Na+ Promotes the Dissociation between GαGDP and Gβγ, Activating G Protein-gated K+ Channels*

G protein-gated K+channels (GIRK, or Kir3) are activated by the direct binding of Gβγ or of cytosolic Na+. Na+ activation is fast, Gβγ-independent, and probably via a direct, low affinity (EC50, 30–40 mm) binding of Na+ to the channel. Here we demonstrate that an increase in intracellular Na+ concentration, [Na+]in, within the physiological range (5–20 mm), activates GIRK within minutes via an additional, slow mechanism. The slow activation is observed in GIRK mutants lacking the direct Na+ effect. It is inhibited by a Gβγ scavenger, hence it is Gβγ-dependent; but it does not require GTP. We hypothesized that Na+ elevates the cellular concentration of free Gβγ by promoting the dissociation of the Gαβγ heterotrimer into free GαGDP and Gβγ. Direct biochemical measurements showed that Na+ causes a moderate decrease (∼2-fold) in the affinity of interaction between GαGDP and Gβγ. Furthermore, in accord with the predictions of our model, slow Na+ activation was enhanced by mild coexpression of Gαi3. Our findings reveal a previously unknown mechanism of regulation of G proteins and demonstrate a novel Gβγ-dependent regulation of GIRK by Na+. We propose that Na+ may act as a regulatory factor, or even a second messenger, that regulates effectors via Gβγ.

GIRK 1 (Kir3) channels are crucial for the regulation of heartbeat and for inhibitory actions of many neurotransmitters in the brain. They are activated by direct binding of G␤␥ released from heterotrimeric G proteins following activation of G protein-coupled receptors (GPCR) (1)(2)(3). GIRK activity also crucially depends on the presence of membrane phosphatidylinositol 4,5-bisphosphate (4). Cytosolic Na ϩ has been shown to activate GIRK by a direct, G protein-independent mechanism. The direct activation by Na ϩ is fast and exhibits low affinity for Na ϩ with an EC 50 of 30 -40 mM Na ϩ (4 -7).
GIRKs are usually heterotetramers composed of two pairs of homologous subunits. GIRK1/GIRK4 is predominant in the heart; GIRK1/GIRK2 is abundant in the brain. An aspartate, which is absent in GIRK1 but present in the proximal C terminus of GIRK2 (Asp-226) and GIRK4 (Asp-223), is crucial for fast direct gating by Na ϩ (6,8).
We have noticed an additional, slow activating effect of Na ϩ on GIRK channels in excised patches of Xenopus oocytes. The slow activation occurred both in wild-type (WT) GIRK channels and, surprisingly, in GIRK mutants that lack the fast direct Na ϩ regulation. It did not require GTP but was blocked by a G␤␥ scavenger, suggesting mediation by G␤␥. We hypothesized that Na ϩ promotes dissociation of the heterotrimeric G␣ GDP ␤␥ complex into free G␣ GDP and G␤␥; the latter activates GIRK. This hypothesis was supported by direct biochemical measurements. Our findings shed new light on mechanisms of regulation of G proteins and GIRK channels by Na ϩ and suggest that Na ϩ may act as a second messenger that regulates effectors via G␤␥.
Xenopus Oocyte Preparation and Electrophysiology-Xenopus oocytes were prepared, injected with RNA, and incubated in NDE-96 solution (96 mM NaCl, 2 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , 2.5 mM sodium pyruvate, 50 g/ml gentamycin, 5 mM Hepes/NaOH, pH ϭ 7.6). Patch-clamp measurements of GIRK activity were done at Ϫ80 mV as described previously (10). The electrode solution contained 146 mM KCl, 1 mM MgCl 2 , 1 mM CaCl 2 , 2 mM NaCl, 10 mM Hepes/KOH. The bath (500 l) contained a Na ϩ -free solution (130 mM KCl, 2 mM MgCl 2 , 10 mM Hepes/KOH, 1 mM EGTA, 2 mM Mg-ATP). The pH of all solutions was 7.4 -7.6. NaCl was added in 50 l of bath solution and mixed manually. Patches that showed Ͼ50% rundown of activity within 6 min after addition of Na ϩ (about 10% of all patches) were excluded from the study. Currents were filtered at 2 kHz and sampled at 5 kHz using the Axotape software (Axon Instruments). The results were analyzed as described previously (10). In each patch, the Na ϩ -induced changes in activity (-fold activation) were calculated as -fold change in total open probability, NP o , relative to basal NP o measured during the last minute before the addition of Na ϩ .
