Molecular determinants for activation of G-protein-coupled inward rectifier K+ (GIRK) channels by extracellular acidosis.

Synaptic cleft acidification occurs following vesicle release. Such a pH change may affect synaptic transmissions in which G-protein-coupled inward rectifier K(+) (GIRK) channels play a role. To elucidate the effect of extracellular pH (pH(o)) on GIRK channels, we performed experiments on heteromeric GIRK1/GIRK4 channels expressed in Xenopus oocytes. A decrease in pH(o) to 6.2 augmented GIRK1/GIRK4 currents by approximately 30%. The channel activation was reversible and dependent on pH(o) levels. This effect was produced by selective augmentation of single channel conductance without change in the open-state probability. To determine which subunit was involved, we took advantage of homomeric expression of GIRK1 and GIRK4 by introducing a single mutation. We found that homomeric GIRK1-F137S and GIRK4-S143T channels were activated at pH(o) 6.2 by approximately 20 and approximately 70%, respectively. Such activation was eliminated when a histidine residue in the M1-H5 linker was mutated to a non-titratable glutamine, i.e. H116Q in GIRK1 and H120Q in GIRK4. Both of these histidines were required for pH sensing of the heteromeric channels, because the mutation of one of them diminished but not abolished the pH(o) sensitivity. The pH(o) sensitivity of the heteromeric channels was completely lost when both were mutated. Thus, these results suggest that the GIRK-mediated synaptic transmission is determined by both neurotransmitter and protons with the transmitter accounting for only 70% of the effect on postsynaptic cell and protons released together with the transmitter contributing to the other 30%.

The G-protein-coupled inward rectifier K ϩ (GIRK) 1 channels are important players in cellular communications in several excitable tissues (1)(2)(3). The GIRK channels are activated by ␤␥-subunits of G-proteins, which are dissociated from the ␣␤␥trimer as a result of receptor binding to neurotransmitters or hormones (3). Four members of GIRK channels have been identified in mammals with GIRK1/GIRK4 expressed abundantly in the heart and brain (4). GIRK channels are modulated by several intracellular signal molecules such as Na ϩ , ATP, and phospholipids (5)(6)(7)(8)(9)(10)(11)(12). Extra-cellular molecules including hormones, neurotransmitters, and integrins directly or indirectly modulate GIRK channel activity through signaling transduction pathways (13)(14)(15)(16)(17)(18). GIRK channels are also the major targets of ethanol, anesthetics, and opioids (19 -23). Another potentially important modulator of the GIRK channels is hydrogen ion. Increasing evidence indicates that H ϩ can act as a messenger modulating multiple cellular functions (24). In the central nervous system, protons have been shown to modulate synaptic transmission, neuronal plasticity, and membrane excitability (25). It is known that the pH level in synaptic vesicles is ϳ1.5 pH units lower than in cytosol (26). These protons are released from synaptic vesicles together with neurotransmitters during synaptic transmission, leading to extracellular acidification in the synaptic cleft (27). If the GIRK channels are sensitive to extracellular pH (pH o ), such extracellular acidification can have a major impact on synaptic transmission. Indeed, certain unidentified GIRK channels in the brainstem that play a part in the generation and control of central respiratory activity have been suggested being sensitive to hypercapnic acidosis (28 -31).
It is possible that the GIRK channels are pH-sensitive, because several inward rectifier K ϩ channels are directly gated by intracellular and/or extracellular protons. To test the hypothesis that the GIRK channels are modulated by pH o , we performed experiments on the heteromeric GIRK1/GIRK4 channels. Our results indicate that these channels are augmented by extracellular acidification through an increase in single channel conductance, and such activation relies on a histidine residue in the extracellular loop.

