Molecular properties of neuronal G-protein-activated inwardly rectifying K+ channels.

Four cDNA-encoding G-activated inwardly rectifying K+ channels have been cloned recently (Kubo, Y., Reuveny, E., Slesinger, P. A., Jan, Y. N., and Jan, L. Y.(1993) Nature 364, 802-806; Lesage, F., Duprat, F., Fink, M., Guillemare, E., Coppola, T., Lazdunski, M., and Hugnot, J. P. (1994) FEBS Lett. 353, 37-42; Krapivinsky, G., Gordon, E. A., Wickman, K., Velimirovic, B., Krapivinsky, L., and Clapham, D. E. (1995) Nature 374, 135-141). We report the cloning of a mouse GIRK2 splice variant, noted mGIRK2A. Both channel proteins are functionally expressed in Xenopus oocytes upon injection of their cRNA, alone or in combination with the GIRK1 cRNA. Three GIRK channels, mGIRK1-3, are shown to be present in the brain. Colocalization in the same neurons of mGIRK1 and mGIRK2 supports the hypothesis that native channels are made by an heteromeric subunit assembly. GIRK3 channels have not been expressed successfully, even in the presence of the other types of subunits. However, GIRK3 chimeras with the amino- and carboxyl-terminal of GIRK2 are functionally expressed in the presence of GIRK1. The expressed mGIRK2 and mGIRK1, −2 currents are blocked by Ba2+ and Cs+ ions. They are not regulated by protein kinase A and protein kinase C. Channel activity runs down in inside-out excised patches, and ATP is required to prevent this rundown. Since the nonhydrolyzable ATP analog AMP-PCP is also active and since addition of kinases A and C as well as alkaline phosphatase does not modify the ATP effect, it is concluded that ATP hydrolysis is not required. An ATP binding process appears to be essential for maintaining a functional state of the neuronal inward rectifier K+ channel. A Na+ binding site on the cytoplasmic face of the membrane acts in synergy with the ATP binding site to stabilize channel activity.

Inward rectifier K ϩ channels were first described in skeletal muscle and egg-cell membranes (1,2). They are now found in many cell types and are characterized by the following properties: (i) an activation by hyperpolarization negative to the re-versal potential for K ϩ ( E K ), (ii) an activation potential shifting with E K , (iii) a blockade by Cs ϩ and Ba 2ϩ (3).
A class of inward rectifier K ϩ channels is gated via G-proteins (GIRK). In atrial cells, acetylcholine released by stimulation of the vagal nerve causes the opening of a GIRK channel (I KACh ) via the activation of a m2-muscarinic receptor. The induced hyperpolarization results in a slowing of cardiac frequency (4,5). GIRK channels also exist in a variety of neuronal cells and the modulation of such channels generates slow synaptic potentials (6,7). They are coupled to various neurotransmitter receptors such as the muscarinic cholinergic, , ␦, and opioid, ␣ 2 -adrenergic, somatostatin, substance P, and GABA B receptors (8 -10).
A GIRK channel, termed GIRK1 (11) or KGA (12), was cloned from rat heart, and two structural homologs were cloned from mouse brain, mGIRK2 and mGIRK3 (13). Another close structural parent of the GIRK family (rcK ATP ) was described initially as an ATP-sensitive K ϩ channel (14), i.e. a channel for which activity is controlled by intracellular ATP (15). However, it has been shown recently that the functional I KACh channel stimulated by the G-protein ␤␥ subunits is a heteromultimer composed of two GIRK subunits, GIRK1 and CIR, a channel subunit which is nearly identical with rcK ATP (16).
A mouse analog of rcK ATP /CIR that we have cloned, and designed in this paper as mGIRK4, presents the characteristic features of a G-protein-activated inward rectifier K ϩ channel. In contrast to mGIRK1, which is present both in heart and brain, mGIRK4/CIR is specifically localized in the heart, whereas mGIRK2 and mGIRK3 are expressed mainly in the brain. By using different strategies, co-localization of transcripts, immunoprecipitation, and electrophysiology, we present evidences for a heterologous GIRK subunit assembly in the brain. The paper also describes the main electrophysiological properties and the modulation by ATP and Na ϩ of the currents expressed by mGIRK2 and the mGIRK1 ϩ mGIRK2 combination in Xenopus oocyte.

EXPERIMENTAL PROCEDURES
Isolation of mGIRK2A and mGIRK4/CIR Clones and mGIRK2/3/2 Chimera Construction-The mGIRK2A clone was isolated by screening a mouse brain library with a GIRK1 probe as described (13). The mouse GIRK4/CIR clone was amplified by polymerase chain reaction (PCR) 1 using primers corresponding to the published rat sequence (16) and subcloned into the pEXO plasmid (17). cDNA clones were sequenced on both strands by using the dye terminator method on an automatic Sequencer (Applied Biosystems model 373A).
