin-1 and complexin-2 and used rescue experiments to test the function of various complexin proteins (Maximov et al., 2009; Yang et al., 2010). These KD experiments only targeted complexin-1 and complexin-2 because complexin-3 and complexin-4 are expressed at only very low levels in the cultured cortical neurons used here. In the experiments for the present paper, we examined a series of complexin-1 mutants that were introduced into the KD neurons via the same lentivirus used for the KD and that were designed to test the functional role of the complexin-1 C-terminal sequence that accounts for nearly half its size. The various complexin mutants were robustly expressed in the cultured neurons as examined by immunoblotting (Fig. 1B). Immunocytochemistry showed that complexin is normally present throughout the neuron and is not a synapse-specific protein, consistent with its high expression levels and soluble nature (Fig. 1C; McMahon et al., 1995). As expected, the complexin KD resulted in a loss of complexin immunofluorescence signal, which was restored by expression of wild-type complexin (Fig. 1C).
In the first set of experiments, we compared the relative rescue efficiencies of wild-type complexin-1 (CpxWT) and of C-terminally truncated complexin-1 (Cpx11–86) that lacks almost the entire C-terminal complexin-1 sequence (Fig. 1A). Measurements of spontaneous mini exocytosis, quantified as the frequency of mEPSCs in the presence of 1 μm tetrodotoxin, revealed that C-terminally truncated complexin-1 was unable to reverse the increase in mEPSC frequency induced by the complexin KD, whereas wild-type complexin-1 fully reversed this increase (Fig. 2A,B). No effect on mEPSC amplitude was observed by any condition.
C-terminally truncated complexin-1 did, however, rescue the decrease in action-potential-evoked exocytosis in complexin-deficient neurons, demonstrating that Ca2+ triggering by synaptotagmin-1—which is impaired in complexin-deficient synapses (Maximov et al., 2009)—was normal (Fig. 3A,B).
Next, we measured priming by determining the size of the readily-releasable pool (RRP) of vesicles as the charge transfer mediated by synaptic responses that are induced by an application of hypertonic sucrose (Rosenmund and Stevens, 1996). Surprisingly, C-terminally truncated complexin-1 was unable to rescue the decrease in vesicle priming in complexin-deficient neurons (Fig. 4A–C). This result was unexpected because Ca2+ triggering was normal, suggesting that the Ca2+ triggering and priming functions of complexin-1 are mechanistically distinct. Thus, the complexin-1 C-terminal sequence is essential for its clamping and priming functions but is not required for its Ca2+-triggering function. The C-terminal truncation mutant of complexin is the first mutant that dissociates its priming and Ca2+ functions.
Complexin-3 mediates vesicle priming and Ca2+ triggering of exocytosis but not clamping of spontaneous exocytosis
Complexin-1 and complexin-3 exhibit significant homology throughout their sequences except for their C-terminal region in which complexin-3, similar to complexin-4 but different from complexin-1 and complexin-2, contains an isoprenylation sequence (Fig. 1A; McMahon et al., 1995; Reim et al., 2005). Given this difference, we examined whether complexin-3 could functionally substitute for complexin-1 in complexin-deficient neurons. Previous studies showed that complexin-3 rescues Ca2+-triggered release in complexin-deficient neurons but did not address the priming and clamping functions of complexin (Reim et al., 2005). Thus, in view of the essential role of the complexin-1 C terminus in vesicle clamping and priming, we systematically examined these functions in complexin-deficient neurons that express complexin-3 (Figs. 2⇑–4).
Strikingly, complexin-3 was unable to reverse the increase in mini frequency in complexin-deficient neurons (Fig. 2A–C) but completely rescued their impairment in Ca2+ triggering (Fig. 3). Moreover, complexin-3 rescued the decrease in priming induced by the complexin KD (Fig. 4). Thus, complexin-1 and complexin-3—despite a modest degree of sequence identity (Fig. 1A)—perform similar functions, except that complexin-3 does not clamp spontaneous mini release.
The difference in clamping functions of complexin-1 and complexin-3 localizes to their C-terminal sequences
To determine whether complexin-3 is unable to clamp mini release because of sequence differences with complexin-1 in the accessory α-helix that was shown previously to be essential for clamping (Maximov et al., 2009; Xue et al., 2010) or in the C-terminal region that we here showed to be essential for clamping (Fig. 1), we generated chimeric complexin-1 and complexin-3 proteins in which the three N-terminal regions (the N-terminal activation domain and the two α-helices) are derived from one complexin isoform and the C-terminal sequence from the other complexin isoform (Fig. 5A). Rescue experiments revealed that complexin-1/3 composed of the three N-terminal complexin-1 regions and the C-terminal complexin-3 sequence aggravated the loss of clamping in complexin-deficient neurons, resulting in an even higher mini frequency (Fig. 5B–D). In contrast, complexin-3/1, composed of the three N-terminal complexin-3 regions and the C-terminal complexin-1 sequence, fully reversed the increase in mini frequency, i.e., was fully capable of clamping spontaneous release. Both chimeric complexin proteins fully rescued the decrease in Ca2+-triggered exocytosis in complexin-deficient neurons (Fig. 6). Thus, the inability of complexin-3 to clamp mini release is based on its C-terminal sequence, possibly because the membrane anchoring of complexin-3 via its C-terminal isoprenylation blocks its clamping function by an unknown mechanism.