MATERIALS AND METHODS
Experiments were performed as we described previously (32)(33)(34). Oocytes from Xenopus laevis were used in this study. Frogs were anesthetized by bathing them in 0.3% 3-aminobenzoic acid ethyl ester. A few lobes of ovaries were removed after a small abdominal incision (ϳ5 mm). The surgical incision then was closed, and the frogs were allowed to recover from the anesthesia. Xenopus oocytes were treated with 2 mg/ml collagenase (Type IA, Sigma) in the OR2 solution (in mM) as follows: 82 NaCl, 2 KCl, 1 MgCl 2 , and 5 HEPES, pH 7.4, for 60 min at room temperature. After three washes (10 min each) of the oocytes with the OR2 solution, cDNAs (25-50 ng in 50 nl water) were injected into the oocytes. The oocytes were then incubated at 18°C in the ND-96 solution containing (in mM) 96 NaCl, 2 KCl, 1 MgCl 2 , 1.8 CaCl 2 , 5 HEPES, and 2.5 sodium pyruvate with 100 mg/liter Geneticin added, pH 7.4.
Rat GIRK1 (Kir3.1) cDNA (GenBank TM accession number U01071) and rat GIRK4 (Kir3.4) cDNA (GenBank TM accession number X83584) are gifts from Dr. Norman Davidson (California Institute of Technology). These cDNAs were subcloned into a eukaryotic expression vector pcDNA3.1 (Invitrogen) and used for Xenopus oocyte expression without in vitro cRNA synthesis. PCR was used to generate GIRK1/GIRK4 dimer. Two PCR fragments encoding the entire length of GIRK1 and GIRK4 were joined with an XbaI restriction site created using primers with the GIRK1 at the 5Ј end and then cloned to the pcDNA3.1. Five extra amino acids (RCQQQ) were created between the C terminus of GIRK1 and the N terminus of GIRK4. We did not find any detectable effect of these additional residues on channel expression, current pro-file, and pH sensitivity. Site-specific mutations were made using a site-directed mutagenesis kit (Stratagene, La Jolla, CA). Correct constructions and mutations were confirmed with DNA sequencing.
Whole-cell currents were studied on the oocytes 2-4 days after injection. Two-electrode voltage clamp was performed using an amplifier (Geneclamp 500, Axon Instruments Inc., Foster City, CA) at room temperature (ϳ24°C). The extracellular solution contained (in mM): 90 KCl, 3 MgCl 2 , and 5 HEPES, pH 7.4. Extracellular acidification was done by perfusing the oocytes with the extracellular solution containing PIPES buffer, pH 5.2 and 6.2. HEPES buffer was used for extracellular alkalization, pH 8.4. These buffers were chosen because their buffering ranges are suitable for these pH levels and none of them is membranepermeable as shown in our previous studies (33,34). Several control experiments were done in which inward rectifying currents of oocytes receiving an injection of the expression vector or Kir2.1 channel were not affected by these PIPES and HEPES buffers. Currents were also measured at various pH points in oocytes receiving an injection of the expression vector or the pH-insensitive Kir2.1 as controls. Intracellular acidification was produced using 90 mM KHCO 3 to replace all KCl (90 mM) in the extracellular solution, which acidified the cytosol to pH 6.6 (see "Results"). This solution was titrated to pH 7.4 immediately before use (33,34,50).
pH o and intracellular pH (pH i ) were measured using ion-selective microelectrodes as we detailed previously (34,50). Two single-barreled microelectrodes were employed. One of them (ion-selective) was exposed to hexamethyldisilazan vapor (Fluka Chemie AG, Buchs, Switzerland) for 30 min and then baked at 125°C for 8 h. The tip of the ion-selective microelectrode was filled with H ϩ liquid exchanger (Hydrogen Ion Ionophore l-Mixture A, Fluka Chemie AG), and the remainder of the microelectrode was backfilled with phosphate buffer, pH 7.0, for both pH i and pH o measurements. This ionophore is greatly selective for H ϩ (e.g. H ϩ :K ϩ , Na ϩ or Ca 2ϩ Ͼ 1,000,000:1). The other microelectrode was filled with 3 M KCl. Electrodes were used only if they had high frequency response (90% response time Յ 5 s) and showed an excellent sensitivity (a voltage change Ͼ 55 mV when pH changed from 6.0 to 7.0). A high input-resistance amplifier (Duo773, World Precision Instruments, Inc., Sarasota, FL) was used for pH measurements. The ion-selective electrode was connected to the high input-resistance channel (10 15 ohms), and the KCl electrode was connected to the other (10 12 ohm). Voltage was removed by subtracting records between these two channels. Serial calibrations of ion-selective microelectrodes were made with potassium phosphate buffer at pH 6.0, 7.0, and 8.0 (Fisher Scientific).
