Molecular Determinants for Sodium-dependent Activation of G Protein-gated K+ Channels*

G protein-gated inwardly rectifying K+ channels (GIRKs) are activated by a direct interaction with Gβγ subunits and also by raised internal [Na+]. Both processes require the presence of phosphatidylinositol bisphosphate (PIP2). Here we show that the proximal C-terminal region of GIRK2 mediates the Na+-dependent activation of both the GIRK2 homomeric channels and the GIRK1/GIRK2 heteromeric channels. Within this region, GIRK2 has an aspartate at position 226, whereas GIRK1 has an asparagine at the equivalent position (217). A single point mutation, D226N, in GIRK2, abolished the Na+-dependent activation of both the homomeric and heteromeric channels. Neutralizing a nearby negative charge, E234S had no effect. The reverse mutation in GIRK1, N217D, was sufficient to restore Na+-dependent activation to the GIRK1N217D/GIRK2D226N heteromeric channels. The D226N mutation did not alter either the single channel properties or the ability of these channels to be activated via the m2-muscarinic receptor. PIP2 dramatically increased the open probability of GIRK1/GIRK2 channels in the absence of Na+ or Gβγ but did not preclude further activation by Na+, suggesting that Na+ is not acting simply to promote PIP2binding to GIRKs. We conclude that aspartate 226 in GIRK2 plays a crucial role in Na+-dependent gating of GIRK1/GIRK2 channels.

The G protein-gated inwardly rectifying K ϩ channels (GIRKs) 1 were the first channels to be shown to be gated by a direct interaction with the ␤␥ subunits of GTP-binding proteins (1,2,3). This mechanism mediates the coupling of m2-muscarinic receptors with GIRK channels in the atria, and the generation of inhibitory postsynaptic currents by neurotransmitters acting on GABA B , 5HT 1A and A 1 receptors, in hippocampal neurons (4). More recently, other regulators of GIRK channels have been identified, which can activate these channels in the absence of G proteins. Recombinant GIRK1/GIRK4 channels and the native atrial channels were shown by Sui et al. (5) to be activated by a rise in cytosolic [Na ϩ ], with a threshold for activation of 3-10 mM and an EC 50 of ϳ40 mM. In isolated membrane patches, activation by Na ϩ required the hydrolysis of ATP. Subsequently, Sui et al. (6) showed that phosphatidylinositol 4,5-bisphosphate (PIP 2 ) mimics the ATP effects and that depletion or block of PIP 2 inhibits the activation of GIRK channels by both G␤␥ and Na ϩ . Huang et al. (7) reported similar findings and showed also that activation of GIRK channels, by applying PIP 2 in the presence of Na ϩ , precluded further activation by G␤␥.
Activation by internal Na ϩ is not unique to atrial GIRK1/ GIRK4 channels. GIRK1/GIRK2 heteromeric channels have also been shown to be activated by 20 mM Na ϩ in the presence of ATP (8), and a mutation in the pore region of GIRK2, which renders these channels permeable to both Na ϩ and K ϩ , causes a large increase in their G␤␥-independent activity (9,10,11).
Some of the molecular aspects of G protein gating of GIRKs have already been elucidated. G proteins have been shown to interact with both the cytoplasmic N-and C-terminal regions of GIRK subunits. Huang et al. (12) made a series of GIRK1 C-terminal fusion proteins and reported that the region between amino acids 318 and 462 was involved in binding purified G␤␥. A recent study by Krapivinsky et al. (13) identified a single region within the C-terminal tail of GIRK4 that lies in close proximity to the second transmembrane segment (TM2) (amino acids 209 -245) as being critical for G␤␥ binding and channel activation. A point mutation within the proximal C terminus of GIRK4, substituting threonine for cysteine 216, drastically reduced the potency of G␤␥ in activating not only GIRK4 homomeric channels but also GIRK1/GIRK4 heteromeric channels (13).
