Mutual action by Gγ and Gβ for optimal activation of GIRK channels in a channel subunit-specific manner

The tetrameric G protein-gated K+ channels (GIRKs) mediate inhibitory effects of neurotransmitters that activate Gi/o-coupled receptors. GIRKs are activated by binding of the Gβγ dimer, via contacts with Gβ. Gγ underlies membrane targeting of Gβγ, but has not been implicated in channel gating. We observed that, in Xenopus oocytes, expression of Gγ alone activated homotetrameric GIRK1* and heterotetrameric GIRK1/3 channels, without affecting the surface expression of GIRK or Gβ. Gγ and Gβ acted interdependently: the effect of Gγ required the presence of ambient Gβ and was enhanced by low doses of coexpressed Gβ, whereas excess of either Gβ or Gγ imparted suboptimal activation, possibly by sequestering the other subunit “away” from the channel. The unique distal C-terminus of GIRK1, G1-dCT, was important but insufficient for Gγ action. Notably, GIRK2 and GIRK1/2 were not activated by Gγ. Our results suggest that Gγ regulates GIRK1* and GIRK1/3 channel’s gating, aiding Gβ to trigger the channel’s opening. We hypothesize that Gγ helps to relax the inhibitory effect of a gating element (“lock”) encompassed, in part, by the G1-dCT; GIRK2 acts to occlude the effect of Gγ, either by setting in motion the same mechanism as Gγ, or by triggering an opposing gating effect.

Fluorescently (YFP or CFP)-labeled Gγ constructs are often used instead of wild-type (WT) Gγ, e.g. for imaging. We tested how such Gγ-based protein constructs affect GIRK1* currents. We tested Gγ, YFP-Gγ, YFP A207K -Gγ (the A207K mutation prevents dimerization of YFP or CFP 63 ) and CFP A207K -Gγ. Expression of all Gγ constructs caused a significant increase in GIRK1*'s I basal and I evoked (Fig. 3A,B). Effect on I basal was quantified as fold increase in I basal . CFP A207K and YFP A207K -tagged Gγ increased I basal similarly to WT Gγ. Interestingly, the YFP-Gγ lacking the A207K mutation caused the strongest activation of GIRK1*, ~7 fold (Fig. 3A). To address the possibility that dimerization of this YFP-fused construct somehow contributes to increased potency of Gγ activation of GIRK1*, we generated a Gγ concatemeric construct (Gγ tandem) consisting of two Gγ subunits joined head-to-tail. Coexpression of the Gγ tandem strongly activated GIRK1*, similarly to YFP-Gγ (Fig. 3A,C); the dose dependency on RNA dose was bell-shaped, like in the WT Gγ (Fig. 3C). These results indicate that formation of dimers enhances the potency of Gγ. Interestingly, I evoked was similarly, mildly potentiated by all Gγ constructs tested ScieNTific REPORTS | (2019) 9:508 | DOI: 10.1038/s41598-018-36833-y ( Fig. 3B), underscoring the complexity of underlying mechanisms(s). In the following we routinely used YFP-Gγ and Gγ tandem, which produce a better channel activation than the WT Gγ, and YFP tag allows measuring Gγ expression if needed. Key experiments have been repeated with WT Gγ to verify the authenticity of the observed phenomena. Next, we set to test whether YFP-Gγ recruits GIRK1* to the PM (which could cause an increase in the whole-cell GIRK current). Oocytes were injected with two concentrations of RNA GIRK1*, 0.2 and 1 ng, with or without YFP-Gγ (2.5 ng RNA). Giant excised PM patches were prepared from the oocytes, stained with an Representative records of GIRK currents showing that expression of Gγ (0.2 and 2 ng RNA/oocyte) increases I basal and I evoked of GIRK1*. GIRK1* was expressed at 0.2 ng RNA/oocyte together with m2R, 1 ng RNA/oocyte. Currents were first measured in a low-K+ solution (ND96) which was switched to the high K+ solution (hK, 96 or 24 mM K + , see Methods) resulting in an inward basal current, I hK . Then the oocyte was perfused with hK solution containing 10 µM ACh, to produce I evoked . At the end, 5 mM BaCl 2 was added to the solution to block GIRK currents and to reveal the residual non-GIRK current, I residual . I basal is defined as I hK -I residual , I evoked as the net additional inward current evoked by ACh. Here I basal and I evoked are shown graphically for the control (GIRK1*) record. In Gγ-or Gβγ-expressing oocytes, the basal current is termed I γ or I βγ , respectively, and defined as I hK -I residual . (B,C) Dose-dependence of the Gγ effect on I basal and I evoked . Increasing doses of Gγ RNA were injected, together with fixed amounts of RNAs of GIRK1* and m2R (same experiment as in A). Each point shown mean ± SEM, n = 9 to 16 cells, N = 1 experiment. *p < 0.05; **p < 0.01, ***p < 0. antibody against GIRK1, and the expression was measured using a confocal microscope (Fig. 3D). Coexpression of YFP-Gγ did not increase the level of GIRK1* in the PM (Fig. 3D,E), whereas GIRK1* basal currents measured in the oocytes of the same batch were increased (Fig. 3F). We conclude that the increase of GIRK1* current, caused by coexpressed of Gγ, is not due to an increase in the surface level of GIRK1* channels. Interestingly, activation of GIRK1* (fold increase in I basal ) by both YFP-Gγ and YFP-Gβγ was milder for higher expression levels of GIRK1* (1-2 ng RNA) than for the lower expression level (0.2 ng RNA; Supplementary  Fig. S2). The YFP-Gγ -induced increase in I evoked was 1.48 ± 0.05 fold (n = 152; Figs 3B, S2) for low GIRK1* expression levels, and no increase in I evoked was observed with high GIRK1* levels (0.95 ± 0.08, n = 34; Figs 3F, S2). Poor activation by GPCR agonists and Gβγ at high levels of channel expression has been reported previously for GIRK1* 18 and GIRK1/2 64 . In GIRK1/2, this phenomenon has been attributed to recruitment of free Gβγ by the channel, which results in increased basal activity and correspondingly reduced evoked responses 34,62 . We assume that a similar process may take place in GIRK1* which also recruits Gβγ 34 , but have not further pursued this subject here.
