Biophysical and Molecular Mechanisms Underlying the Modulation of Heteromeric Kir4.1–Kir5.1 Channels by Co₂ and Ph

Inward rectifier K⁺ channels (Kir channels) are the primary regulators of membrane excitability and are themselves also regulated by several intra- and extracellular factors (Nichols and Lopatin 1997). One of these factors is a hydrogen ion that has been known to affect specific members in the Kir channel family (Coulter et al. 1995; Tsai et al. 1995; Doi et al. 1996; Shuck et al. 1997; Pearson et al. 1999; Zhu et al. 1999; Xu et al. 2000). The modulation of Kir channels by protons is significant because changes in intra- and extracellular pH are seen in a number of physiological and pathophysiological conditions, and because such a modulation may allow cells to produce appropriate responses to these conditions. Indeed, Kir channels with pH sensitivity have been shown to be involved in the maintenance of pH and K⁺ homeostasis by renal epithelial cells (Schlatter et al. 1994; Zhou and Wingo 1994; Tsai et al. 1995; Doi et al. 1996) and CO₂ chemosensitivity of brainstem neurons (Pineda and Aghajanian 1997). In brainstem neurons, the inhibition of inward rectifying K⁺ channels causes depolarization and an increase in membrane excitability, which may constitute an initial event in CO₂ chemoreception (Mitchell and Berger 1975). Thus, detailed studies of these Kir channels may give rise to an understanding of a group of CO₂-sensitive molecules that are located on plasma membranes of nerve cells, detect neuronal ambient PCO₂ levels, and couple the PCO₂ fluctuation to a corresponding change in membrane excitability.

Several cloned Kir channels respond to CO₂ and pH similarly to the K⁺ currents seen in brainstem neurons. Kir1.1 and Kir1.2 are inhibited by a decrease in intracellular pH (Tsai et al. 1995; Doi et al. 1996; Fakler et al. 1996b; Choe et al. 1997; McNicholas et al. 1998) and hypercapnic acidosis (Chanchevalap et al. 2000; Zhu et al. 2000). A lysine residue near to the first membrane-spanning domain or TM1 (Lys80 in Kir1.1 and Lys61 in Kir1.2) is critical for the pH sensitivity (Fakler et al. 1996b; McNicholas et al. 1998). Hypercapnia and acidosis also inhibit Kir2.3 (Coulter et al. 1995; Zhu et al. 1999, Zhu et al. 2000). A threonine residue also located at the immediate vicinity of the TM1 domain (Thr53) plays an important role in channel sensitivity to intracellular protons (Qu et al. 1999). Although these Kir channels are sensitive to CO₂ and pH, they may not be the optimum candidates of the CO₂-sensing molecules for the simple reason that none of them has a high level of expression in the brainstem. Thus, it is uncertain whether these Kir channels are the CO₂/pH sensors in the brainstem neurons.

In contrast, Kir4.1 is expressed predominantly in the brainstem (Bredt et al. 1995). Although its pH sensitivity remains low (apparent pK or pKa ∼ 6.0), coexpression of Kir4.1 with the brain Kir5.1 greatly enhances CO₂ and pH sensitivities of the heteromeric Kir4.1–Kir5.1 channels (Xu et al. 2000). Interestingly, the Kir5.1 that is also expressed in the brainstem (our unpublished observations) does not form functional channels in its own homomultimers. Its only known function is to produce heteromeric channels, particularly with Kir4.1 (Pessia et al. 1996). The high CO₂/pH sensitivities and brainstem-specific expression indicate that these channels are preferential candidates as CO₂-sensing molecules in brainstem neurons. To understand their potential role in CO₂ chemoreception, it is necessary to understand the biophysical and molecular mechanisms underlying the modulation of Kir4.1–Kir5.1 channels by CO₂ and pH.

Several mechanisms can underlie the modulation of the heteromeric Kir4.1–Kir5.1 by CO₂. Our previous studies have shown that Kir channel inhibition during hypercapnia is mediated by protons rather than molecular CO₂ (Xu et al. 2000; Zhu et al. 2000). The proton sensing, however, can be an inherent property of the channel proteins, or it can be carried out by another intermediate molecule; the modulation of Kir4.1–Kir5.1 currents can be a result of a depression of single-channel conductance or a reduction in channel open-state probability (P_open); and the proton-sensing mechanisms can be on the Kir4.1 or Kir5.1 subunit, since the lysine and threonine residues found in Kir1.1 and Kir2.3 are also seen in Kir4.1 (Lys67) and Kir5.1 (Thr68), respectively. To examine these possibilities, we performed experiments in which cell-free excised patch and molecular genetical techniques were used to answer these questions.