The aim of our study was to identify the region of GIRK1/ GIRK2 channels that mediates their activation by internal Na ϩ . Using a chimeric approach, combined with site directed mutagenesis of the GIRK subunits, we have identified the proximal C-terminal region of GIRK2 as playing a crucial role in the Na ϩ -dependent activation of both GIRK2 homomeric and GIRK1/GIRK2 heteromeric channels. Within this region, GIRK2 has an aspartate at position 226, which is conserved in GIRK4, but is an asparagine in GIRK1. Substituting asparagine for aspartate 226 in GIRK2, abolished Na ϩ -dependent activation of both the homomeric and GIRK1/GIRK2 heteromeric channels without affecting their ability to be activated via the m2-muscarinic receptors. The reverse mutation in the GIRK1 subunit (N217D) was sufficient to recover Na ϩ -dependent activation of the heteromeric channel.

EXPERIMENTAL PROCEDURES
cDNA Clones-Standard PCR procedures were used to construct a strong Kozak consensus sequence GCCGCCACC immediately upstream of the ATG initiation codon in cDNA clones for GIRK1 and GIRK2. We used the longer of the GIRK2 splice variants, which contains 423 amino acids (14). These constructs were then subcloned into the EcoRI site of the pBG7.2 vector, which provides the 5Ј-and 3Јuntranslated regions of the Xenopus ␤-globin gene. GIRK4 was also subcloned into the pBG7.2 vector, and the human muscarinic m2 re-FIG. 1. Activation of GIRK1/GIRK2 heteromeric channels by intracellular Na ؉ and PIP 2 . Multichannel recordings from oocytes coexpressing GIRK1 and GIRK2. A, channel activity recorded in the cell-attached mode ran down upon formation of an inside-out, isolated patch ceptor was inserted into pBluescript KS(II)ϩ.
Construction of Chimeras and Point Mutants-The construction of chimeras 1211 and 1222 was as described in Stevens et al. (15). Briefly, silent restriction sites were introduced for AflII and BssHII into the GIRK2 cDNA at positions corresponding to amino acids 87 and 199, respectively. These sites lie at the beginning of the first proposed membrane-spanning segment (TM1) and just downstream of the second membrane-spanning segment (TM2). GIRK1 has a BssHII site at the equivalent position (amino acid 190), downstream of TM2, and in addition, an AflII site was introduced at the DNA sequence corresponding to amino acid position 78. Standard subcloning, in which the N-and C-terminal hydrophilic regions of GIRK1 and GIRK2 were substituted individually and together, generated 1211 and 1222 chimeric constructs. For construction of 1221 and 1212 chimeras, an NheI restriction site was introduced into chimera 1211 chimera at the position corresponding to amino acid 361 using standard PCR methods. GIRK2 cDNA has an NheI site at the equivalent position (amino acid 370). 1221 and 1212 chimeric constructs were generated by exchanging the transmembrane and proximal C-terminal regions of 1211 and 1222 using AflII and NheI sites. Point mutants, GIRK2D226N and GIRK2E234S, were generated by oligonucleotide-mediated mutagenesis using standard PCR methods. 1211N217D mutant was generated using GeneEditor™ in vitro site-directed mutagenesis system (Promega). GIRK1N217D mutant was made by exchanging the transmembrane regions of GIRK1 and 1211N217D using AflII and BssHII sites. The sequence of all PCR-amplified products and point mutations were verified by DNA sequence analysis.
Preparation and Microinjection of Oocytes-Oocytes were surgically removed from female Xenopus laevis anesthetized with 0.3% (w/v) 3-amino benzoic acid (Sigma) and dissociated from connective tissue using 0.3% (w/v) collagenase (Sigma) in Ca 2ϩ -free buffer (mM): 82.5 NaCl, 2.5 KCl, 1 MgCl 2 , 5 HEPES, pH 7.6 with NaOH). Isolated oocytes were microinjected with 50 nl cRNAs dissolved in water. The in vitro transcription of cRNAs was as described previously in Stevens et al. (15). A similar total amount of channel subunit cRNA was injected into each oocyte. In some experiments, the human m2-muscarinic receptor cRNA was coinjected. Oocytes were incubated in ND96 at 18°C.