Effect of Gγ on I basal requires ambient Gβ. Since Gβ is considered as the main GIRK-interacting partner and activating moiety, we sought to investigate the possible involvement of Gβ in Gγ-induced activation of GIRK1*. To this end, we used phosducin, a Gβγ-binding protein which is widely used as a Gβγ "scavenger" [65][66][67] . Phosducin interacts with Gβγ via contacts mainly in Gβ subunit 68,69 , therefore it is not expected to sequester any Gγ that is not associated with Gβ. We purified His-tagged phosducin (His-phosducin) and verified that it binds Gβγ ( Supplementary Fig. S3A). We injected His-phosducin into the oocytes to a final concentration of ~23 µM within the cell, at least 40-50 minutes before measuring currents (Fig. 4A). When GIRK1* was expressed alone, phosducin did not significantly reduce I basal (Fig. 4D). We assume that, although Gβγ significantly contributes to I basal in this channel 18 , the expected reduction in I basal was obscured because of the relatively low I basal observed in this experiment. In all other test groups, injection of His-phosducin into oocytes decreased GIRK1* current ( Fig. 4B-D): by 74% for GIRK1* coexpressed with Gβγ, by 79% for GIRK1* coexpressed with YFP-Gγ, and by 63% for GIRK1* coexpressed with Gβ + YFP-Gγ (Fig. 4D). The inhibition of Gγ-YFP -induced GIRK1* activity suggests that the effect of Gγ depends on the presence of endogenous (ambient) Gβ.
An additional way of using phosducin was coexpression of myristoylated phosducin (myr-phosducin) by the injection of its RNA into the oocytes (Fig. 4E). The myristoylation tag at the N-terminus of myr-phosducin targets it to the membranes, including the PM 65 . Expression of myr-phosducin significantly decreased I basal of GIRK1* alone (Fig. 4H), as reported previously 18 . Myr-phosducin also inhibited ~90% of the GIRK1* current activated by coexpressed YFP-Gγ (Fig. 4G,H). Strikingly, when GIRK1* was activated by YFP-Gγ with coexpressed Gβ (5 ng RNA), expression of myr-phosducin increased the current (Fig. 4F,H), or had no effect ( Supplementary  Fig. 3B). The possible reason for this seemingly paradoxical effect became clear only later, after titration of Gβ concentrations (see Fig. 5 and the Discussion). We assumed that the expressed amount of phosducin is not sufficient to fully sequester all expressed Gβ and therefore does not inhibit GIRK1* activation. In support, when we injected a lower dose of Gβ (0.5 ng instead of 5 ng as in Fig. 4) together with YFP-Gγ, activation of GIRK1* was very strong but phosducin almost completely inhibited it ( Supplementary Fig. S3B). In summary, inhibition of YFP-Gγ -induced activation by coexpressed or added (as purified protein) phosducin, supports the notion that Gγ -induced activation of GIRK1* requires ambient Gβ.
Complex stoichiometric relationships between Gγ and Gβ. To better understand the mutual dependence of actions of Gβ and Gγ on GIRK1*, we titrated Gβ RNA in the presence of a fixed concentration of YFP-Gγ RNA (2.5 ng/oocyte). As before, YFP-Gγ increased both I basal and I evoked ( Fig. 5A-C). Activation of GIRK1* by YFP-Gγ was strongly affected by coexpression of Gβ in a complex manner. A low dose of Gβ (0.5 ng) enhanced the effect of YFP-Gγ but higher doses of Gβ reversed this effect. Gβγ still activated GIRK1*, but with higher levels of expressed Gβ the activation was even lower than with YFP-Gγ alone (Fig. 5B). Strikingly, expression of even the lowest dose of Gβ, 0.5 ng RNA, suppressed I evoked (Fig. 5C). Figure 5D,E summarize the effects of coexpression of YFP-Gγ alone and with 5 ng Gβ from all experiments, showing that the reduction of YFP-Gγ -induced activation of GIRK1* by a high dose of Gβ, and the suppression of I evoked , were highly reproducible and significant. Similarly, low doses of Gβ potentiated activation of GIRK1* induced by WT Gγ or YFP A207K -Gγ, and this effect was diminished when the dose of Gβ was increased ( Supplementary Fig. S4). These results support the notion of mutual dependence of action of Gγ and Gβ. It appears that overexpression of either Gβ (Fig. 5) or Gγ ( Fig. 1) is counterproductive for optimal channel activation, possibly through sequestration of one subunit by an excess of the other one (see Discussion).