Oocytes from female frogs (Xenopus laevis) were used in the present studies. 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 (∼5 mm). The surgical incision was closed and the frogs were allowed to recover from the anesthesia. Xenopus oocytes were treated with 2 mg/ml of collagenase (Type I; Sigma-Aldrich) in OR2 solution (mM): 82 NaCl, 2 KCl, 1 MgCl₂, and 5 HEPES, pH 7.4, for 90 min at room temperature. After three washes (10-min each) of the oocytes with OR2 solution, cDNAs (25–50 ng in 50 nl double-distilled water) were injected into the oocytes. The oocytes were then incubated at 18°C in ND-96 solution containing (mM): 96 NaCl, 2 KCl, 1 MgCl₂, 1.8 CaCl₂, 5 HEPES, and 2.5 sodium pyruvate with 100 mg/liter geneticin added (pH 7.4).

Brain Kir4.1 (BIRK10) and brain Kir5.1 (BIRK9) cDNAs were generously provided by Dr. John Adelman (Organ Health Science University) (Bond et al. 1994). A vector for eukaryotic expression (pcDNA3.1; Invitrogen) was used to express Kir4.1 and coexpress Kir4.1 and Kir5.1 channels in the Xenopus oocytes. The Kir4.1 cDNA was removed from the pBF vector at EcoRI restriction sites on each end of the cDNA, and then were subsequently subcloned into corresponding EcoRI sites in the pCDNA3.1. For coexpression of Kir4.1 and Kir5.1, a tandem dimer of these two cDNAs was constructed using the overlapping extension technique (Qu et al. 1999), in which the full-length Kir4.1 and Kir5.1 sequences were obtained using Pfu DNA polymerase (Stratagene) chain reaction. The PCR products were joined to each other at the 3′ end of Kir4.1 and 5′ end of Kir5.1, and then cloned in the pcDNA3.1, as detailed by Pessia et al. 1996. Site-specific mutations were produced using a site-directed mutagenesis kit (Stratagene) using the tandem dimer of Kir4.1–Kir5.1 as a template. The orientation and site-specific mutations were confirmed with DNA sequencing. Expression of these channels was examined with cDNA injections.

Whole-cell currents were studied on the oocytes 2–4 d after injection. Two-electrode voltage clamp was performed using an amplifier (Geneclamp 500; Axon Instruments Inc.) at room temperature (23–26°C). The extracellular solution contained (mM): 90 KCl, 3 MgCl₂, and 5 HEPES, pH 7.4. Cells were impaled using electrodes filled with 3 M KCl. The potential leakage of KCl from the recording electrodes was not corrected because of the large volume of oocytes. One of the electrodes (1.0–2.0 MΩ) served as voltage recording, which was connected to an HS-2 × 1L headstage (input resistance = 10¹¹ Ω), and the other electrode (0.3–0.6 MΩ) was used for current recording connected to an HS-2 × 10MG headstage (maximum current = 130 μA). Oocytes were accepted for further experiments only if their leak currents, measured as a difference before and after leak subtractions, were <10% of the peak currents. The leak subtraction was applied to oocytes if their leak currents were 5–10%, which was done by subtracting a sum of currents produced by five small depolarizing prepulses in 1/5 of command potentials. Current records were low-pass filtered (Bessel, four-pole filter, 3 dB at 5 kHz), digitized at 5 kHz (12-bit resolution), and stored on computer disk for later analysis (PCLAMP 6; Axon Instruments, Inc.) (Yang and Jiang 1999; Zhu et al. 1999). Junction potentials between bath and pipette solutions were appropriately nulled.

Patch-clamp experiments were performed at room temperature (∼25°C) as described previously (Yang and Jiang 1999; Zhu et al. 1999). In brief, fire-polished patch pipettes (2–4 MΩ) were made from 1.2-mm borosilicate capillary glass (Sutter Instrument Co.). Single channel currents were recorded from inside-out, outside-out, and cell-attached patches (Hamill et al. 1981). Giant inside-out patches were also employed to study macroscopic currents in a cell-free condition using recording pipettes of 0.5∼1.0 MΩ. Current records were low-pass filtered (2,000 Hz, Bessel, four-pole filter, −3 dB), digitized (10 kHz, 12-bit resolution), and stored on computer disk for later analysis (PCLAMP 6; Axon Instruments, Inc.). Junction potentials between bath and pipette solutions were appropriately nulled before seal formation.