Electrophysiology-Two electrode voltage clamp recordings were performed 3-6 days after microinjection using an OC-725B amplifier (Warner Instruments), interfaced to a Macintosh Quadra 700 computer using an ITC16 AID board (Instrutech) with Pulse acquisition software (version 7.89; HEKA Electronics, Lambrecht, Germany). Microelectrodes filled with 3 M KCl had resistances ranging between 0.5 and 2 megaohms. Oocytes were continually perfused with standard recording solution (mM): 90 KCl, 1 MgCl 2 , 1 CaCl 2 , 1 HEPES (pH 7.4 with KOH). Currents were recorded in the presence of 3 M carbacol (CCh) before and after application of 1 mM Ba 2ϩ . Current records were filtered at 1 kHz and digitized at 5 kHz. Experiments were carried out at room temperature (22°C).
Patch clamp experiments were performed using an Axopatch 200A patch clamp amplifier, and currents were recorded at 10 kHz and filtered at 2 kHz. Patch pipettes had resistance of between 1 and 2 megaohms when filled with pipette solution and were coated with wax. Excised inside-out patch pipettes were transferred to a small chamber with a volume of ϳ100 l. Solutions were exchanged by perfusing through this small chamber a volume of 500 l. Continuous records of channel activity were obtained at a holding potential of Ϫ80 mV. Tensecond ramp tests from Ϫ140 mV to ϩ80 mV were performed throughout the recording to check that there were no noninwardly rectifying channels present in the patch. Data were acquired using Pulse acquisition software (version 8.11; HEKA Electronics).
The pipettes solution contained (mM): 96 KCl, 1 MgCl 2 , 1.8 CaCl 2 and 10 HEPES, pH 7.2 with KOH; while the bath solution contained (mM): 96 KCl, 2 MgCl 2 , 5 EGTA, 10 HEPES, pH 7.2 with KOH. Gadolinium at 100 M was routinely added to the pipette solution to suppress native stretch-activated channel activity in the oocyte membrane. In some experiments, CCh at 3 M was added to the pipette solution. NaCl (1-200 mM) or NMDG (10 -100 mM) was added to 96 mM KCl bath solution without compensation for changes in osmolarity and ionic strength. Solutions containing MgATP were prepared fresh each day and the pH was readjusted to 7.2 after addition of MgATP. The concentration of free Mg 2ϩ was maintained at ϳ1.9 mM and calculated using EQCAL software (BioTools). PIP 2 (a gift from R. Irvine) was purified from pig brain tissue. The stock of purified PIP 2 was in chloroform (8.7 mM) and was kept at Ϫ20°C in an air-tight glass container under nitrogen gas. Test solutions were prepared freshly before each experiment by evaporating the chloroform and redissolving in 100 l of Mg 2ϩ -free 96 mM KCl bath solution under sonication for 1-2 min. The final concentration of PIP 2 was adjusted by adding the appropriate volume of Mg 2ϩ -free 96 mM KCl bath solution.
