The in vivo specificity of synaptic Gβ and Gγ subunits to the α2a adrenergic receptor at CNS synapses

G proteins are major transducers of signals from G-protein coupled receptors (GPCRs). They are made up of α, β, and γ subunits, with 16 Gα, 5 Gβ and 12 Gγ subunits. Though much is known about the specificity of Gα subunits, the specificity of Gβγs activated by a given GPCR and that activate each effector in vivo is not known. Here, we examined the in vivo Gβγ specificity of presynaptic α2a-adrenergic receptors (α2aARs) in both adrenergic (auto-α2aARs) and non-adrenergic neurons (hetero-α2aARs) for the first time. With a quantitative MRM proteomic analysis of neuronal Gβ and Gγ subunits, and co-immunoprecipitation of tagged α2aARs from mouse models including transgenic FLAG-α2aARs and knock-in HA-α2aARs, we investigated the in vivo specificity of Gβ and Gγ subunits to auto-α2aARs and hetero-α2aARs activated with epinephrine to understand the role of Gβγ specificity in diverse physiological functions such as anesthetic sparing, and working memory enhancement. We detected Gβ2, Gγ2, Gγ3, and Gγ4 with activated auto α2aARs, whereas we found Gβ4 and Gγ12 preferentially interacted with activated hetero-α2aARs. Further understanding of in vivo Gβγ specificity to various GPCRs offers new insights into the multiplicity of genes for Gβ and Gγ, and the mechanisms underlying GPCR signaling through Gβγ subunits.


Results
The interaction of α 2a adrenergic receptors and Gβγ. To study the specificity of neuronal Gβγ subunits to synaptic α 2a ARs, we used brain synaptosomes from wildtype, α 2a AR KO, HA-and FLAG-α 2a AR mice. Because no GPCR antibodies are specific enough to co-IP α 2a ARs and Gβγ, we used HA-and FLAG-α 2a ARs expressing mice to overcome this limitation. Wildtype and α 2a -ARs KO mice were used as controls for HA-and FLAG-α 2a ARs mice. Synaptosomes from these mice were resuspended in a buffer with (stimulated) or without (unstimulated) epinephrine. DSP, a lipid-soluble thiol cleavable crosslinker, was added to ensure the receptor and Gβγ remained intact during co-IP experiments. The synaptosomes were then lysed and co-IPed for HA-or FLAG-α 2a ARs and Gβγ (Fig. 1A), which was validated by Western blot. Input represents total proteins present in lysate after the preclear while supernatant (Sup) represents what proteins are left in lysate after the co-IP with HA or FLAG specific antibodies (see Materials and Methods for more details). In wildtype and α 2a ARs KO mice, no α 2a AR and Gβγ interactions were detected following receptor stimulation (Fig. 1B,C). Here, we detected HA-and FLAG-α 2a ARs interacting with Gβγ only following α 2a AR stimulation (Fig. 1B,C).

Limit of Gβ 1 detection and quantification.
To determine the number of co-IPs needed to detect Gβ and Gγ subunits in our MRM method, we used a serial dilution of purified Gβ 1 γ 1 and monitored four non-heavy labeled proteolytic peptides of Gβ 1 to determine the limits of detection and quantitation (LOD/LOQ) (Supplementary Table 1) 53 . Because Gβ 1 γ 1 is easily purified from the bovine retina, we chose it as our standard. It is used as a control to make sure that our method is running correctly and accurately. Previously, we have validated how each Gβ and Gγ are detected in our quantitative method 28 . Because Gγ 1 is not present in the brain but only in photoreceptors, we only monitored Gβ 1 with mass spec. Below 10 pg of Gβ 1 γ 1 , we couldn't confidently identify the presence of Gβ 1 in samples. Between 10 pg to 250 pg, we were able to detect Gβ 1 but total area under the curve (AUC) didn't increase as the amount of purified Gβ 1 γ 1 was increased ( Supplementary Fig. 1). This suggests that we need more than 250 pg of Gβ 1 to detect and quantify proteins using our MRM method. We subsequently found using quantitative Western blots, that ~400-700 ng of Gβγ was pulled down with FLAG-α 2a ARs per half mouse brain used (10 co-IPs/half mouse brain) (data not shown). However, the previous limit of quantification experiment suggests that we need more than 4 ng of Gβγ for quantification 28 . Thus, using a half brain per condition, we can detect and quantify neuronal Gβ and Gγ despite our previously described technical challenges 28 .
