The expanding roles and mechanisms of G protein–mediated presynaptic inhibition

Throughout the past five decades, tremendous advancements have been made in our understanding of G protein signaling and presynaptic inhibition, many of which were published in the Journal of Biological Chemistry under the tenure of Herb Tabor as Editor-in-Chief. Here, we identify these critical advances, including the formulation of the ternary complex model of G protein–coupled receptor signaling and the discovery of Gβγ as a critical signaling component of the heterotrimeric G protein, along with the nature of presynaptic inhibition and its physiological role. We provide an overview for the discovery and physiological relevance of the two known Gβγ–mediated mechanisms for presynaptic inhibition: first, the action of Gβγ on voltage-gated calcium channels to inhibit calcium influx to the presynaptic active zone and, second, the direct binding of Gβγ to the SNARE complex to displace synaptotagmin downstream of calcium entry, which has been demonstrated to be important in neurons and secretory cells. These two mechanisms act in tandem with each other in a synergistic manner to provide more complete spatiotemporal control over neurotransmitter release.

The development and characterization of selective agonists and antagonists for cell surface receptors was instrumental to formulating the current model of G protein receptor signaling. Radioligand antagonist-binding studies for numerous cell surface receptors, including the ␣and ␤-adrenergic receptors (1-7), muscarinic acetylcholine receptors (8), dopaminergic receptors (9), and glucagon receptors (10), demonstrated the existence of two affinity states within the receptor for agonist binding, a high-affinity state and a low-affinity state, but only a single state with respect to antagonist binding. This critical finding enabled Robert Lefkowitz to formulate and propose a ternary complex model in JBC using early computer-modeling techniques, where a high-affinity state for agonist binding consisting of the agonist, the receptor, and a yet-undetermined guanine nucleotide-binding component was needed for downstream adenylate cyclase activation (11).
The guanine nucleotide-binding component, known today as the heterotrimeric G protein, previously thought of as the "regulatory subunit" of adenylyl cyclase, was then purified in sufficiently large quantities to study. It was shown to consist of three polypeptides, one of which bound guanine nucleotides (12, 13). The isolation of purified ␤-adrenergic receptors (14 -16) and heterotrimeric G protein enabled the determination that the presence of these two components was necessary and sufficient for inducing two-state agonist binding in the receptor and nucleotide exchange for the G protein (17,18). Building upon this model, it was the purification of adenylate cyclase that enabled several articles in JBC and elsewhere from Lefkowitz, Gilman, and Smigel. These articles were instrumental for the initial model of GPCR 3 signaling, with critical reconstitution experiments showing that a minimum of three cellular proteinaceous complexes-the ␤ adrenergic receptor, the heterotrimeric G protein, and the adenylate cyclase-were required for cAMP accumulation in response to agonists (19 -21). This effectively reproduced the cellular response in a strictly defined system.

The discovery of G␤␥ as a signaling molecule
This initial model of GPCR signaling was predicated upon the GTP-bound G␣ subunit conveying the transduced signal to the effector, whereas the G␤␥ subunit's only role was to obstruct the G␣ effector-binding domains in the absence of ligand or act as an anchor to the plasma membrane (22). An abrupt challenge to the model came from the critical finding of David Clapham's group (23) that purified G␤␥ subunits from the brain, and not G␣ subunits, were responsible for transducing the signal of muscarinic acetylcholine receptors to activate GIRK channels in cardiac pacemaker and atrial cells. This novel finding was vehemently disputed by the groups of Birnbaumer and Brown (24 -26), who showed evidence that the activating component of the heterotrimer, then termed "Gk" after the potassium channel, was a G␣ subunit and that the purified G␤␥ preparation of Clapham's group may have been slightly contaminated with a high-affinity G␣, or interference was generated somehow from CHAPS detergent. The initial finding of Clapham's group was echoed by other groups (27). The dispute persisted until the work of Jan and Lester (28 -30), facilitated by the development of heterologous expression systems such as Xenopus oocytes, and the cloning of the individual cDNAs responsible for the G proteins ␣ i , ␤, and ␥, which implicated G␤␥ as the signal transducer of GIRK.