Data Analysis-To measure channel activity, the mean current of each continuous recording segment was calculated by first subtracting the baseline current and then summing the amplitudes of all of the sample points and dividing by the number of sample points within the continuous recording segment (Igor Pro, version 3; Wavemetrics, Lake Oswego, OR). We did not correct for the small reduction in the unitary current amplitude at high [Na ϩ ]. For each Na ϩ -activated dose-response experiment, the mean currents at different Na ϩ concentration were normalized to the response obtained in the presence of 100 mM Na ϩ . For each Na ϩ -inhibited dose-response experiment, the mean currents were normalized to the response obtained at 0 mM Na ϩ . Dose-response curves were fitted to the Hill equation:

RESULTS
Gating of GIRK1/GIRK2 Heteromeric Channels by Intracellular Na ϩ and PIP 2 -We measured the effects of raising internal [Na ϩ ] on GIRK1/GIRK2 heteromeric channels heterologously expressed in Xenopus oocytes. Fig. 1A shows a multichannel recording firstly in the cell-attached mode and then following patch excision in a solution lacking ATP, GTP, and Na ϩ . Channel activity ran down over a period of 1-2 min and was partly restored by application of 5 mM MgATP and further increased by application of 20 mM Na ϩ in the continued presence of ATP. In contrast, application of 20 mM Na ϩ in the absence of MgATP had little effect on channel activity (Fig. 1B), and 5 mM AMP-PNP was unable to substitute for ATP in activating the channel. Even in the presence of 100 M GTP␥S plus 5 mM MgATP, application of 20 mM Na ϩ further increased channel activity (Fig. 1C), and the effects of Na ϩ were not blocked by 100 M GDP␤S (Fig. 1D). Thus, Na ϩ -dependent activation of GIRK1/GIRK2 channels appears to require ATP hydrolysis, but is independent of G proteins. Sui et al. (6) suggested that the production of PIP 2 by hydrolysis of MgATP is responsible for the ATP dependence of GIRK1/GIRK4 channel activity. They applied 1 M PIP 2 in the absence of Na ϩ or G␤␥ and showed little change in GIRK1/GIRK4 channel opening frequency, but a large increase in channel activity upon subsequent application of 20 mM Na ϩ . We applied 50 M PIP 2 for 5 min in the absence of Na ϩ or G␤␥ and saw a dramatic increase in the opening frequency of the channel (Fig. 1E), with no apparent change in the single channel conductance. Subsequent application of 20 mM Na ϩ further increased the open probability (P open ) by ϳ2-fold (1.93 Ϯ 0.16, n ϭ 3). Thus PIP 2 appears to be an activator of GIRK channels in the absence of either Na ϩ or G␤␥, but the actions of PIP 2 do not preclude further activation by Na ϩ . Fig. 2 shows the dose-response relationship for Na ϩ -activation of GIRK1/GIRK2, GIRK2, and GIRK1/GIRK4 channels in the presence of 5 mM MgATP. The data were fitted with the Hill (arrow). Application of 5 mM MgATP and 20 mM Na ϩ , for the duration indicated by the bars, re-activated GIRK1/GIRK2 channels. B, application of 20 mM Na ϩ alone and together with 5 mM AMP-PNP was unable to reactivate GIRK channels. C, 100 M GTP␥S in the presence of 5 mM MgATP maintained channel activity in an inside-out patch; subsequent application of 20 mM Na ϩ further enhanced channel activity. D, channel activation by 20 mM Na ϩ was not blocked by 100 M GDP␤S. E, recordings from an inside-out patch, before and after application of 50 M purified PIP 2 . PIP 2 dramatically increased the channel opening frequency in the absence of Na ϩ or G␤␥. Subsequent application of 20 mM Na ϩ further increased the channel open probability. The holding potential was Ϫ80 mV, and the solution in the bath and in the pipette contained 96 mM K ϩ . Hill coefficients were close to 2 for GIRK1/GIRK2 and GIRK1/ GIRK4 and close to 4 for the GIRK2 homomeric channels (Table I). This suggests that all four subunits within the homomeric GIRK2 channel complex bind Na ϩ , whereas in the heteromeric channels, only two of the four subunits are Na ϩsensitive. GIRK1 subunits do not form functional homomeric channels and so to test the possibility that GIRK1 subunits are Na ϩ -insensitive, we looked at the effects of Na ϩ on a chimeric channel (1211) containing the N-and C-terminal tails of GIRK1 and the TM and H5 regions of GIRK2 (Fig. 2C). We have shown previously that this chimeric subunit (previously called 121), expressed with the m2 receptor in Xenopus oocytes, gives rise to large ACh-induced currents (15). In contrast to the GIRK1/GIRK2 or GIRK2 currents the 1211 currents were inhibited by increasing internal Na ϩ , with an IC 50 of ϳ23 mM and a Hill coefficient of ϳ1.0 (Fig. 2D). This inhibition appears to be partly caused by a reduction in the single channel conductance at Na ϩ levels greater than 10 mM (Fig. 2E) and partly by a decrease in the channel P open . Interestinly the 1211 channels were more sensitive to MgATP than were the wild type channels. Application of 2 mM MgATP, in the absence of Na ϩ or GTP, caused a 4.9 Ϯ 0.5 (n ϭ 3)-fold increase in the P open of 1211 channels (Fig. 2F) and only a 2.1 Ϯ 0.5 (n ϭ 3)-fold increase in the P open of GIRK1/GIRK2 channels. Thus, the loss of Na ϩ -dependent activation does not appear to be caused by a decrease in the ATP sensitivity of the channel.
