Gβγ directly modulates vesicle fusion by competing with synaptotagmin for binding to neuronal SNARE proteins embedded in membranes

Gi/o-coupled G protein-coupled receptors can inhibit neurotransmitter release at synapses via multiple mechanisms. In addition to Gβγ-mediated modulation of voltage-gated calcium channels (VGCC), inhibition can also be mediated through the direct interaction of Gβγ subunits with the soluble N-ethylmaleimide attachment protein receptor (SNARE) complex of the vesicle fusion apparatus. Binding studies with soluble SNARE complexes have shown that Gβγ binds to both ternary SNARE complexes, t-SNARE heterodimers, and monomeric SNAREs, competing with synaptotagmin 1(syt1) for binding sites on t-SNARE. However, in secretory cells, Gβγ, SNAREs, and synaptotagmin interact in the lipid environment of a vesicle at the plasma membrane. To approximate this environment, we show that fluorescently labeled Gβγ interacts specifically with lipid-embedded t-SNAREs consisting of full-length syntaxin 1 and SNAP-25B at the membrane, as measured by fluorescence polarization. Fluorescently labeled syt1 undergoes competition with Gβγ for SNARE-binding sites in lipid environments. Mutant Gβγ subunits that were previously shown to be more efficacious at inhibiting Ca2+-triggered exocytotic release than wild-type Gβγ were also shown to bind SNAREs at a higher affinity than wild type in a lipid environment. These mutant Gβγ subunits were unable to inhibit VGCC currents. Specific peptides corresponding to regions on Gβ and Gγ shown to be important for the interaction disrupt the interaction in a concentration-dependent manner. In in vitro fusion assays using full-length t- and v-SNAREs embedded in liposomes, Gβγ inhibited Ca2+/synaptotagmin-dependent fusion. Together, these studies demonstrate the importance of these regions for the Gβγ-SNARE interaction and show that the target of Gβγ, downstream of VGCC, is the membrane-embedded SNARE complex.

target membrane. One group of calcium sensors of particular interest is the synaptotagmins. In neurons, it is currently believed that a major calcium sensor for exocytosis is synaptotagmin 1, containing an N-terminal intraluminal domain, a transmembrane domain, and a tandem set of calcium-binding C2-domains termed C2A and C2B (6 -8). Whereas synaptotagmin can bind syntaxin1A and SNAP25 with low affinity in the absence of calcium, its affinity increases markedly in the presence of elevated calcium levels (9 -11). Synaptotagmin 1 binds lipids in addition to SNAREs, with both phosphatidylserine and PIP 2 essential for the function of the full-length protein (9,(12)(13)(14). The lipid-binding and SNARE-binding functionalities of synaptotagmin are both essential for its role in mediating fast synchronous release at the synapse.
Activation of inhibitory G i/o -coupled GPCRs causes inhibition of exocytosis through several mechanisms. GPCRs signal through heterotrimeric G proteins, with both the guanine nucleotide-binding G␣ subunit and the G␤␥ subunit implicated in discrete signaling pathways (1,15). Although the beststudied mechanism involves the inhibition of adenylyl cyclase by G␣ i , the G␤␥ heterodimer is also capable of inhibiting exocytosis in several ways. A large number of independent groups have shown that G␤␥ can inhibit calcium currents through direct binding to Ca V 2 (N-P/Q-and R-type) voltage-gated calcium channels (16 -19). This mechanism of inhibition is found at a large number of synapses, including certain subtypes of GPCRs thought to work entirely through it (20 -22). Inhibition of exocytosis by G i/o -coupled GPCRs downstream of calcium entry is also known to occur (23)(24)(25)(26). It has been shown by multiple independent groups that G␤␥ binds to the SNARE complex to inhibit exocytosis downstream of calcium entry (27)(28)(29)(30)(31). Electrophysiological and in vitro protein-binding studies have shown that the C terminus of SNAP-25 is a critical binding site for this interaction (30 -32). We propose that G␤␥ competes with synaptotagmin 1 for binding sites on SNAP-25, disrupting the ability of synaptotagmin 1 to promote vesicle fusion in response to elevated intracellular calcium (28 -30). This hypothesis is in line with current crystallographic structures of the synaptotagmin-SNARE interaction, featuring significant overlap between the known synaptotagmin 1-binding site and key G␤␥-binding residues on the second helix of SNAP-25 that would be expected to produce a steric clash (30,33,34). Each inhibitory GPCR may signal via one or more of these mechanisms to achieve precise spatial and temporal control over neurotransmitter release; for example, the GABA B receptor in the CA1 axons of the hippocampus signals via modulation of calcium currents, whereas the 5-HT 1B receptor signals only downstream of calcium entry (20). This mechanism downstream of calcium entry has been reported to occur at a variety of inhibitory G i/o -coupled GPCRs in a variety of secretory cell types (35)(36)(37)(38)(39)(40)(41), demonstrating its importance for the regulation of exocytosis.
Prior protein biochemical studies elucidating the interplay between G␤␥, synaptotagmin, and SNAREs have been conducted in aqueous environments in the presence of detergent, and it has never been determined whether G␤␥ can bind to SNAREs and directly inhibit vesicle fusion in lipid bilayers (28 -31). In addition, the molecular requirements for the binding of SNARE proteins to G␤␥ are not well understood. It has been demonstrated that G␤ 1 ␥ 2 has a higher affinity than G␤ 1 ␥ 1 for t-SNARE complexes and is more efficacious at inhibiting fusion in permeabilized PC12 cells (28), but the regions on G␥ 2 responsible for the 20-fold tighter interaction are not known. Although a number of residues on SNARE have been shown to be important for G␤␥ binding (30,31), the individual residues on G␤␥ implicated in the SNARE interaction have yet to be identified.
Here, we have further explored the complex interplay between G␤␥ and synaptotagmin 1 for the regulation of SNARE-driven fusion. We expand on previous differences noted between G␤␥ isoforms in this mechanism and highlight the importance of a single residue, Trp-332, on G␤, for the interaction of G␤␥ not only with SNAREs but also voltagegated calcium channels. Furthermore, we examine the ability of G␤␥ to compete with synaptotagmin for association with SNARE-containing liposomes as well as demonstrate a role for this inhibition as it relates to fusion in vitro.