To construct the chimera mGIRK2/3/2, the mGIRK3 sequence was mutated at positions 151 (the A of the initiation codon taken as base 1) and 1012 to introduce MunI and NheI restriction sites, respectively, without modification of the amino-acid coded sequence. Site-directed mutagenesis was performed using oligonucleotide primers according to * This work was supported by CNRS, the Association Française contre les Myopathies (AFM), Commission of the European Communities Contract CHRX-CT93-0167, and Bristol-Myers Squibb company for "Unrestricted Award." 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.
The nucleotide sequence(s) reported in this paper has been submitted to the GenBank TM /EMBL Data Bank with accession number(s) U33631.
In Situ Hybridization-All experiments were performed on 10-to 12 week-old (250 -300 g) male Wistar rats (Charles River Laboratories), by using standard procedures (19). Antisense RNA probes were generated by in vitro transcription, using [␣-33 P]UTP (1000 Ci/mmol, Amersham) from linearized plasmids containing a 82-base pair HindIII fragment of GIRK1 cDNA in the 3Ј-coding sequence, and a 244-base pair BamHI fragment of GIRK2 cDNA in the 5Ј-untranslated sequence. Sections (10 m) were treated and probed as described (19) and exposed to Amersham ␤-max Hyperfilm. Selected slides were dipped in Amersham LM1 photographic emulsion and exposed for 2 weeks at 4°C and then developed in Kodak D-19 for 4 min. All slides were counterstained with Cresyl violet. For control experiments, adjacent sections were hybridized with sense probe or digested with RNase before hybridization.
Antibody Preparations, Immunoprecipitations of Transfected GIRK Channels-DNA fragments corresponding to the carboxyl termini of mGIRK2 (44 amino acids) and mGIRK4/CIR (91 amino acids) were amplified by PCR and subcloned into the pGEX3 plasmid behind the glutathione S-transferase (GST) coding sequence. GST-GIRK fusion proteins were prepared and purified according to the manufacturer's protocol (Pharmacia Biotech Inc.). Antibodies directed against these proteins were raised in rabbits and guinea pigs using routine immunization protocols, boosting and bleeding being performed every 4 weeks. Polyclonal rabbit (R) and guinea pig (GP) antibodies, noted R␣GIRK2 and R␣GIRK4 and GP␣GIRK2 and GP␣GIRK4, were first depleted of the anti-GST antibodies by repeated absorption against strips of nitrocellulose saturated with the GST protein. The depleted sera were then affinity-purified against their respective antigen fixed on nitrocellulose. Specificity of antibodies was verified by Western blotting and immunoprecipitation assays on transfected cell microsomes.
TSA201 cells were transfected with mGIRK2 and mGIRK4/CIR subcloned into the expression vector pcDNA (Invitrogen) by the calcium phosphate method. After 48 h, cells were harvested and microsomes were prepared. Briefly, cells were homogenized in 150 mM NaCl, 3 mM EDTA, 0.1 mM phenylmethylsulfonyl fluoride, 0.7 g/ml pepstatin A, and 10 mM Tris-HCl (pH 8) buffer, centrifuged at 1000 ϫ g, and the supernatant was pelleted at 100,000 ϫ g for 30 min. Pellets were dissolved in the homogenization buffer and stored at Ϫ20°C. Aliquots of 50 g of microsomes were solubilized in 100 l of homogenization buffer containing 1% Triton X-100. After 1 h at 4°C, the volume was adjusted to 400 l with homogenization buffer (final Triton X-100 concentration of 0.25%). Preimmune or immune sera were added for 3 h at a 200-fold dilution, followed by addition of 10 l of protein A immobilized on Sepharose CL-4B (Sigma) for 1 h at 4°C under slow rocking. Pellets were washed five times with the homogenization buffer containing 0.25% Triton. Immunoprecipitated proteins were resolved by SDS-polyacrylamide gel electrophoresis (10% polyacrylamide) and transferred onto nitrocellulose membrane (Hybond-C extra, Amersham). Blots were saturated with phosphate-buffered saline containing 3% low-fat dry milk and incubated with affinity-purified guinea pig polyclonal antibodies diluted 400-fold. Blots were revealed with a F(abЈ) 2 -purified horseradish peroxidase-conjugated goat anti-guinea pig antibody (Cappel) and then incubated with substrate for ECL (chemiluminescence method, Boehringer).
Preparation of Xenopus laevis oocytes and cRNA injections have been described previously (20). 33 ng of GIRK cRNAs and 833 pg of ␤ 1 ␥ 2 subunit cRNAs were injected per oocyte.