NMDG had a similar effect to Na ϩ on the 1211 currents, and it also inhibited the GIRK1/GIKR2 currents. Fig. 2G shows GIRK1/GIRK2 currents in the presence of increasing concentrations of NMDG and then following wash out. The inhibition by NMDG was fully reversible and the dose response relation was fitted with a Hill equation with an IC 50 of ϳ20 mM and a Hill coefficient of 1.2 (Fig. 2H). These results suggest that Na ϩ may have a dual action on the GIRK1/GIRK2 channels, an inhibitory action that is masked by the ability of Na ϩ to activate the channel by another mechanism. Na ϩ -dependent activation of GIRK1/GIRK2 channels appears to require the cytoplasmic tail regions of the GIRK2 subunit.
The Proximal C-terminal Region of GIRK2 Confers Na ϩ -dependent Activation-To locate the Na ϩ -sensing region of GIRK2 we looked at the effects of Na ϩ on the 1222 chimeric subunit. This differs from chimera 1211 in that it has the GIRK2 C-terminal tail and only the N-terminal tail of GIRK1. 1222 subunit (previously called 122) does not appear to form functional homomeric channels but does form functional heteromeric channels when coexpressed with GIRK1 in Xenopus oocytes (15). Increasing internal Na ϩ activated this channel with an EC 50 of ϳ43 mM and a Hill coefficient of 2.3 (Fig. 3, A  and B), suggesting that the C-terminal tail of GIRK2 is sufficient to restore the Na ϩ -dependent activation of GIRK channels. To further narrow down the region crucial for Na ϩ gating we generated two additional chimeric subunits, 1221 and 1212, which divided up the GIRK2 C terminus into two segments: amino acids 199 to 369, which are highly conserved within the GIRK family, and the poorly conserved region downstream of residue 369 (Fig. 3C). Two electrode voltage clamp measurements showed that both of these chimeric subunits produced much larger whole cell currents when coexpressed with GIRK1 than when expressed alone, indicating that they formed heteromeric complexes with GIRK1. We compared the effects of Na ϩ , in the presence of MgATP, on the homomeric chimeric channels and the heteromeric channels. The 1221 and GIRK1/ 1221 channels were both reversibly activated by Na ϩ , whereas the 1212 and GIRK1/1212 channels were inhibited (Fig. 3). The Na ϩ dose-response curve for 1221 had a Hill coefficient of 3.8, whereas the curve for GIRK1/1221 had a Hill coefficient of 1.6 (Table I). Thus, the region of the GIRK2 C-terminal tail that lies proximal to TM2 appears to be responsible for the Na ϩ -dependent activation of GIRK1/GIRK2 channels.
Aspartate 226 Is Important for Na ϩ -dependent Activation of GIRK2 and GIRK1/GIRK2 Channels-We compared the sequences of GIRK1, GIRK2, and GIRK4 within the proximal C-terminal region and looked for positions where there is a negative charge in GIRK2 and GIRK4 that is not present at the equivalent position in GIRK1 (Fig. 4A). There are 7 acidic residues conserved between GIRK2 and GIRK4 but not GIRK1. Two of these residues are located within the first 45 amino acid segment downstream of the TM2 region (Fig. 4A). There is an aspartate at position 226 in GIRK2, which is also present in GIRK4, but is an asparagine in GIRK1. There is a glutamate at position 234, where GIRK1 has a serine. We generated two GIRK2 mutants, GIRK2D226N and GIRK2E234S, and expressed these individually and together with GIRK1, plus the m2-muscarinic receptor. The mutants displayed similar characteristics to wild type GIRK2 in that they produced small whole cell currents when expressed individually but much larger currents when coexpressed with GIRK1 indicating the formation of functional heteromeric channels (Fig. 4B). The time course of the GIRK1/GIRK2E234S and GIRK1/  a Group I is Na ϩ -activated heteromeric channels. b Group II is Na ϩ -activated homomeric channels. c Group III is Na ϩ -inhibited channels.