Gγ activates GIRK1/3 but not GIRK2 or GIRK1/2. We explored whether Gγ affects neuronal GIRK channels of different subunit composition, starting with GIRK2. In contrast to GIRK1*, GIRK2 was not activated by Gγ and showed a canonical activation by Gβγ, with greater extent of activation with higher Gβ concentrations  Fig. 6A,B). YFP-Gγ also did not activate GIRK2 (Fig. 6C). Co-expression of Gβ with YFP-Gγ increased I βγ as expected and to a similar extent as Gβ coexpressed with WT-Gγ (Fig. 6C). The small decrease in I basal of GIRK2 by YFP-Gγ ( Fig. 6C) may be due to changes in channel expression. These results indicate a specific role of GIRK1 subunit in mediating the effect of Gγ. Next, we tested the effects of Gγ on GIRK1/2 and GIRK1/3, the most abundant neuronal GIRK channels. Fig. 7A,B shows a representative experiment in which GIRK1/3 was co-expressed with increasing doses of Gγ RNA. Gγ significantly increased I basal up to a ~2 fold at 2 ng Gγ RNA (Fig. 7A,B). Interestingly, unlike in GIRK1*, here we did not observe the bell-shaped dose-response relationship, but we have not tested higher doses of Gγ. YFP-Gγ and Gγ tandem also significantly increased GIRK1/3 I basal , about 5-and 2.5 fold, respectively (Fig. 7C,D). I evoked was not significantly affected by WT-Gγ (104 ± 9% of control, n = 12, N = 2, not shown) but it was significantly increased by YFP-Gγ (153 ± 11%, p < 0.001, n = 46, N = 6, not shown). In a separate experiment, we have monitored the effect of Gγ tandem on surface expression of GIRK1/3 in giant excised PM patches ( Supplementary  Fig. S5). Titrated expression of RNA of the Gγ tandem indicated that at concentration that produced maximal activation of GIRK1/3 (0.25 ng RNA, Supplementary Fig. S5C), the Gγ tandem did not affect the level of GIRK1/3 in the PM. These experiments suggest that coexpression of Gγ increases GIRK1/3 currents without affecting the channel's surface levels.
Unlike GIRK1* or GIRK1/3, the basal activity of GIRK1/2 was not affected by coexpression of Gγ or YFP-Gγ ( Fig. 7E-G). I evoked was also unaffected (not shown). Note that I basal of GIRK1/2 was > 4 µA (Fig. 7E,F) with a low dose of RNA, 0.05 ng RNA of each subunit. I basal of GIRK2 was 0.36 ± 0.05 µA (n = 45) with a 40-fold higher dose of RNA, 2 ng/oocyte (for review on I basal differences in GIRK channels, see ref. 29 ). Therefore, >90% of I basal in this case originated from GIRK1/2 heterotetramers rather than from any incidentally present GIRK2 homotetramers. We conclude that GIRK2 is not activated by Gγ, and it also appears to prevent Gγ-induced activation of GIRK1 in a GIRK1/2 heterotetrameric context. Distal C-terminus of GIRK1 (dCT) is important for Gγ-induced activation. We hypothesized that a structural difference between GIRK1* and GIRK2 may explain the divergent effects of Gγ on GIRK1* and GIRK2 in the homomeric context. One candidate structural element is the unique distal C-terminus of GIRK1, G1-dCT, which contributes to Gβγ anchoring and channel gating 29 . Both GIRK2 and GIRK1* Δ121 (which is a deletion mutant of GIRK1* without the dCT) lack this element and cannot recruit Gβγ. Accordingly, they have low basal activity and stronger relative activation by Gβγ compared to GIRK1* and GIRK1/2, suggesting that the Gβγ activation site in channel's core is intact 18,34 .  To assess the possible role of G1-dCT, we coexpressed YFP-Gγ with GIRK1* Δ121 . Figure 8A,B shows an exemplary experiment, Fig. 8C summarizes two experiments of this kind. YFP-Gγ had no effect on GIRK1* Δ121 expressed at low density (0.2 ng RNA), whereas YFP-Gγ increased GIRK1* I basal as expected (Fig. 8A-C). When GIRK1* Δ121 was expressed at a higher level, 2 ng RNA, WT Gγ did not enhance I basal , but YFP-Gγ appeared to produce a residual 2-fold activation of GIRK1* Δ121 that did not reach statistical significance in one-way ANOVA test (Supplementary Fig. S6A). Notably, both WT Gγ and YFP-Gγ significantly reduced I evoked of GIRK1 Δ121 *, further underscoring the importance of G1-dCT for Gγ regulation of GIRK1* (Supplementary Fig. S6B). We also tested GIRK1 Δ121 as a heterotetramer with GIRK3, GIRK1 Δ121 /3. This channel expressed well and its level in the PM was similar to that of GIRK1/3 containing the full-length GIRK1 subunit (Fig. 8D,E). Coexpression of Gγ tandem did not change the PM level of GIRK1/3; the PM level of GIRK1 Δ121 /GIRK3 was reduced by ~20% by the higher dose of Gγ tandem used, 2 ng RNA/oocyte (Fig. 8D,E). Unlike GIRK1/3, GIRK1 Δ121 /3 was not activated by coexpression of Gγ tandem (Fig. 8F). Gβγ strongly activated both GIRK1/3 and GIRK1 Δ121 /3 (Fig. 8F). Thus, removal of the G1-dCT significantly reduces the activating effect of Gγ on GIRK1* and GIRK1/3, suggesting a role for G1-dCT in this action of Gγ.