For single channel recordings, the oocyte vitelline membranes were mechanically removed after being exposed to hypertonic solution (400 mOsm) for 15 min. The stripped oocytes were placed in a petri dish containing regular bath solution (see below). Recordings were performed using solutions containing equal concentrations of K⁺ applied to the bath and recording pipettes. The bath solution contained (mM): 40 KCl, 75 potassium gluconate, 5 potassium fluoride, 0.1 sodium vanadate, and 10 potassium pyrophosphate, 1 EGTA, 0.2 ADP, 10 PIPES, 10 glucose, and 0.1 spermine (FVPP solution, pH 7.4). The pipette was filled with the same FVPP solution used in the bath or a solution containing (mM): 40 KCl, 110 potassium gluconate, 0.2 ADP, 1 EGTA, 10 HEPES, 10 glucose, 2 MgCl₂, pH 7.4. This bath solution was chosen after several others had been tested regarding channel rundown in excised patches. In a control experiment, we found that macroscopic currents recorded from giant inside-out patches were very well maintained, showing <10% reduction over a 17-min period of recordings in such a bath solution. This period is sufficient for all our single-channel recording protocols that were designed to be completed generally within 10 min.

CO₂ exposures were performed in a semi-closed recording chamber (BSC-HT; Medical System), in which oocytes were placed on a supporting nylon mesh, and the perfusion solution bathed both the top and bottom surface of the oocytes. The perfusate and the superfusion gas entered the chamber from two inlets at one end and flowed out at the other end. There was a 3 × 15-mm gap on the top cover of the chamber, which served as the gas outlet and the access to the oocytes for recording microelectrodes. The perfusate contained (mM): 90 KCl, 3 MgCl₂, and 5 HEPES, pH 7.4. At baseline, the chamber was ventilated with atmospheric air. Exposure of the oocytes to CO₂ was carried out by switching a perfusate that had been bubbled for at least 30 min with a gas mixture containing CO₂ (5, 10, or 15%) balanced with 21% O₂ and N₂, and superfused with the same gas. The high solubility of CO₂ resulted in a detectable change in intra- or extracellular acidification as fast as 10 s in these oocytes. Thus, in most experiments, only the superfusion air was switched to CO₂, in which similar results were produced.

A parallel perfusion system was used to administer agents to patches or cells at a rate of ∼1 ml/min with no dead space (Zhu et al. 1999). Low pH exposures were carried out using the same bath solutions that had been titrated to various pH levels, as required by experimental protocols. PIPES buffer was used because of its appropriate buffering range and its impermeability to plasma membranes. Indeed, we have done a control experiment in which a recombinant Kir2.3 with Kir2.1 sequence in its NH₂ terminal and Kir2.3 sequence in the rest part of the channel protein was expressed in the oocytes (Qu et al. 1999). Currents studied in inside-out giant patches did not show any detectable change to intracellular acidifications, although whole-cell currents of this Kir2.3 mutant were inhibited by extracellular acidification (Qu et al. 1999).

For single-channel analysis, data were further filtered (0–1,000 Hz) with a Gaussian filter. This filtering causes events shorter than 100 μs to be ignored. No correction was attempted for the missed events. Single channel conductance was measured as a slope conductance with at least two voltage points. P_open was calculated by first measuring the time, t_j, spent at current levels corresponding to j = 0,1,2,…n channels open (Yang and Jiang 1999; Zhu et al. 1999). The P_open was then obtained as P_open = (∑ⁿ_{j = 1} t_j j)/tn, where n is the number of channels active in the patch and t is the duration of recordings. P_open values were calculated from stretches of data having a total duration of 20–200 s. Open and closed times were measured from records in which only a single channel was active. The open- and closed-time distributions were fitted using the method of maximum likelihood (Yang and Jiang 1999; Zhu et al. 1999). The current amplitude was described using Gaussian distributions and the difference between two adjacent fitted peaks was taken as unitary current amplitude.

Data are presented as means ± SEM (n = number of patches). Differences in means were tested with the Student's t test and were accepted as significant if P ≤ 0.05.

Effects of CO₂ on K⁺ currents were studied using a high K⁺ (90 mM) bath solution. Membrane potential was held at 0 mV and stepped from −160 to 100 mV in 20-mV increments. Under such a condition, clear inward rectifying currents were observed in the oocytes receiving an injection of Kir4.1 or a tandem-linked Kir4.1 and Kir5.1 (Kir4.1–Kir5.1) cDNA 2–3 d earlier. Exposure of these oocytes to 15% CO₂ for 4–6 min produced an inhibition of Kir4.1 and Kir4.1–Kir5.1 currents. The degree of the inhibition was different between these two currents. At the maximal inhibition (measured at −120 mV), the Kir4.1–Kir5.1 currents were suppressed by 59.4 ± 3.7% (n = 9) and the Kir4.1 by only 23.3 ± 4.8% (n = 8). The effect of CO₂ on these currents was reversible and depended on CO₂ concentrations. Evident inhibition of these Kir currents was seen with 5% CO₂. Higher concentrations of CO₂ (10–15%) induced a much stronger inhibition of these currents. This inhibition was likely to be mediated by intracellular acidification, since lowering intra- but not extracellular pH to levels measured during 15% CO₂ produced a similar inhibition to 15% CO₂. These results are therefore consistent with our previous observations (Xu et al. 2000).