FIG . 3. The proximal region of the GIRK2 C-terminal tail is required for Na ؉ -dependent activation. A, activation of GIRK1/1222 heteromeric channels by internal Na ϩ in the presence of 5 mM MgATP. B, the dose-response relationship fitted with a Hill equation with EC 50 of GIRK2D226N heteromeric currents and their degree of inward rectification were indistinguishable from wild type GIRK1/ GIRK2 currents (Fig. 4C and Ref. 15). The mutant GIRK2 homomeric currents also rectified strongly, suggesting that neither glutamate 234 nor aspartate 226 plays a key role in determining the rectification properties of these channels. Fi-43 mM and Hill coefficient of 2.3 (n ϭ 4). C, diagram to illustrate the structure of the chimeric subunits. D, the mean amplitudes of the Ba 2ϩ -sensitive basal and CCh-induced currents recorded using the two-electrode voltage clamp method from oocytes injected with the subunit cRNAs indicated. All of the values shown were obtained from the same batch of oocytes and represent the mean Ϯ S.E. for at least six oocytes. Coexpression of GIRK1 with the chimeric subunits dramatically increases the size of the currents, indicating the formation of functional GIRK1/1212 and GIRK1/1221 heteromers. E and G, GIRK1/1221 currents (E) and GIRK1/1212 currents (G), recorded from inside-out macropatches in the presence of increasing internal Na ϩ . Recovery represents return to 0 mM Na ϩ . F and H, Na ϩ dose-response relationships fitted with Hill equations. nally, all of the mutant channels were activated by application of CCh, indicating that coupling to the m2-muscarinic receptor was not disrupted by the mutations.
GIRK2E234S and GIRK1/GIRK2E234S channels were activated by increasing internal Na ϩ , similar to the wild type channels; the dose-response relationships were fitted with Hill functions with EC 50 values of ϳ42 and ϳ30 mM, respectively, and Hill coefficients of 4.0 and 1.6, respectively (Fig. 5, A and  B). In contrast both GIRK2D226N channels and GIRK1/ GIRK2D226N channels were inhibited by increasing internal Na ϩ (Fig. 5, C and D), similar to the 1211 channels. Thus the aspartate at position 226 in GIRK2 appears to play a crucial role in Na ϩ -dependent activation of both GIRK2 homomeric and GIRK1/GIRK2 heteromeric channels. Two other mutations at this position, D226K and D226E, failed to produce functional channels, either when expressed alone or coexpressed with GIRK1. As a result, we were not able to further investigate the relationship between the structure of the side group of this residue and activation by Na ϩ .
The reverse mutation in the 1211 chimera, N217D, was able to confer Na ϩ -dependent activation to this channel. The doseresponse relationship was best fitted with a Hill coefficient with an EC 50 of ϳ37 mM and a Hill coefficient of 3.6 (Fig. 5, E and F, and Table I). This is very similar to the dose-response relationship for Na ϩ -dependent activation of the GIRK2 homomeric channel. The reverse mutation in the GIRK1 subunit, N217D, was also sufficient to recover Na ϩ -dependent activation to GIRK1N217D/GIRK2D226N heteromeric channels (Fig. 5, G and H, and Table I).