We next wanted to test whether addition of G1-dCT to GIRK2 will render the channel sensitive to Gγ. For this purpose, we used the chimera GIRK2HA/GIRK1dCT, which contains the G1-dCT (a.a. 371-501) fused to dCT-less GIRK2 (a.a. 1-381), as well as an extracellular HA tag (see cartoon in Fig. 9B). We have previously shown that C-terminal fusion of G1-dCT confers upon GIRK2HA the ability to recruit Gβγ to the PM; accordingly, this chimera has a much greater I basal than GIRK2HA 18,34 . However, YFP-Gγ did not activate the GIRK2HA/ G1-dCT channel. In the same experiment, YFP-Gγ increased GIRK1* I basal as expected (Fig. 9A,B). This result indicates that although G1-dCT is necessary for Gγ effect, it is not sufficient and possibly requires additional structural elements of the channel.

Discussion
In performing its cellular functions, Gβγ acts as an obligatory, stable dimer, dissociated only by denaturation 43,70 . Gβ and Gγ are synthesized separately 59,71 ; dimerization greatly increases the stability of each subunit [72][73][74] . Folding of Gβ requires the chaperone CCT complex, from which Gβ is released by the co-chaperone PhLP1 59,71 . Mature Gβγ dimer is formed only after the binding of Gγ 75 . Lipid modification (prenylation) of Gγ is not required for association with Gβ, but is crucial for PM targeting of Gβγ and for activation of GIRKs 45,46 . In accord with these cellular mechanisms, we found that heterologous expression of Gβ alone did not activate GIRK1* or GIRK2 channels (Figs 2, 6, S6). Unlike Gβ, Gγ can fold separately and is produced in cells in the absence of Gβ 71,76 , especially when fused to GFP, though still less than with Gβ 60 . We found that overexpression of Gγ or Gγ tagged with CFP or YFP enhances the activity of homomeric GIRK1* and heterotetrameric GIRK1/3 channels. Interestingly, the strongest activation was achieved by YFP-Gγ without the A207K mutation, where YFP is prone to dimer formation 63 . Gγ tandem (concatemer) consisting of two fused Gγ subunits activated GIRK1* and GIRK1/3 similarly to YFP-Gγ (Figs 3, 7, 8), excluding a role for the xFP moiety in channel activation. No natural Gγ dimer formation has been reported; but we hypothesize that artificial dimerization can protect against degradation and increase Gγ's stability. Alternatively, dimerization could be related to some steric aspect of the mechanism of action of Gγ.
To address the action of Gγ, we first considered the possibility that expression of Gγ facilitates synthesis or trafficking to PM of GIRK1, which could account for the observed increase in I basal and I evoked . However, direct immunocytochemical measurements in giant excised PM patches consistently showed no change in the amount of GIRK1* and GIRK1/3 channel protein in the PM (Figs 3, 8, S5).