We examined the Gβ and Gγ subunits interacting with α 2a ARs to distinguish which Gβ and Gγ subunits interact with auto-vs. hetero-α 2a ARs. In Figs 2 and 3, we applied the quantitative MRM method 28 to co-IP samples of wildtype (WT) and HA-α 2a ARs mouse synaptosomes. Using SDS-PAGE gel, we excised Gβ and Gγ bands and added the heavy labeled proteolytic peptides to quantify each neuronal Gβ and Gγ subunit 28 (see Materials and Methods). Because Gβγ can be sticky, we built in a number of negative controls. To identify nonspecific interactions of Gβ and Gγ subunits, we used both unstimulated WT (WT no epi) and HA-α 2a AR (HA-α 2a AR no epi) samples as controls. In addition, we used stimulated WT (WT +epi) samples to detect nonspecific interactions with other receptors (non-HA-α 2a AR-mediated interactions). Thus the first three conditions in each graph in Figs 2 and 3 were to detect non-specific interactions of Gβγ, while the last detected interaction of Gβγ isoforms with epi-stimulated HA-α 2a AR. Gβ 2 and Gβ 4 were significantly enriched with HA-α 2a ARs stimulated with epi ( Fig. 2B,C). More Gβ 4 was detected than Gβ 2 In contrast, Gβ 5 did not interact with HA-α 2a ARs. Next, we examined the specificity of Gγ subunits to α 2a ARs to determine possible Gβγ dimer interactions with α 2a ARs. From the 6 detectable and quantifiable neuronal Gγ subunits 28 , Gγ 2 , Gγ 3 , Gγ 4 , and Gγ 12 were significantly enriched with HA-α 2a ARs upon epinephrine stimulation (Fig. 3A-C and E). We detected Gγ 2 > Gγ 3 ≈ Gγ 4 > Gγ 12 . Gγ 7 and Gγ 13 in stimulated HA-α 2a ARs + epi samples were equal to, or less, than corresponding control samples, suggesting these Gγs are present nonspecifically (Fig. 3D,F). From the subunits we have detected, we postulate that there may be as many as 8 different combinations of Gβγ dimers in vivo (Gβ 2 γ 2 , Gβ 2 γ 3 , Gβ 2 γ 4 , Gβ 2 γ 12 , Gβ 4 γ 2 , Gβ 4 γ 3 , Gβ 4 γ 4 , and Gβ 4 Gγ 12 ) which may interact with α 2a ARs in adrenergic and non-adrenergic neurons. Based on their detection levels, Gβ 2 γ 2 , Gβ 2 γ 3 , and Gβ 2 γ 4 may be more likely to interact with α 2a ARs than other Gβγ dimers. Gβ 2 γ 12 , Gβ 4 γ 2 , Gβ 4 γ 3 , Gβ 4 γ 4 , and Gβ 4 Gγ 12 are less abundant Gβγ dimers interacting with α 2a ARs. Further biochemical analysis will be needed to validate the presence of these Gβγ dimers and their specificities with α 2a ARs in both adrenergic and non-adrenergic neurons. Gβ 2 , Gγ 2 , Gγ 3 , and Gγ 4 specifically interact with auto-adrenergic α 2a receptors. After identifying the specificities of Gβ and Gγ for α 2a ARs in both adrenergic and non-adrenergic neurons, we decided to examine the specificity to auto-α 2a ARs which are only present in adrenergic neurons. In previous studies, auto-α 2a ARs and hetero-α 2a ARs were shown to have very different physiological functions 37 . We wondered if these different physiological functions may be mediated by unique Gβ and Gγ specificities for the different receptor types or through specific effector interactions. We again applied a quantitative MRM method to TCA-precipitated and trypsin-digested co-IP samples of α 2a ARs KO and FLAG-α 2a ARs mouse synaptosomes. FLAG-α 2a ARs only express auto-α 2a ARs at the sympathetic presynaptic terminal, allowing us to study Gβ and Gγ subunit specificities to autoreceptors uniquely in sympathetic neurons. Similar to the previous experiment, α 2a ARs KO no epi and FLAG-α 2a ARs no epi samples were used as controls to identify nonspecific interactions, and α 2a ARs KO + epi samples were used to detect non-α 2a ARs associations. Here, Gβ 2 but not Gβ 4 , showed a significant enrichment with auto-α 2a ARs (FLAG-α 2a ARs) (Fig. 4B). Again, Gβ 1 and Gβ 5 did not specifically interact with auto-α 2a ARs upon stimulation (Fig. 4A,D).