Results
Previous studies of G␤␥-SNARE interactions used recombinant soluble SNARE complexes in aqueous solution to show that G␤␥ binds to ternary SNARE complexes, t-SNARE heterodimers, and the monomeric SNARE proteins SNAP25, syntaxin1A, and VAMP2 (28 -30). To examine binding of G␤␥ to full-length t-SNARE complexes embedded in lipid bilayers, we developed an assay using total internal reflection (TIRF) fluorescence intensity and anisotropy (Fig. 1A). Purified G␤␥ subunits fluorescently labeled at primary amine residues with Alexa Fluor 488 N-hydroxysuccinimide ester were applied from a pipette over a t-SNARE-containing (syntaxin1A and SNAP-25) bilayer on a coverslip illuminated with TIRF or epifluorescence. During pulses of G␤ 1 ␥ 1 , there was a small increase in TIRF fluorescence measured with a photomultiplier in the absence of t-SNARE complexes; if t-SNAREs were present, this increase was an order of magnitude larger (Fig. 1B). To confirm that this interaction represented binding of G␤␥ to a target in the lipid bilayer, anisotropy of the fluorescence TIRF signal was measured (Fig. 1B). Laser TIRF excitation was polarized, and emission polarization was detected parallel and orthogonal to this excitation polarization. Immediately after G␤␥ pressure ejection (which lasted for 1 s; Fig. 1B, gray bar) no increase in anisotropy of the fluorescence signal was observed if no t-SNARE was present, whereas in the presence of t-SNAREs there was a large increase in anisotropy (Fig. 1B, inset). The difference between the two conditions was significant.
The amplitude of the TIRF fluorescence response to G␤␥ pressure application over the bilayer was used to quantify t-SNARE interactions of various G␤␥ subtypes. G␤␥ concentration at the lipid bilayer was calculated using epifluorescence of G␤␥ at known concentrations in the solution above the lipid bilayer and comparing that value to epifluorescence of pressure-ejected G␤␥ (Fig, 1Ci). Dosing to saturation was obtained by increasing concentrations of G␤␥ in the pressure ejection pipette. A concentration-response curve was constructed comparing G␤ 1 ␥ 1 to G␤ 1 ␥ 2 . We observed a 14-fold difference in affinity between G␤ 1 ␥ 1 and G␤ 1 ␥ 2 isoforms. These results con-firmed the results obtained for binding in aqueous solution (28,29). G␤␥ subunits containing Ala mutations of two residues on the G␣-binding surface of G␤, Lys-78 and Trp-332, inhibited exocytosis at a significantly higher potency than wild type (28). To determine if this was due to increased affinity for t-SNAREs, we tested this mutant G␤ 1 ␥ 2 subunit in the TIRF assay. G␤ 1 K78A/W332A ␥ 2 (Fig. 1D, red curve) exhibited a significantly higher 1.7-fold increase in affinity for t-SNARE complexes compared with wild-type G␤ 1 ␥ 2 . Consistent with saturation of binding, leaving excess unbound G␤␥, anisotropy of these fluorescence signals decreased as the pressure-ejected G␤␥ concentrations were increased and the intensity signal saturated (data not shown).