To understand channel biophysical properties underlying the effect of protons, single-channel activity was studied in inside-out patches, after the expression of Kir currents was identified in the two-electrode voltage-clamp mode. These patches were exposed to symmetric concentrations of K⁺ (150 mM) on both sides of plasma membranes; command potentials from −120 to 100 mV were applied to these patches. When inward-rectifying currents were seen, the slope conductance was first measured. Fig. 1 A shows a single-channel current recorded from a Kir4.1–Kir5.1-injected oocyte. The current showed a clear inward rectification with the slope conductance of 59 pS in the inside-out patch configuration (Fig. 1 B). In comparison with the homomeric Kir4.1, the single channel conductance of the Kir4.1–Kir5.1 was much larger, which averaged 59.2 ± 1.4 pS (n = 17) vs. 22.2 ± 0.5 pS (n = 22) in the Kir4.1 currents. Although the Kir4.1 showed a high baseline channel activity (P_open = 0.890 ± 0.021, n = 5), the heteromeric Kir4.1–Kir5.1 channels had a mean P_open of only 0.416 ± 0.095 (n = 6) at pH 7.4. When the pH level in the internal solution increased to pH 8.5, a large increase in channel activity was seen (Fig. 2 A), suggesting that channel activity was partially inhibited at physiological pH level. The P_open was not affected by a change in membrane potential from −100 to −40 mV in both Kir4.1 and Kir4.1–Kir5.1. At pH 8.0 and 8.5, single channel Kir4.1–Kir5.1 currents showed two sublevels of conductance, which were ∼1/4 and 3/4 of the conductance of the full open state (Fig. 2A and Fig. B). No bursting activity was observed at pH 8.0 and 8.5.

At pH 8.5, single-channel Kir4.1–Kir5.1 currents showed long periods of openings and short periods of closures (Fig. 2). The mean open time of these currents was 49.4 ± 3.6 ms (n = 4), and the mean closed time was 12.6 ± 4.5 ms (n = 4). Fig. 3A and Fig. B, illustrates dwell-time histograms of the Kir4.1–Kir5.1 currents, in which two components of time constant are revealed for the channel open state and three for the closed state (τ_O1 = 0.9 ± 0.2 ms; τ_O2 = 57.0 ± 3.8 ms; τ_C1 = 0.3 ± 0.1 ms; τ_C2 = 4.0 ± 0.3 ms; τ_C3 = 135.5 ± 53.0 ms; n = 5).

It is known that activity of Kir6, Kir3, and Kir1 members is also subject to the modulation of phosphatidylinositol-4,5-bisphosphate (PIP₂) (Baukrowitz et al. 1998; Huang et al. 1998; Shyng and Nichols 1998; Leung et al. 2000). Because of the similarity between Kir1.1 and Kir4.1, we asked if PIP₂ affects baseline activity of Kir4.1 and Kir4.1–Kir5.1. Exposure to 10 μM PIP₂ (Fluka Chemie AG) for 20–30 s enhanced baseline currents of Kir4.1–Kir5.1 by 19.9 ± 11.5% (n = 5) at pH 8.5, but had no significant effect on Kir4.1 at pH 7.5 (−3.1 ± 2.9%, n = 3). The effect of PIP₂ on Kir4.1–Kir5.1 was due to a selective increase in P_open (29.8 ± 13.1%, n = 5, at pH 7.5) without significant changes in single-channel conductance (59.9 ± 3.3 pS, n = 5, P > 0.05, at pH 7.5).

The effect of pH_i on macroscopic Kir currents was studied using giant patches under conditions similar to that described above. At pH 8.5, inward-rectifying currents as large as 2 nA were seen in most patches obtained from Kir4.1–Kir5.1-injected oocytes at a membrane potential of −100 mV. Fig. 4 illustrates modulations of the Kir4.1–Kir5.1 currents by exposures of the internal patch membrane to solutions of various pH levels. The current amplitude remained roughly the same at pH 8.0 and 8.5, started decreasing at pH 7.5, and reached almost zero at pH 7.0–6.5 (Fig. 4 A). This effect was fast, reversible, and dependent on pH levels. The relationship of the current amplitude to pH_i was described using the Hill equation with pKa (pH level at a half of the maximal currents) 7.45 and the Hill coefficient (h) 2.3 (n = 9; Fig. 4 B). When the affected currents obtained by subtraction of the remaining currents at pH 7.5 from those at pH 8.5 were scaled to the same amplitude of the baseline currents, slopes of these two current recordings were almost identical, suggesting that the pH effect is not a voltage-dependent process (not shown). In comparison, the Kir4.1 was only inhibited at much lower pH levels with pKa 5.99 and h 2.0 (Fig. 4 B).

