Permeation Properties of Inward-Rectifier Potassium Channels and Their Molecular Determinants

Permeation is the defining property of an ion channel and has been a subject of intensive studies (Hille 1992). It is now well established that ions move through the pore by an electrodiffusive mechanism whose molecular determinants are beginning to be understood. One of the fundamental questions concerning permeation is how an ion channel can have a simultaneous high conductance and high selectivity since selectivity implies strong interactions between the ions and the pore (Latorre and Miller 1983).

Almost all K⁺ channels are >100-fold times as permeable to K⁺ as to Na⁺ or Li⁺ (Hille 1992). This high selectivity can be accounted for by the close-fit hypothesis based on the size of Na⁺ (1.90 Å) and K⁺ (2.66 Å) (Mullins 1959; Bezanilla and Armstrong 1972; Hille 1973). The crystal structure of KcsA channel supports this hypothesis, and furthermore reveals that the K⁺ ion is coordinated by backbone carbonyl oxygens in the selectivity filter (Doyle et al. 1998).

Other monovalent cations, NH₄⁺, Rb⁺, and Tl⁺, have sizes (3.00, 2.96, and 2.80 Å, respectively; Hille 1973) similar to that of K⁺, and indeed many K⁺ channels conduct those ions with different rates. The selectivity among these ions is thought to occur by different affinities at binding sites within the pore and different rates of translocation to and from these sites.

The inward-rectifier K⁺ channel (Kir) is a tetramer (Glowatzki et al. 1995; Yang et al. 1995b), where each subunit has two transmembrane domains: M1 and M2 (for reviews, see Jan and Jan 1997; Nichols and Lopatin 1997). Both NH₂ and COOH termini are located in the cytoplasm, and the extracellular loop (ECL) contains a P region that forms a selectivity filter. Sequence similarity suggests that the Kir channel is homologous to the core (S5, S6, and H5) structure of voltage-gated K⁺ channels (Kv).

There is evidence that both the COOH-terminus and the H5 (pore) region of the channel are important for determining the K⁺ conductance of the Kir and Kv families (Lopez et al. 1994; Taglialatela et al. 1994; Zhou et al. 1996; Choe et al. 1999). It is not known whether these two regions form independent barriers to ion movement or whether the structure of one could modulate the properties of the other.

Previously, we have described the gating mechanism of ROMK2 (Kir1.1b) (Choe et al. 1998) and the structural determinants of the Kir channel gating (Choe et al. 1999). Here we report the permeation properties of Kir channels and explore the molecular determinants of channel conductance and permeation. Guided by the finding that two members of the Kir channel family, ROMK2 and IRK1, have different conductances and different selectivities for the cations (K⁺, NH₄⁺, Rb⁺, and Tl⁺ ions), we used chimeras of these two channels to investigate which parts of the inward rectifier structure are responsible for differences in ion permeation.

The construction of the chimeras between ROMK2 (Zhou et al. 1994) and IRK1 (Kubo et al. 1993) has been previously described (Choe et al. 1999). The designs of the chimeras were based on division of the ROMK and IRK subunit structures into seven homologous and interchangeable regions. These are indicated in Fig. 1 as: (a) the N-terminus, (b) the M1 transmembrane segment, (c) the MP region, (d) the highly conserved pore (P) domain, (e) the PM segment, (f) the M2 transmembrane segment, and (g) the COOH terminus. The pore alpha-helix includes both the latter part of the MP segment and the first part of the P region. The amino acids of the seven segments are defined in Figure 4 of Choe et al. 1999.

Primers were synthesized by Operon Technologies, Inc. (Alameda, CA). Nucleotide sequences between two restriction enzyme sites were checked using an ABI 377 automated DNA sequencer at The Rockefeller University DNA Technology Center (New York, NY), or an ABI 377XL automated DNA sequencer at The Cornell University Bioresource Center (Ithaca, NY).

Plasmids were linearized with NotI restriction enzyme and transcribed in vitro with T7 RNA polymerase in the presence of the GpppG cap using mMESSAGE mMACHINE kit (Ambion, Inc.). cRNA pellet was dissolved in Ultraspec water containing DEPC (Biotecx) and stored at −70°C before use. Stage V–VI oocytes were obtained by partial ovariectomy of female Xenopus laevis oocytes (Xenopus-I or NASCO), anesthetized with tricaine methanesulfonate (1.5 g/liter, adjusted to pH 7.0 with NaOH). Oocytes were defolliculated by incubation in OR2 solution (82.5 mM NaCl, 2 mM KCl, 1 mM MgCl₂, and 5 mM HEPES, adjusted to pH 7.5 with NaOH) containing 2 mg/ml collagenase type II and 2 mg/ml hyaluronidase type II (Sigma Chemical Co.) for 90 min and another 90 min in a fresh enzyme solution at room temperature. Oocytes were injected with 0.5–1 ng of cRNA and incubated at 19°C in 2× diluted Leibovitz medium (Life Technologies, Inc.) for 1–4 d before measurements were made. For patch-clamp experiments, oocytes were subjected to a hypertonic shrinking solution containing 200 mM sucrose, thereby allowing the vitelline membrane to be easily removed.