Gβ 4 and Gγ 12 may specifically interact with heteroreceptors. Only a subset of Gβ and Gγ subunits from the HA-α 2a ARs study exhibited specificity to auto-α 2a ARs, suggesting that hetero-α 2a ARs may utilize those Gβ and Gγ subunits not associated with auto-α 2a ARs to regulate unique downstream signaling pathways. Without a transgenic tagged hetero-α 2a ARs mouse; however, we cannot directly measure the Gβ and Gγ subunits specific to hetero-α 2a ARs. However, in this study, we can infer the Gβ and Gγ specific to hetero-α 2a ARs by comparing and subtracting the results of our HA-and FLAG-α 2a ARs studies. By comparing the Gβ and Gγ subunits detected each set of experiments (which represent overall synaptic α 2a ARs and presynaptic α 2a ARs at the sympathetic terminal, respectively), we determined that Gβ 4 (Figs 2 and 4C) and Gγ 12 (Figs 3 and 5E) may be heteroreceptor specific. As a result, it is possible that Gβ 2 γ 12 , Gβ 4 γ 2 , Gβ 4 γ 3 , Gβ 4 γ 4 , and Gβ 4 γ 12 dimers may be left to interact with hetero-α 2a ARs.

Discussion
It is well defined that Gβγ dimers are released upon the activation of G i/o -coupled GPCRs, such as the α 2a AR, and act as important signaling units to various downstream signaling cascades to ultimately mediate various physiological functions 54-61 . It is not known whether all 32 possible neuronal Gβγs (combined from the known expression of 4 neuronal Gβs and 8 neuronal Gγs 28 ), are functional in vivo, however, how such sorting may take place to In vivo specificity of α 2a ARs for Gβγ. In this study, we have addressed the in vivo specificity of Gβ and γ interaction with the α 2a AR using MRM proteomics. We demonstrate that α 2a ARs preferentially interact with a subset of Gβ and Gγ subunits at synaptic terminals in vivo. Neuronal α 2a ARs (both auto-and hetero-α 2a ARs) interacted with Gβ 2 , Gβ 4 , Gγ 2 , Gγ 3 , Gγ 4 , and Gγ 12 while auto-α 2a ARs interacted with Gβ 2 , Gγ 2 , Gγ 3 , and Gγ 4 only. These findings suggest that Gβγs may shape signaling pathway specificity and that receptor and Gβγ interactions may be important in determining specific effector interactions. In our previous study, we found Gβ 1 as the most abundant Gβ subunit in whole synaptosomes as well as at both pre-and post-synaptic fractions 28 . Interestingly, however, in this study we did not find a statistically significant interaction between Gβ 1 and HA-α 2a ARs upon receptor activation ( Fig. 2A). Interestingly, we found Gβ 2 and Gβ 4 with activated α 2a AR instead, though there was more than 1,000-fold more Gβ 1 present at synapses. Despite the low abundance of Gβ 4 at the membrane 28 , Gβ 4 binding to α 2a ARs, as well as the exclusion of the highly abundant Gβ 1 , suggests a high specificity of this interaction. The numbers of receptors and effectors that specifically bind to unique Gβ and Gγ subunits may influence the abundance of certain Gβ and Gγ subunits at the membrane. For example, Gβ 1 may be specific to other receptors that are more abundant than α 2a ARs at synaptic terminals. Further studies are needed to determine these specificities, but these findings suggest that each receptor may utilize a unique set of Gβγ dimers to finely regulate receptor-specific downstream signaling. Moreover, we detected a minor interaction between Gγ 12 and HA-α 2a ARs but not with auto-α 2a ARs (Figs 3 and 5E). Although Gγ 12 was one of most abundant Gγ subunits at the membrane fraction in our previous study 28 , it was not specifically associated with auto-α 2a ARs, providing evidence for high specificity of the Gγ 12 subunit at the hetero-α 2a ARs. This suggests a Gβ 4 γ 12 dimer at hetero-α 2a ARs. In addition, Gβ 5 showed no specific interaction with α 2a ARs (Figs 2 and 3D), which supports previous studies that demonstrate it preferentially forms a stable dimer with the RGS R7 subfamily in vivo to modulate postsynaptic Gα i -mediated signal transduction pathways [20][21][22][23][24] .