We then hypothesized that the expressed Gγ may "recruit" Gβ to the PM. Absence of GIRK activation by expression of Gβ alone suggested that there was no free ambient (endogenous) Gγ to form functional dimers with the expressed Gβ. It was plausible, however, that expression of exogenous Gγ could enhance the release of  75 , and in this way elevate the total Gβγ in the PM. However, several lines of evidence argue against the Gβ recruitment hypothesis. First, we directly measured the PM levels of endogenous Gβ using Western blots of manually separated PMs. Expression of Gγ at doses that produced robust activation of GIRK1* did not significantly change PM levels of Gβ (Fig. 1). Second, Gγ and its derivatives did not decrease or (for low GIRK1* expression levels) increased I evoked of GIRK1*, whereas expression of Gβγ always decreased it (Figs 2, 5, S2). This is incompatible with recruitment: even a small addition of Gβ (if it were recruited by Gγ) on top of Gγ would reduce I evoked in GIRK1* (see Fig. 5). Third, expression of Gγ or YFP-Gγ did not activate either GIRK2, which is highly sensitive to expressed Gβγ 18 (Fig. 6), or GIRK1* Δ121 that retains the GIRK1 Gβγ-activation site and is strongly activated by Gβγ 34 , or GIRK1 Δ121 /3 (Fig. 8). The inability of Gγ to activate these channels strongly argues against the Gβ-recruitment hypothesis. All evidence considered, we conclude that changes in PM levels of either GIRK or Gβγ cannot explain the activating effect of Gγ. We therefore propose that Gγ regulates the gating of GIRK1* and GIRK1/3. How can Gγ regulate GIRK gating? A key insight into the mechanism of Gγ action comes from the finding that Gγ action is blocked by phosducin (Fig. 4). Phosducin interacts with Gβ but not with Gγ 68,69 , implicating Gβ in the effect of Gγ. Another indication of the involvement of Gβ is the enhancement of Gγ effect by coexpression of low doses of Gβ (Figs 5, S4). Therefore, we suggest that the presence of ambient Gβ is essential for the activation of GIRK1* and GIRK1/3 by Gγ. Since the presence of free Gβ in cells is unlikely, we propose that the effect of Gγ requires the presence of ambient Gβγ, which is dynamically associated with the GIRK1-containing channels 29 .
Another key insight came from the lack of GIRK2 activation by Gγ (Fig. 6). It implied the possible involvement of G1-dCT, which is unique to GIRK1, in Gγ regulation of GIRK1* and GIRK1/3. In support, deletion of G1-dCT greatly reduced the Gγ-induced activation of GIRK1* and GIRK1/3 (Figs 7, 8, S6) and completely abolished the Gγ-induced increase in I evoked , causing a decrease instead (Fig. S6). To further address the function of G1-dCT, we used the chimera G2HA/G1-dCT in which the short dCT of GIRK2 (~34 amino acids) is replaced by G1-dCT. Addition of G1-dCT endows this chimera with an enhanced Gβγ binding compared to GIRK2; it also recruits Gβγ to the PM 34 . However, the G2HA/G1-dCT channel was not affected by Gγ (Fig. 9), suggesting that G1-dCT alone is not sufficient to confer Gγ activation to GIRK2. Core elements in GIRK1 may also be involved; indeed, cross-talk between gating effects of G1-dCT and core of GIRK1 has been proposed 31 . Furthermore, since the G2HA/G1-dCT channel recruits Gβγ, we posit that Gγ activation is not related to recruitment of Gβγ. We hypothesize that it may be related to the second function of the G1-dCT, which is an inhibitory one, as explained below.
In light of these considerations, we propose a model (Fig. 10) in which Gγ aids Gβ to drive the opening of GIRK1* and GIRK1/3 by acting on a gating element within the channel, rather than by participating in binding to the activation site. We envision that this function is normally carried out by Gγ from within the Gβγ dimer, e.g. when Gβγ is released from the Gαβγ heterotrimer following GPCR-catalyzed GDP-GTP exchange (Fig. 10, "activated GIRK1*). We propose that the coexpressed, properly prenylated Gγ can reach the vicinity of the channel and act on the same gating element, further helping Gβγ to shift the closed-open equilibrium in favor of the open Figure 10. A hypothetical scheme of regulation of a GIRK channel containing the GIRK1 subunit by Gγ. In GIRK1-containing channels, Gβγ or the Gαβγ heterotrimer (shown in the figure) may be anchored to GIRK. In the resting state, the interaction surface of Gβ is occluded by Gα and cannot contact the activation site of GIRK. Lock element (encompassing the G1-dCT and other unknown parts of the channel) is closed, reducing channel activity. Upon activation by agonist, the GPCR (not shown) activates the G protein causing dissociation of Gα from Gβγ, exposing the GIRK-interacting surface of Gβ. Gβγ may now bind to the activation site. We propose that, at the same time, Gγ interacts with a channel's element and helps to release the inhibitory effect of the "lock". Exogenous Gγ may mimic this action without activating the channel by itself, but only if Gβγ is present.
ScieNTific REPORTS | (2019) 9:508 | DOI:10.1038/s41598-018-36833-y state. Since GIRK1/2 is not regulated by Gγ, we propose that GIRK2 either counteracts or occludes the action of Gγ; this will be discussed separately. The nature of the gating element affected by Gγ is currently unclear. As a working hypothesis, we put forward the involvement of the hypothetical "lock" encompassed, in part, by the G1-dCT. A peptide corresponding to the last 20 amino acids of GIRK1 reduced the open probability (P o ) of Gβγ-activated GIRK1/5 and GIRK1/4 channels by a non-competitive mechanism, suggesting a gating effect rather than competition for Gβγ binding 36 . Accordingly, the "lock"-deficient GIRK1* Δ121 has a higher maximal Gβγ-induced P o that GIRK1* 34 . Thus, one of the possible scenarios of Gγ action could be the removal of the inhibitory effect of the "lock", which would increase both I basal and I evoked by increasing the P o of Gβγ-activated channels. We emphasize that other mechanisms in which Gγ allosterically regulates (enhances) the P o will result in the same action. The proposed mechanism helps to explain, and is supported by, the findings obtained used the "expression pharmacology" approach 62,77 (titration of protein expression in oocytes by injecting a range of RNA doses).