. Supported lipid bilayer assay of G␤␥-t-SNARE interactions in membranes.
A, fluorescently labeled G␤␥ was assayed over a lipid bilayer using TIRF to isolate the fluorescence field to a band of ϳ100 nm above a coverslip. Anisotropy of the fluorescence emission was also assayed to confirm a G␤␥ interaction occurred. B, G␤ 1 ␥ 1 pressure ejected in a 1-s pulse (gray bar) over the bilayer containing t-SNAREs (produced from liposomes containing 130 t-SNAREs/vesicle) evoked a sustained increase in TIRF fluorescence. A similar pressure application over a bilayer lacking t-SNAREs showed a much smaller signal. Inset, increase in anisotropy after the pressure application ceased as flow in the chamber removed aqueous G␤␥ but no increase in anisotropy after application over a bilayer lacking t-SNAREs (response 10.5 Ϯ 1.2%). The experiment was repeated 23 times. The mean change in anisotropy post-injection was r ϭ 0.13 Ϯ 0.04 for 8.8 mM G␤ 1 ␥ 1 for membranes containing SNAREs. For membranes lacking SNARE, the magnitude of the change was Dr ϭ 0.016 Ϯ 0.004, with a significance value of p ϭ 0.02 between the two conditions; Student's t test. C, the amplitude of fluorescence transients were measured first under epifluorescence to calibrate G␤␥ concentrations applied from pressure ejection pipettes. Ci, output gave a signal proportional to the concentration of G␤␥ applied, and the signal ceased as the pressure application stopped. Cii, in TIRF, pressure application of G␤␥ revealed a longer-lived response. This signal was normalized to the pressure applied concentration of G␤␥ by dividing by the signal in Ci. D, dose-response curves were constructed from this normalized data (e.g. Cii) plotted against the dose calculated from Ci. G␤ 1 ␥ 1 is depicted in blue (EC 50 ϭ 2.08 Ϯ 0.023 M, n ϭ 4 technical replicates), G␤ 1 ␥ 2 is depicted in green (EC 50 ϭ 147 Ϯ 71 nM, n ϭ 5), and G␤ 1 K78A/W332A ␥ 2 is depicted in red (EC 50 ϭ 86 Ϯ 50 nM, n ϭ 5, p ϭ 0.045, one-way ANOVA with Tukey's HSD test for each curve).
selectively on the lone Cys residue present in the sequence (Cys-277). Alexa Fluor 488-synaptotagmin 1 was applied to coverslips coated with t-SNARE complexes consisting of purified recombinant syntaxin1A and SNAP-25 embedded in lipid membranes illuminated by TIRF. The anisotropy of the fluorescence TIRF signal produced via the binding of fluorescent synaptotagmin 1 to t-SNARE was recorded ( Fig. 2A). Application of Ca 2ϩ increased the anisotropy of the fluorescent TIRF signal in a saturable manner produced by application of fluorescent synaptotagmin 1 to the t-SNARE containing bilayer (Fig. 2B). The rise in anisotropy produced via the binding of synaptotagmin 1 to t-SNARE complexes could be reversed in a concentration-dependent manner by purified G␤ 1 ␥ 1 (Fig. 2C).
To investigate whether the ability of G␤ 1 K78A/W332A ␥ 2 and its corresponding single-Ala mutants to inhibit voltage-gated calcium currents was correspondingly enhanced, we conducted electrophysiological studies in HEK cells expressing Ca V 2.2 (N-type) calcium channels along with wild-type G␤ 1 ␥ 2 , G␤ 1 K78A/W332A ␥ 2 , G␤ 1 K78A ␥ 2 , or G␤ W332A ␥ 2 (Fig. 3). G␤␥mediated inhibition of Ca V 2.2 channels is voltage-dependent; the inhibition is less pronounced at more depolarized test potentials and is transiently reversed by a conditioning prepulse to very depolarized potentials (42). This reversal (also called pre-pulse facilitation) is thought to reflect transient dissociation of G␤␥ from the channel at the depolarized membrane potential. The magnitude of pre-pulse facilitation can, therefore, be used to quantify the extent of G␤␥-mediated inhibition of I Ca . We used a double-pulse protocol (Fig. 3A) in which cells voltage-clamped at Ϫ80 mV were stimulated by two identical test pulses (P1 and P2) to various potentials (Ϫ10, 0, ϩ10, and ϩ20 mV). The second test pulse (P2) was preceded by a conditioning prepulse to ϩ120 mV to maximally reverse any G␤␥-mediated inhibition. In control conditions lacking G␤␥ overexpression, there was a slight prepulse facilitation due to the presence of endogenous G␤␥ (43). Robust prepulse facilitation and inhibition of I Ca was observed with co-expression of G␤␥. The magnitude of prepulse facilitation was diminished at more depolarized test potentials (Fig. 3B), consistent with the known reduction of G␤␥-mediated inhibition at depolarized test potentials (42). Compared with cells expressing wild-type G␤␥, prepulse facilitation was only modestly affected in cells expressing the K78A mutant but was abolished in cells expressing the W332A mutant (Fig. 3B). The K78A/W332A double mutant produced slight prepulse facilitation of I Ca that was not statistically different from control cells (Fig.  3B). These studies, in tandem with those conducted in Fig. 1, suggest that the W332A mutation and the K78A/W332A double mutant dramatically reduce the ability of G␤ 1 ␥ 2 to interact with Ca V 2.2 channels and that the increased potency of the G␤ 1 K78A/W332A ␥ 2 double mutant in the cracked PC12 cell assay is due to the enhanced interaction of G␤ 1 K78A/W332A ␥ 2 with SNAREs.
To identify important regions on G␤ and G␥ for the binding of SNAP25, we utilized a peptide-competition approach to examine whether peptides derived from the primary sequence of G␤ 1 , G␥ 1 , or G␥ 2 could disrupt the interaction between fulllength G␤␥ and SNAP25. We measured the ability of peptides to disrupt the interaction between full-length G␤ 1 ␥ 2 with SNAP25 using the Alphascreen competition-binding assay (31). G␤ 1 peptide 328 -337, which contains Trp-332, is a reasonably potent inhibitor of G␤ 1 ␥ 2 SNAP25 interaction (Fig.  4A); however, peptides corresponding to G␤ 1 peptides 86 -98 and 243-251 did not inhibit the interaction. Peptides from G␥ subunits could also compete with G␤ 1 ␥ 2 -SNAP-25 interactions; a peptide corresponding to the N-terminal 2-24 residues of G␥ 2 was much more potent than a corresponding peptide on G␥ 1 (Fig. 4, B and C) . Peptides corresponding to residues 8 -25 on G␥ 1 inhibited the interaction with equivalent potency to residues 9 -28 on G␥ 2 . In contrast, no inhibition was observed with peptides corresponding to residues 32-48 on G␥ 1 or 29-45 on G␥ 2 . These studies suggest that the region 328 -337, including Trp-332, on G␤ 1 is important for SNAP-25 binding, and the N terminus of G␥ 2 is responsible for G␤ 1 ␥ 2 's increased affinity for t-SNARE heterodimers. In Fig. 5, regions on G␤ 1 , G␥ 1 , and G␥ 2 shown to be involved in interactions with SNAP-25 are illustrated in red in the 3-dimensional X-ray crystallographic structures of the G␤␥ subunit (44 -46), whereas individual residues Lys-78 and Trp-332 are depicted in blue.
To determine whether G␤␥ subunits can inhibit SNAREcatalyzed liposome fusion in the presence of synaptotagmin-1, we performed reconstituted fusion assays similar to those conducted in previous studies (47)(48)(49)(50). Liposomes containing v-SNAREs were prepared using recombinant full-length VAMP2 and a FRET donor-acceptor pair consisting of NBDphosphatidylethanolamine (PE) and rhodamine-PE. t-SNARE liposomes were prepared containing full-length recombinant rat syntaxin 1A and SNAP-25B heterodimers. As the quenched FRET-pair-containing liposomes fuse with the unlabeled liposomes, the concentration of the FRET acceptor rhodamine in the liposome is lowered, resulting in an increase in NBD donor fluorescence, as measured with an excitation wavelength of 460 nm and an emission wavelength of 538 nm (Fig. 6A). The cytoplasmic domain of synaptotagmin 1 (C2AB, 10 M; Ref. 50) stimulated fusion in the presence of 1 mM Ca 2ϩ (Fig. 6B, black lines). The extent and maximum rate of fusion obtained in the presence of both Ca 2ϩ and synaptotagmin 1 was substantially greater than in the absence of either component (data not shown). Purified bovine G␤ 1 ␥ 1 as well as recombinant Histagged G␤ 1 ␥ 2 purified from SF9 cells inhibited Ca 2ϩ -synaptotagmin-1 triggered fusion in a concentration-dependent manner with reductions in both the slope and maximal levels of fusion (Fig. 6B, blue and green lines). The potency for G␤ 1 ␥ 2 inhibition of Ca 2ϩ and SNARE-dependent synaptotagmin 1-driven lipid mixing was 14.5-fold higher than for G␤ 1 ␥ 1 (Fig.  6C). Maximal concentrations of G␤ 1 ␥ 1 reduced fusion to a baseline level, comparable with conditions lacking synaptotag-min 1 or Ca 2ϩ . In the absence of synaptotagmin 1, G␤ 1 ␥ 1 did not additionally inhibit fusion (Fig. 6D), indicating that the effect of G␤␥ is to inhibit synaptotagmin 1-stimulated fusion.
To investigate whether G␤␥ had higher potency at alternative concentrations of synaptotagmin 1, we performed reconstituted fusion assays similar to those in Fig. 6C with a lower concentration of synaptotagmin 1 (3.16 M) at 1 mM Ca 2ϩ . G␤ 1 ␥ 1 was found to inhibit fusion 1.8-fold more potently in a concentration-dependent manner at this reduced concentration of synaptotagmin 1 (Fig. 6E).