As previously addressed 28 , we experienced some technical challenges in detecting and quantifying Gγ subunits with this method. The amount of detected Gγ subunits was not similar to the amount of detected Gβ subunits. This difference may be due to the differences in peptide yield, which could stem from post-translational modifications, sample preparation artifacts, and differences in peptide re-solubilization efficiencies, all of which can lead to systematic errors in quantification 62 . Because of these, we are unable to calculate absolute protein quantities, but we can accurately determine the expression pattern of neuronal Gβ and Gγ subunits and compare within Gβ and Gγ subunits.
No evidence for pre-coupling of α 2a AR GPCRs in vivo. The association of receptor and G protein prior to receptor activation ("pre-coupling") has been suggested in some studies, but still remains unclear 1,63-68 . For example, in in vitro FRET assay, activated α 2a ARs were found to interact with Gβ 1 32,33 . However, in our study using synaptosomes from brain tissue, we do not see significant basal association between α 2a ARs and Gβ and Gγ. And we see only non-specific interaction between Gβ 1 and α 2a AR, even though it is highly abundant pre-synaptically. By contrast, we saw significant interactions of Gβ 2 and Gβ 4 with α 2a ARs, but only after epinephrine activation of α 2a ARs. α 2a AR autoreceptors vs. heteroreceptors. Our findings suggest that unique Gβγ combination may play specific roles in mediating interactions with receptors. We found different Gβ and Gγ subunits in FLAG-tagged autoreceptors as compared to total HA-tagged α 2a ARs. This suggests that Gβγ specificities to receptors may change based on the cell type and localization of receptors. We estimate Gβ and Gγ subunit interactions with hetero-α 2a ARs by subtraction of presynaptic autoreceptor-associated Gβs and Gγs from total HA-α 2a AR-associated Gβs and Gγs, yielding the finding that Gβ 2 may be auto-α 2a AR specific, while Gβ 4 may be hetero-α 2a ARs specific. For Gγ subunits, Gγ 2 , Gγ 3 and Gγ 4 were determined to be auto-α 2a ARs specific, while Gγ 12 was hetero-α 2a ARs specific. (Table 1). Overall, hetero-α 2a ARs may associate with G protein heterotrimers paired with Gβ 4 γ 12 to mediate hetero-α 2a AR-specific phenotypes such as sedation and anesthetic sparing 37 . One difference between these two mice is that heteroreceptors may be found either pre-or post-synaptically, whereas autoreceptors are only pre-synaptic. We were not able to separate these two populations of heteroreceptors to determine whether this localization makes a difference. We were able to compare the results of these two studies side-by-side as similar levels of proteins were detected for most Gβ and Gγ subunits, however, one limitation of our studies is that we were unable to determine the differences in co-IP efficiency of HA-and FLAG-antibodies and the number of receptors in digested samples to calculate the relative Gβ and Gγ enrichment with hetero-α 2a ARs. Again, future studies with refined methodologies are needed to determine the functional consequences of identified specificities. Because HA-α 2a ARs represent both auto-and heteroreceptors and are found throughout the brain, we did not specify the neuronal type nor the location of receptors in the synaptosomes. Gβ 2 and Gβ 4 were previously identified to interact with α 2a ARs 30 , and in this study these Gβ subunits are identified to interact with Gγ 2 , Gγ 3 , Gγ 4 , Gγ 12 subunits. The rank order of Gγ specificity to overall neuronal α 2a ARs is similar to the Gγs found in whole and fractionated synaptosomes in the previous study 28 . It still remains unclear which Gγ subunits associate with each Gβ subunit. Though the rules for specificity determination are unknown, we assume that multiple factors affect the specificity: the preference of these Gβ subunits for Gγ subunits, the localization of receptors, and effector availability. The protein abundance and location of Gγ subunits will affect the Gβγ dimerization and their specificity to α 2a ARs.