Finding #1: Gγ increases I basal and I evoked of GIRK1* and GIRK1/3, but expression of free Gβγ suppresses I evoked (Figs 1-5). Explanation: Gγ acts by enhancing Gβγ-induced activation, irrespective of the source of Gβγ (a pre-associated "basal" Gβγ or Gβγ released from Gαβγ through the activation of GPCR). Hence the increase both in I basal and I evoked . In contrast, coexpressed Gβγ suppresses I evoked because it already maximally activates GIRK channels 21 , and also sequesters Gα away from the GIRK-G protein complex, reducing activation by the GPCR 18 .
Finding #2: Expression of small amounts of Gβ enhances the Gγ-induced activation better than higher amounts of Gβ (Figs 5, S2). Explanation: GIRK1-containing channels are associated with excess of Gβγ over Gα 29 , possibly already with 3-3.5 Gβγ molecules per channel 62  Finding #3: Expression of high doses of Gγ produces a smaller activation of GIRK1* than an optimal, lower dose (Figs 1, 3). Explanation: similarly to the excess of Gβ, excess Gγ may be sequestering Gβ "away from the channel". However, more complex mechanisms cannot be excluded.
Finding #4: In the presence of coexpressed high dose of Gβγ, expression of phosducin further increased I basal (Fig. 4H). Rather counterintuitive, this finding is consistent with the proposed mechanistic framework. With 5 ng phosducin RNA, sequestration of Gβγ is probably incomplete 56,65 and leaves a small amount of "extra" expressed Gβγ. The latter, together with added coexpressed Gγ, activates the channel to a high level, as explained above. In line with this, with low dose of Gβ RNA (0.5 ng) phosducin effectively counteracted the action of Gγ, supporting the role of Gβ and underscoring the importance of stoichiometric considerations. Interestingly, the injected purified phosducin decreased channel activation in all conditions, suggesting a more complete Gβγ sequestration. Other possibilities, such as a long-term effect of coexpressed myr-phosducin on cellular levels of Gβγ or interactions within the signaling complex, cannot be ruled out.
In addition to mutual sequestration of Gβ, Gγ and phosducin, mechanisms that could contribute to the reduction in Gγ activation at high Gγ or Gβ levels include the formation of inactive Gβγ oligomers or Gγ aggregates 74 , or variation in stoichiometry of Gα that may alter channel's activity. In summary, we have found that stoichiometric relationships between expressed proteins crucially determine the observed regulations of GIRKs by Gβ and Gγ. Mutual sequestration or formation of protein oligomers of inadequate stoichiometry probably explain the complex, bell-shaped dose-response relationships of Gγ and Gβ effects (Figs 1, 3, 5, S2). The new insights obtained here underscore the power of the "expression pharmacology" approach for studies of complex regulatory mechanisms in heterologous expression systems.
Importantly, GIRKs of different subunit composition showed diverse regulation by Gγ. Only GIRK1/3 was regulated similarly to GIRK1*; GIRK2 and GIRK1/2 were not affected by Gγ. These findings carry a potential physiological relevance because of the specific and diverse subunit composition of GIRKs in the brain. Furthermore, they may provide new insights as to the mechanism of Gγ action and, generally, of the regulation of GIRKs by G protein subunits. If, as we have proposed, Gγ allosterically regulates GIRK by acting on a gating element (such as the "lock" present in GIRK1), then GIRK2 may contain a structural element that exerts the same effect, whereas GIRK3 does not. In this way GIRK2 may occlude the effect of Gγ. This would explain why GIRK1/3 is regulated by coexpression of Gγ whereas GIRK1/2 is not. Another possibility is that GIRK2 counteracts the effect of Gγ by acting on a different structural element within the GIRK1/2 heterotetramer.
Absence of activation of GIRK1/2 by coexpressed Gγ seems to be at odds with the previous finding that Gγ is essential for Gβ to activate GIRK1/2 54 . However, a more detailed examination of our and Kawano's data reveals that there is no controversy. Kawano et al. 54 showed that mutated Gβ that does not bind Gγ can still associate (at least it is co-immunoprecipitated) with GIRK subunits, but it cannot activate GIRK1/2. These results are supported by our data where Gβ alone cannot activate any of the GIRK channels tested, and corroborate the notion that, without prenylation of Gγ, Gβγ cannot reach the PM and cannot activate any GIRK channel 45,46 . In comparison, our approach with expression of Gγ reveals unknown mechanistic aspects of Gγ action with physiologically relevant, functional Gβ and Gγ proteins. Our results show that Gγ not only is essential for GIRK activation by the Gβγ dimer, but also actively supports the Gβγ-induced transition to open state.

Conclusions.