Discussion
Multiple independent groups have reported that the G␤␥-SNARE interaction is one of several important mechanisms through which G i/o -coupled GPCRs inhibit exocytosis (35)(36)(37)(38) along with inhibition of Ca 2ϩ influx by G␤␥ through voltagegated calcium channels. Mutagenesis studies have provided some of the strongest arguments in favor of this hypothesis; mutant forms of SNAP-25 that are unable to efficiently bind G␤␥ are also unable to support G i/o -coupled GPCR-mediated inhibition of exocytosis (30), and partial loss-of-function SNAP-25 mutants, with decreased G␤␥ binding, show concomitant partially reduced G i/o -coupled GPCR-mediated inhibition in the same populations of neurons (31). The heterogeneous nature of cellular systems prevents ruling out the possibility that an alternative effector for exocytosis that uses the same residues in SNAP-25 is similarly perturbed. Here, we provide strong evidence against this idea by showing that G␤␥ inhibits synaptotagmin-1-regulated, SNARE-catalyzed liposome fusion in a purified system lacking any other proteinaceous components. A concentration dependence for G␤ 1 ␥ 2 Figure 3. G␤ 1 mutants that bind better to SNARE are less potent at inhibiting voltage-gated calcium channel currents. HEK cells were transiently transfected with Ca V 2.2 channels alone (control) or Ca V 2.2 channels with wild-type G␤ 1 , G␤ 1 K78A (K78A), G␤ 1 W332A (W332A), or G␤ 1 K78A/W332A (K78A/W332A) in addition to G␥2. A, representative whole cell currents (I Ca ) recorded from a cell expressing Ca V 2.2 channels and wild-type G␤ 1 ␥ 2 . The upper panel shows the voltage protocol consisting of two identical 20-ms steps to 0 mV (P1 and P2) and the conditioning prepulse (50 ms step to ϩ120 mV) before P2. The lower panel shows I Ca . Note the current after the prepulse (P2) was much larger than before the prepulse (P1) due to reversal of the tonic G␤␥-mediated inhibition. Also note the slow activation kinetics in P1, another characteristic of G␤␥-mediated inhibition that is reversed by the prepulse. The inset (bottom) shows superimposed P1 and P2 currents (normalized to the peak of P2) from a control cell and cells expressing the indicated G␤␥. B, prepulse facilitation was determined at several test potentials (potential of P1 and P2 step) in control cells (no exogenous G␤␥) (n ϭ 7) and cells transfected with G␤␥ variants. Box and whisker plots denote the range of the individual data points (box denotes 25%, median, 75%; whiskers denote standard deviation of mean). At a test potential of Ϫ10 mV there was statistically significant prepulse facilitation produced by wild-type G␤␥ (n ϭ 12, p Ͻ 0.001) and K78A (n ϭ 5, p Ͻ 0.05) but not by W332A (n ϭ 5) or K78A/W332A (n ϭ 13) (one-way ANOVA with Dunnett's multiple pairwise comparison to control cells); at 0-mV test potential there was significant prepulse facilitation produced by wild type (n ϭ 12, p Ͻ 0.001) and K78A (n ϭ 5, p Ͻ 0.05) but not by W332A (n ϭ 5) or K78A/W332A (n ϭ 13) (one-way ANOVA with Dunnett's multiple pairwise comparison to control); at ϩ10-mV test potential only wild-type G␤␥ produced significant prepulse facilitation (n ϭ 12, p Ͻ 0.01) but not K78A (n ϭ 5), W332A (n ϭ 5), or K78A/W332A (n ϭ 13) (one-way ANOVA with Dunnett's multiple pairwise comparison to control); at ϩ20-mV test potential only wild-type G␤␥ produced significant prepulse facilitation (n ϭ 12, p Ͻ 0.01) but not K78A (n ϭ 5), W332A (n ϭ 5), or K78A/W332A (n ϭ 13) (one-way ANOVA with Dunnett's multiple pairwise comparison to control). ns denotes not significantly different; *, p Ͻ 0.05; **, p Ͻ 0.01; ***, p Ͻ 0.001; one-way ANOVA followed by Dunnett's multiple pairwise comparison to control).
inhibition was observed, with an IC 50 of 157 nM. Other components, such as voltage-gated calcium channels and adenylyl cyclases, are entirely absent. From these and prior studies, it can be concluded that G␤ 1 ␥ 1 and G␤ 1 ␥ 2 both bind SNAREs, including monomeric syntaxin1A, SNAP25, t-SNARE heterodimers, and/or VAMP2, incorporated into lipid bilayers, as determined here by fluorescence polarization or previously in aqueous solution (28,29), and inhibits membrane fusion. The inhibitory effect of G␤␥ on liposome fusion is only observed in the presence of Ca 2ϩ -synaptotagmin 1. The binding sites of G␤␥ on SNAP-25 and synaptotagmin 1 on ternary SNAREs are known, encompassing overlapping but not identical regions (10,30,33). It has previously been shown that G␤␥ competes with synaptotagmin 1 for binding sites on t-SNARE heterodimers and SNAP-25 in solution (28,29). From these prior studies in tandem with the competition-binding assay studies conducted in Fig. 2 and our data showing that G␤␥ has minimal effects on fusion in the absence of synaptotagmin 1 (Fig. 6D), we conclude that the inhibition of liposome fusion occurs directly via G␤␥ binding to membrane-embedded SNAREs in a competitive manner with synaptotagmin 1. This observation corroborates and extends our previously published work (28 -32). Despite this, it is clear that inhibition of voltage-gated calcium currents is an important mechanism through which G␤␥ inhibits exocytosis for some, but not all, G i/o -coupled GPCRs (20,42).