These technological advances paved the way for the group of Catterall (31) to determine that the active G protein subunit responsible for G i/o -coupled GPCRs' inhibitory action upon Nand P/Q-type voltage-gated calcium channels was G␤␥ and not G␣. This finding was echoed by Ikeda (32) for N-type channels in the same issue of Nature. The contemporaneous findings of Harden and Gilman (33) that phospholipase C␤ and type II adenylyl cyclase (34) were activated by G␤␥ provided more widespread acceptance that G␤␥-effector interactions were a critical aspect of G protein signaling. Many more direct G␤␥binding effectors would be identified in that period, including the GRK family of kinases (35) and phosphoinositide-3 kinase (36), the molecular basis for many of which were characterized through identification of key residues on the G␣-binding surface of G␤␥ (37). The widespread acceptance of G␤␥ as a critical signaling protein with a multitude of effectors was instrumental to our studies of its role in the presynaptic regulation of neurotransmitter release. The development and characterization of selective agonists and antagonists for cell surface receptors was instrumental to formulating the current model of G protein receptor signaling, including the field of signaling that modulates neurotransmitter release from the presynaptic terminal.

The historical basis for presynaptic regulation of neurotransmitter release
Physiological mechanisms of control over neurotransmitter release are essential for the orderly transduction of signals from presynaptic to postsynaptic neurons. Our understanding of these mechanisms has expanded considerably over the past half-century. Presynaptic regulation of transmitter release is vital to the plasticity of a neuron and the signaling network in which it functions. In various studies, the phenomenon of presynaptic inhibition studies in vertebrates started from analysis of sensory inputs to the spinal cord. As early as the 1930s, presynaptic depolarization from sensory stimulation was shown as a mechanism of physiological modulation of these inputs (38). This was later characterized as primary afferent depolarization (PAD) (39) and first described as a presynaptic mechanism of inhibition by Frank and Fuortes (40). Eccles et al. (41,42) went on to firmly link the phenomenon of PAD to presynaptic inhibition as well as to demonstrate that higher brain centers use this mechanism to control sensory inflow (43). It was also determined by Eccles et al. (44) that this inhibition is mediated by GABA receptors, although the identity of GABA as an amino acid neurotransmitter was only later proven in vertebrates (45) following extensive work in invertebrate models (46 -49). Ultrastuctural and immunohistochemical data showing GABAergic axo-axonic synapses later reinforced the existence of axo-axonic synapses that mediate presynaptic inhibition (50 -52).
Supporting evidence for the role of GABA in presynaptic inhibition was provided by studies in the periphery on sympathetic nerve terminals and ganglia in which GABA receptors caused depolarization of sympathetic ganglia neurons. In these neurons, the first example of GPCR-mediated presynaptic inhibition was found in response to noradrenaline (53)(54)(55)(56). Strikingly, however, GABA-mediated inhibition at these terminals was not mediated by the ionotropic GABA A receptor, but rather by a G␣ i/o -coupled GPCR (57)(58)(59)(60), the GABA B receptor (61), which responded to the anti-convulsant (Ϫ)-baclofen (62). GABA B receptors are distributed throughout the central nervous system, where they also mediate presynaptic inhibition. Indeed, it is now clear that GABA B receptors, along with many other GPCRs, including but not limited to opioid receptors (63,64), cannabinoid receptors (65), ␣ 2 adrenergic receptors (66), and metabotropic glutamate receptors (67), inhibit glutamate release at spinal sensory synapses, but also throughout the central nervous system.
A critical result in determining a mechanism by which GPCRs mediate presynaptic inhibition came from recordings from dorsal root ganglion cells in culture. These are the cells that give rise to spinal sensory synapses, which demonstrate PAD, thus understanding their cellular biology provides insight into the biochemistry of their presynaptic terminals. Various GPCR agonists (68,69) that target G proteins in dorsal root ganglia cells (59,60,70) directly inhibited Ca 2ϩ currents, including both noradrenaline ␣ 2 -adrenergic and GABA B receptors. This led to the hypothesis that GPCRs cause presynaptic inhibition by inhibiting presynaptic Ca 2ϩ entry, although this has not been demonstrated directly at spinal sensory synapses. Nevertheless, it is now clear that presynaptic inhibition is ubiquitous and is found throughout the central nervous system (61,(71)(72)(73). Whereas some presynaptic GPCRs clearly inhibit neurotransmission by inhibiting presynaptic Ca 2ϩ entry (74 -77), it is also clear that other mechanisms targeting the release machinery directly are also important (78,79).

