Functional and Biochemical Evidence for G-protein-gated Inwardly Rectifying K+ (GIRK) Channels Composed of GIRK2 and GIRK3*

G-protein-gated inwardly rectifying K+ (GIRK) channels are widely expressed in the brain and are activated by at least eight different neurotransmitters. As K+ channels, they drive the transmembrane potential toward EK when open and thus dampen neuronal excitability. There are four mammalian GIRK subunits (GIRK1–4 or Kir 3.1–4), with GIRK1 being the most unique of the four by possessing a long carboxyl-terminal tail. Early studies suggested that GIRK1 was an integral component of native GIRK channels. However, more recent data indicate that native channels can be either homo- or heterotetrameric complexes composed of several GIRK subunit combinations. The functional implications of subunit composition are poorly understood at present. The purpose of this study was to examine the functional and biochemical properties of GIRK channels formed by the co-assembly of GIRK2 and GIRK3, the most abundant GIRK subunits found in the mammalian brain. To examine the properties of a channel composed of these two subunits, we co-transfected GIRK2 and GIRK3 in CHO-K1 cells and assayed the cells for channel activity by patch clamp. The most significant difference between the putative GIRK2/GIRK3 heteromultimeric channel and GIRK1/GIRKx channels at the single channel level was an ∼5-fold lower sensitivity to activation by Gβγ. Complexes containing only GIRK2 and GIRK3 could be immunoprecipitated from transfected cells and could be purified from native brain tissue. These data indicate that functional GIRK channels composed of GIRK2 and GIRK3 subunits exist in brain.

While most brain regions express GIRK1, GIRK2, and GIRK3 mRNAs (13)(14)(15), there are regions where only GIRK2 and GIRK3 mRNA transcripts are present or expression of GIRK1 is very low. We combined electrophysiological and biochemical approaches to determine if a GIRK heteromultimer composed of GIRK2 and GIRK3 subunits alone was functional and if such a channel exists in vivo. GIRK2 and GIRK3 subunits had been previously coexpressed in Xenopus oocyte expression studies (16 -19), but the presumed GIRK2/GIRK3 currents observed were small (Ͻ0.1 A). In this study, we examine this subunit combination in mammalian cells. We demonstrate that GIRK2/GIRK3 forms a functional channel in heterologous cells and that a GIRK2/GIRK3 complex can be purified from both heterologous expression systems and native brain tissue.
Electrophysiology-Single channel recordings were made from patches in the inside-out configuration in symmetrical high K ϩ (140 mM KCl, 2 mM MgCl 2 , 5 mM EGTA, and 10 mM HEPES, pH 7.2). Patch electrodes were prepared from borosilicate capillary glass (Warner Instrument Corp., Hamden, CT) and polished to a resistance of 5-12 * 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.
Records of channels in inside-out patches activated by 40 M GTP␥S were collected at voltages ranging from Ϫ100 to Ϫ40 mV. Unitary current values from all patches were averaged, and the slope conductance was determined as the best fit to the mean data. Channel open times were determined for 10-s records of inside-out patches activated by 40 M GTP␥S collected at Ϫ100 mV. Log analysis (21) was used for determination of channel mean open time ( o ).
The concentration dependence of channel activation by G␤␥ was examined by adding increasing concentrations of G␤␥ to the bath while recording the channel activity. Patches were excised into the standard bath solution, and basal activity was recorded for 2.5 min before G␤␥ was added to an initial concentration of 2 nM. Additional G␤␥ was added incrementally at 2.5-min intervals until a final concentration of 60 or 90 nM was reached. 2.5 min after the highest G␤␥ concentration was attained, 100 M GTP␥S was added to the bath. NP o was determined for the final minute of each G␤␥ concentration step. NP o values were normalized to the final activity level 1.5 min after the addition of GTP␥S for each patch in order to derive relative NP o values. Relative NP o values from all patches were averaged, and the dose-response curve was determined as the best fit to the mean data.