We demonstrate that the Gγ subunit contributes to Gβγ-induced activation of GIRK channels in a GIRK subunit-specific manner. Expression of Gγ alone activated homotetrameric GIRK1* and heterotetrameric GIRK1/3 channels, but not GIRK2 or GIRK1/2. In GIRK1* and GIRK1/3, Gγ increases both I basal and I evoked , without affecting surface expression of the channels. Our results suggest that, besides its known role in targeting Gβγ to the plasma membrane, Gγ regulates the gating of GIRKs, in concert with Gβ. The unique distal ScieNTific REPORTS | (2019) 9:508 | DOI:10.1038/s41598-018-36833-y C-terminus of GIRK1, G1-dCT, is important but not sufficient for Gγ action. As a working hypothesis, we propose that Gγ regulates GIRK1* and GIRK1/3 channels by relaxing the inhibitory effect of the "lock" which is encompassed, in part, by the G1-dCT. We further hypothesize that, within the GIRK1/2 heterotetramer, GIRK2 acts to occlude the effect of Gγ, either by operating through the same mechanism as Gγ, or by triggering an opposing gating effect.

Ethical approval and animals. Experiments were approved by Tel Aviv University Institutional Animal
Care and Use Committee (permits M-08-081 and M-13-002). All experiments were performed in accordance with relevant guidelines and regulations. Xenopus laevis female frogs were maintained and operated as described 78 . Frogs were kept in dechlorinated water tanks at 20 ± 2 °C on 10 h light/14 h dark cycle, anesthetized in a 0.17% solution of procaine methanesulphonate (MS222), and portions of ovary were removed through a small incision in the abdomen. The incision was sutured, and the animal was held in a separate tank until it had fully recovered from the anesthesia and then returned to post-operational animals' tank. The animals did not show any signs of post-operational distress and were allowed to recover for at least 3 months until the next surgery. Following the final collection of oocytes, after 4 surgeries at most, anesthetized frogs were killed by decapitation and double pithing.
DNA constructs, RNA and purified phosducin. cDNA constructs of YFP-or CFP-labeled and unlabeled GIRK subunits, Gβ 1 , Gγ 2 , m2R and myristoylated phosducin (myr-Phosducin) constructs were cloned into pGEM-HE, pGEM-HJ or pBS-MXT vectors, which are high expression oocyte vectors containing 5′ and 3′ untranslated sequences of Xenopus β-globin 79 , as described 65,80 . Constructs are described in Table 1. All PCR products were fully sequenced. Fluorescent xFP proteins (CFP A207K and YFP A207K ) usually contained the A207K mutation that prevents their dimerization 63 ; however, when indicated, YFP was also used without the A207K mutation. Point mutations were introduced using PCR with the Pwo Master polymerase (Roche) according to manufacturer's instructions, with primers containing the desired mutation. Afterwards, DpnI (New England Biolabs, R0176) was added to the reaction in order to degrade the template. The cDNA constructs were fully sequenced. RNA was transcribed in vitro essentially as described 78 but precipitated overnight at −20 °C with 4 M LiCl instead of the standard ethanol/salt precipitation. RNA was divided into 1-2 μl aliquots and stored at −80 °C. The amounts of RNA injected per oocyte were varied according to the experimental design and are indicated in the results or in figure legends.
For His-phosducin protein production, the coding sequence of bovine phosducin cDNA (see Table 1) was subcloned into pETMII vector which adds an N-terminal His-tag. Protein was amplified in BL-21 E. coli. Protein purification was done with Ni-NTA column using the following buffer: 50 mM KH 2 PO 4 , 20 mM Tris-HCl, 100 mM NaCl, 5 mM β-mercaptoethanol, 250 mM Imidazole. The size of His-phosducin is ~29 kDa and its initial concentration was 13.5 mg/ml, or 465 µM. The injection volume per oocyte was 50 nl, therefore the final concentration of His-phosducin was ~23 µM in each oocyte (assuming oocyte volume of ~1 µl).  Fig. 1A. Net basal GIRK currents were calculated by subtracting the residual Ba 2+ -insensitive current recorded in each cell at the end of the recording protocol, or, on rare occasions, by subtracting average current recorded in naïve oocytes of the same experiment. Data acquisition and analysis were performed using the pCLAMP 9 or pCLAMP 10 software (Molecular Devices, Sunnyvale, CA, USA).

Measurement of Gβγ in plasma membrane (PM) by Western blotting.