The pH_i sensitivity of the Kir4.1–Kir5.1 was affected by PIP₂. The titration curve left-shifted by 0.23 pH U in the presence of 10 μM PIP₂ (Fig. 4 B, Table). Similar exposure to PIP₂, however, had no evident effect on the pH_i sensitivity of Kir4.1 currents (Fig. 4 B, Table).

Fig. 5 shows modulations of single channel activity in a conventional inside-out patch when the internal surface of the patch was exposed to solutions with different pH levels. The Kir4.1–Kir5.1 current showed a high channel activity at pH 8.5 (P_open 0.934) and pH 8.0 (P_open 0.935). Channel activity started being inhibited at pH 7.5. This inhibition is due to the appearance of repetitive bursting activity with no detectable reduction in the current amplitude (Fig. 5). The channel was shut off at pH 6.5. The P_open of the Kir4.1–Kir5.1 currents also can be expressed as a function of pH_i with pKa 7.48 and h 2.3 (n = 6) (Fig. 6 A).

The inhibition of P_open was due to a decrease in the channel mean open time and an increase in the mean closed time. At pH_i 7.5, single-channel Kir4.1–Kir5.1 currents had a mean open time of 25.0 ± 9.2 ms (n = 4), and the mean closed time 48.6 ± 7.4 ms (n = 4). These figures are significantly different from those obtained at pH 8.5 (see above, P < 0.05). At pH_i 7.5, the dwell-time histograms still showed two components of time constant in the open state (τ_O1 = 1.3 ± 0.4 ms; τ_O2 = 30.3 ± 8.4 ms; n = 4) and three in the closed state (τ_C1 = 0.4 ± 0.1 ms; τ_C2 = 7.5 ± 4.3 ms; τ_C3 = 471.9 ± 173.8 ms; n = 4) (Fig. 3C and Fig. D). Among these components, the τ_O2 was significantly reduced, and the τ_C3 increased at pH 7.5 in comparison with those at pH 8.5 (P < 0.05). The reduction in the long-lasting openings and the increase in long-lasting closures are consistent with the observed bursting activity emerging at pH 7.5.

The single-channel conductance was examined at these pH levels. Fig. 6 B shows that the single channel conductance is fairly constant at a pH range of 7.0–8.5, indicating that protons selectively inhibit P_open without affecting the single-channel conductance. At pH 7.0 and 7.5, when channel activity was markedly inhibited, the substate conductances were still clearly seen (Fig. 7), suggesting these sublevels of conductance are independent of intracellular protons.

To understand the molecular mechanisms underlying the pH sensitivity of the Kir4.1–Kir5.1 and Kir4.1 channels, we studied the Lys67 in Kir4.1 and Thr68 in Kir5.1 using site-directed mutagenesis. The reason for choosing these residues is that, at corresponding positions, a lysine residue (Lys80) in Kir1.1 and a threonine in Kir2.3 have been known to be critical in pH sensing of these channels (Fakler et al. 1996b; Qu et al. 1999). After the Lys67 was substituted with methionine, a counterpart residue found in the pH-insensitive Kir2.1, we found that 15% CO₂ inhibited the mutant Kir4.1 channel by only 2.3 ± 3.9% (n = 7) (Fig. 8A and Fig. B), and the heteromeric Kir4.1–Kir5.1 by 4.1 ± 1.3% (n = 12) (Fig. 8C and Fig. D). These figures were not significantly different from that obtained from the pH-insensitive Kir2.1 channels (2.9 ± 1.7%, n = 5; P > 0.05) (Fig. 8 E). In contrast to the K67M mutation, replacement of Thr68 in Kir5.1 with alanine (T68A) had no significant effect on CO₂ sensitivity of the heteromeric Kir4.1–Kir5.1 (69.6 ± 8.3, n = 3; P > 0.05 in comparison with wild-type Kir4.1–Kir5.1), suggesting that the enhanced pH sensitivity in Kir4.1–Kir5.1 is not due to introducing this residue.

The effects of the Lys67 mutations on channel sensitivity to pH_i were examined in inside-out patches. Consistent with our data obtained from the whole-cell voltage clamp, the K67M mutation completely eliminated pH_i sensitivity of the homomeric Kir4.1 (Fig. 8 E). Interestingly, the heteromeric Kir4.1–Kir5.1 with the K67M mutation remained pH_i sensitive. Such pH_i sensitivity, however, was eliminated in the presence 10 μM PIP₂ (Fig. 8 E); i.e., this mutant was only modestly inhibited at extremely acidic pH_i with pH sensitivity similar to the K67M-mutant Kir4.1. Since there are micromolar concentrations of PIP₂ in oocytes (Huang et al. 1998), these results suggest that heteromeric Kir4.1–Kir5.1 with the K67M mutation loses its sensitivity to CO₂ and pH_i when PIP₂ is available in the cytosol.