Biochemistry and Immunocytochemistry-The DNA of GST-G␣ i3 was constructed by inserting the coding sequence of human G␣ i3 into the EcoRI-NotI sites of pGEX-4T-1 vector (Amersham Biosciences). The protein was amplified and purified from Escherichia coli using glutathione-Sepharose affinity beads. For pull-down assays, GST-G␣ i3 was incubated for 30 min at 4°C in binding buffer (50 mM Tris, 5 mM MgCl 2 , 1 mM EDTA, 0.05% Tween 20, pH 7.0 with 150 mM KCl or 150 mM NaCl) with either 100 M GDP or GTP␥S. [ 35 S]Methionine-labeled G␤␥ was synthesized in reticulocyte lysate, diluted 1:4, and incubated with GST-* This work was supported by National Institutes of Health Grants GM56260 (to N. D.), GM60419 (to C. D.), and GM60019 (to V. S.) and grants from the United States-Israel Binational Science Foundation (to N. D., V. S., and C. D.). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Surface plasmon resonance (SPR) analysis of GST-G␣ i3 -G␤␥ interaction was carried out on Biacore 2000 (Biacore AB, Uppsala, Sweden) using anti-GST antibodies (90 ng) covalently attached to the surface of a CM5 sensor chip as described previously (13). Following the capture of GST-G␣ i3 , G␤␥ was injected across the sensor surface at a flow rate of 15 ml/min in series of concentrations in a buffer containing 50 mM Tris-Cl, 5 mM MgCl 2 , 1% CHAPS buffer, and either 100 mM KCl or 100 mM KCl. Intermediate concentrations of Na ϩ were obtained by mixing the two buffers. The kinetic data were analyzed using BIAEvaluation 3.1 software (Biacore AB) and Sigmaplot (SPSS Inc.).
The level of GIRK1 F137S protein in plasma membrane ( Fig. 4C) was measured as described previously (10). Briefly, large oocyte membrane patches were attached to coverslips, fixated, stained by a specific GIRK1 antibody (Alomone Labs, Jerusalem, Israel), and visualized using a Cy3-conjugated rabbit IgG (Jackson Immunoresearch Laboratories) using a Zeiss LSM 410 confocal microscope.
Statistics-Data are presented as mean Ϯ S.E.; pairwise comparisons were done using a two-tailed t test.

RESULTS
WT GIRK1/GIRK2 channels were expressed in Xenopus oocytes, and Na ϩ -induced activation was studied in excised patches (Fig. 1A). GIRK activity was first recorded in cellattached configuration. Then the patch was excised ("insideout") with its intracellular membrane surface facing the bath solution, which was Na ϩ -and GTP-free but contained Mg-ATP to preserve membrane phosphatidylinositol 4,5-bisphosphate (4). After excision, the basal activity declined to a new steady level within 0.5-2 min (10). 3-4 min after excision, 40 mM NaCl was added to the bath solution, causing a 6.7 Ϯ 2.5-fold increase (n ϭ 8) in channel activity. In agreement with previous studies (7), this fast Na ϩ activation was not affected by coexpression of the G␤␥ scavenger c␤ARK (myristoylated C terminus of ␤-adrenergic receptor kinase; data not shown). In addition, in many WT GIRK1/GIRK2 patches, a slow, late activation was observed. It started after about 1 min and developed over the next 3-6 min (Fig. 1A).