Plasma membranes were separated from the rest of the oocyte ("cytosol") as described 61,83 . In brief, PM together with the vitelline membranes (extracellular collagen-like matrix) was removed manually with fine forceps after a 5-15 min incubation of the oocyte in a low osmolarity solution (5 mM NaCl, 5 mM HEPES, and protease inhibitors (Roche Complete Protease Inhibitors Cocktail, 1 tablet/50 ml), pH = 7.5). The remainder of the cell ("cytosol") was processed separately, after removing of nuclei by centrifugation for 10 min at 700 × g at 4 °C. PMs (18-25 per lane) were solubilized in 35 μl running buffer (2% SDS, 10% glycerol, 5% β-mercaptoethanol, 0.05% Bromophenol Blue, 62.5 mM Tris-HCl, pH 6.8). Samples were electrophoresed on 12% polyacrylamide-SDS gel and transferred to nitrocellulose membranes for Western blotting with previously characterized 62 antibody against Gβ at 1:500 or 1:1000 dilution (Santa Cruz Biotechnology, SC-378). Goat Anti-Rabbit IgG Antibody, (H + L) HRP conjugate secondary antibody at 1:40,000 dilution was applied (Merck Millipore, AP307P). The signals were visualized using the SuperSignal kit (Thermo, 15168) and images were obtained with the fluorescent imager Fusion FX7 (Vilber Lourmat, Germany) and quantitated using the ImageJ software (National Institutes of Health, USA). Oocytes of stage 6 84 ) were used. These large cells have a rather constant size (~1 mm in diameter); the total amount of protein in the PM is considered uniform, and in Western blots of oocyte's protein number of loaded oocytes rather than µg protein is routinely reported, and no normalization of measured amount of Gβ to a housekeeping protein is considered necessary (e.g. refs 61,62,83,85 ). The same amount of cells was used for each lane on the gel (Fig. 1D, Supplementary Fig. S1).
Giant excised PM patches. Giant excised PM patches were prepared, stained with antibodies and imaged as described 86 . Oocytes were mechanically devitellinized using fine forceps in a hypertonic solution (in mM: NaCl 6, KCl 150, MgCl 2 4, HEPES 10, pH 7.6). The devitellinized oocytes were transferred onto a Thermanox ™ coverslip (Nunc, Roskilde, Denmark) immersed in a Ca 2+ -free ND96 solution, with their animal pole facing the coverslip, for 10-20 minutes. The oocytes were then suctioned using a Pasteur pipette, leaving a giant membrane patch attached to the coverslip, with the cytosolic face toward the medium. The coverslip was washed thoroughly with fresh ND96 solution and fixated using 4% formaldehyde for 30 minutes. Fixated giant PM patches were immunostained in 5% milk in phosphate buffer solution (PBS). Non-specific binding was blocked with Donkey IgG 1:200 (Jackson ImmunoResearch, West Grove, PA, USA). Anti-Kir3.1 (GIRK1) antibody (Alomone labs, APC-005) or Anti-Kir3.3 (GIRK3) antibody (Alomone labs, APC-038) were applied at 1:200 or 1:100 dilution respectively, for 45 minutes at 37 °C. Anti-rabbit IgG DyLight650-labeled secondary antibody 1:400 (Abcam, ab96886) was then applied for 30 minutes in 37 °C, washed with PBS, and mounted on a slide for visualization. Immunostained slides were kept in 4 °C for no more than a week.

Confocal imaging.
Confocal imaging and analysis were performed as described 34,80 , with a Zeiss 510 META confocal microscope, using a 20x objective. In whole oocytes, the image was focused on oocyte's animal (dark) hemisphere, at the equator. Images were acquired using spectral (λ)-mode: CFP was excited with a 405 nm laser and emission was collected at 481-492 nm. YFP was excited with the 514 nm line of the argon laser and emission was collected at 535-546 nm. Fluorescent signals were averaged from 3 regions of interest (ROI) at the PM and 3 similar ROIs from the coverslip outside the oocyte's image, using Zeiss LSM Image Browser. The average background signal was subtracted from the average PM signal in each oocyte, and then the average net signal from the membrane of uninjected (naïve) oocytes was subtracted as well.
Imaging of proteins in giant PM patches was performed using the confocal microscope in λ-mode. DyLight650 was excited using 633 nm laser and emission was collected at 663-673 nm. Images centered on edges of the membrane patches, so that background fluorescence from coverslip could be seen. Two ROIs were chosen: one comprising most of the membrane patch within the field of view, and another comprising background fluorescence, which was subtracted from the signal obtained from the patch. The signal from giant PM patches of naive oocytes' membranes, immunostained using the same protocol, was subtracted from all groups.
Statistical analysis and data presentation. Imaging data on protein expression, as well as GIRK currents data collected from several experiments, have been normalized as described previously 87 . Fluorescence intensity or current in each oocyte was normalized to the average signal in the oocytes of the control group of the ScieNTific REPORTS | (2019) 9:508 | DOI:10.1038/s41598-018-36833-y same experiment. This procedure yields average normalized intensity or current, as well statistical variability (e.g. SEM), in all treatment groups as well as in the control group. Statistical analysis was always performed on raw data with SigmaPlot 11 or SigmaPlot 13 (Systat Software Inc., San Jose, CA, USA). Two-group comparisons were performed using t-test if the data passed the Shapiro-Wilk normality test and the equal variance test, otherwise we used the Mann-Whitney Rank Sum Test. Multiple group comparisons were done using one-way ANOVA (ANOVA on ranks was performed whenever the data did not distribute normally). Tukey's or Dunnet's tests were performed for normally distributed data and Dunn's test otherwise. The data in the graphs are presented as mean ± SEM or as raw data with superimposed box plots indicating 25-75 percentiles (box borders), median, mean (usually red line), and for some sets of data also 5-95 percentiles (whiskers).

Data Availability Statement
The data that support the findings of this study are available from the corresponding author, N.D, upon reasonable request.