Gβ and Gγ subunit specificity to α 2a ARs studied in vitro. Numerous in vitro studies have attempted to determine the specificity of Gβγ dimerization and their selectivity in interacting with various GPCRs and effectors 11,69,70 . Similar to our observations, Gβ 2 , Gβ 4 , Gγ 2 , Gγ 3 , and Gγ 4 were previously shown to be strongly associated with α 2a ARs 32,71 . Using FRET, Gibson and Gilman demonstrated that endogenous α 2a ARs preferentially stimulated Gα i1 heterotrimers paired with Gβ 1 or Gβ 4 , and Gα i3 heterotrimers paired with Gβ 2 32 . They also found that Gβ 2 association permitted 2-fold higher receptor activation, which was lost when Gβ 2 was replaced with Gβ 1 . This result and our studies suggest that α 2a ARs with Gα i3 β 2 γ heterotrimers may be most likely to be present at the in vivo synaptic terminals. Moreover, Gβ 2 γ and Gβ 4 γ dimers were determined to interact with adrenergic and opioid GPCRs, while Gβ 1 γ and Gβ 3 γ dimers, particularly Gβ 1 γ 3 and Gβ 3 γ 4 , may preferentially couple with somatostatin and muscarinic M4 GPCRs [29][30][31] . However, no specificity was identified based on the localization of receptors. In addition to the identify of Gα and Gγ subunits, the localization of receptor may play a role in α 2a AR selectivity of Gβ 2 and Gβ 4 over Gβ 1 . Depending on the localization of receptor, α 2a ARs may also preferentially interact with specific effectors. Based on our results and previous biochemical studies, Gβ 2 γ 2 , Gβ 2 γ 3 , and Gβ 2 γ 4 may be auto-α 2a ARs specific, while Gβ 4 γ 12 may be hetero-α 2a ARs specific.
Other in vitro G protein specificity studies 71-74 depict a different Gβ and Gγ specificity than seen in our study. The gap between in vitro and in vivo detection of G protein specificity may be explained by tissue-specific determinants of specificity that are not present in heterologous expression systems, or difference in expression and availability of Gβ and Gγ subunits for in vitro studies. It is clear that Gβγ subunits are sticky, and this is why we provided multiple controls for non-specific effects. Future studies will be needed to address these differences.
Role of Gα subunits in determining Gβγ specificity to α 2a AR receptors. In addition to Gβγ, Gα may also define the selectivity of G i/o -coupled GPCRs such as α 2a ARs. Unlike Gα s , much less is known about how GPCRs selectively activate inhibitor Gα i1-3 and Gα o subunits. Recent cryo-electron microscopy (cryoEM) studies reporting the structures of G i/o bound GPCRs, such as μ-opioid 75 , adenosine A 1 76 , 5HT 1B 77 , and light receptor rhodopsin 78 , determine the interaction of these receptors with G i or G o and suggest the conformational re-arrangements on the GPCR cytoplasmic site may affect the binding of specific G proteins. Interestingly, they found different interactions of G i/o bound GPCRs and Gβ subunits 79 . However, the role of Gβγ in GPCRs-G protein specificity is unclear in these studies due to the modification of the proteins and the resolution of cryoEM structures. Moreover, the studies of GABA B heteromeric receptors with GABA B1 and GABA B2 have suggested hetero-dimerization of GPCRs may also affect the binding interactions of Gβγ with the receptor 80,81 . Further studies are needed to determine how Gα subunits affect the specificity of Gβγ.
As a G i/o -coupled GPCR, α 2a ARs couple to Gα i1-3 and Gα o1-2 . In a previous study by Richardson and Robishaw, Gα i -containing heterotrimers were highly coupled to α 2a ARs 71 . Further, Gα i subunits were demonstrated to mediate sedative anesthetic-sparing effects, but not inhibition of evoked release 82 , and Gαi 1 were found to preferentially associate with Gβ 1 γ 3 over Gβ 1 γ 1 or Gβ 1 γ 10 71 . This suggests that Gα−mediated selectivity additionally contributes to the specificity of α 2a AR signaling through G proteins and their physiological functions. Further studies will be needed to understand the specific associations of Gα subunits with the Gβ and Gγ subunits observed here and their roles in known α 2a AR-mediated physiological effects.