Identification of the G␤␥-SNARE interaction as a critical inhibitory mechanism of exocytosis downstream of Ca 2؉ entry
Many G i/o -coupled GPCRs were shown to inhibit exocytosis via the action of G␤␥ on voltage-gated calcium channels to inhibit Ca 2ϩ fluxes (80 -84). Other researchers found that inhibition occurred at a distinct site downstream of Ca 2ϩ entry (78,85,87). The large size of the sea lamprey (Petromyzon marinus) giant axon preparation, utilized by our group as a model of synaptic transmission, facilitated the direct axonal injection of proteins, such as G␤␥, or critical G␤␥-chelating peptides, such as ␤ARK-CT. Injection of G␤␥ inhibited EPSCs without perturbing voltage-gated Ca 2ϩ currents, whereas the chelating peptide blocked the inhibitory effect of the lamprey serotonin receptor, a receptor homologous to human 5-HT 1B (79). Later studies expanded upon this initial finding, showing that the step inhibited by G␤␥ was late in the vesicular docking and priming cycle and could occur within tens of milliseconds after the uncaging of agonist (88). In addition, 5-HT action on the lamprey serotonin receptor reduced quantal size and prevented full fusion (89), reinforcing the notion that G␤␥ was acting at a late step and not influencing vesicular pool sizes via this mechanism.
No data were shown pertaining to the effector target of G␤␥ in the original study from 2001 (79), but in 2005, the presynaptic release machinery was shown to be the target (88), because botulinum neurotoxin type A (BoNT/A) cleaves the nine C-terminal residues of the peripheral membrane t-SNARE SNAP-25 (90,91), leaving a truncated SNAP-25⌬9 that is still fusogenic (92,93), albeit to a greatly reduced extent. BoNT/A completely eliminated any inhibition by 5-HT. This was confirmed in mammalian tissue, by recording effects of 5-HT 1B receptors on glutamate release from CA1 pyramidal neurons (76). We also showed direct in vitro binding of G␤␥ to the t-SNARE proteins SNAP-25 and syntaxin 1A and synaptobrevin, as well as t-SNARE and ternary SNARE complexes (94,95). These studies showed further that competition occurred between G␤␥ and the synaptic calcium sensor synaptotagmin 1 for binding sites upon SNAP-25. From this, it was hypothesized that G␤␥ could displace synaptotagmin 1 at docked and primed SNARE complexes, preventing it from performing its fusogenic lipid-mixing activity. At higher levels of Ca 2ϩ , the fully Ca 2ϩ -occupied synaptotagmin would bind more tightly to the membranebound SNARE complex (94 -96) and be more able to overcome the inhibition of G␤␥ (94,95,97).
Lipid-mixing assays in reconstituted systems with pure components lent critical support to this hypothesis. G␤␥ inhibited Ca 2ϩ -synaptotagmin and SNARE-dependent lipid mixing in a defined system of v-and t-SNARE-harboring liposomes, in a concentration-dependent manner (97). This inhibitory action of G␤␥ on lipid mixing was more potent at lower concentrations of synaptotagmin 1 C2AB and was greatly diminished in the complete absence of synaptotagmin 1 C2AB. In addition, G␤␥ could displace fluorescent synaptotagmin 1 C2AB from t-SNAREharboring supported lipid bilayers (97). These studies, much like the original GPCR reconstitution experiments of Lefkowitz and Gilman's groups (19 -21) published in JBC, show the requirements for a critical GPCR signaling pathway in a defined system.