Western and Coimmunoprecipitation Assays Using Recombinant GIRK2/GIRK3-Membrane proteins were isolated from single 100mm 2 dishes of CHO-K1 cells transfected with GIRK1 plus GIRK2, GIRK1 plus GIRK3, and GIRK2 plus GIRK3. To serve as a control, membrane proteins were also isolated from untransfected cells. The cells were washed twice with PBS before being lifted from the dish with 5 mM EDTA in PBS. The cells were pelleted by centrifugation at 1200 ϫ g for 5 min. Cold hypotonic 5/5/5 lysis buffer (5 mM each of Tris-HCl, EDTA, and EGTA; pH 8) with protease inhibitors (1 mM phenylmethylsulfonyl fluoride and 2 g/ml each of aprotinin, pepstatin, and leupeptin) was added to the cells, which were lysed on ice by drawing them through a 23-gauge needle 10 times and a 27-gauge needle four times. Crude membranes were obtained by centrifugation at 120,000 ϫ g for 15 min at 4°C. The crude membranes were resuspended in 5/5/5 buffer with protease inhibitors, and the proteins were precipitated with 100% (w/v) trichloroacetic acid (Sigma). The precipitated proteins were dissolved in 1ϫ SDS sample buffer and assayed by Western blot with anti-GIRK1 fusion protein antibody ␣CSh (anti-CSh serum, generated against the carboxyl-terminal 156 amino acids of GIRK1) or with anti-GIRK1 antibody (Alomone Laboratories, Jerusalem, Israel).
Epitope-tagged GIRK2-Myc and GIRK3-AU1 constructs were transfected in the recombinant coimmunoprecipitation experiments. The GIRK2-Myc construct has been described previously (20). The AU1 epitope, DTYRYI, was inserted after the initiator methionine in the pcDNA3.1-GIRK3 plasmid using the QuickChange TM mutagenesis kit (Stratagene, La Jolla, CA). The primer sequences used were 5Ј-CCCG-CTGCGGCCGCCATGGACACTTACCGCTATATTGCGCAGGAGAAC-G-3Ј and 5Ј-GCGTTCTCCTGCGCAATATAGCGGTAAGTGTCCATGG-CGGCCGCAGCGG-3Ј. Positive clones were identified by the introduction of a unique FspI restriction site, and the integrity of the tagged GIRK3 sequence was verified by DNA sequencing. COS-7 cells were used for the recombinant coimmunoprecipitation experiments because they produced significantly more expressed protein than the CHO-K1 cells. COS-7 cells have been used to examine GIRK channels in previous functional studies (7,8,10,17,22,23) and are known to produce functional GIRK channels. Single 100-mm 2 dishes of COS-7 cells expressing GIRK2-Myc and GIRK3-AU1 were washed twice with PBS (22°C) before being lifted from the dish with 5 mM EDTA in PBS. The cells were pelleted by centrifugation at 1200 ϫ g for 5 min. Hypotonic 5/5/5 lysis buffer with protease inhibitors was added to the cells, which were lysed on ice by drawing them through a 23-gauge needle 10 times. The lysed cells were centrifuged at 150,000 ϫ g for 15 min at 4°C to obtain a crude particulate fraction. This fraction was resuspended in 5/5/5 with protease inhibitors and centrifuged through a 20% sucrose cushion at 150,000 ϫ g for 35 min at 4°C. The resulting pellet was solubilized in 1% CHAPS, 10 mM HEPES, 300 mM NaCl, 5 mM EGTA, and 5 mM EDTA with protease inhibitors at pH 8.0.
For immunoprecipitation, samples were precleared with Gammabind-G Sepharose (Amersham Pharmacia Biotech) for 1 h at 4°C. Centrifuged ascites fluid for AU1 monoclonal antibodies (BabCO, Richmond, CA) was used at a 1:200 dilution. Samples were incubated for 4 h at 4°C on a rotator. Bound proteins were eluted with 1ϫ SDS sample buffer at 55°C for 20 min. The eluted proteins were analyzed by Western blot with anti-GIRK2 antibody (Alomone Laboratories) and rabbit anti-GIRK3 antibody C-312 (Cocalico Biologicals, Inc., Reamstown, PA) against a carboxyl-terminal peptide antigen (SIPSRLDEKVEEE-GAGEGGR). The specificity of anti-GIRK3 antibody C-312 was confirmed against a panel of recombinant GIRK proteins expressed in COS-7 cells.