Although a considerable amount is known regarding the SNAP-25 residues that mediate binding to G␤␥ (29 -32), comparatively little is known concerning which residues of G␤␥ mediate binding to SNARE proteins. It has been hypothesized that the SNARE-binding residues are located on the G␣-binding surface of G␤, as heterotrimeric G␣␤␥ is incapable of binding SNAREs (29). Here, we validate and expand upon those results by demonstrating the role of Trp-332 of G␤ in interaction with the SNARE complex and modulation of voltage-dependent calcium channels. This is a key G␣-binding residue required for heterotrimer assembly and receptor interaction as well as adenylyl cyclase activation (51). Strikingly, however, Ala mutation of this residue increases the potency of G␤␥ inhibition of fusion in permeabilized PC12 cells (28), while abrogating modulation of voltage-dependent calcium channels. Strong linkage between this observation and the G␤␥-SNARE interaction has been generated here with the K78A/W332A double mutant of G␤ 1 ␥ 2 -binding lipid-embedded t-SNAREs with significantly higher affinity than wild-type G␤ 1 ␥ 2 . Peptide binding studies expanded upon these studies, with a peptide G␤ 1 328 - 337 corresponding to this region on G␤ 1 -binding t-SNARE. Importantly, this peptide also inhibits the interaction of G␤ 1 ␥ 2 with SNAP25 in a concentration-dependent manner, whereas peptides corresponding to other regions of G␤ 1 do not.
Two key regions of interaction were identified on G␥. In both this article (Fig. 1) and previous work (28), G␤ 1 ␥ 2 was shown to bind to t-SNARE complexes consisting of Stx1A and SNAP25 20-fold more tightly than G␤ 1 ␥ 1 . In Fig. 4, peptide competition data suggests that there is a SNAP-25-binding site in residues 9 -28 of G␥ 1 and 8 -25 of G␥ 2 , but a larger 2-24-residue peptide is 4.66-fold more potent than the 8 -25-residue peptide. This suggests that the N-terminal residues 2-7 of G␥ 2 may be partially responsible for the increased ability of G␤ 1 ␥ 2 to bind t-SNARE.
The reduced ability of W332A mutants to inhibit voltagegated calcium channels further supports the role of the G␤␥-SNARE binding in inhibition of exocytosis. Our results echo prior studies showing the W332A mutant dramatically reduced the G␤␥-mediated inhibition of I Ca (43, 51) (Fig. 3). The double mutant K78A/W332A produced only slight inhibition of I Ca , with facilitation ratios not significantly different from control (no exogenous G␤␥) at any of the voltages tested (Fig. 3B). The K78A mutant has also been reported to reduce the inhibition of I Ca (51), although we found only a modest effect (Fig. 3B). That said, the inhibition produced by K78A was more sensitive to membrane potential; it produced significant inhibition (prepulse facilitation) at hyperpolarized test potentials but not at the more depolarized test pulses (Fig. 3B). This might suggest a modest reduction in binding affinity with K78A, but we did not investigate this further in the present study.
The bulky tryptophan residue Trp-332 has previously been shown to be inhibitory to certain classes of G␤ effectors, with enhanced ␤ARK interaction previously reported for W332A mutants (51). In the X-ray crystallographic structure of ␤ARK in complex with G␤ 1 ␥ 2 (46,52), Trp-332 is in close proximity to the ␣-helical C terminus of ␤ARK. Potentially, steric clashes may occur between Trp-332 and Lys-K663 or Asn-666 of ␤ARK, which may be relieved in the W332A mutant. The enhanced binding affinity of W332A-containing mutant G␤␥ for SNARE in in vitro binding assays supports a hypothesis that the side chain of Trp-332 and/or Lys-78 inhibits the interaction through steric clash and/or electrostatic repulsion with one of the Lys or Arg residues on the SNARE complex, such as Arg-135, Arg-136, Arg-142, Arg-161, Arg-198, or Lys-201 of SNAP25 (30). In line with this hypothesis, a peptide corresponding to residues 548 -671 of ␤ARK blocks G␤␥-SNAREmediated inhibition of exocytosis (27). Charge-reversal of G␤␥binding residues from Lys/Arg to Glu at the C terminus of SNAP25 is far more destructive than neutral mutation to Ala (30,31), implying that negatively charged residues on G␤ or G␥ in addition to Lys-78/Trp-332 may contribute to the interaction.
Both direct binding assays and cell-based studies highlight the contribution of G␥ to the interaction, as G␤ 1 ␥ 2 binds SNAREs and inhibits exocytosis with an order of magnitude higher affinity than G␤ 1 ␥ 1 (28) despite the presence of identical G␤ in each complex. Two explanations for this phenomenon were hypothesized: specific residues on G␥ 2 have higher affinity for SNARE than those on G␥ 1 , or the C-terminal geranylgeranyl modification of G␥ 2 (53,54) contributes to the interaction more so than the farnesyl modification of G␥ 1 . We have localized the SNARE-binding residues to the N terminus of G␥ 2 (Fig.  5); thus, our data are supportive of the former hypothesis. Despite these studies, we have limited insights as to the specific binding mode of SNARE upon G␤␥. It is clear that X-ray crystallographic studies of the complete G␤␥-SNARE complex would yield tremendous insights as to the specific interplay of individual residues for the interaction.
In summary, we show that G␤␥ subunits bind SNARE complexes in a lipid environment and inhibit fusion in a system containing only G␤␥, SNAREs, synaptotagmin 1, calcium, and lipids. We highlighted residues and regions of importance on G␤ and G␥ for SNARE binding. These studies provide further evidence for the G␤␥-SNARE hypothesis and highlight its importance in lipid bilayer-containing environments that more closely approximate the environment of the presynaptic active zone.