To understand how PIP₂ affects pH_i sensitivity of the Kir4.1–Kir5.1, we mutated the potential PIP₂-binding site Arg178 in Kir5.1. The Arg178 is a conserved residue found in most Kir channels, which has been shown to be a potential PIP₂-binding site in Kir6.2 (Fan and Makielski 1997). This mutation (R178Q) completely abolished the effect of PIP₂, and the mutant showed identical pH sensitivity with or without PIP₂ in the internal solution (Fig. 9). Combined mutations of K67M in Kir4.1 and R178Q in Kir5.1 rendered a channel that showed the pH_i sensitivity similar to the Kir4.1–Kir5.1 with single K67M mutation, and its pH sensitivity was not influenced by PIP₂ (Fig. 9). These results thus indicate that PIP₂ decreases pH_i sensitivity of the heteromeric Kir4.1–Kir5.1 by interacting with the Kir5.1 protein at Arg178.

In our current studies, we have demonstrated the first K⁺ channel that responds to either an increase or a decrease in intracellular pH from its physiological level, shown the modulation of pH sensitivity of Kir channels by PIP₂, and provided information about the biophysical and molecular mechanisms underlying the modulation of the heteromeric Kir4.1–Kir5.1 channels by CO₂ and pH.

Baseline single-channel properties of the homomeric Kir4.1 and heteromeric Kir4.1–Kir5.1 have not been well studied previously. The only known single-channel property as yet is the conductance. In cell-attached patches, the Kir4.1 shows a single-channel conductance of 16 pS and the Kir4.1–Kir5.1 has 40 pS (Bond et al. 1994; Pessia et al. 1996). In a separate study, we looked at single-channel properties of the Kir4.1 (Yang and Jiang 1999). In excised inside-out patches, the Kir4.1 shows a single-channel conductance of 22 pS with a high P_open at baseline (pH 7.4). In the present study, we show that the Kir4.1–Kir5.1 has a single-channel conductance of 59 pS. Both of these figures obtained from excised patches are ∼30% larger than those measured in cell-attached patches, which seems to result from different K⁺ concentrations used in these two experimental conditions (150 mM in the present study vs. 90 mM by Pessia et al. 1996). At pH 7.4, the Kir4.1–Kir5.1 channels have a moderate baseline activity (P_open ∼ 0.4) that increases with an increase in pH_i and decreases with a drop in pH_i, suggesting that these channels are potentially down or up regulated by either acidic or alkaline pH. At pH 8.0–8.5, when channel activity reaches its maximum, the Kir4.1–Kir5.1 channels show two open and three closed states with a mean open time of ∼50 ms. Another interesting feature of these channels is the substate of conductance. These K⁺ channels show two sublevels of conductance a ∼1/4 and 3/4 of the conductance of the full openings at pH 8.0–8.5. The sublevels of conductance have also been observed in outward currents of Kir2.1 (Omori et al. 1997) and inward currents of Kir2.3 (Zhu et al. 1999). In Kir2.1, the substates of conductance are related to internal Mg²⁺, while the causes for their appearance in the inward currents of Kir2.3 and the Kir4.1–Kir5.1 are not clear. Mg²⁺ is not involved since there is not Mg²⁺ in the internal solutions in both experiments on Kir2.3 (Zhu et al. 1999) and Kir4.1–Kir5.1. Although the substate conductance has been seen in several proton-gated K⁺ channels, including Kir2.3 and Kir4.1–Kir5.1, it is not found in Kir1 channels (Choe et al. 1997; McNicholas et al. 1998). Thus, the substate conductance may not be a prerequisite for the proton gating of K⁺ channels. Supporting this idea are our data showing that these substates of conductance are not affected by lowering pH_i in Kir4.1–Kir5.1. The presence of substate conductance as well as the relatively large conductance and moderate baseline P_open makes the heteromeric Kir4.1–Kir5.1 clearly different from the homomeric Kir4.1 (Yang and Jiang 1999), further supporting that they are indeed two distinct K⁺ channels (Pessia et al. 1996).

It is known that baseline channel activity of Kir channels can be modulated by PIP₂ (Baukrowitz et al. 1998; Huang et al. 1998; Shyng and Nichols 1998; Leung et al. 2000). We also looked at the modulation of Kir4.1 and Kir4.1–Kir5.1 by PIP₂. We found that baseline Kir4.1–Kir5.1 currents were enhanced by PIP₂, which was due to an augmentation of P_open without affecting single-channel conductance. Interestingly, PIP₂ did not affect Kir4.1, suggesting that PIP₂ interacts with Kir5.1 rather than Kir4.1. Supporting this idea are our mutagenesis data showing that the R178Q mutation completely eliminated the PIP₂ effect. Thus, it is possible that PIP₂ binds to the arginine residue through electrostatic charges, affecting the protein conformation and channel activity, as suggested previously (Fan and Makielski 1997; Huang et al. 1998).