The time profile of current response to pH o was studied using a perfusion chamber with a total volume of 1 ml (RC-3Z, Warner Instruments, New Haven, CT). The current amplitude was plotted against time together with pH o measured using a H ϩ -selective electrode (34,45). The pH electrode was positioned in the extracellular solution near the cell. When the extracellular solution was switched to a low pH perfusate, the pH o started to change in ϳ1 min and reached a plateau level in ϳ3 min.
Single channel currents were studied in outside-out patch configuration using solutions containing equal concentrations of K ϩ applied to the bath and recording pipettes. The bath solution contained (in mM): 140 KCl, 10 Na 2 H 2 P 2 O 7 , 5 NaF, 0.1 Na 3 VO 3 , 0.2 ATP, 0.2 GTP, 10 EGTA, and 10 HEPES, pH 7.4. The pipette was filled with the same solution (32). The open-state probability (P o ) was calculated as we described previously (32). The single channel conductance was measured in negative membrane potential using a ramp command potential (from Ϫ100 to 100 mV). To change pH o , the patch membrane was perfused with solutions in different pH levels with no dead space.
Data are presented as the means Ϯ S.E. ANOVA or Student's t test was used. The differences of pH effects before versus during exposures were considered to be statistically significant if p Յ 0.05.

RESULTS
Base-line Activity of Heteromeric GIRK1/GIRK4 -Wholecell currents were studied in the two-electrode voltage-clamp mode using an extracellular solution containing 90 mM K ϩ (see "Materials and Methods"). Depolarizing and hyperpolarizing command pulses were given to the cell in a range from Ϫ140 mV (or Ϫ160 mV) to 100 mV with a 20-mV increment at a holding potential of 0 mV. Under this condition, inward rectifying currents were observed 2-4 days after coinjection of GIRK1 and GIRK4 cDNAs. The GIRK1/GIRK4 currents showed a clear inward rectification (Fig. 1A) and were sensitive to micromolar concentrations of Ba 2ϩ (data not shown). These currents should be recorded from the heteromeric GIRK1/ GIRK4 rather than the homomeric channels, because similar inward rectifying currents were seen following an injection of a tandem dimer of GIRK1/GIRK4 cDNA (data not shown).
Response to Extracellular pH-Exposure of the oocytes to a perfusate of pH 6.2 augmented the heteromeric GIRK/GIRK4 currents (Fig. 1A). Evident increase in the GIRK/GIRK4 currents was seen within 1 min into the exposure and reached the maximal level in 3-4 min (Fig. 1B). Washout with the pH 7.4 perfusate led to a clear recovery (Fig. 1A). When the current amplitude was plotted together with pH o against time, the current response curve was almost identical to the pH o profile with only an ϳ0.5-min delay (Fig. 1B). This effect was not produced by a change in pH i , because the perfusate containing membrane-impermeable PIPES buffer does not change pH i as shown in our previous studies (33,34). To strengthen this FIG. 1. Activation of heteromeric GIRK1/GIRK4 channels by extracellular acidification. The GIRK1 and GIRK4 cDNAs were cloned in a eukaryotic expression vector pcDNA3.1 and expressed in Xenopus oocytes. Whole-cell currents were studied in two-electrode voltage clamp using an extracellular solution containing 90 mM K ϩ . A, inward rectifying currents were recorded from an oocyte 3 days after a coinjection of the GIRK1 and GIRK4 cDNAs. Exposure to a perfusate of pH 6.2 for 10 min increased the currents by 38%. The current activation was reversible as washout at pH 7.4 led to almost complete recovery. B, the current amplitude together with pH o was plotted against time. The pH o (gray line) started to change in ϳ0.5 min with acid exposure and reached a plateau in ϳ3 min. The current response curve (black circle) was almost the same as the pH o profile, although it showed ϳ0.5-min delay. Note that the data are obtained from A, and the currents are measured at Ϫ120 mV. C, concentration dependence of the heteromeric GIRK1/GIRK4 channels on pH o levels. The heteromeric GIRK1/GIRK4 channels were studied at different pH o levels as indicated in the x axis. The currents were inhibited at alkaline pH and stimulated at acidic pH levels with the maximal activation at pH o 6.2. I, current. Data are presented as the means Ϯ S.E. (n ϭ 4 -11).