Electrophysiology-The two-microelectrode electrophysiological measurements were performed as described (20). For patch-clamp experiments, devitellinized oocytes were placed in a bath solution containing 140 mM KCl, 1.8 mM CaCl 2 , 2 mM MgCl 2 , 5 mM HEPES at pH 7.4 with KOH. Pipettes were filled with a high K ϩ solution (40 mM KCl, 100 mM potassium methanesulfonate, 1.8 mM CaCl 2 , 2 mM MgCl 2 , and 5 mM HEPES adjusted at pH 7.4 with KOH). 100 M GdCl 3 was added to the pipette solution to inhibit the activity of the stretch-activated channels. Inside-out patches were perfused with a solution containing 140 mM KCl, 10 mM MgCl 2 , 5 mM HEPES adjusted at pH 7.2 with KOH and 5 mM EGTA added daily. Single-channel signals were filtered at 3.5 kHz and analyzed with the Biopatch software (Bio-Logic).

RESULTS
Cloning of a Splice Variant of mGIRK2-Screening a mouse brain cDNA library at low stringency with a GIRK1 probe resulted in the isolation of the clone mGIRK2 (13). During this screening, a splice variant of mGIRK2, noted herein as mGIRK2A, was also isolated. This clone had an overall size of 2.7 kilobases, identical with mGIRK2. It also displayed exactly the same 5Ј sequence up to the GGG glycine codon located just upstream of the TGA stop codon of mGIRK2. From that point, the mGIRK2A sequence totally diverged, leading to an extra reading frame of 11 amino acids at the COOH end (Fig. 1a). The entire 3Ј-untranslated sequence was then found to be unique as judged by sequence determination of its 300 first nucleotides and by analysis of its restriction map (not shown). It is likely that the two mGIRK2 clones are alternatively spliced products of the same gene. The amplification by reverse transcription-PCR of DNA fragments specific for each splice variant from mouse brain messenger RNA excluded the possi-FIG. 1. Sequences and PCR detection of mGIRK2 splice variants. a, nucleotide and deduced amino acid sequences of the carboxyl termini of mGIRK2 splice variants. Nucleotides are numbered from the first initiation ATG codon, and amino acids are numbered beginning with the initiating Met. Nucleotides that differ between mGIRK2 and mGIRK2A are printed in italics. The carboxyl-terminal 11 amino acids specific to the mGIRK2A variant are shown in bold. Sequences corresponding to the oligonucleotides sense (P1 and P2) and antisense (P3 and P4) used in b are boxed. b, reverse transcription-PCR amplification of both splice variants. Specific DNA fragments were amplified from mouse brain cDNA by using P1 and P3 primers for mGIRK2 and P1 and P4 primers for mGIRK2A. PCR products were blotted and probed with the 32 P-labeled P2 oligonucleotide. bility that one clone is an artifactual chimeric DNA produced during the library preparation and confirmed the existence of the two forms of mGIRK2 transcripts (Fig. 1b).
The entire coding region of the mGIRK4/CIR cDNA was cloned from heart mouse cDNA (GenBank TM accession number U33631). It shares 95% and 97.6% sequence identity with the rat CIR (16)/rcKATP (14) at the nucleotide and amino acid levels, respectively. The percentages of amino acid identities between mGIRK4/CIR and the other mGIRKs were 64.3% (mGIRK1), 71% (mGIRK2A), and 70% (mGIRK3). These values fall down to 44% and 46% in comparison between mGIRK4/CIR and ROMK1 and IRK1, respectively. The highest degree of sequence conservation between all these channels was found in a central core, starting approximately 50 residues upstream of the first transmembrane domain and ending 175 residues downstream of the second transmembrane domain.
Distribution of the mGIRK mRNAs-The Northern blot analysis presented in Fig. 2a compares the relative expressions of the different mGIRKs transcripts in mouse brain and heart. While mGIRK1 was found at almost the same level in both tissues, the other subunits were differentially expressed. mGIRK2 and mGIRK3 were abundant in brain and were apparently absent in heart. The reverse situation was observed for mGIRK4/CIR. A more extensive Northern blot analysis which included mRNAs from other mouse tissues such as skeletal muscle, kidney, lung, and liver showed that mGIRK2 and mGIRK3 were expressed only in brain with the exception of a low mGIRK3 expression in skeletal muscle ((13) and not shown). Hence, these two channel subunits may represent the specific components of the neuronal G-protein-activated inward rectifiers.