Single channel records of the wild type and mutant channels in the cell-attached configuration are shown in Fig. 6. The E234S mutation, while not affecting the Na ϩ -dependent activation of the channel, did alter the intrinsic gating of the GIRK2 homomeric channels; all of the traces show a rapid flickering between the open and closed states of the channel. However, the D226N mutation did not appear to change either the kinetic behavior or the unitary conductance of the homomeric and heteromeric channels. GIRK2 and GIRK2D226N openings were very brief; the open time distributions were fitted with single exponential with time constants of 0.51 and 0.50 ms, respectively. The heteromeric channels displayed longer openings; there was an additional slower component to the open time distribution with a time constant of 2.8 ms (36%) for GIRK1/GIRK2D226N and 3.8 ms (23%) for GIRK1/GIRK2. Thus the loss of Na ϩ -dependent activation of the mutant channels does not appear to be caused by a change in their intrinsic gating behavior. Interestingly the reverse mutation in the 1211 chimera, N217D, did alter the intrinsic gating of the channel. In the cell-attached recording mode, this mutant displayed very brief opening events ( ϭ 0.21 ms), and correspondingly very small whole cell currents (Fig. 4B), unlike the 1211 chimera, which produces large whole cell currents (15) and displayed much longer openings; 1 ϭ 0.85 (61%), 2 ϭ 2.5 ms (39%). DISCUSSION We have shown that the aspartate residue at position 226 within the proximal C-terminal region of GIRK2 plays a crucial role in the Na ϩ -dependent activation of GIRK1/GIRK2 heteromeric channels and GIRK2 homomeric channels. When D226 in GIRK2 was substituted for an asparagine, activation of GIRK2D226N homomeric channels and GIRK1/GIRK2D226N heteromeric channels by Na ϩ was lost, and instead P open decreased with increasing internal [Na ϩ ]. The wild type GIRK1 subunit was not able to support Na ϩ -dependent activation of the heteromeric channel. The reverse mutation, N217D, in the GIRK1 subunit was sufficient to recover Na ϩ -dependent activation of the GIRK1N217D/GIRK2D226N channel, and similarly this mutation in the 1211 chimera introduced Na ϩ -dependent activation to 1211N217D channels. The D226N mutation in GIRK2 did not inhibit the activation of the channel via the m2 receptor, suggesting that G protein gating of GIRKs occurs via a different mechanism to Na ϩ -dependent activation.
Substituting asparagine for aspartate 226 in GIRK2 could be disrupting Na ϩ -dependent activation of the homomeric and heteromeric channels in at least three different ways: 1) it could reduce the binding affinity of Na ϩ to the channel; 2) it could interfere with the mechanism by which Na ϩ binding, to either GIRK2 or an accessory protein, is coupled to channel opening; 3) it could disrupt the intrinsic gating of the channel in such a way as to prevent the channel from opening in response to any stimulus. We can rule out the third possibility, because the mutant and chimeric channels that were not activated by Na ϩ were still activated by both internal MgATP and external CCh. Also, the single channel properties of GIRK2D226N channels and GIRK1/GIRK2D226N channels in FIG. 5. Effects of intracellular Na ؉ on GIRK2E234S, GIRK2D226N, 1211N217D, and GIRK1N217D mutants. A, C, E, and F, inside-out patches from oocytes injected with the cRNAs indicated, in the presence of increasing concentrations of Na ϩ and 5 mM Mg ATP. B, D, F, and H, Na ϩ dose-response relationships for the channels indicated. Values for EC 50 and Hill coefficients are given in Table I. the cell-attached recording configuration did not differ substantially from the wild type homomeric and heteromeric channels. Thus, the intrinsic gating of these channels does not appear to have been dramatically altered.
We have no direct evidence that Na ϩ acts by binding to aspartate 226 or anywhere on the GIRK2 subunit. However, Na ϩ -dependent activation is observed in isolated membrane patches, indicating that other cytosolic proteins are not required for mediating its actions. Also, other negatively charged residues within the proximal C-terminal region of Kir subunits are involved in the binding of Mg 2ϩ and polyamines (16,17), indicating that this region of Kir subunits is accessible to internal cations. Aspartate residues have also been shown to mediate the allosteric actions of internal Na ϩ on other transmembrane proteins, for example the D 2 -dopamine receptor (18).
Aspartate 226 in GIRK2 and aspartate 223 in GIRK4 are only 7 residues downstream of the cysteine that appears to be crucial for G␤␥-mediated activation of GIRK4 and GIRK1/ GIRK4 channels (13). A peptide from this region of GIRK4 has also been shown to compete with binding of the G␤␥ to the native channel (13). The binding site for PIP 2 also appears to be located within this proximal C-terminal region. PIP 2 regulates the activity of several of the inward rectifiers, including ROMK1 (Kir1.1) and the ATP-sensitive K ϩ channel (SUR/ Kir6.2), as well as GIRK channels. Neutralization of the arginines at positions 176 and 177 in Kir6.2, and position 188 in Kir1.1, reduce PIP 2 sensitivity (19,7,20). Thus, the proximal C-terminal region immediately following the second transmembrane segment appears to be an important domain for mediating the effects of internal ligands.