argument, we selectively reduced pH i to 6.6 without changing pH o using 90 mM bicarbonate as demonstrated previously (34,50). Such intracellular acidification for 10 min did not produce any significant increase in the GIRK1/GIRK4 currents (n ϭ 6, data not shown). Thus, these results indicate that extracellular acidification augments the GIRK1/GIRK4 currents. Fig. 1C shows concentration dependence of the GIRK/GIRK4 currents on the pH o levels. The currents were moderately stim-ulated at pH o 6.8 and inhibited at pH o 8.4. The maximal activation occurred at pH o 6.2, whereas a further drop in pH o to 5.2 did not have any additional effect, suggesting that the most sensitive pH range of the GIRK/GIRK4 channels is between pH 7.0 and 7.4. At the maximal activation with pH o 6.2, the amplitude of heteromeric GIRK/GIRK4 currents was enhanced by 33.4 Ϯ 3.6% (n ϭ 11). Similar channel activation was also observed in the tandem-dimeric GIRK1/GIRK4 channel whose FIG. 2. Effects of acidic pH o on single-channel properties and voltage dependence of heteromeric GIRK1/GIRK4 channels. A, single-channel currents were recorded from an outside-out patch using symmetric concentrations of K ϩ (140 mM) applied to both sides of the patch membrane. Two active channels are seen at pH o 7.4. The single channel conductance indicated by straight lines is 27 picoSiemens for both of these channels. B, at pH o 6.2, their single channel conductance becomes 37 picoSiemens. C-F, the channel P o was studied on a stretch of record of 20 s with a holding potential of Ϫ80 mV using the same solutions as A. The P o at pH o 6.2 (D) did not show significant difference from that at pH o 7.4 (C). Note that E and F are obtained from areas indicated by an arrow in C and D, respectively. Label C, closure; label 1, first opening; and label 2, second opening. G-I, the voltage dependence was studied in whole-cell recording. G, at base line (pH o 7.4), the heteromeric GIRK1/GIRK4 currents showed almost linear conductance at negative membrane potentials (V m ). H, the currents affected by low pH were isolated by subtracting base-line currents from those recorded during pH o 6.2. The isolated currents showed conductance similar to G at negative V m . I, the I-V relationship of the isolated currents (triangle) is similar to that of the base-line currents (circle). amplitude increased by 33.9 Ϯ 3.6% (n ϭ 6).
To understand how single channel properties and voltage dependence underlie the change in whole-cell currents, we performed single channel recordings in the outside-out patch configuration. The heteromeric GIRK/GIRK4 channels showed low base-line activity at pH 7.4 in the absence of exogenous G-protein. These currents had single channel conductance of 28.0 Ϯ 0.4 picoSiemens (n ϭ 6) ( Fig. 2A). Exposure of external patch membranes to a perfusate of pH 6.2 enhanced the single channel conductance by 29.8 Ϯ 2.9% (n ϭ 6) (Fig. 2B). Unlike the single channel conductance, the P o did not show significant change with pH o 6.2 (p Ͼ 0.05, n ϭ 9) (Fig. 2, C-F). In wholecell recordings, the increase in GIRK1/GIRK4 currents at acidic pH o did not show any voltage dependence in a voltage range from Ϫ160 to 0 mV (Fig. 2, G-I). Thus, the increase in whole-cell currents at acidic pH o is likely to be produced by augmentation of the single channel conductance.