Conclusions
With the quantitative MRM method 28 , we now can further elucidate the in vivo Gβ and Gγ specificities to other GPCRs as well as Gβγ effectors, and validate previous in vitro studies of the Gβγ dimerization and their selectivity in interacting with various GPCRs and effectors 11,69,70 . In the CNS, numerous Gβ and Gγ subunits exhibit interesting subcellular localizations 28,83 . We do not yet fully understand the importance of these localizations and their physiological role, however. This study begins to piece together the puzzle why multiple different isoforms of Gβ and Gγ subunits exist. Further efforts and development of tools, such as knockout or tissue-specific knockout animals, will be needed to determine the specificity and roles of each unique Gβγ dimer in regulating various GPCR signaling cascades, and their impacts on neurological diseases and GPCR targeted drug mechanisms. Eventually this will allow us to determine how cells precisely regulate multiple downstream mechanisms to modulate signal intensity and specificity. GPCR specificity to G proteins is defined by the Gα subunit preferred by a given GPCR. Whether GPCRs also have preference for Gβ and Gγ subunits is not well investigated. Here, we measured the in vivo specificity of presynaptic α 2a ARs to a subset of neuronal Gβ and Gγ subunits using a previously published proteomic approach. We found that Gβγ dimers, other than the most abundant Gβ 1 γ 2 , are also involved in α 2a ARs-mediated signaling cascades in vivo. In addition, auto-and hetero-α 2a ARs exhibit specificity to different Gβ and Gγ subunits. The variety of potential Gβγ dimers identified implies that the specificity of Gβγs to signaling pathways could be in part mediated through the receptors and their locations on particular types of neurons.

Materials and Methods
See supplementary for more details. Synaptosome. Crude synaptosomes were isolated from mouse brain tissue, as described previously 53,84,85 and stimulated with 100 µM epinephrine (epi). This mimics the local synaptic concentration of epinephrine and it is a commonly used concentration in alpha2a adrenergic receptor studies [86][87][88] . They were frozen in liquid nitrogen and stored at −80 °C.

Co-immunoprecipitation (Co-IP).
Crude synaptosomes were gently resuspended in 4 mL of RIPA buffer using a 25-gauge needle to lyse membranes and diluted to 1 mg/ml. Homogenates were centrifuged to separate the triton-soluble and insoluble fractions. Triton-soluble fractions were used for co-IP by incubating with either an anti-HA or FLAG antibody and Protein G agarose beads overnight. For elution, 100 µL of 1X sample buffer with DTT and 5% βME were used for HA-α 2a ARs and wildtype samples while 15.09 µg FLAG peptide was used for FLAG-α 2a ARs and α 2a ARs KO samples. Elutants were TCA precipitated and resuspended in 100 µL of 1x sample buffer with DTT and 5% βME. All samples were stored at −80 °C freezer for Western blot or MRM analysis.

Immunoblot analysis.
To examine the results of IP, Western blot analysis was performed on equal volumes of input, co-IP, and supernatant samples using 10% SDS-PAGE gels. Using Western Lightning ™ Chemiluminescence Reagent Plus (Perkin-Elmer) and Bio-rad Western blot imager, Western blots were developed.
Heavy labeled peptide cocktail. A heavy labeled peptide cocktail was made as described previously 28 .
Quantitative MRM of Gβ and Gγ subunits. Co-IP samples containing Gβ and Gγ subunits were separated, digested, and analyzed by a TSQ Vantage triple quadrupole mass spectrometer (Thermo Scientific) 28 . To allow comparisons between G proteins co-IPed from multiple mice, quantitative Gβ and Gγ subunits detected (fmol) were normalized by the amount of protein (mg) used in co-IPs. The amount of protein used in co-IPs was calculated using the volume of precleared lysate used and the protein concentration of precleared lysate from BCA assay. Statistical analysis. One-way analysis of variance (ANOVA) with a Tukey post hoc test was used to account for differences in protein expression of Gβ and Gγ subunits ( * p < 0.05, ** p < 0.01, *** p < 0.001). All statistical tests were performed using GraphPad Prism v.7.0 for Windows, (GraphPad Software, La Jolla, California, USA, www. graphpad.com).

Data Availability
All data generated or analyzed during this study are included in this published article (and its Supplementary  Information files).