Plasmids
The open reading frames for the SNARE component proteins were subcloned into the glutathione S-transferase (GST) fusion vector, pGEX6p1 (GE Healthcare), for expression in bacteria. The dual-expression vector pRSF-Duet1 with subcloned N-terminal His-tagged full-length rat SNAP-25 and full-length rat syntaxin1A was previously described (55). Full-length bovine G␤1 and His-tagged G␥2 were incorporated in Sf9 vectors, described previously (51). The plasmid of a GST fusion with rat synaptotagmin1 (56) and the plasmid for the highaffinity G␤ 1 ␥ 2 (K78A/W332A) were previously described (30). Voltage-gated calcium channel plasmids were as follows: bovine N-type Ca V 2.2 (GenBank TM number NM174632; rat brain ␤ 2a (GenBank TM number M80545); rat ␣ 2 ␦ (GenBank TM number M86621).

Chemicals
Unless otherwise specified, chemicals were obtained from Sigma. Accudenz A.G. Cell Separation Media was obtained from Accurate Chemical & Scientific Co.

Preparation and purification of recombinant proteins in Escherichia coli
Recombinant bacterially expressed syntaxin1A and His 6 -SNAP-25 were expressed in pRSF-Duet vector was trans- formed into E. coli strain BL21. His 6 VAMP2 was expressed in the plasmid pTW2. After the initial starter culture overnight in LB media with 50 g/ml kanamycin (or ampicillin for pTW2), 4 liters of LB with antibiotic were inoculated with the starter culture and placed on a shaker at 37°C until an A 600 of 0.8 was obtained. Protein expression was induced with 0.4 mM isopropyl ␤-D-thiogalactoside for an additional 4 h at 37°C. Bacterial cultures were pelleted and then resuspended in 10 -15 ml of resuspension buffer (25 mM HEPES-KOH, pH 8.0, 400 mM KCl, 10 mM imidazole, and 5 mM ␤-mercaptoethanol). After adding protease inhibitors (aprotinin, leupeptin, pepstatin, and phenylmethylsulfonyl fluoride, cells were sonicated with a sonic dismembrator at 4°C for 2 cycles of 45 s with 45 s of rest in between cycles. After sonication, 3 ml of 25% Triton X-100 was added to each tube and then incubated at 4°C on a rotator for 3-4 h. These samples were then centrifuged to remove the insoluble material (25 min in a Beckman SW-34 rotor at 26,000 rpm). The supernatant was then purified via bulk affinity chromatography by mixing at 4°C overnight with cobalt resin (Talon), prewashed, and equilibrated in resuspension buffer with the addition of 1% Triton X-100. The next day the beads were pelleted and washed twice with 5 volumes of OG wash buffer (25 mM HEPES-KOH, pH 8.0, 400 mM KCl, 20 mM imidazole, 5 mM ␤-mercaptoethanol, and 1% n-octylglucoside). After the second pelleting, the beads were then mixed at 4°C on a rotator with 5 volumes of elution buffer (25 mM HEPES-KOH, pH 8.0, 400 mM KCl, 200 mM imidazole, 5 mM ␤-mercaptoethanol, 1% n-octylglucoside, and 10% glycerol) for 2 h. The samples were pelleted to remove beads from the eluted protein. The protein concentrations were approximately determined with the Pierce 660 nM Protein Assay (#22660, Thermo Scientific) and then confirmed for purity and concentration by SDS/PAGE analysis and comparison to a bovine serum albumin standard curve from the same gel. For synaptotagmin 1 purification, recombinant synaptotagmin 1 C2AB was prepared as a GST fusion according to previously published methods (47), substituting the imidazole for recombinant human rhinovirus 3C protease to liberate the synaptotagmin 1 from the GST tag.

G␤␥ purification
G␤ 1 ␥ 1 was purified from bovine retina as described previously (57). Recombinant G␤ 1 ␥ 2 was expressed in Sf9 cells and purified according to the method of Kozasa and Gilman (58) with the following exceptions; frozen Sf9 cell pellets were lysed via sonication with a duty cycle of 10 s followed by a resting period of 20 s for 3 min at 30% intensity at 0°C. G␤ 1 ⅐His 6 -␥ 2 dimers were affinity-purified twice from detergent-solubilized crude cell membrane using Talon cobalt resin (Clontech) followed by 3 rounds of dialysis for a minimum of 2 h in the following buffer: 20 mM Na-HEPES, pH 8.0, 100 mM NaCl, 10 mM 2-mercaptoethanol, 0.8% n-octylglucoside, 10% glycerol.
For v-SNARE-containing liposomes, 1.5% 1.5% N-(7-nitro-2-1,3-benzoxadiazol-4-yl)-1,2-dipalmitoyl phosphatidylethanolamine (NBD-PE) and 1.5% N-(lissamine rhodamine B sulfonyl)-1,2-dipalmitoyl phosphatidylethanolamine (rhodamine-PE) were added to a mixture of 55% POPC/15% DOPS/27% POPC before drying down and 0.095 mg of recombinant His 6tagged VAMP2 was added in elution buffer. These mixtures were agitated on a tabletop vortex until the lipids were dissolved (10 -15 min). After lipids were dissolved, 2 volumes of reconstitution buffer (25 mM HEPES-KOH, pH 7.8, 100 mM KCl, 10% glycerol, 1 mM DTT) was added dropwise, and the sample was vortexed gently for another 10 min. After 10 min, the sample was transferred to dialysis tubing (10,000 molecular weight cutoff, ThermoScientific) and dialyzed overnight with two dialysis buffer exchanges after 6 h. The dialysis buffer contained 25 mM HEPES-KOH, pH 7.8, 100 mM KCl, 10% glycerol, and 1 mM DTT in two 4-liter volumes. The next morning the samples were retrieved and mixed in equal volumes with a solution of 80% iohexol (Accudenz, Accurate Chemical Co.) in 25 mM HEPES-KOH, pH 7.8, 100 mM KCl, 10% glycerol, and 1 mM DTT. Gradients were assembled in thin-walled centrifuge tubes (#344057, Beckman Coulter) for a SW-55 swinging bucket rotor (Beckman Coulter) with 1.5 ml of the lipid/Accudenz mixture at the bottom, 1.5 ml of 30% Accudenz, and finally 450 l of 0% Accudenz on the top. Liposomes were floated by centrifugation at 55,000 rpm for 2 h at 4°C with minimal brake. Liposomes can be visualized as a thin uniform layer of opacity at the 0 -30% interface of the gradient. Approximately 0.4 ml of liposomes were removed from each layer by direct puncture with a 27-gauge needle at the 0 -30% Accudenz interface. All tubes from each preparation were mixed, aliquoted, and flashfrozen with an ethanol/dry ice bath. Liposomes were stored at Ϫ80°C. Lipid concentrations and recovery rates were quantified using the Beer-Lambert law from NBD-PE absorbance values at 460 nm from liposomes containing NBD-PE that were diluted with dodecyl maltoside to 0.5%. SNARE protein concentrations in liposomes were determined by Coomassie Brilliant Blue R-250 staining of SDS-PAGE gels containing a standard curve of known concentrations of bovine serum albumin (Thermo Scientific) followed by densitometric analysis using the Fiji distribution of ImageJ software (59,60). SNARE copy number was determined according to previously published methods (47).