In situ hybridization studies have shown that the gene ex-pression patterns of GIRK1, GIRK2, and GIRK3 are widely distributed in the brain and are very similar ((21, 22) and results not shown). The highest expression levels appeared in the neo-and allocortical regions, hippocampus, olfactory bulb, and cerebellum. The mGIRK4/CIR expression was very low in the adult rat brain (not shown). To determine the potential significance of heteromultimeric formation in the brain, a high resolution study obtained by microscopic analysis of emulsiondipped sections has been performed. Fig. 2b shows an example of the high degree of overlap of the expression patterns of mGIRK1 and mGIRK2 transcripts. More than 80% of the neurons are labeled with the mGIRK1 and mGIRK2 probes in the CA3 pyramidal cell layer of the hippocampus (Fig. 2b). Coexpression of both transcripts in the same neuron type was also observed in most of the other strongly labeled central nervous system areas and was particularly evident in the granule cells of the dentate gyrus, in granular layers of the cerebellum, and in the mitral cells in the olfactory bulb (not shown). No area expressing only one GIRK mRNA could be clearly detected in the whole brain.

Properties of Macroscopic mGIRK Currents Expressed in Xenopus Oocytes-Both
Northern blot analysis and in situ hybridization have shown that mGIRK1, mGIRK2, and mGIRK3 are highly represented in brain while mGIRK4/CIR is not. Thus, it is unlikely that the mGIRK1 ϩ mGIRK4/CIR (mGIRK1,-4) combination that has been shown to form the I KACh channel in atrial cells (16) is an abundant GIRK channel in the brain. The most probable combinations are made of the assembly of the three other subunits. All attempts to express the mGIRK3 subunit have been unsuccessful, either alone or in combination with mGIRK1, mGIRK2, mGIRK4/CIR, or mGIRK1,-4 (8, 7, 4, and 4 independent batches of oocytes for the combinations mGIRK1,-3, mGIRK2,-3, mGIRK3,-4, and mGIRK1,-3,-4, respectively, with at least 3 oocytes per batch). Consequently, the electrophysiological study was restricted to currents expressed by the mGIRK1, mGIRK2, and mGIRK4/CIR cRNAs.
Similarly to I KACh (23), expressed mGIRK channels are stimulated by the G-protein ␤ 1 ␥ 2 dimer. Although the requirement for these G-protein subunits was not systematically observed for heteromultimeric channels (not shown), ␤ 1 ␥ 2 were always co-injected in the following part of the work. Injections of mGIRK2 or mGIRK1 ϩ mGIRK2 (mGIRK1,-2) into Xenopus oocytes resulted in the expression of inwardly rectifying currents. The expression of mGIRK2 in the absence of other mGIRK subunits was successful only in 35% of the oocyte batches tested. This low expression frequency was nevertheless sufficient to allow a detailed characterization of the biophysical properties of the current. The GIRK current expression frequency reached 100% with combined injections of mGIRK1 and mGIRK2. Fig. 3, a and b, shows superimposed current traces evoked by voltage steps ranging from Ϫ135 to ϩ45 mV in 30-mV increments from a holding potential of 0 mV (K ϩ equilibrium potential). mGIRK2 (Fig. 3a) and mGIRK1,-2 ( Fig. 3b) currents in response to hyperpolarizing voltage steps display different kinetics. mGIRK2 currents activate rapidly, in less than 5 ms, and then partially inactivate with a time constant of 243 Ϯ 15 ms (n ϭ 8) at Ϫ130 mV. Activation/inactivation kinetics of mGIRK1,-2 currents were very different, with a slower time constant (81 Ϯ 5 ms at Ϫ130 mV, n ϭ 14) and no inactivation.
As expected for a K ϩ -selective inward rectifier (24), the activation potential of mGIRK2 became more negative as external K ϩ concentration decreased and the amount of shift (52.6 Ϯ 0.8 mV, n ϭ 6, for a 10-fold change in external [K ϩ ] was close to the K ϩ equilibrium value (59 mV) estimated from the Nernst equation (Fig. 3, c and d). The shifts in the threshold of activation for mGIRK1,-2 and mGIRK2 ϩ mGIRK4/CIR (mGIRK2,-4) were, respectively, 50.9 Ϯ 2.3 mV (n ϭ 4) and 50.6 Ϯ 3.3 mV (n ϭ 3) for a 10-fold change in external [K ϩ ] (not shown), consistent with a predominant K ϩ selectivity for these channels.
As for the majority of K ϩ selective channels, external application of Ba 2ϩ or Cs ϩ blocked mGIRK2 currents in a concentration-dependent manner (Fig. 3, e-h). The Cs ϩ block was voltage-dependent giving rise to typical bell-shaped I/V curves for potential values negative to Ϫ50 mV (Fig. 3f). The mechanism of Ba 2ϩ block, for concentrations less than 1 mM, is probably of the "open channel block" type (24) as suggested by the pronounced fast inactivation component of the resulting current (Fig. 3g). The IC 50 values for the Cs ϩ inhibition were 94.  (15), such as pinacidil and P1060, both at 100 M, were without effect on mGIRK2, mGIRK1,-2, and mGIRK2,-4 currents. On the other hand, verapamil and bepridil, two L-type Ca 2ϩ channel blockers, partially inhibited these currents, up to 60% and 40%, respectively, at 100 M (not shown).