Huang et al. (7) suggested that G␤␥ activates GIRK channels by stabilizing interactions between PIP 2 and the channel. They showed that in the presence of G␤␥, the rate of channel activation by PIP 2 was increased, and the rate of dissociation of PIP 2 was decreased. They also showed that activation of GIRK channels by PIP 2 , in the presence of 20 mM Na ϩ , precluded further activation by G␤␥. One possible mechanism for Na ϩdependent activation of GIRK channels is that Na ϩ interacts with aspartate 226 (or aspartate 223 in GIRK4) to promote the binding of the anionic PIP 2 to a nearby region of the C terminus. However, the results of Sui et al. (6) suggest that PIP 2 and Na ϩ activate GIRK channels via different mechanisms. They reported that 1 M PIP 2 increased the mean open time of GIRK1/GIRK4 channels, whereas subsequent application of Na ϩ specifically increased the opening frequency. In our experiments, a 5-min application of 50 M PIP 2 produced a dramatic increase in the apparent open frequency of GIRK1/GIRK2 channels, but it did not preclude further activation of the channels by Na ϩ . Thus, we propose that Na ϩ increases the P open of PIP 2bound channels. It remains to be tested whether or not it also increases the affinity of PIP 2 binding to GIRK channels.
For those mutant GIRK channels lacking aspartate 226, the decrease in the open probability as a result of increasing internal [Na ϩ ] might be caused by a reduction in PIP 2 binding. Shyng and Nichols (20) compared the ability of different phospholipids to activate ATP-sensitive K ϩ channels (SUR1/Kir6.2) and concluded that a negatively charged head and a lipid tail are necessary to stimulate these channels. Screening of negative charges by application of polycations, such as Ca 2ϩ or polylysine, inhibited PIP 2 -stimulated K ATP channel activity. In our experiments, increasing internal [NMDG] appeared to be as effective as increasing [Na ϩ ] in reducing the open probability of these mutant GIRK channels, and it also inhibited GIRK1/ GIRK2 currents. The Na ϩ and NMDG were added to the bath solution without compensation for changes in ionic strength. The resultant increase in the ionic strength of the internal solution might have reduced long range electrostatic interactions between PIP 2 in the membrane and the GIRK channels.
Neither the D226N nor E234S mutation in GIRK2 appeared to change the rectification of the whole cell currents. The E234S mutation did alter the channels gating characteristics, causing it to flicker rapidly between an open and shut state. Mutating the equivalent glutamate in IRK1 channels has been shown to produce a similar flickering behavior, but it also decreases the sensitivity of IRK1 to the internal blockers Mg 2ϩ and polyamines (16,17), thus reducing rectification. Our results, together with the fact that GIRK2 does not possess a negatively charged residue within its TM2 region, suggest that the residues involved in determining the rectification of GIRK2 differ from those shown to be important in the rectification of other inward rectifier channels.
GIRK channels are located in postsynaptic neurons in several areas of the central nervous system (21,11), where the influx of Na ϩ occurs through both ligand-gated receptors and voltage-gated Na ϩ channels (22). During periods of rapid firing, large increases in subplasmalemmal [Na ϩ ] are likely to occur in the vicinity of these channels. Yu et al. (23) recently reported that NMDA receptors are activated by [Na ϩ ] i over a similar range of concentrations as required for the activation of GIRK channels. They showed in hippocampal neurons that Na ϩ entry through either NMDA receptors or voltage-gated Na ϩ channels could activate neighboring channels that were isolated within a cell-attached membrane patch. Thus, it seems likely that GIRK channels will be activated by Na ϩ influx into dendrites under physiological conditions, and this may provide a negative feedback mechanism for suppressing neuronal firing following periods of high activity.