Point mutations in the C-terminal complexin-1 sequence inhibit its clamping function
To further understand how the C terminus of complexin contributes to the regulation of spontaneous mini release, we searched for point mutations within the C terminus that can alter this function. The complexin C terminus is predicted to be primarily unstructured, except for a possible amphipathic α-helix that is conserved among human, mouse, and C. elegans complexins (Fig. 1A; Seiler et al., 2009). Based on previously published mutants that alter complexin–liposome binding (Seiler et al., 2009), we generated four point mutations in or near the putative amphipathic α-helix: leucine-117 is replaced by either a bulky hydrophobic tryptophan (CpxL117W) or a charged lysine (CpxL117K), valine-120 is replaced by a proline (CpxV120P) to disrupt the α-helix, and isoleucine-101 is replaced with a tryptophan (CpxI101W; Fig. 1A).
mEPSC recordings showed that the CpxI101W and CpxL117W mutations did not impair rescue of the increase in spontaneous mini release in Cpx KD neurons, whereas the CpxL117K and CpxV120P mutations blocked this rescue (Fig. 7A–C). However, all four mutants fully rescued EPSC amplitude (Fig. 8).
Function of the C-terminal complexin sequence in release induced by 10 Hz stimulus trains
To test the role of the complexin-1 C-terminal sequence in release induced by stimulus trains and to clarify whether the clamping function of complexin-1 that requires its C-terminal sequence is also involved in limiting delayed release (a form of asynchronous release that occurs after the last action potential in a stimulus train, when synchronous release has ended; Maximov and Südhof, 2005), we analyzed NMDA-receptor-mediated EPSCs under blockage of AMPA receptors. AMPA-receptor-mediated release induced by stimulus trains cannot be easily analyzed in cultured neurons because of the strong network activity in these cultures (Fig. 6A), which can be blocked when NMDA-receptor-mediated EPSCs are examined in postsynaptically depolarized neurons (Fig. 9A). To ensure that measurements of AMPA- and NMDA-receptor-mediated release produces the same overall conclusions, we first examined the effect of the Cpx KD with rescue by truncated complexin-1 (Cpx11–86) and two complexin-1 point mutants (Cpx1L117W and Cpx1L117K) on release induced by isolated action potentials (Fig. 9A,B). We found that the complexin KD caused the same decrease in the EPSC when measured via NMDA- or AMPA-receptor-mediated responses and that this decrease was similarly rescued by the various complexin-1 mutants (Fig. 9B). Thus, the NMDA receptor response precisely correlates with the AMPA receptor response under the present conditions.
Next, we examined the function of the complexin-1 C-terminal sequences on release induced by a 10 Hz stimulus train (Fig. 9C,D). The synaptic response to the first action potential and to most of the following action potentials is synchronous, although an asynchronous component increasingly contributes to the responses during the train; in addition, all release after the end of the train (the so-called delayed release) is asynchronous (Maximov and Südhof, 2005). Asynchronous and synchronous release are derived from the same RRP and therefore dependent on the size of RRP (Hagler and Goda, 2001; Otsu et al., 2004).
All three C-terminal mutants tested (Cpx1–86, CpxL117W, and CpxL117K) rescued the decreased release during the entire train in complexin-deficient neurons (Fig. 9C,D). A similar pattern was observed for delayed release, whereas the ratio of delayed-to-total release was modestly but significantly increased by the complexin KD and after Cpx1–86 rescue (Fig. 9C,D). It is thought that both total release and delayed release depend on the RRP size, that the total release is primarily dependent on synchronous Ca2+-triggered release, and that delayed release is primarily dependent on asynchronous Ca2+-triggered release. Thus, the increased ratio of delayed-to-total release in both complexin KD and Cpx1–86 rescue neurons is consistent with the notion that the complexin C-terminal sequences clamp asynchronous Ca2+-triggered release but prime overall release, leading to a relatively higher ratio of delayed-to-total release in the stimulus train experiments (Fig. 9D).
In the present study, we have explored the function of the C-terminal sequences of complexins, relatively large sequences that account for nearly half of the complexin protein but nevertheless have remained functionally incompletely characterized. Because complexin-1 and complexin-2, and complexin-3 and complexin-4 form homologous pairs, we tested the effect of deleting or mutating the C-terminal sequences of complexin-1 and complexin-3, asking among others whether complexin-1 and complexin-3 perform similar functions in synaptic exocytosis and whether the functional differences between these complexins localize to their C-terminal regions. We made the following principal observations.