To explore the slow effects of Na ϩ , we utilized the GIRK1 pore mutant GIRK1 F137S , which forms functional homotetrameric channels (9) and lacks the C-terminal aspartate crucial for the direct Na ϩ effect. Fast activation of homotetrameric GIRK1 F137S channels by 40 mM Na ϩ was very weak, 1.82 Ϯ 0.36-fold (n ϭ 7) (Fig. 1B). It might reflect a weak direct effect of Na ϩ on the GIRK1 F137S channel. In seven of eight patches, the activity continued to increase over the next several minutes, reaching a maximum after 4 -7 min. This slow activation, measured during a 3-min period between 4 and 6 min after addition of Na ϩ , was 3.9 Ϯ 1-fold (n ϭ 7) above basal (Fig. 1D). It was fully blocked by coexpression of c␤ARK ( Fig. 1, C and D), implying a G␤␥-dependent mechanism. We assumed that the additional 2-fold slow increase in GIRK activity (compared with the 1st min after addition of Na ϩ ) reflects an increase in cellular concentration of free G␤␥, [G␤␥]. Comparable slow activation (2.62 Ϯ 0.45-fold, n ϭ 6) was observed already at 10 mM Na ϩ , which causes little fast activation in WT GIRK channels (5,14) and no fast activation in GIRK1 F137S (1.31 Ϯ 0.16fold, see Fig. 4).
How could Na ϩ activate GIRK in a G␤␥-dependent manner? GPCRs activate G proteins by promoting the exchange of GDP for GTP at the G␣ subunit followed by dissociation of G␣ GTP from G␤␥ (15). Cl Ϫ also promotes GTP binding to G␣ in vitro (16), but Cl Ϫ concentration in our bath solution was already saturating for this effect before the addition of NaCl. Most importantly the activation of GIRK1 F137S by NaCl was achieved in a GTP-free solution, ruling out a mechanism involving GDP-GTP exchange. Another source of free G␤␥, which does not require GTP, is the basal equilibrium between the G␣␤␥ heterotrimer and free G␣ GDP and G␤␥ (17): We hypothesized that Na ϩ promotes the dissociation of the G␣␤␥ heterotrimer into free G␣ GDP and G␤␥ by increasing the K D of Reaction 1. This will increase the concentration of free G␤␥, [G␤␥], activating GIRK in the absence of GTP. The ob- served time course of Na ϩ activation (minutes) is consistent with the slow dissociation rate constant of the G␣␤␥ heterotrimer (ϳ0.001 s Ϫ1 ) (17).
To scrutinize this hypothesis, we examined the effect of Na ϩ on binding of G␤ 1 ␥ 2 to a GST-fused G␣ i3 protein, GST-G␣ i3 ( Fig. 2A), using a pull-down assay. G␤␥ was synthesized in vitro in reticulocyte lysate. Only GST-G␣ i3-GDP , but not GST or GST-G␣ i3-GTP␥S , bound G␤␥ in this assay (Fig. 2B), confirming specificity and functional activity of GST-G␣ i3 . In three individual experiments like that shown in Fig. 2C (a and b), we examined the dose dependence of binding of G␤␥ to GST-G␣ GDP . The binding was dose-dependent and reached saturation at about 20 l of G␤␥-containing lysate. The affinity of binding was lower in high Na ϩ than in high K ϩ solution, but the differences were too small for a statistically reliable estimation of the affinity shift. Therefore, we compared the binding at a low [G␤␥] (2 l of G␤␥-containing lysate) and at saturating [G␤␥] (40 l of lysate) (summarized in Fig. 2C, c). With 40 l of lysate, the extent of binding in 150 mM NaCl was the same as in 150 mM KCl (102.6 Ϯ 8.2%, n ϭ 10). However, at 2 l of lysate, the binding in the high Na ϩ solution was reproducibly weaker than in high K ϩ (65.9 Ϯ 4.2%, n ϭ 14, p Ͻ 0.001 compared with 40 l of lysate). At doses of G␤␥ lower than 2 l of lysate (where a greater effect of Na ϩ was expected), the signal in the autoradiograms was too weak for a reliable measurement. The ϳ34% decrease in G␤␥ binding caused by Na ϩ at 2 l of lysate (which gave ϳ37% of maximum binding in high K ϩ ) corresponds to a 1.9-fold change in K D . The effect of Na ϩ was dose-dependent (Fig. 2D, a and b) with a K i of 13.6 mM Na ϩ (Fig. 2D, c).