Effect of pH o on Homomeric Channels-To elucidate which subunit is responsible for proton sensing, homomeric expression of these GIRK channels was carried out. It is known that the homomeric GIRK1 and GIRK4 produce only small currents with or without G␤␥-subunits, whereas the small currents are probably derived from heteromeric channels formed by Xenopus endogenous GIRK5 with the exogenous GIRK subunits (35,36). However, previous studies have demonstrated that the mutation of a single amino acid residue in the pore-forming sequence (P) allows both GIRK1 and GIRK4 to be expressed homomerically (35). This homomeric expression technique has been conducive to elucidating the modulation of GIRK channels by several modulators (11,35,37,38). Therefore, we constructed the GIRK1 with Phe-137 mutated to serine (GIRK1-F137S) and GIRK4 with Ser-143 mutated to threonine (GIRK4-S143T) as shown previously by Vivaudou et al. (35). Consistent with their results, functional expressions of both these homomeric GIRK1 and GIRK4 channels were seen with the mutations (Fig. 3, A and B). When the channels were challenged with extracellular acidosis (pH o 6.2), the amplitude of the homomeric GIRK1 and GIRK4 currents increased by 21.0 Ϯ 2.9% (n ϭ 9) and 66.7 Ϯ 6.2% (n ϭ 6), respectively (Fig. 3C). These data indicated that both GIRK1 and GIRK4 possess protonsensing mechanism with the one in GIRK4 more prominent.
Identification of Potential Proton Sensors-If these GIRK channels are inherently sensitive to protons, there should be specific protein domain or amino acid residue that is accessible to extracellular protons and responsible for the pH o -induced channel activation. To test this hypothesis, we performed sitedirected mutagenesis on potentially titratable histidine residues, an amino acid with its side-chain pK of 6.04 most close to pH o levels for the channel activation. Therefore, we examined amino acid sequences from the first membrane-spanning helix (M1), the P-loop, to the second membrane-spanning sequence FIG . 4. Identification of proton sensors in the homomeric GIRK1 and GIRK4 channels. A, alignment of amino acid sequence of several inward rectifier K ϩ (Kir) channels in the extracellular domains and P-loop. A histidine residue is found in this region of GIRK1 and GIRK4. This histidine is conserved in all GIRK channels but absent in other Kir channels. B, mutation of the His-116 eliminated the pH o sensitivity of the homomeric GIRK1 channels. C, similarly, the His-120 mutation made the GIRK4 channels pH o -insensitive.
(M2) helix and found a histidine in the M1-P linker (Fig. 4A). The sequence alignment using the BLAST 2 sequences shows that this histidine is conserved in all GIRK channels but not seen in any other Kir channels (Fig. 4A). Thereafter, we sitespecifically mutated this histidine to a neutral polar glutamine. We found that the histidine mutation totally abolished the pH o sensitivity in both homomeric GIRK1-F137S and GIRK4-S143T channels (Figs. 4, B and C, and 5A), indicating that the histidine residue is the proton sensor in these homomeric channels.
Because the homomeric GIRK4-S143T is more sensitive to pH o than the GIRK1-F137S, it is possible that the His-120 in GIRK4 plays a more important role in the pH o sensitivity of the heteromeric GIRK1/GIRK4 channels. To test this hypothesis, we studied the heteromeric channels with the mutation of the histidine residue in GIRK1 and/or GIRK4. We found that heteromeric channels carrying any one of the histidines (GIRK1-H116Q/ GIRK4, GIRK1/GIRK4-H120Q) remained pH o -sensitive, although their pH o sensitivity was significantly lower than the wild-type channels. Simultaneous mutations of the histidine in both GIRK1 and GIRK4 (GIRK1H116Q-GIRK4H120Q) completely eliminated the pH o sensitivity of the heteromeric channels (Figs. 5B and 6). Thus, these results suggest that all four histidine residues in both GIRK1 and GIRK4 subunits are involved in proton sensing in their heteromeric channels. DISCUSSION This is the first demonstration of the pH o sensitivity in the GIRK channels. We have found that the heteromeric GIRK1/ GIRK4 are stimulated by extracellular acidification and inhibited by alkalization. A decrease in pH o to 6.2 enhances the GIRK1/GIRK4 currents by ϳ30%, which has been observed in both GIRK1/GIRK4 coexpression and the tandem-dimeric channel. The increase in whole-cell currents is attributed to augmentation of the single-channel conductance without changes in P o and voltage dependence. A histidine residue in the extracellular domain is crucial, a mutation of which eliminates the pH sensitivity of the homomeric channels, whereas the pH sensitivity in the heteromeric GIRK1/GIRK4 channels requires this histidine in both subunits.