Finally, activation of protein kinase C by the phorbol 12myristate 13-acetate (30 nM), the diacylglycerol analog, OAG (100 M), or arachidonate (100 M), and activation of protein kinase A by forskolin or 8-chloro-cAMP (3 and 300 M) were without effect on GIRK currents (not shown).
Single-channel Analysis and Rectification Properties-Single-channel properties of the two splice variants, mGIRK2 and mGIRK2A co-expressed with mGIRK1 in Xenopus oocytes (mGIRK1,-2 and mGIRK1,-2A), were compared by examining the dependences on internal Mg 2ϩ concentration of their inward rectification and by measuring their unitary conductances and their open-time distributions (Fig. 4, a-f). In the presence of 10 mM Mg 2ϩ , the mGIRK1,-2 and mGIRK1,-2A channels recorded in inside-out patches showed similar inward rectification which could be removed in Mg 2ϩ -free internal solution. The unitary conductances of mGIRK1,-2 and mGIRK1,-2A were 37 Ϯ 8 pS (n ϭ 5) and 39 Ϯ 6 pS (n ϭ 5), respectively. The open-time distribution in steady-state conditions for mGIRK1,-2 and mGIRK1,-2A at Ϫ80 mV was fitted by a single exponential characterized by a time constant of 0.21 ms and 0.16 ms, respectively (Fig. 4, c and f). In conclusion, no difference in the channel properties of the two forms of mGIRK2 transcripts was detected.
Since a highly voltage-dependent block both by intracellular Mg 2ϩ and by the polyamine spermine have been shown to underlie strong inward rectification in cloned inward rectifiers (25-27), we tested the internal spermine dependence of the inward rectification of our expressed channels. In the experi-  4. Single-channel properties of mGIRK1,-2 (a, b, and c)  ment illustrated in Fig. 4, g and h, inside-out patches containing mGIRK1,-2 channels maintained at ϩ80 mV were first perfused with a Mg 2ϩ -free internal solution leading to an immediate removal of the inward rectification. Then, application of 100 M spermine led to a complete blockade of the outward current which promptly reappeared after spermine removal. The bar graph (Fig. 4i) shows that the spermine block was dose-dependent with an IC 50 of about 10 M. Essentially the same spermine effects were obtained on oocytes co-expressing mGIRK1 and the splice variant mGIRK2A (not shown).
Aspartic acid, a negatively charged amino acid present in the second transmembrane domain of inward rectifiers which are not regulated by G-proteins such as IRK1 and BIR10 (positions 172 and 158) has been shown to be implicated in their Mg 2ϩ and spermine sensitivities (25,(27)(28)(29). The corresponding residue is an aspartate in position 173 in mGIRK1 and a neutral asparagine in position 185 in mGIRK2. To evaluate the importance of the charge at this position for the rectification characteristics of the heteropolymeric G-protein-activated channel mGIRK1,-2, we took advantage of the presence of an asparagine (instead of an aspartate) residue in the corresponding position in mGIRK4/CIR (position 180) as in mGIRK2 (13). The mGIRK2,-4 channel has no negative charge in the positions that have been considered as crucial for Mg 2ϩ -and polyamineinduced inward rectification in IRK channels. Similarly to mGIRK1,-2, mGIRK2,-4 presents an inward rectification in the presence of 10 mM Mg 2ϩ (Fig. 5, a and b). In addition, Fig. 5 (d  and e) shows that the outward current recorded in a Mg 2ϩ -free solution at ϩ80 mV was totally abolished in the presence of 100 M spermine. In symmetrical 140 mM K ϩ , the unitary conductance was 39 Ϯ 5 pS (n ϭ 5), and the time constant of the open-time distribution in steady-state conditions at Ϫ80 mV was 0.51 ms (Fig. 5c).
Immunochemical Demonstration that mGIRK Proteins Form Heteromultimers-To demonstrate biochemically the effective association of mGIRK2 and mGIRK4/CIR proteins in a multimeric complex, specific antibodies were raised against these two subunits and used in immunoprecipitation-immunoblot studies. The R␣GIRK4 antibodies are specific for the mGIRK4/ CIR subunit as they do not immunoprecipitate the mGIRK2 subunit in mGIRK2-TsA transfected cells. However, the two proteins were coprecipitated by R␣GIRK4 in cells cotransfected with both plasmids (Fig. 5f, RK4 immunoprecipitated). The two subunits did not coprecipitate when they were expressed in separate cells and mixed afterward during the solubilization process, demonstrating that mGIRK2 and mGIRK4/CIR subunits cannot co-aggregate during the immunoprecipitation reaction. These data strongly suggest that the observed coprecipitation is indeed due to the biosynthetic formation of heteromultimeric channels. The specificity of the guinea pig revealing antibodies (GP␣GIRK2 and GP␣GIRK4) was demonstrated by using a combination of cells transfected with the different GIRK plasmids (Fig. 5f, Microsomes).