The effect of Na ϩ on G␣ GDP -G␤␥ interaction was further studied by SPR. Recombinant G␤ 1 ␥ 2 was injected in buffers with different concentrations of NaCl (substituted for KCl) across the sensor surface with GST or GST-G␣ i3 immobilized via anti-GST antibody (13). G␤␥ reveals a relatively high degree of nonspecific binding to GST in this assay (13,18). However, the extent of G␤␥ binding to GST-G␣ i3-GDP was always greater than to GST (Fig. 3, A and B). Also the wash-out of G␤␥ from GST was fast as expected for a low affinity process, whereas unbinding of G␤␥ from GST-G␣ i3-GDP displayed a slower kinetic component, characteristic for a high affinity interaction. The net specific signal of G␤␥-G␣ i3 binding, obtained by subtraction of the GST signal (see Ref. 13), displayed rise and decay phases that were reasonably well fitted by single exponential functions (Fig. 3C). As expected, GST-G␣ i3-GTP␥S bound G␤␥ much weaker than GST-G␣ i3-GDP (Fig. 3D). When the Na ϩ concentration was increased, the net binding of G␤␥ to GST-G␣ i3-GDP was reduced in a dose-dependent manner with a K i of 8 mM Na ϩ (Fig. 3, E and F). Although these data clearly show specific binding of G␤␥ to G␣ i3 , the contribution of nonspecific binding, particularly in the first few seconds postinjection and sample wash-out, impaired the precise determination of the kinetic constants in Na ϩ and K ϩ . A limited kinetic analysis on the subtracted traces showed that the dissociation constant of Reaction 1, K D , was increased by 10 mM Na ϩ by ϳ2-fold (two separate sets of experiments). The calculated 2-fold change is in very good agreement with the pull-down results. The SPR experiments support the conclusion that Na ϩ reduces the affinity of interaction between G␣-GDP and G␤␥.
The findings so far support the hypothesis that the slow activation of GIRK by Na ϩ is due to elevated [G␤␥] that dissociated from G␣ GDP . This hypothesis not only explains the block of the slow effect of Na ϩ by ␤ARK but also allows testable predictions. Thus, mass law dictates that coexpression of G␣ should shift the equilibrium in Reaction 1 to the left, increasing [G␣␤␥] and reducing free [G␤␥]. This should reduce the basal activity of GIRK, but now addition of Na ϩ should cause a larger relative increase in [G␤␥] than in control conditions. This prediction is independent of initial conditions chosen to describe channel activation by G␤␥. 2 However, it is important not to overexpress G␣ to very high levels at which free G␣ GDP can 2 D. Yakubovich, I. Rishal, and N. Dascal, unpublished results. scavenge G␤␥ and blunt any G␤␥-induced effects. We produced mild overexpression of G␣ i3 , 3-4-fold over the resting cellular level (by injecting 5 ng of G␣ i3 RNA or less), that does not hinder the activation of GIRK by G␤␥ (10).
In agreement with the prediction, in G␣ i3 -expressing oocytes, slow activation of GIRK1 F137S by 10 mM Na ϩ was 10.3 Ϯ 2.3-fold (n ϭ 10), significantly greater than without G␣ i3 ( Fig.  4A and summarized in Fig. 4B). Slow activation by Na ϩ was still fully blocked by coexpression of c␤ARK (Fig. 4, A and B). Another GIRK channel mutant known to lack the direct Na ϩ activation, GIRK1/GIRK2 D226N (14), was also slowly activated by 10 mM Na ϩ in G␣ i3 -expressing cells (11.6 Ϯ 2.8-fold, n ϭ 9), and this effect was inhibited by c␤ARK ("activation" by 1.04 Ϯ 0.13-fold, n ϭ 4). The dramatic effect of G␣ i3 on slow Na ϩinduced activation strongly supports the G protein-dependent character of this phenomenon. The extent of slow activation by 10 mM Na ϩ , in the presence of G␣, is comparable to the direct effect of 40 mM Na ϩ , although it is still severalfold smaller than the 20 -800-fold activation by saturating doses of G␤␥ (10). A simple kinetic model based on our hypothesis and on known or estimated parameters of G␣ GDP -G␤␥ and G␤␥-GIRK interactions correctly described the slow GIRK activation by Na ϩ . The calculations showed that a ϳ1.8-fold increase in K D of Reaction 1 by 10 mM Na ϩ fully accounted for the 2-fold slow activation of GIRK1 F137S under control conditions and for an ϳ8-fold activation after coexpression of G␣ i3 . 2 It appeared (although a systematic study has not been conducted) that coexpression of G␣ i3 did not cause the expected reduction in basal activity of GIRK1 F137S (e.g. compare the cell-attached levels of channel activity in Fig. 4A, a and b). The unexpectedly high basal activity could reflect an increase in membrane levels of GIRK1 F137S caused by G␣ i3 as described for WT GIRK1/GIRK2 (10). Indeed immunostaining of the ex-pressed channels in large plasma membrane patches (Fig. 4C) showed a net 6.4-fold increase of GIRK1 F137S protein by G␣ i3 (after subtraction of background observed in uninjected oocytes) compared with channels expressed alone.