The molecular requirements for the G␤␥-SNARE interaction
As stated above, the initial insight that the formed SNARE complex was the effector of G␤␥ downstream of Ca 2ϩ entry came from studies of the action of BoNT/A. After BoNT/A treatment, 5-HT 1B receptor-mediated presynaptic inhibition was lost in both lamprey giant synapses (88) and in outputs from CA1 pyramidal neurons (76). Subsequent binding studies with recombinant SNAP-25⌬9 showed only a partial 1.5-fold loss of affinity for G␤␥ relative to full-length SNAP-25, however (95), implying that other residues were involved.
In light of this finding, we utilized a peptide-mapping approach to find residues on SNAP-25 that were instrumental to bind G␤␥ and confirmed these studies with scanning Ala mutagenesis (98). We found nine residues located in two clusters upon SNAP-25. The first cluster was adjacent to the N terminus, whereas the second cluster was adjacent to the C terminus. Mutation of eight of the nine residues to Ala abolished both the ability of G␤␥ to bind SNAP-25 and the ability of the lamprey serotonin receptor to inhibit neurotransmission, creating a key mechanistic linkage between G␤␥ binding and G i/o -coupled GPCR-mediated inhibition of exocytosis (98).
Despite this, mutation of the two residues highlighted by the peptide mapping screen within the C-terminal nine residues to Ala did not phenocopy the effect of BoNT/A with respect to G i/o -coupled GPCRs (99), implying that more of the C terminus was involved. Screening of truncated SNAP-25 constructs for G␤␥ binding showed that the extreme C-terminal three residues were involved in the mechanism (99). Given the two glycine residues within the extreme C terminus, we speculated that the high flexibility of this region may be required to stabilize G␤␥ binding.
The binding of G␤␥ to SNARE may be splice variant-dependent; the two splice variants of SNAP-25, SNAP-25A and SNAP-25B, differ by 9 residues via differential splicing of exon 5 within the palmitoylation domain of the linker region between the two SNARE-forming helices (100 -102). The two splice variants exhibit time-dependent expression, with SNAP-25A predominating in embryonic development and early postnatal life and SNAP-25B being the predominant isoform in adults, although SNAP-25a expression is retained in the adult hypothalamus and cortex (102)(103)(104). These two splice variants appear to differ in their ability to interact with certain G␤␥ subunits. Co-immunoprecipitation studies show that genetic ablation of SNAP-25B expression partially reduces the ability to co-immunoprecipitate G␤ 2 , but not G␤ 1 , with SNAP-25 in the hippocampus (105). The region of SNAP-25 where the two splice variants differ is adjacent to only one of the nine residues critical for G␤␥ binding, residue Glu-62 (98). Scanning Ala mutagenesis showed no reduction in G␤␥ binding after mutagenizing any of the residues that differ between each splice variant. Two possible explanations could account for this phenomenon: first, that this differential binding is due to perturbations in the local environment adjacent to residue Glu-62 and, second, that isoform-dependent palmitoylation results in more efficient localization of SNAP-25 to the G-protein/GPCR complex within the active zone.

Differential G i/o -signaling mechanisms in secretory cells: synergy between G␤␥'s action on VGCC and SNARE
Subsequent to the initial discovery of the G␤␥-SNARE mechanism in the lamprey giant synapse, numerous studies have demonstrated its applicability to several families of G i/ocoupled GPCRs in a variety of mammalian secretory cells. Over the past 18 years of G␤␥-SNARE research, the development of pharmacological and genetic tools for dissection of presynaptic GPCR mechanisms has allowed researchers to demonstrate its importance in an increasingly sophisticated manner. The release of GABA from Purkinje cells in the cerebellum was shown to occur via two mechanisms, including a spontaneous and cAMP-dependent form of release that was inhibited by group II metabotropic glutamate receptors (mGluRs) under conditions where Ca 2ϩ entry from the extracellular space and internal release are both blocked (106).