Membrane TIRF imaging
Bilayers were prepared from 55% phosphatidylcholine, 15% phosphatidylserine, 29% PE, and 1% DiO liposomes with or without t-SNARE complexes as above except DiO (1%) was added. Coverslips were cleaned by soaking in 2% Hellmanex II solution, sonicated at 50°C, rinsed in 18 megohms of deionized water and in 100% ethanol and stored in ethanol. For recording, coverslips were rinsed, dried under filtered compressed air, and placed in a microscope-recording chamber. 650 l of HEPES-KCl was added to 100 l of proteoliposome mix, and then 750 l of HEPES-KCl with 10 mM CaCl 2 was added. 25 l of this mixture was placed in the coverslip chamber and sat for 1 h to allow a lipid bilayer to settle on the coverslip. This lipid bilayer was washed with a superfusate of 290 mM HEPES-KCl and 10 mM EGTA titrated with CaCl 2 to a final free Ca 2ϩ concentration of 100 nM). The DiO fluorescent bilayer was excited using laser TIRF microscopy under a 60ϫ 1.45 NA lens (Olympus Plan-APO-N). This enabled the correct focal plane to be attained and alignment of the TIRF laser angle to be made. DiO fluorescence was then bleached to extinction by continuous excitation (20 min). G␤␥ labeled with Alexa 488 (see above) was pressure-ejected from a pipette over the bilayer to obtain a transient exposure of G␤␥ to the lipid bilayer, and the intensity of fluorescence excited with TIRF illumination was then measured with photomultipliers, whereas G␤␥ concentrations and subtype were varied in the pipette. The concentration of G␤␥ over the bilayer was measured by conventional epifluorescence through the coverslip, and the resulting signal amplitude was compared with known concentrations of labeled G␤␥ in the recording chamber. In some experiments fluorescence anisotropy was also recorded to confirm protein-protein interactions occurred. This was achieved by measuring fluorescence through a polarizing beamsplitter with a detector parallel and orthogonal to the plane of the TIRF laser plane polarization angle.

HEK cell culture and transfection
HEK293 cells were maintained in an incubator at 37°C and 5% CO 2 and were passaged every 3-4 days for up to 15 passages. Cells were cultured in Minimum Essential Media (MEM; Life Technologies) supplemented with fetal bovine serum (10%; GE Healthcare), L-glutamine (2 mM; Life Technologies), and penicillin/streptomycin (100 unit ml Ϫ1 /100 g ml Ϫ1 ; Mediatech Inc., Manassas, VA). 24 h before transfection, cells were plated in 35-mm dishes according to manufacturer's guidelines for Lipofectamine 2000 (Invitrogen) transfection. Cells were transiently transfected with voltage-gated calcium channel subunits Ca V 2.2, ␤ 2a , an ␣ 2 ␦ alongside G␤ 1 (either WT, K78A, W332A, or K78A/W332A) and G␥ 2 subunits in a 1:1:1:3:3 ratio, respectively. Some control cells were transfected with only the voltage-gated calcium channel subunits. Transfected cells were visually identified by GFP expressed downstream of an IRES (internal ribosome entry site) sequence in the ␤ 2a subunit plasmid. Cells were re-plated onto poly-L-lysine-coated coverslips 48 -60 h after transfection and left to adhere for 2 h before patch clamp electrophysiology experiments.

Patch clamp electrophysiology experiments
Transfected HEK cells were recorded in the whole cell patch clamp configuration, and all experiments were performed at room temperature. Patch pipette electrodes were pulled from borosilicate glass capillary tubes (World Precision Instruments, Sarasota, FL) using a Sutter P-97 pipette puller (Sutter Instruments, Novato, CA), coated with dental wax (Electron Microscopy Services, Hatfield, PA), and fire-polished using a Narishige MF-830 micro forge (Narishige, Amityville, NY). Pipette resistance was ϳ2 megohms when filled with an internal patch pipette solution containing 110 mM CsCl, 4 mM MgCl 2 ⅐6H 2 O, 20 mM HEPES, 10 mM EGTA, 4 mM MgATP, 0.35 mM Na 2 GTP, 14 mM creatine phosphate, pH 7.3, osmolarity ϳ305-310 mosmol. Coverslips were placed in a recording bath continually perfused with extracellular solution at a rate of ϳ3 ml/min, and cells were viewed on a Nikon TE2000 inverted microscope. Cells were initially washed in a NaCl-based extracellular solution consisting of mM 145 NaCl, 2 mM KCl, 1 mM MgCl 2 ⅐6H 2 O, 10 mM glucose, 10 mM HEPES, 2 mM CaCl 2 , pH 7.3, osmolarity ϳ305 mosmol. After obtaining the whole cell recording configuration, the extracellular solution was a tetraethylammonium chloride-based (TEACl) solution containing 145 mM TEACl, 10 mM HEPES, 10 mM glucose, 1 mM MgCl 2 ⅐6H 2 O, 5 mM BaCl 2 , pH 7.3, osmolarity ϳ305 mosmol. Transfected HEK cells were voltage-clamped using an Axopatch 200B amplifier, Digidata 1400A interface, and PClamp10 (Clampex) acquisition software (Molecular Devices, Sunnyvale, CA). A double-pulse protocol was used to evoke voltage-gated calcium channel currents (I Ca ). Cells were stimulated by two identical 20-ms-step depolarizations (P1 and P2) to various potentials (Ϫ10, 0, 10, and ϩ 20mV) from a holding potential of Ϫ80 mV. The second test pulse (P2; 270 ms after P1) was preceded by a 30-msconditioning prepulse to ϩ120 mV. Analog data were filtered at 2 kHz and sampled at 20 kHz. Series resistance was partially compensated (ϳ60 -70%) using the Axopatch circuitry, and I Ca were subject to linear capacitance and leak subtraction using P/Ϫ8 protocols with leak pulses applied after test pulses. Raw data were analyzed in PClamp (Clampfit) software with I Ca amplitude measured 5 ms after P1 or P2 test-pulse onset. Graphing and statistical analysis were performed using Origin-Pro 7 (Originlab Corp., Northampton, MA) and Prism 5 (GraphPad Software, Inc., La Jolla, CA) software. Statistical sig-nificance was determined using ANOVA with Dunnett's multiple pairwise comparison.