(1) Deletion of the complexin-1 C-terminal sequence revealed that it is essential for both its priming and clamping function but not for its synaptotagmin-activating function, thereby dissociating the two “positive ” functions (synaptotagmin-activating and priming) of complexin.
(2) Complexin-1 and complexin-3 have similar priming and Ca2+-triggering functions but dramatically differ from each other in that only complexin-1 but not complexin-3 mediates synaptic vesicle clamping of spontaneous mini release. This result is consistent with studies performed in Drosophila (Cho et al., 2010).
(3) Swapping the C-terminal sequences of complexin-1 and complexin-3 showed that the complexin-3 C-terminal sequence inactivates the complexin-1 clamping function, whereas the complexin-1 C-terminal sequence enables synaptic clamping by complexin-3. Thus, the functional differences between these two isoforms are entirely based on the differences between their C-terminal sequences.
(4) Two point mutations in the predicted amphipathic C-terminal α-helix of complexin-1 that binds to phospholipids (Seiler et al., 2009) also block the clamping function of complexin-1, suggesting that phospholipid binding by complexin is important for its clamping function.
Viewed together, our data demonstrate that the C-terminal complexin-1 sequence critically contributes to its overall function and show that complexin-3 performs similar functions as complexin-1, except that, at least in our system, complexin-3 does not clamp fusion and that this functional difference localizes to the C-terminal sequence of complexin-1 and complexin-3.
It is surprising that a small molecule such as complexin-1 with only 134 residues mediates three independent functions during synaptic exocytosis (activation of synaptotagmin function, clamping, and priming) and that these functions within the already small complexin molecule selectively depend on distinct short modular sequences (Table 1). The overall design of complexin appears to involve the linkage of different peripheral functional modules that mediate activation of Ca2+ triggering, priming, and clamping to the same central functional module, namely the SNARE-binding α-helix that is essential for all of these functions (Chen et al., 2002; Tang et al., 2006; Maximov et al., 2009).
Our results extend some of the previous studies but are inconsistent with others. Experiments with liposome fusion suggested that the C-terminal complexin-1 but not complexin-3 sequence facilitates membrane fusion and that this function can be transferred from complexin-1 to complexin-3 in complexin-3/1 chimera (Seiler et al., 2009). In our more physiological assay system that monitors synaptic exocytosis in cultured primary neurons, however, complexin-3 activates fusion as well as complexin-1, confirming previous results (Reim et al., 2005). Thus, the liposome results may be particular to liposomes as a reduced system. Moreover, we find that complexin-3 in mouse neurons does not clamp mini release, whereas in Drosophila the same complexin isoform was found to clamp release (Cho et al., 2010), possibly as a result of species differences. Furthermore, our conclusion that the C terminus of complexin-1 is essential for its clamping function agrees well with studies in C. elegans that arrived at the same conclusion based on a genetic approach (Martin et al., 2011). Finally, a long-term unresolved issue about complexin has been the question why KD of complexins in cultured neurons produces an unclamping of mini release (Maximov et al., 2009; Yang et al., 2010), whereas knock-out of complexins in autapses or in slices does not (Reim et al., 2001; Xue et al., 2007; Strenzke et al., 2009), although more recent studies have also discovered a clamping function for complexins in autapses (Xue et al., 2010). This issue remains unresolved, because it is still unclear why the “unclamping ” phenotype is so much less pronounced in autapses from knock-out neurons than in synapses between cultured neurons after KD of complexins. However, there appears to be general agreement that the activating functions of complexin are more important than its clamping functions, and it is possible that the clamping function is more dependent on the experimental context than the activating functions of complexin.
We do not know the precise mechanisms by which the C-terminal complexin sequences contribute to vesicle priming and clamping. A plausible hypothesis, however, is based on the finding that the amphipathic α-helix in the C-terminal complexin-1 sequence binds to lipid bilayers (Malsam et al., 2009; Seiler et al., 2009). It is tempting to speculate that such binding blocks fusion and that mutations in the α-helix reverse this block. However, this hypothesis does not explain the role of the complexin-1 C-terminal sequence in priming and raise additional questions, such as why the accessory α-helix is additionally required for clamping (Maximov et al., 2009). Future studies will have to address these important issues.
This study was supported by NIH Grants MH086403 and NS077906 (T.C.S.). We thank Ira Huryeva and Timothy Galfin and for excellent technical support, and members of the Südhof lab for valuable comments and discussions.
- Correspondence should be addressed to Thomas C. Südhof, Department of Molecular and Cellular Physiology, Stanford University, 265 Campus Drive, Palo Alto, CA 94305-5453.