ATP Prevents the Rundown of mGIRK Channel Activity-In the cell-attached conformation, the expression of the mGIRK currents was stable for periods of time as long as 1 h. However, when patches were excised in ATP-free internal solution, channel activities quickly ran down. The presence of an internal solution containing ATP but not ADP partially prevented this run down. Fig. 6 illustrates the effect of 10 mM disodium ATP (ATP/2Na) on the cytoplasmic face of the patch excised from an oocyte expressing the mGIRK1,2 channel. The final Na ϩ concentration of the internal solution was kept constant at 20 mM. Fig. 6, a and b, shows that the channel activities were strongly dependent on the presence of ATP. Because the internal solution contained 10 mM MgCl 2 , it appeared that ATP was prob-ably mainly associated with Mg 2ϩ , suggesting a possible involvement of a kinase in the rundown process. However, in the presence of 10 mM ATP, the perfusion of 40 units/ml protein kinase A catalytic subunit did not modify channel activity and did not reverse or prevent the slow rundown (Fig. 6c). Moreover, neither the protein kinase C inhibition with protein kinase C fragment 530 -558 (0.5 M) or with the protein kinase inhibitor PKI (20 M), nor the protein kinase C activation with OAG (100 M) modified the rundown and/or the effect of ATP (not shown). Alkaline phosphatase (100 units/ml) did not prevent the effect of ATP (Fig. 6d). Finally, the nonhydrolyzable ATP analog (AMP-PCP) was as effective as ATP itself in reversing the rundown (Fig. 6, e and f). These experiments suggest that the activity of the mGIRK1,-2 channel does not require a phosphorylation/dephosphorylation process but rather the binding of ATP without hydrolysis. In the experiment shown in Fig. 6f Surprisingly, channel activities were maximal when disodium ATP (ATP/2Na) was used instead of Mg-ATP. Fig. 7a shows that mGIRK1,-2 channel activity could be partly re- f, immunoprecipitation of mGIRK2 and mGIRK4/CIR subunits from TsA201-transfected cells with pcDNA carrying the indicated GIRK coding sequences. Corresponding microsomes were solubilized in nondenaturing conditions and analyzed by immunoblotting directly or after immunoprecipitation with anti-mGIRK4/CIR rabbit antibodies (lane RK4). Antibodies used to reveal the Western blots were prepared from guinea pig (revealing antibodies K2 and K4). stored by application of a 20 mM NaCl in an ATP-free internal solution. To reach maximal channel activity, the simultaneous presence of ATP and Na ϩ ions was required. Channel activity was only 20% of the maximal activity on application of 10 mM Mg-ATP in Na ϩ -free solution (Fig. 7b). Fig. 7c presents mean results from 5 experiments and clearly shows the synergy of action of ATP and Na ϩ in the restoration of channel activity. In the presence of ATP, Li ϩ could replace Na ϩ for the activation of mGIRK1,-2 channels but with less efficacy (Fig. 7d). The sensitivity to internal Na ϩ led us to check if the cloned mGIRK channels could be related to K ϩ channels activated by internal Na ϩ which have been described in cardiac and neuronal cells (30,31). In fact, the only similarity is the requirement of a high concentration of Na ϩ (Ͼ30 mM) to activate these channels. The high unitary conductance (Ͼ100 pS), the impossibility to replace Na ϩ by Li ϩ for activation, and the specific blockade by R56865 which characterizes the Na ϩ -sensitive K ϩ channel (32) were not found for mGIRK channels (not shown).
Expression of a mGIRK3-mGIRK2 Chimera-As previously indicated, mGIRK3 failed to express alone or in combination with GIRK1, mGIRK2, or mGIRK4/CIR. However, the chimeric protein (Fig. 8a) in which the central core domain, including the transmembrane segments and the putative K ϩ pore of mGIRK3, is linked to the cytoplasmic amino and carboxyl termini sequences of mGIRK2 is functional. In association with mGIRK1, it expresses an inward rectifying K ϩ channel activity which is very similar to that produced by the mGIRK1,-2 channel (Fig. 8b). Co-injection of mGIRK2/3/2 with mGIRK2 also increases the mGIRK2 expression (not shown).

DISCUSSION
Four proteins with structures corresponding to G-proteingated inward rectifier (11)(12)(13)(14)16) have been cloned to date. They are designated as mGIRK1, mGIRK2, mGIRK3, and mGIRK4/CIR. mGIRK4/CIR seems to be specific to the heart, and it is not detected in the brain. Conversely, mGIRK2 and mGIRK3 transcripts are specifically present in the brain. mGIRK1 is present at similar levels in heart and brain. In situ hybridization experiments have shown that the distribution of the three mGIRKs is very similar if not identical. Moreover, the colocalization of distinct GIRK transcripts in the same neuronal cells is in agreement with the hypothesis of heteromeric formation. This hypothesis is strongly supported by the tremendous increase of functional expression of GIRK channels  -out configuration (a-c).