The dose dependence of the slow Na ϩ effect was studied in oocytes coexpressing GIRK1 F137S and G␣ i3 . Maximal activation was observed at 10 mM Na ϩ (Fig. 4D). At higher [Na ϩ ] the effect was smaller, consistent with the previously reported inhibition, at high [Na ϩ ], of GIRK channel constructs lacking the fast Na ϩ activation (6). The mechanism of this inhibition is unknown. The half-maximal activation dose (EC 50 ) appeared to be slightly above 5 mM Na ϩ , but considering the interference of the inhibitory effect at high [Na ϩ ], the actual EC 50 of slow Na ϩ activation is probably higher.

DISCUSSION
Na ϩ Regulates Basal Equilibrium between Free and G␣ GDPbound G␤␥-Our results strongly suggest that Na ϩ reduces the affinity of interaction between G␣ GDP and G␤␥. This hypothesis initially arose following the observation of slow, G␤␥-dependent activation by Na ϩ of mutant GIRK channels lacking the direct Na ϩ modulation. We then demonstrated weakening of G␣ GDP -G␤␥ binding using direct in vitro binding assays. Although the detected change in affinity was small, the results obtained by two independent methods (a pull-down assay and surface plasmon resonance) were in good agreement. Both methods, as well as estimates obtained from patch-clamp data, indicate that Na ϩ causes a ϳ2-fold decrease in the affinity of interaction between G␣ i3-GDP and G␤␥. The EC 50 or K i of the Na ϩ effect estimated by electrophysiological and the two biochemical methods are also in good agreement (ϳ6 -14 mM), falling within the physiological range of [Na ϩ ] in . The proposed hypothesis made it possible to explain the observed inhibition FIG. 3. Analysis of G␤ 1 ␥ 2 interaction with GST-G␣ i3-GDP or GST-G␣ i3-GTP␥S by SPR. GST-G␣ i3-GDP (150 ng) was captured on one channel of Biacore, GST-G␣ i3-GTP␥S was captured on another channel, and recombinant GST was immobilized on a reference channel. A and B, representative traces showing the SPR signal (in resonance units, RU) upon binding of 350 nM G␤ 1 ␥ 2 to the GST and GST-G␣ i3-GDP channels in 0 mM Na ϩ (A) and in 10 mM Na ϩ (B). The different Na ϩ concentrations were obtained by mixing the parental buffers containing 100 mM KCl (0 Na ϩ ) or 100 mM NaCl (0 K ϩ ). C, net GST-G␣ i3-GDP SPR signal after subtraction of the nonspecific binding on the GST channel. Rise and decay phases were fitted to monoexponential functions (solid lines). D, comparison of SPR signal at the GST-G␣ i3-GDP and GST-G␣ i3-GTP␥S channels (nonspecific binding to GST on the reference channel was subtracted). E, net binding of G␤␥ to GST-G␣ i3-GDP is reduced by Na ϩ : a summary of three separate measurements. F, the Na ϩ dose-response curve, fitted to the Michaelis-Menten equation with a K i of 8 mM (solid line). The effect of Na ϩ (decrease in specific G␤␥ binding), at each [Na ϩ ], was presented as percentage of maximal decrease in binding. In two measurements, maximal effect occurred at 50 mM Na ϩ and in one measurement at 20 mM. of the slow Na ϩ -induced activation by the G␤␥ scavenger, c␤ARK, and to predict a novel physiological effect: a dramatic enhancement of slow Na ϩ activation of GIRK by coexpressed G␣. Interestingly Xenopus oocytes possess an unusually high basal level of free G␤␥ (19,20), and coexpression of G␣ may actually mimic the "usual" cellular condition of low [G␤␥]. Na ϩ has long been known to modulate the binding of agonists to many GPCRs, uncoupling these receptors from G proteins. Initially Na ϩ had been suspected to regulate G proteins directly, but later studies demonstrated a pivotal role of Na ϩ interaction with a conserved aspartate residue located in the transmembrane region of many GPCRs (for review, see Ref. 21). Finally, although NaCl has been shown to promote GDP-GTP exchange at the G protein, the active agent was Cl Ϫ rather than Na ϩ (16). Thus, no direct effects of Na ϩ on G proteins are known, and this report is the first demonstration of such regulation.