Several members of the Kir4 family have been demonstrated to be pH sensitive. Shuck et al. 1997 have cloned a Kir channel from the kidney (named Kir1.2) that has a 97% identity to the brain Kir4.1. This renal K⁺ channel is inhibited by pH. The pH sensitivity of this Kir (pKa 6.2) is slightly higher than the brain Kir4.1. Kir4.2 cloned from the liver is also pH sensitive. It has 64% amino acid sequence homology to the brain Kir4.1 (Pearson et al. 1999). This channel is markedly inhibited by intracellular acidification produced by 50 mM HCO₃⁻, although its pKa value is still unknown. In our previous studies, we have demonstrated that the brain Kir4.1 is sensitive to CO₂ and intracellular pH (Yang and Jiang 1999). Interestingly, we found that channel sensitivity to CO₂ and pH is dramatically enhanced when the Kir4.1 is coexpressed with brain Kir5.1 (Xu et al. 2000). It is known that the Kir5.1 tends to form heteromeric channels with other Kir members, particularly with Kir4.1 (Fakler et al. 1996a; Pessia et al. 1996), in which several new biophysical properties appear. Our current data indicate that such a heteromeric expression also changes in a major way channel modulations by intracellular protons.

Our previous studies have indicated that a decrease in pH during CO₂ exposure is the primary cause for the Kir channel inhibition (Xu et al. 2000), which is consistent with our observations in the current studies. Furthermore, our results have shown that the inhibition of the Kir4.1–Kir5.1 by low pH is mediated by selective suppression of P_open without evident effects on single-channel conductance, leading to a decrease in the channel mean-open time and an increase in the mean-closed time. The biophysical mechanisms are more like those underlying Kir1.1 channel inhibition by intracellular protons (Choe et al. 1997; McNicholas et al. 1998) than those for the Kir2.3 (Zhu et al. 1999) and the homomeric Kir4.1 inhibitions (Yang and Jiang 1999). In the Kir4.1, we have shown that intracellular protons inhibit P_open and augment the single-channel conductance (Yang and Jiang 1999). The opposite effect of pH on these two different single-channel properties, however, is not seen in the heteromeric Kir4.1–Kir5.1, even though it apparently has two Kir4.1 subunits in its channel protein. What causes this difference is unclear. It is possible that introducing the Kir5.1 subunits not only alters the architecture of the heteromeric K⁺ channels, but also alters the channel modulation by intracellular ligands.

Interestingly, we found that the pH sensitivity of the heteromeric Kir4.1–Kir5.1 is modulated by PIP₂. Channel sensitivity to intracellular protons is reduced in the presence of PIP₂. Considering the presence of micromolar concentrations of PIP₂ in an intact cell (Huang et al. 1998), our results suggest that the more realistic pKa level of the heteromeric Kir4.1–Kir5.1 is likely to be pH 7.2–7.3, instead of pH 7.45 as seen in cell-free excised patches. Therefore, a sharp change in channel activity appears to occur in pH_i 7.5–7.0, allowing the heteromeric Kir4.1–Kir5.1 to detect pH fluctuations in most physiological and pathophysiological conditions.

The inhibition of Kir channels by low pH_i is unlikely to be caused by changes in concentrations of cytosolic soluble factors such as second messengers, polyamines, or Mg²⁺ since it is seen in cell-free excised patches. Also, our results do not support the idea that protein phosphorylation is responsible for the modulation of these K⁺ currents by intracellular protons. Several blockers of phosphatase and phosphodiesterase such as vanadate, fluoride, and pyrophosphate were used in the intracellular solution. These chemicals tend to inhibit protein dephosphorylation. In addition, there was no Mg²⁺ or ATP in this intracellular solution. Under such a condition, the turnover of protein phosphorylation and dephosphorylation should not occur, at least in the time frame of our low pH experiments (0.5–2.0 min). Therefore, the modulation of the Kir4.1–Kir5.1 during low pH may not be a result of protein phosphorylation.