The discovery that the G␤␥-SNARE mechanism could be bypassed via treatment with BoNT/A or the delivery of the SNAP-25 193-206 C-terminal peptide (88) was critical for distinguishing this mechanism from other pathways that regulate presynaptic release. Biochemical studies showed that the SNAP-25⌬9 produced by BoNT/A had lower affinity for G␤␥ than full-length SNAP-25 (95). Single fiber inputs from the nociceptive pontine parabrachial nucleus form synapses upon the neurons of the central amygdala, and neurotransmission at this synapse is regulated presynaptically by norepinephrine acting on the ␣ 2 -adrenergic receptor (107). The action of this receptor was abolished after treatment with BoNT/A, phenocopying that of the lamprey serotonin receptor. In the paraventricular nuclei of the hypothalamus, -opioid receptors regulate release from magnocellular neurosecretory cells with a mechanism that the authors attributed to the action of G␤␥ on the release machinery; however, intriguingly, quantal size was not affected by -opioid receptor agonism, differing from results observed with the lamprey serotonin receptor (108). Moreover, this -opioid receptor agonism was abolished via treatment with ionomycin to permeabilize the plasma membrane to 2.5 mM Ca 2ϩ .
These latter results using ionomycin are consistent with effects in synapses, where high presynaptic Ca 2ϩ prevents G␤␥mediated SNARE interactions (76,88), but create an intriguing contrast to results observed with amperometry in chromaffin cells, in which -opioid agonism remains inhibitory to release subsequent to treatment with ionomycin in 5 mM Ca 2ϩ (109). Nevertheless, the charge of the chromaffin cell amperometric spike, a measure homologous to quantal size in neurons, was reduced by overexpression of G␤␥, as was quantal size itself by the action of serotonin receptors in the lamprey giant synapse (89). In peripheral tissue, the G␤␥-SNARE mechanism also inhibits release. This was first demonstrated to occur in the ␤ cells of the islets of Langerhans, where the ␣ 2 adrenergic receptor inhibited insulin release in a manner that could be overcome by BoNT/A or via intracellular chelation of G␤␥ with the SNAP-25 193-206 peptide (110).
In the hippocampus at CA1-subicular synapses, the 5-HT 1B receptor was shown to inhibit exocytotic release without alter-ing Ca 2ϩ fluxes. Destruction of the exocytotic apparatus with BoNT/C was shown to liberate G␤␥ from its site on SNARE; G␤␥ is then capable of modulating Ca 2ϩ entry through VDCC (76). GABA B receptors on the same presynaptic terminals acted to reduce Ca 2ϩ fluxes (Fig. 1). Also, in the hippocampus, longterm depression of synaptic strength in presynaptic Schaffer collateral terminals, as well as presynaptic inhibition mediated by group II mGluRs, was similarly shown to be dependent upon G␤␥ and the C terminus of SNAP-25 (111).
From this group of studies as a whole, we hypothesize that G␤␥-SNARE and G␤␥-VGCC actions are synergistic, receptor-specific mechanisms to provide more complete temporal and scalar control over vesicular release. In hippocampal CA1subicular paired recordings, greater extent of inhibition was observed by treating with both a 5-HT 1B agonist and a GABA B agonist (Fig. 2). The addition of a GABA B agonist prevents recovery of EPSC amplitude subsequent to treatment with the 5-HT 1B agonist during repetitive stimulation. We interpret these data as the GABA B agonist blocking the influx of Ca 2ϩ into the terminal, preventing Ca 2ϩ from binding to synaptotagmin and thus reducing its competition with G␤␥ on the SNARE complex and allowing augmented inhibition. This paradigm of G i/o -coupled GPCRs working in tandem to synergistically inhibit release may be present for a multitude of receptor systems at many different synapses.