The GIRK1/GIRK4 channels are activated by extracellular but not intracellular acidification for the following reasons. 1) In our previous studies, we have previously shown that a decrease in extracellular pH does not cause intracellular acidification using the same PIPES buffer (32,33). 2) Our data indicate that augmentation of these currents is associated with a change in pH o . 3) A decrease in pH i does not enhance the currents. 4) Similar current augmentation was seen in excised patched. Thus, the increase in current amplitude seen in this study should be produced by pH o indeed. Biophysical mechanisms underlying the change of whole-cell currents were examined in this study. Our data indicate that the increase in whole-cell currents is produced by selective augmentation of single-channel conductance without affecting the open-state probability and the voltage dependence. The channel activation is reversible and dependent on pH levels. The linear working range of the GIRK1/GIRK4 heteromeric channels is between pH 7.0 and 7.4, suggesting that the heteromeric GIRK1/GIRK4 channels can detect pH changes in most physiologic and pathophysiologic conditions. Several members of K ϩ channels in the Kir family are known to be pH-sensitive, such as Kir1. 1  which make them similar only to the Kir2.3 (32,40), and 2) stimulation by acidic pH, which renders them more like Kir6 channels (34). In contrast, most of the Kir channels are inhibited by acidic pH with the proton sensors mostly located on the cytosolic side of the plasma membranes (39 -45).
The proton detection in the GIRK1/GIRK4 channels depends on a histidine residue in the extracellular protein domain. The pH o sensitivity of GIRK1 and GIRK4 homomeric channels is lost when this histidine residue is mutated to a non-titratable amino acid. Thus, the pH sensing in these GIRK channels is similar to several other Kir channels as shown previously (46 -48). This histidine residue is conserved among all GIRK channels, suggesting that other GIRK subunits (i.e. GIRK2, GIRK3) are probably pH-sensitive as well. Because this histidine does not exist in other Kir channels, these results may explain why most of them, with the exception of Kir2.3 that is inhibited by both intracellular and extracellular acidifications and is also regulated by Gproteins (see Refs. 32,40,and 49), are insensitive to pH o .
One interesting finding from our current studies is the graded pH o sensitivity of the heteromeric GIRK1/GIRK4 channels with the histidine mutation. We have found that mutation of the histidine in either GIRK1 or GIRK4 leads to a reduction but not elimination of the pH o sensitivity, whereas the heteromeric channel loses the pH o sensitivity only when both histidines in GIRK1 and GIRK4 are simultaneously mutated. Based on this observation, it is possible that the GIRK channel modulation by extracellular protons requires protonation of this histidine residue in all four subunits and perhaps movements of these extracellular protein domains as well.
The pH o sensitivity of these GIRK channels has a profound impact on synaptic transmission. In the presynaptic terminals, neurotransmitter uptake relies on ATP-dependent proton pumps whose activity results in acidification inside synaptic vesicles up to 1.5 pH units lower than cytosolic pH (26). Following each action potential at presynaptic terminals, protons are released with neurotransmitters into the synaptic cleft. This has been shown to lead to significant extracellular acidification in the synaptic cleft (27). As the GIRK channels are major targets of the neurotransmitters on the postsynaptic membranes, the pH o sensitivity provides these channels with an important modulatory mechanism by which they are first activated by neurotransmitters, and subsequently, channel activity is further enhanced by protons released from presynaptic terminals. Such a modulatory mechanism appears to enhance greatly the synaptic efficiency, because according to our data, the neurotransmitter accounts for only 70% of the synaptic transmission and protons released together with the transmitter contribute to the other 30%.
How does the pH o sensitivity affect specificity of synaptic transmission? Our results show that extracellular protons do not seem to affect GIRK channel gating, because they act on single channel conductance and augment the GIRK channels by 33% at most. Therefore, the synaptic cleft acidification should have very little effect on the GIRK channels before they are opened by specific neurotransmitters. Such a modulation instead of channel gating may prevent these channels from being activated by protons released with irrelevant neurotransmitters. This is remarkable, because protons released from presynaptic terminals may contribute significantly to the GIRK-mediated synaptic transmission without compromising synaptic specificity. Thus, the demonstration of the modulation of GIRK1/GIRK4 channels by extracellular acidification contributes valuable information to the understanding of these GIRK channels and their function in synaptic transmission.