Our studies suggest that amino acid sequences and tertiary structures in Kir channel proteins are the molecular basis underlying the modulation of these K⁺ channels by protons. We have found that Lys67 in the Kir4.1 is critical in the modulation. Channel sensitivity to CO₂ is completely eliminated when this lysine residue is mutated to methionine or glutamine. A lysine residue found at the same position of Kir1.1 and Kir1.2 channels has been demonstrated to play a crucial role in pH sensing of these channels (Fakler et al. 1996b; McNicholas et al. 1998). Interestingly, our data have shown that this lysine residue is not only critical in pH modulation of the homomeric Kir4.1, but is also responsible for proton detection in the heteromeric Kir4.1–Kir5.1, as the K67M and K67Q mutations also totally abolish CO₂ sensitivity of the Kir4.1–Kir5.1. It is known that lysine is titratable only at extremely high pH levels with pKa 10.5. Previous studies have suggested that adjacent residues to the Lys80 in Kir1.1 may be involved in reducing its pKa value to a physiological pH range (Fakler et al. 1996b; Choe et al. 1997). If this is also the case in the Kir4.1–Kir5.1, then the influence of amino acid residues appears to be stronger in Kir5.1 than in Kir4.1 (Schulte et al. 1999). Such close interaction between these two subunits may underlie the high CO₂/pH sensitivity of the heteromeric channels. Since PIP₂ seems to favor membrane association of the intracellular protein domains (Shyng and Nichols 1998), the interaction of Kir5.1 with Kir4.1 may be limited in the presence of PIP₂, leading to a shift of the titration curve to a lower pH level. The tendency of reducing pH_i sensitivity persists and is even exaggerated in the K67M mutant through certain unknown processes. Despite the fact that the methionine does not exist in the wild-type channel and the pH sensitivity of the K67M-mutant Kir4.1–Kir5.1 appears to be an artifact, our results imply that there are other proton-sensing sites in addition to Lys67. These potential titration sites seem to be located on Kir5.1 and may not be functioning in the wild-type channel. They begin to play a part only when the Lys67 is mutated and there is a lack of PIP₂ in the cytosol. Since a large number of residues are involved in pH sensing (Schulte et al. 1999; Chanchevalap et al. 2000), these secondary titration sites appear to play a less important role in pH sensing of the Kir4.1–Kir5.1 channel.

We understand that this lysine residue may also be a part of the gating mechanisms in these Kir channels. With an interruption of the gating mechanism, the Kir4.1 and Kir4.1–Kir5.1 may not be able to close appropriately during hypercapnia or intracellular acidification, and thus show a marked decrease in pH sensitivity. Since the binding versus gating has been a common problem in studies of all ligand-gated ion channels (Colquhoun 1998), the demonstration of a convergent site for proton gating in Kir4.1 and Kir4.1–Kir5.1 in our current studies may contribute to the understanding of the gating mechanisms for these K⁺ channels by intracellular protons.

Inward-rectifier K⁺ channels are important players in the maintenance of plasma membrane excitability and the control of intra- and extracellular K⁺ ionic homeostasis. In the brainstem where the Kir4.1 and Kir5.1 channels are expressed, the inhibition of heteromeric Kir4.1–Kir5.1 channels by hypercapnia can have a major impact not only on cells expressing these channels, but also on local neuronal networks. The inhibition of these K⁺ channels produces depolarization and increases membrane excitability (Pineda and Aghajanian 1997; Zhu et al. 2000), which may lead to a spread of the excitation to other brainstem neurons such as those responsible for cardio-respiratory controls. Consequently, the hypercapnic information is coupled to a corresponding change in excitability of the cardio-respiratory system (von Euler 1986). Thus, expression of these Kir channels in brainstem neurons that have direct or indirect connections with cardio-respiratory neuronal networks may enable these cells as well as the cardio-respiratory system to detect PCO₂ levels in the cerebral spinal fluid and blood circulation. In the kidney, several inward K⁺ channels are known to be modulated by CO₂ and pH, with some of them showing high pH sensitivity (Schlatter et al. 1994; Zhou and Wingo 1994). These K⁺ channels may play a role in the regulation of K⁺ and pH homeostasis (Wang et al. 1997). Since the heteromeric Kir4.1–Kir5.1 channels are highly pH sensitive and are also expressed in the kidney, they may be another potential candidate of the renal K⁺ channels responsible for pH sensing and regulating K⁺ secretion in the kidney. A common characteristic of these K⁺ channels, whether in brainstem neurons or in renal tubular cells, is the sensitivity to pH changes around the physiological level, by which channel activity can be modulated whenever pH is shifted from the physiological level. Therefore, the finding of a K⁺ channel with pKa 7.2–7.3 and the demonstration of the biophysical and molecular mechanisms underlying the modulation of these K⁺ channels by CO₂ and pH constitute a significant step towards the understanding of their potential functions in CO₂ and pH sensing in these cells.

We thank Dr. John Adelman for his generosity in sharing with us the Kir4.1 and Kir5.1 cDNAs.

This work was supported by the National Institutes of Health (RO1 HL58410-01) and a Grant-in-Aid award (9950528N) from the American Heart Association.