Isolation and Solubilization of Mouse Brain Membranes-Whole brains were removed from wild-type and GIRK3 knockout mice. 2 The tissue was minced in homogenization buffer (320 mM sucrose, 4 mM HEPES-NaOH, pH 7.4) with protease inhibitors (50 g/ml leupeptin, 100 g/ml phenylmethysulfonyl fluoride, 1 g/ml aprotinin, and 2 g/ml pepstatin) on ice before homogenization with a Polytron homogenizer (2 ϫ 15 s). The homogenized tissue was centrifuged for 10 min at 1000 ϫ g to remove nuclei and large cellular fragments. The supernatant was centrifuged for 30 min at 120,000 ϫ g at 4°C to pellet the membranes. The pelleted membranes were resuspended in immunoprecipitation buffer (10 mM HEPES, 100 mM NaCl, 5 mM EDTA, pH 7.53) with 1% Triton X-100 (Sigma). Protein was quantified using the Bio-Rad Protein Assay (Bio-Rad). Solubilized membranes were stored at Ϫ80°C.
Immunoprecipitation of GIRK Complexes from Mouse Brain-GIRK2/GIRK3 complexes were purified by serial immunoprecipitation. Any potential GIRK1-containing complexes were eliminated in the first two rounds of immunoprecipitation so that any complex containing GIRK1, GIRK2, and GIRK3 would not be mistakenly identified as one containing only GIRK2 and GIRK3. The third round was designed to immunoprecipitate remaining complexes that contained only GIRK2 and/or GIRK3 subunits.
Each serial experiment tested three wild-type samples against three GIRK3 knockout mouse samples. One tube in each group was a control containing no antibody. Anti-GIRK1 serum ␣CSh was prebound for 8 h to protein A-Sepharose beads (Amersham Pharmacia Biotech) at a 1:92 dilution. The resin was washed 7ϫ with 1-ml aliquots of immunoprecipitation buffer before 200 g of whole brain membrane proteins from wild-type or GIRK3 knockout mice were added. Samples were incubated at 4°C overnight on a rotator and centrifuged, and the supernatants were transferred to fresh tubes containing additional ␣CSh bound to protein A-Sepharose. For the second round of GIRK1 removal, the samples were incubated at 4°C for 3 h on a rotator. Afterward, supernatant samples were assayed for GIRK1 by Western blotting with ␣CSh. Supernatants were transferred to fresh tubes containing anti-GIRK2 antibody (Alomone Laboratories) at a 1:174 dilution or anti-GIRK3 antibody C-312 at a 1:67 dilution. The antibodies were preincubated with protein A-Sepharose and washed seven times with 1-ml aliquots of immunoprecipitation buffer prior to sample addition and incubation at 4°C overnight. Supernatants were removed before the resin was washed, and bound proteins were eluted with 1ϫ SDS sample buffer with ␤-mercaptoethanol (Sigma; 55°C for 20 min). Eluted proteins were run on 10% Bis-Tris NuPAGE gels in MOPS buffer (Novex, San Diego, CA), providing maximal separation of proteins in the 35-55-kDa range. To avoid confusing signals from the GIRK proteins with the heavy chains of the immunoglobulins used to purify them, the blots were probed with biotinylated anti-GIRK2 (Alomone Laboratories) and biotinylated C-312 anti-GIRK3 antibodies, followed by incubation with strepavidin-horseradish peroxidase. The antibodies were biotinylated using E-Z-Link NHS-LC-Biotin (Pierce).