Peptide synthesis
Peptide array synthesis was performed using the ResPep SL peptide synthesizer (Intavis AG, Koeln, Germany) according to previously published automated SPOT synthesis methods (30,61). Peptides correspond to the primary amino acid sequence of the protein of interest on the GenBank TM database for human GNB1, GNGT1, or GNG2.

Alphascreen competition-binding assays
Alphascreen luminescence measurements were performed in an EnSpire multimode plate reader (PerkinElmer Life Sciences) at 27°C. Biotinylated SNAP-25 was diluted to a 5ϫ concentration of 100 nM in assay buffer (20 mM HEPES, pH 7.0, 10 mM NaCl, 40 mM KCl, 5% glycerol, and 0.01% Triton X-100). An EC 80 concentration of 180 nM purified His 6 -G␤ 1 ␥ 2 was made in assay buffer. Peptide stocks in DMSO were spotted onto 384well white OptiPlates (PerkinElmer Life Sciences) at concentration ranges of 1 nM to 100 M using a Labcyte Echo 555 Omics acoustic liquid handler (Labcyte, Sunnyvale, CA), with DMSO being back-added to a final concentration of 0.1%. 4 l of G␤ 1 ␥ 2 solution was incubated with peptide for 5 min while shaking. After 5 min, 1l of biotinylated SNAP25 was added to a final concentration of 20 nM. Subsequent to incubation while shaking for an additional 5 min, 10 l of Alphascreen Histidine Detection Kit (nickel chelate) acceptor beads were added to a final concentration of 20 g/ml in assay buffer. The assay plate was agitated in dim light for 30 min. At that point, Alphascreen Streptavidin Donor Beads were added to a final concentration of 20 g/ml in dim light. All aqueous solutions in this assay were manipulated by a Velocity 11 Bravo liquid handler (Agilent Automation Solutions, Santa Clara, CA). The final volume in the assay plate was 25 l. After being spun down briefly to settle all fluid at the bottom of the well, plates were incubated for an additional 1 h at 27°C before being read in the EnSpire. 20 nM biotinylated recombinant glutathione S-transferase (Schistosoma japonicum) in place of SNAP-25 with G␤ 1 ␥ 2 were used as a negative control for nonspecific binding in each assay. IC 50 concentrations for each peptide were determined by sigmoidal dose-response curve-fitting with variable slope. To have a strong signal in the Alphascreen competition-binding assay that could still be competed with a peptide, we used an EC 80 concentration (180 nM) of His 6 -tagged G␤ 1 ␥ 2 combined with 20 nM recombinant human SNAP25 biotinylated on primary amine residues with NHS-biotin and increasing concentrations of peptides corresponding to a region on G␤ 1 or G␥ 1 in the Alphascreen bead assay. As a negative control, peptides were tested for their ability to disrupt a second Alphascreen assay in which donor and acceptor beads were reacted with 50 nM concentrations of a peptide consisting of a biotinylation site and a His-tag peptide (PerkinElmer Life Sciences).

Reconstituted fusion/lipid mixing assay
45 l of t-SNARE liposomes were reacted with 5 l of v-SNARE liposomes in a total of 75 l of assay buffer (25 mM HEPES-KOH, pH 7.8, 100 mM KCl, 10% glycerol, 1 mM DTT, 0.2 mM EGTA) in white 96-well FluoroNunc plates (Thermo Fisher, Waltham, MA). Purified synaptotagmin 1 (10 M) was added to the buffer along with a concentration-response curve of purified bovine G␤ 1 ␥ 1 followed by the addition of the t-SNARE liposomes. All components of the fusion reaction, except v-SNARE vesicles, were combined and pre-warmed to 37 C for 15 min. NBD fluorescence (excitation 460 nm/emission 538 nm) was continuously monitored over 80 min in a BioTek Synergy plate reader under continuous agitation, with fluorescence intensity being read every 8 s. After 20 min, CaCl 2 was added to a final concentration of 1.2 mM in the assay with 0.2 mM being chelated by EGTA to yield an effective concentration of 1.0 mM. After 80 min, dodecyl maltoside was added to a final concentration of 0.5% to maximally dequench NBD via infinite distance of the NBD:rhodamine FRET pair.

Protein structure visualization
All representatives of protein structure were made using PyMOL.

Statistics
Two-tailed Student's t tests and all concentration-responsecurve-fitting sigmoidal dose-response with variable slope were performed using GraphPad Prism v.4.03 for Windows, (GraphPad Software, La Jolla, CA). Sigmoidal dose-response curves were plotted using 1500 line segments. For the purposes of IC 50 estimation in sigmoidal concentration-response curves, all IC 50 values were calculated using an additional maximal point representing an order of magnitude above the highest concentration tested at the lowest signal value obtained in the study. For two-tailed Student's t tests, p values Ͻ 0.05 were considered to be statistically significant. *, p Ͻ 0.05; **, p Ͻ 0.01.