Two Mechanisms of Regulation of GIRK by Na ϩ -Na ϩ regulation of GIRK channels is considered of high physiological importance since it is believed to determine part of their basal activity in intact cells and to underlie the negative chronotropic effect of cardiac glycosides in the heart (2,5). Our data imply that, in GIRK channels, Na ϩ acts both directly and indirectly (via G␤␥) to modulate basal activity. The EC 50 of the well characterized, direct, G␤␥-independent activation by Na ϩ is 30 -40 mM. The slow, G␤␥-dependent activation of GIRK by Na ϩ has a definitely lower EC 50 : maximal effect is attained at 10 mM Na ϩ . Despite the small magnitude of G␣ GDP -G␤␥ affinity change by Na ϩ , 10 mM Na ϩ caused an impressive 4 -10-fold activation of GIRK. Therefore, the slow mechanism may contribute substantially to the basal activity of the channel in intact cells where the resting [Na ϩ ] in is 5-10 mM.
The observed features of slow Na ϩ modulation of GIRK conform to the model in which [G␤␥], elevated because of the direct effect of Na ϩ on G␣␤␥ dissociation, activates the channel. Yet at present we cannot exclude contribution of additional mechanisms, for instance activation of a nucleotide diphosphate kinase (NDPK), which catalyzes the transfer of phosphate from ATP to GTP. This might, in principle, promote G␣␤␥ dissociation and activate GIRK (see the discussion in Ref. 2). However, Na ϩ has been reported to inhibit NDPK (22) and should have inhibited GIRK if NDPK were involved.
Na ϩ as a Second Messenger or a Servo-type Intracellular Regulator-Sodium ions are crucial for neuronal activity as carriers of depolarization and also regulate many physiological processes (fluid balance and secretion, cardiac contraction, glutamate-induced neuronal excitation, etc.) by a variety of molecular mechanisms from a direct binding of Na ϩ (GIRK and Na ϩ -dependent transporters and exchangers (23)) to activation of Src by Na ϩ via an unknown intermediate (N-methyl-D-aspartate receptors (24)). Basal [Na ϩ ] in is tightly regulated by pumps and exchangers, being maintained in resting cells below 10 mM. Yet [Na ϩ ] in often widely fluctuates. In neurons, even relatively short periods of synaptic activity produce very large increases in [Na ϩ ] in , reaching ϳ30 mM in apical dendrites and Ͼ100 mM in dendritic spines (25). These considerations raise the possibility that Na ϩ may, in some cases, act as a second messenger that regulates intracellular targets when a substantial change in concentration occurs.
Our findings suggest a new mechanism of Na ϩ -dependent regulation of cellular processes: via G␤␥. The effective range of Na ϩ concentrations that regulate the G␣␤␥ % GG␣ GDP ϩ G␤␥ equilibrium is close to the resting physiological range of [Na ϩ ] in . Therefore, regulation of G proteins by Na ϩ may be a servo-type mechanism that sensitively responds to bidirectional changes in [Na ϩ ] in rather than to increases alone. We propose that Na ϩ regulation of the dynamic equilibrium between bound and free G␣ GDP and G␤␥ can have a substantial biological effect by regulating a host of effectors of G␤␥ (1), some of which are second messenger-generating enzymes that may further amplify Na ϩ effects. This establishes a possible novel way of communication between electrical activity and other cellular processes.