The development of mouse genetic models deficient in the G␤␥-SNARE interaction
The idea of G␤␥-SNARE-dependent GPCRs and G␤␥-VGCC-dependent GPCRs acting in a synergistic manner was expanded upon with the development of the SNAP-25⌬3 mouse model, which carries a point mutation in SNAP-25 (Gly- Figure 1. G␤␥-SNARE and G␤␥-VGCC act synergistically to inhibit vesicle release. Serotonin (blue) from paracrine sources acts on inhibitory presynaptic 5-HT 1B receptors (blue) to liberate G␤␥ to bind SNAREs, displacing the fusogenic calcium sensor synaptotagmin I and inhibiting the release of glutamate (green) This mechanism is distal to and synergistic with GABA B heteroreceptors, which in the presence of axo-axonic GABA (red), inhibit calcium fluxes into the terminal (75). Class 2/3 mGluRs, acting as autoreceptors, may also signal via G␤␥-SNARE, as has been demonstrated to occur at the synapse between cone photoreceptors and horizontal cells of the inner nuclear layer of the retina (86). 204*), truncating it by three residues. The mutation partially disables the inhibitory G␤␥-SNARE mechanism (99,112) while maintaining full exocytotic activity. This mouse is viable in the homozygous state, enabling researchers to assess the importance of the G␤␥-SNARE mechanism in all SNAP-25expressing secretory cell populations. Critically, the model permits researchers to investigate the mechanism in the absence of toxic reagents, such as BoNT/A or Cd 2ϩ , that could render behavioral or physiological studies impossible. Field potential studies at the CA1-subicular synapse in the SNAP-25⌬3 homozygote phenocopy the results (112) obtained with BoNT/A in WT animals (76), indicating a role for G␤␥-SNARE in the 5-HT 1B receptor, but not the GABA B receptor.
The animal model was also used to identify a new role for the G␤␥-SNARE mechanism on attenuating stress in the bed nuclei of the stria terminals (BNST), where presynaptic inhibition upon excitatory parabrachial inputs onto BNST neurons was shown to occur via the ␣ 2A adrenergic receptor, but not the GABA B receptor. Moving forward, we anticipate that the SNAP-25⌬3 model will be an important tool for investigators to assess the role of the G␤␥-SNARE mechanism in vivo. This animal model will be invaluable to assess chronic, whole-organism effects of G␤␥-SNARE disruption. In the future, we also plan to develop genetic models that permit tissue-specific ablation of the mechanism to more directly assess its importance in individual circuits. This would allow researchers the ability to identify specific neuronal subtypes where G␤␥-SNARE could be a key regulatory mechanism.

Specificity of G␤␥ signaling
A growing body of work supports the notion of subunit specificity for a subset of effectors of the five known G␤ and 12 G␥ subunits, most of which can heterodimerize with each other in vitro, although in vivo evidence is more scarce (113,114). Genetic studies show that despite considerable homology, G␤ and G␥ subunits perform specific roles in development. For example, genetic ablation of G␤ 1 leads to microencephaly or neural tube closure deficits: homozygous null mutants died within 2 days of birth (115). Correspondingly, human de novo loss-of-function mutations in G␤ 1 were associated with hypotonia and seizures along with prominent neurodevelopmental disability (116). G␤ 2 is a mediator of neuronal excitability through inhibition of T-type Ca 2ϩ channels (117,118).
G␤ 5 , which shows the least homology to the other G␤ subunits, has a unique capacity to form heterodimers with the G␥-like domain of the regulator of G protein signaling (RGS) proteins, such as RGS9, in addition to several studies showing heterodimer formation with G␥ 2 (119 -121). Genetic ablation of G␤ 5 shows a very different pattern of phenotypes to G␤ 1 ; G␤ 5 knockout mice showed growth deficits, abnormal motor behavior ("tiptoe-walking" in the unstressed state, and hyperactivity (122,123), despite being viable. G␥ 3 -null mice exhibit handling-induced seizures with associated mortality, while being less susceptible to diet-induced obesity than WT littermates, whereas G␥ 7 -null mice have deficits in adenylyl cyclase activity and increased startle response, but normal locomotion (124,125). These variable phenotypes resulting from genetic ablation of individual G␤ and G␥ subunits suggest isoform dependence for G␤␥ dimerization and specificity in signaling roles (126 -128). Multiple independent groups have observed heterodimer or isoform-dependent coupling of receptors to effectors (129,130). This may be attributable to variations in expression level, localization, or affinity for effectors within each G␤ and G␥ subunit (131); for example, somatostatin receptors in RINm5F cells were specifically shown to inhibit VGCC via ␤ 1 ␥ 3 (131)(132)(133)(134). In GH3 cells, a similar result was observed for somatostatin receptors coupling to VGCC with G␥ 3 , but the muscarinic receptor was shown to couple to VGCC via G␥ 4 (135). In addition, G␤␥ subunits have distinct expression patterns in the brain (131,136,137).