Cotransfection of GIRK2 and GIRK3 Yields a Functional
Channel in CHO-K1 Cells-In order to examine the properties of a potential channel composed of GIRK2 and GIRK3 subunits, CHO-K1 cells were cotransfected with GIRK2 and GIRK3 and assayed for channel activity by patch clamp. Insideout patches of transfected cells contained channels that could be activated by application of GTP␥S to the cytoplasmic face of the patch and which displayed inward rectification (Fig. 1). The slope conductance of the channel was 31 Ϯ 1.7 picosiemens (Ϯ S.E.; n ϭ 14), smaller than that observed for GIRK1/GIRK2, GIRK1/GIRK3, and GIRK1/GIRK4 channels (35,39, and 37 picosiemens, respectively) in our previous study (1) but still within the range of conductance values reported for GIRK channels (24). The o at Ϫ100 mV for the GIRK2/GIRK3 chan-nel was 1.3 ms (n ϭ 11), similar to that for GIRK1/GIRK2, GIRK1/GIRK3, and GIRK1/GIRK4 channels at Ϫ80 mV (1.0, 1.1, and 1.2 ms, respectively) (1). These results were surprising, considering that any one of the GIRK2, GIRK3, or GIRK4 subunits expressed alone yields spiky, irregular channels with very brief open times (1, 3-6).
Since activation by G␤␥ is a hallmark feature of GIRK channels, the concentration dependence of GIRK2/GIRK3 channel activation was determined. Inside-out patches containing the channel were exposed to increasing concentrations of G␤␥ at regularly timed intervals until a final concentration was reached (see Ref. 25). At a set time interval after reaching the final G␤␥ concentration, a saturating concentration of GTP␥S was added to the bath. The channel activity, quantitated as NP o at each concentration of G␤␥, was determined and normalized to the NP o after the addition of GTP␥S in order to derive relative values. The relative activity was plotted against log concentrations of G␤␥ to determine the concentration required to elicit half-maximal activation of the channel (Fig. 2). While GIRK1/GIRK4 and GIRK1/GIRK3 are activated by G␤␥ with a half-maximal concentration of 11 nM (1, 23), the GIRK2/GIRK3 channel displays a half-maximal concentration of 53 nM, a 4.8-fold lower sensitivity to G␤␥. Two-way factorial analysis of variance revealed no significant difference in the slopes of the two dose-response curves, but GIRK1/GIRK3 is significantly more sensitive to G␤␥ than GIRK2/GIRK3; F (1, 25) ϭ 8.77, p ϭ 0.007.

GIRK1 Is Not Expressed in CHO-K1 Cells Cotransfected with GIRK2 and GIRK3-CHO-K1 cells do not normally express
GIRK channel subunits. To rule out the possibility that cotransfection of GIRK2 and GIRK3 could induce the cells to express GIRK1, membrane proteins were harvested from cotransfected cells and tested for the presence of GIRK1 protein by Western blotting. CHO-K1 cells transfected with GIRK1/ GIRK2 and GIRK1/GIRK3 were used as positive controls for GIRK1 expression, while untransfected CHO-K1 cells were used as a negative control. Two separate anti-GIRK1 antibodies were used to test for GIRK1 expression. In both cases, GIRK1 protein could be detected only in cells that had been transfected with GIRK1 cDNA (Fig. 3).
GIRK2 and GIRK3 Form a Complex in Transfected Cells-To examine whether or not cotransfected GIRK2 and GIRK3 subunits were able to form heteromultimeric complexes, we coimmunoprecipitated the GIRK channel proteins from COS-7 cells transfected with epitope-tagged GIRK subunits. After cotransfection with the GIRK2-Myc and GIRK3-AU1 constructs, membrane proteins from COS-7 cells were harvested, and ascites fluid for monoclonal AU1 antibodies was used to immunoprecipitate GIRK3-AU1 and any associated proteins. These proteins were eluted and transferred to a membrane for Western blot analysis with subunit-specific anti-GIRK2 and anti-GIRK3 antibodies. The specific recognition of their respective GIRK proteins by these antibodies can be seen in Fig. 4, A and B. Analysis of the eluted proteins by Western blot demonstrated the presence of both GIRK3 and GIRK2 in the AU1 ascites pellet (Fig. 4C). These coimmunoprecipitation data strongly support the notion that GIRK2 and GIRK3 subunits can combine to form a heteromultimeric complex in transfected mammalian cells.