In this series of experiments, the ATP used was the ATP-Mg 2ϩ salt. GIRK1,-2 activities were recorded at Ϫ80 mV and filtered at 5 Hz. In a and b, initially, patches were internally bathed with an ATP-and Na ϩ -free solution. a, effect of 20 mM Na ϩ in the absence of ATP. b, effects of 10 mM ATP and 10 mM ATP ϩ 20 mM Na ϩ . After the Na ϩ removal, note the instantaneous reduction of activity to the level reached in the presence of ATP alone. c, bar graph (n ϭ 5) indicating the respective increase of GIRK1,-2 activities in the presence of Na ϩ , ATP, and ATP ϩ Na ϩ (taken arbitrary as 100% in each experiment). d, effects of 20 mM Li ϩ followed by 20 mM Na ϩ in the presence of 10 mM ATP.
FIG. 8. Expression of a mGIRK2/mGIRK3 chimeric construct. a, scheme showing the contribution of mGIRK2 sequences (black) in the chimeric construct mGIRK2/3/2. b, bar graph showing the averaged currents recorded at Ϫ130 mV in oocytes injected with mGIRK1 and the ␤ 1 ␥ 2 G-protein subunits together with mGIRK2, mGIRK3, or the mGIRK2/3/2 chimeric assembly (respectively, n ϭ 32, 10, and 12, in 3 to 5 different batches of oocytes). The vertical bars indicate the S.E. when they are co-injected in the same oocyte as compared to single injections.
K ϩ channels expressed after the injection of the mGIRK2 cRNA alone or in combination with mGIRK1 cRNA (mGIRK1,-2) or with mGIRK4/CIR cRNA (mGIRK2,-4) in the presence of ␤ 1 ␥ 2 had the hallmarks of inward rectifier channels: (i) an activation by hyperpolarization negative to the reversal potential for K ϩ (E K ), (ii) an activation potential shifting with E K , and (iii) a blockade by Cs ϩ and Ba 2ϩ . However, in voltage-clamp conditions, there were some differences between expressions of mGIRK2 and mGIRK1,-2. mGIRK2 currents displayed a rapid activation (Ͻ5 ms) and partial inactivation, whereas mGIRK1,-2 channels had a slow activation and did not inactivate. Single-channel analysis of mGIRK2, mGIRK1,-2, and mGIRK2,-4 currents clearly demonstrated that there was only one population of channels with very similar properties characterized by a unitary conductance of about 40 pS and a flickering activity with a mean open time duration of less than 1 ms. The single-channel parameters of mGIRK2 and mGIRK1,-2 were very similar although their activation kinetics at the whole oocyte level were distinct.
A splice variant of mGIRK2 (mGIRK2A) has also been cloned. It has the same sequence as mGIRK2 but contains 11 additional amino acids in the carboxyl-terminal end. mGIRK2A transcripts are also specifically located in the brain. Electrophysiological results have not shown any significant difference between the two forms. Therefore, it is not easy to suggest any specific new function for mGIRK2A. A first possibility would be that the mGIRK2A subunit could associate with other mGIRKs which are not yet discovered. Another possibility is that the different carboxyl-terminal sequences could serve to impose different cellular localizations. Interestingly, the mGIRK2A terminal sequence SKV is very similar to the microbody targeting signal motif SKL (33).
How do expressed neuronal GIRK channels compare with "native" channels? Native GIRK channels recorded in different neuronal cell types have unitary conductances varying from 38 to 55 pS and a time constant of their open-time distribution which is of the order of 2 ms (7,34). It then appears that their conductances are similar, but flickering is more rapid for the cloned channels expressed in Xenopus oocytes. However, it should be noted that detailed literature describing neuronal GIRK channel properties at the single-channel level is not yet available. One possibility is that flickering GIRK channels are difficult to record in neuronal membranes where numerous other K ϩ channel activities might coexist. Another likely possibility is that some subunit which normally slows down the gating kinetics in native channels is still missing in cloned heteropolymeric channels.