Furthermore, unique G␤␥ isoforms play specific roles in mediating interactions with both receptors and effectors (126 -128, 130), and isoforms exhibit tissue-dependent specificity for individual receptors or effectors (125, 139 -141). This has been observed for the t-SNARE complex, where G␤ 1 ␥ 2 has a 14-fold higher affinity than G␤ 1 ␥ 1 (97) and a 20-fold higher potency at inhibiting exocytosis (94). However, the lack of subunit-specific antibodies for each G␤ and G␥ subunit limits the understanding of G␤␥ specificity. Using a powerful quantitative proteomic approach (131), we are beginning to get a glimpse of the specificity by which different G␤␥ subunits are recruited to both GPCRs (131) and the SNARE complex (142). For example, we determined the in vivo G␤␥ specificity of presynaptic ␣ 2a -adrenergic receptors (␣ 2a ARs) in both adrenergic (auto-␣ 2a ARs) and nonadrenergic neurons (hetero-␣ 2a ARs) (142). This study suggested that auto-and hetero-␣ 2a ARs utilize different G␤␥ subunits to regulate their downstream signaling mechanisms. We are currently using similar approaches to delineate mechanisms of G␤␥ subunit specificity to SNARE complexes to continue to pursue the remarkable specificity of G␤ and G␥ subunits.

Future directions
Negative regulators of signaling of many types are very important for shaping signals. For example, RGS proteins and phosphatases both have very important roles in controlling the turnoff rate of signal transduction. The inhibition of secretion is no different. This type of regulation is like a rheostat; it does not affect GPCR activity per se, or signaling through the various G protein ␣ and most ␤␥ subunit pathways, except it blunts SNARE-mediated exocytosis. The G␤␥-SNARE interaction can be a convergent point for various hormones and neuromodulators known to participate in inhibition of exocytosis.
Although G␣ subunit signaling and its dysregulation in diseases has been studied in great detail, understanding how the different G␤␥ subunits show as much specificity to receptors as well as effectors will prove to be an important insight. It will be of interest to explore whether and how G␤␥-SNARE interaction plays a role in neural and endocrine disorders.
We noticed that the phenotypes resulting from disabling G␤␥-SNARE interaction are of variable strength, with, for example, subtle effects on locomotion but dramatic effects on stress responses (112). This leads us to suggest that targeting the G␤␥-SNARE interaction might have some level of specificity. In addition, specificity might be improved by combining an inhibitor with an antagonist of a presynaptic G i/o -coupled receptor (Fig. 3). The two agents would be on the same pathway, and we hypothesize that untoward side effects of each could be mitigated. A great deal is known about regulation of exocytosis by G␤␥-VDCC interaction, and this has led to a large class of pharmaceuticals that work as calcium channel modulators (138). The G␤␥-SNARE interaction downstream of calcium entry is much less understood and has been studied in relatively few cases. Enhancement of the release of hormones and neurotransmitters by inhibiting the G␤␥-SNARE interaction could be a novel pharmaceutical strategy. This is an underappreciated locus that has great potential to differentiate various modulatory pathways and to provide fine tuning of hormone or neurotransmitter release within a range of physiological activities. Similarly, regulation of exocytosis through the two G␤␥-mediated modulatory mechanisms enhanced each other's activity (Fig. 2), leading to a more profound level of inhibition (Fig. 3).

Figure 3. Potential synergy between GPCR antagonists and G␤␥-SNARE inhibitors.
Potential mechanism for a synergistic relationship between existing G i/o -coupled GPCR antagonists (orange) used clinically to inhibit release presynaptically and hypothetical to-be-developed inhibitors of the G␤␥-SNARE interaction (yellow). The usage of two different inhibitors acting on different targets within the same pathway should more potently drive release than either drug used alone and may permit lower concentrations of antagonist to be utilized, minimizing potential side effects from specificity.