GIRK2/GIRK3 Complexes Are Present in Native Brain-To test whether heteromultimeric complexes of GIRK2 and GIRK3 subunits might exist in vivo, we performed coimmunoprecipitation experiments using mouse brain as the tissue source. Membranes solubilized in 1% Triton X-100 buffer were prepared from wild-type and GIRK3 knockout (control) mouse brains. 2 Unlike the heterologous expression system, the native system is complicated by the presence of GIRK1. A straightforward GIRK2/GIRK3 coimmunoprecipitation experiment would have ambiguous results because it would be uncertain whether or not the complex containing GIRK2 and GIRK3 also contained GIRK1. To avoid this ambiguity, the solubilized membranes were first immunodepleted of GIRK1 by two rounds of GIRK1 immunoprecipitation before immunoprecipitation of GIRK2 and GIRK3. After the second round of GIRK1 immunodepletion, small supernatant samples were tested for GIRK1 by Western analysis. As Fig. 5A shows, GIRK1 protein was successfully removed, leaving a GIRK1-free protein pool from which we could immunoprecipitate GIRK2 and GIRK3. The final immunoprecipitates and control supernatants were analyzed by Western blot for GIRK2 and GIRK3 coimmunoprecipitation. pernatants probed with anti-GIRK2 antibody. The GIRK2 signal from native brain membranes consistently appeared as a doublet, suggesting that at least two GIRK2 isoforms are present (26,27). GIRK2 protein was recovered from the control supernatant and both the anti-GIRK2 and anti-GIRK3 immunoprecipitates of the wild-type samples but was recovered only from the control supernatant and the anti-GIRK2 immunoprecipitate of the GIRK3 knockout samples. The presence of GIRK2 in the anti-GIRK3 wild-type immunoprecipitate together with its absence from the anti-GIRK3 knockout immunoprecipitate indicates that GIRK2 was pulled down by the anti-GIRK3 antibody most likely because it was complexed with GIRK3. Fig. 5C shows the final immunoprecipitates and control supernatants probed with an anti-GIRK3 antibody. GIRK3 was recovered from the control supernatant and both the anti-GIRK2 and anti-GIRK3 immunoprecipitates of the wild-type samples. As expected, no GIRK3 signal was seen for the GIRK3 knockout samples. The presence of GIRK3 in the anti-GIRK2 wild-type immunoprecipitate reinforces the evidence of coimmunoprecipitation seen with the anti-GIRK3 antibody. These results demonstrate that complexes containing GIRK2 and GIRK3 without GIRK1 can be immunoprecipitated from native brain. DISCUSSION We have demonstrated that GIRK2/GIRK3 channels are functional, exist in native brain tissue, and are less sensitive to G␤␥ than GIRK1-containing channels. When GIRK2, GIRK3, or GIRK4 subunits are transfected alone, each gives rise to "spiky" channels with very brief open times (Ͻ0.5 ms) and variable single channel conductance (20 -40 picosiemens) (1,3,5,6). Co-transfecting either subunit with GIRK1 yields a heteromultimeric channel with characteristics like the prototypical GIRK channel, I KACh . The uniqueness of GIRK1 has attracted a series of hypotheses to account for its long carboxylterminal tail, including the idea that it modifies the gating of the GIRK heteromultimer to yield longer open times. As we have shown, co-transfecting GIRK2 and GIRK3 into CHO cells verifiably lacking GIRK1 gives rise to an inwardly rectifying current with single channel properties much like those described for the GIRK1-containing channels. It thus appears that the longer open time does not depend on the presence of GIRK1 but that heteromultimerization places the subunits in the proper conformation to achieve a longer open time and quantized conductance level.