It has been shown previously that the functional cardiac G-protein-activated inward rectifier is in fact composed of an assembly of rat GIRK1 and GIRK4/CIR (16). The K ϩ current expression described above suggests that mGIRK2 can also form heteromultimeric assemblies with mGIRK1 and mGIRK4/ CIR. This was actually directly demonstrated by immunoprecipitation studies in the case of the mGIRK2,-4 complex. Experiments using coexpression of mGIRK1 with chimeras of mGIRK3 (which do not express alone or co-injected with mGIRK1, mGIRK2, or mGIRK4/CIR) with the amino-and carboxyl-terminal sequences of mGIRK2 also tend to lead to the same conclusion. The apparent co-localization of mGIRK1 and mGIRK2 in the brain, particularly in CA3 pyramidal cells, is a strong indication that the mGIRK1,-2 complex is a major neuronal GIRK channel. The case of mGIRK3 is not clear. Its lack of expression suggests that it might need a partner that still has to be discovered. One possible partner is the sulfonylurea receptor (35). ATP-sensitive K ϩ channels are present in the brain (36 -38). They have inward-rectifying properties (23), are regulated by G-proteins (39,40), and may be constituted by the assembly of the protein that binds antidiabetic sulfonylureas (35) and an inward rectifier-type K ϩ channel.
After this work was submitted, it was published that the GIRK3 subunit can assemble with GIRK1 and with GIRK2 to either increase (GIRK1) or decrease (GIRK2) their activities (41). These effects were never seen in our own experiments. These apparently conflicting observations might be explained by assuming that a third, not yet identified, subunit is endogenously present in oocytes and confers the expression properties observed by Kofuji et al. (41). This component would not be present in our oocytes.
The inward rectification in cloned inward rectifiers (25-27) is due to a highly voltage-dependent block by intracellular Mg 2ϩ and by polyamines. Mutagenesis experiments have strongly suggested that aspartic acid in position 172 in the inward rectifier IRK1 is pivotal for the effects of Mg 2ϩ and spermine on the inward rectification (25)(26)(27). This Asp residue is present at corresponding positions in sequences of a number of cloned inward rectifier such as IRK1, mGIRK1, and BIR10 (29), but this residue is replaced by an asparagine in mGIRK2, mGIRK4/CIR, and also in ROMK1 (42). Although they lack this Asp residue, both mGIRK2 and mGIRK4/CIR, when they are expressed independently or when they co-expressed, possess all the hallmarks of inward rectifiers, contrary to ROMK1 which also has an Asn in the corresponding position 171 and which presents a quasilinear I-V relationship. The fact that replacement of Asn-171 by Asp in ROMK1 results in the appearance of a Mg 2ϩ -dependent inward rectification (43) would tend to confirm the important role of an Asp for Mg 2ϩ -dependent inward rectification. However, the fact that the expression of GIRK2,-4, with Asn in the sequences instead of Asp, also leads to a Mg 2ϩ -dependent inward rectifier K ϩ channel pleads for the importance of other residues and questions the unique role of this Asp for inducing this inward rectification.
One particularly interesting observation is the requirement of a high concentration of internal ATP (10 mM) in excised patches to prevent a fast rundown of both mGIRK1,-2 and mGIRK2,-4 activities. This ATP dependence would immediately suggest an important role of phosphorylation. However, results presented in this paper show that a kinase activity involving ATP hydrolysis is not implicated as it is for IRK1 and ROMK1 channels (44,45). Treatments capable of activating or inhibiting protein kinase A or protein kinase C activity were without effect on the rundown and/or the reactivating action of ATP. Moreover, alkaline phosphatase which would produce a dephosphorylation did not modify the response to ATP. Finally, the activating effects of the nonhydrolyzable ATP analog AMP-PCP on channel activity were similar to if not identical with those of ATP. All these results taken together show that mGIRK1,-2 and mGIRK2,-4 channels are ATP-regulated channels. They require ATP binding to be functional, but ATP hydrolysis is not necessary. ATP binding might occur at the nucleotide-binding site represented by the consensus Walker type A sequence G(X) 4 GK(X) 7 (V/I). This exact motif is missing in mGIRK sequences, but two motives that share similarities with the Walker A consensus sequence are present in the carboxyl-terminal extremities of mGIRK1, mGIRK2, and mGIRK3 subunits. The I(X) 4 GK(X) 6 V motif is present in mGIRK1 and mGIRK2, the V(X) 4 GR(X) 6 V sequence is present in mGIRK3. It has been suggested that similar motives could be implicated in ATP binding (44). The mGIRK4/CIR sequence does not possess such an ATP consensus sequence.
This paper also shows that internal Na ϩ is a regulator of the neuronal mGIRK1,-2 channel activity. This type of property has in fact been observed before with the inward rectifier K ϩ channel which is present in starfish eggs (46). ATP and Na ϩ are synergistic in their activating effects. The ATP and Na ϩ dependences of neuronal mGIRK activities might be important in neurological diseases. In ischemic situations, or in epileptic seizures, the intracellular ATP concentration drops rapidly while the internal Na ϩ concentration increases massively. It is then possible that the function of neuronal GIRK channels will be affected drastically, leading to changes of membrane polarization that might be an important component in a cascade of events leading to very deleterious effects.