Despite its similarities with respect to open time and conductance, the GIRK2/GIRK3 channel differed from the GIRK1/ GIRKx channels in being nearly 5-fold less sensitive to activation by G␤␥. A recent study has shown that residues proximal to the membrane in the carboxy tail region of GIRK4 are important for binding by G␤␥ and critical for activation of I KACh (10). The similarity of this region (e.g. C216) in GIRK2 and GIRK3 probably accounts in part for the similar responses of the GIRK1/GIRK4 and GIRK1/GIRK3 channels to G␤␥ (1). However, the same study and others have demonstrated that GIRK1 also binds G␤␥ (11,12,28). Thus, G␤␥ binding to GIRK1 may have a potentiating effect, or perhaps GIRK1 and GIRK4 both contribute to the binding site for G␤␥, and GIRK1's contribution may be to increase the affinity for G␤␥. Alternatively, GIRK1 may contribute to or affect the binding site for a modulating molecule such as inositol 1,4,5-bisphosphate, which has been shown to play an important role in the activation of I KACh by G␤␥ and by intracellular Na ϩ (29 -33).
The lower sensitivity of GIRK2/GIRK3 channels to activation by G␤␥ may partly explain why the GIRK currents in Xenopus oocytes co-injected with cRNAs for these subunits are so small (16 -19). Additionally, GIRK3 apparently expresses at low levels in oocytes compared with the other GIRK subunits (16 -19, 34). The combination of poor expression of the channel with its decreased sensitivity to G␤␥ may account for the small peak currents observed for GIRK2/GIRK3 in Xenopus oocytes. In contrast, GIRK3 and GIRK2/GIRK3 express well in mammalian cells.
In addition to providing enhanced G␤␥ sensitivity, GIRK1 may also provide a second localization mechanism via its unique carboxy tail. Additional diversity comes from the fact that GIRK3 and the long form of GIRK2 both have potential PDZ-binding motifs at their carboxyl termini. The variety of GIRK channel combinations may allow for specificity of localization of channels in the neuron, at pre-or postsynaptic sites for example, or for linking of channel subtypes to specific neurotransmitter receptors.
Most regions in the brain have the potential to produce GIRK2/GIRK3 channels because they express GIRK1, GIRK2, and GIRK3 mRNAs (13)(14)(15). The expression of GIRK1 does not preclude the formation of complexes that do not contain GIRK1, and cells may produce more than one type of GIRK channel. This is clearly illustrated by the presence of GIRK4 homomeric complexes in atrial cells expressing I KACh (35). It is possible that a neuron that produces GIRK1/GIRK2 and GIRK1/GIRK3 channels may also produce GIRK2/GIRK3 channels. GIRK2/GIRK3 channels have a different sensitivity to activation by G␤␥ and may be localized differently than channels containing GIRK1. Thus, there may be regions within neurons where GIRK2/GIRK3 channels are more common than GIRK1/GIRK2 or GIRK1/GIRK3 channel complexes.
In summary, we have demonstrated the existence in brain of a heteromultimeric GIRK channel composed of GIRK2 and GIRK3 subunits. These channels have a similar open time to the GIRK1/GIRKx channels but have a slightly smaller conductance and less sensitivity to activation by G␤␥. The discovery of this channel, with its lesser sensitivity to G␤␥ and hence FIG. 5. Purification of GIRK2/GIRK3 complexes from native brain tissue by serial immunoprecipitation. A, after two rounds of immunodepletion (see "Experimental Procedures"), GIRK1 protein had been effectively removed. B, the GIRK1-free samples of membrane proteins were immunoprecipitated with anti-GIRK2 or anti-GIRK3 antibodies (Ab) to yield the final immunoprecipitates. Among the wild type (WT) samples, GIRK2 was recovered from the control supernatant (sup) and both the anti-GIRK2 and anti-GIRK3 immunoprecipitates (ppt). Among the GIRK3 knockout samples, GIRK2 was recovered only from the control supernatant and the anti-GIRK2 immunoprecipitate. C, among the wild type samples, GIRK3 was recovered from the control supernatant and both the anti-GIRK2 and anti-GIRK3 immunoprecipitates. Among the GIRK3 knockout samples, no GIRK3 signal was detected as expected. lower threshold of activation, along with its potential for differential localization from the GIRK1/GIRKx channels, broadens the functional scope of the GIRK family of channels.