Whole-oocyte current–voltage relationships were obtained in intact oocytes, using a two-electrode voltage clamp with 1 M KCl-filled current and voltage electrodes. Currents were recorded for 50 ms at each voltage, using a pulse protocol in which membrane potential was stepped by 10 mV, interspersed with a return to the resting membrane potential. The 5-K bath solution contained (mM): 105 NaCl, 5 KCl, 2 CaCl₂, 1 MgCl₂, and 5 HEPES, pH 7.4. The 5-Rb bath solution contained 5 mM RbCl instead of KCl.

Oocytes were bathed in a solution containing (mM): 110 KCl, 2 CaCl₂, 1 MgCl₂, and 5 HEPES, pH 7.4. Patch-clamp pipettes were pulled from borosilicate glass (7052; Richland Glass Co.) using a three-stage process and were coated with Sylgard. They were filled with solutions containing (mM): 110 KCl and 5 HEPES with or without 5 EDTA, pH 7.4. In some cases, KCl was substituted with NH₄Cl, RbCl, or TlNO₃. When Tl⁺ was the major conducted ion, NO₃⁻ was substituted for Cl⁻ in the bath solution. All experiments were performed at room temperature (19–21°C). Pipette resistances were 1–3 MΩ. Currents were recorded with a patch-clamp amplifier (EPC-7; List Electronik) and stored, unfiltered, on videotape. For off-line analysis, current records were replayed from videotape, filtered at 1 kHz, and sampled at 5 kHz, using an Atari-based data acquisition system (Instrutech Corp.). Construction of open- and closed-time histograms and fitting with exponential distributions were carried using the TAC program (Sigworth and Sine 1987).

We constructed a two-ion three-site four-barrier permeation model using matrix methods for steady state I-V curve computation (Begenisich and Cahalan 1980). The transition rates were constructed based on the transition state theory. The transition rate from state i to j, k_jⁱis given by :

where ν and ΔE_A represent the Kramers' preexponential factor and the voltage-independent component of the transition state energy (between the binding site and the barrier peak), respectively. Other symbols have their usual meanings. No Coulombic interaction between ions was considered.

There are 27 ion-occupied states in this model. Since we are interested in the steady state, the transition-rate equations of the 27 states are given by :

with the conservation equation:

where W_i represents occupancy, a probability of finding the system in state i.

Since the leftmost matrix is singular, for computational purposes, a nonhomogeneous equation and a nonsingular matrix are produced by dropping the equation from the last row of the matrix and substituting . Then ():

The inverse matrix was obtained using the “MINVERSE” function, a built-in function of Microsoft Excel (Microsoft Corp.). A specific occupancy of state i can be obtained using the “INDEX” function; for example, W_i= INDEX [MINVERSE (array), i, 27].

The current of a particular ion is its steady state net flux crossing over any one barrier multiplied by its valence. Therefore ():

where z_A and z_B represent charge of ion A and ion B, respectively. The symbol EAE indicates the state that the second energy well is occupied by ion A, and the symbol EAA indicates the state that the second and the third energy wells from outside are occupied by ion A, and so on. Other symbols have their usual meanings.

All the fitting procedures were done using the “Solver” function, a built-in optimization function of Microsoft Excel. Results of the fitting process were displayed as both energy profile diagrams and transfer rates between energy wells. The energy profiles diagrams with peaks and wells have the advantage of providing a pictorial representation of the permeation process. Transfer rates have the advantage of being independent of model assumptions about absolute barrier height, or, equivalently, of the “preexponential factor.”

Channels were expressed by injecting cRNA into Xenopus oocytes, and patch-clamp recordings of single channels were obtained in the cell-attached mode. We were interested in whether the COOH terminus and the pore region contribute independently to the channel conductance. To address this question, we systematically measured the K⁺ conductance of various chimeras. Fig. 2 shows current traces of ROMK2, IRK1, and some of the chimeras at −100 mV. ROMK2 has a larger open-channel current (−3.7 pA) than IRK1 (−2.8 pA) under these conditions. The K⁺ conductance of IRK1 in this study is 29 pS, which is consistent with a previous study (Sabirov et al. 1997), but slightly larger than some other measurements; e.g., 23 pS (Kubo et al. 1993). These differences can be explained by different concentration of Mg²⁺ in the pipette solution: in our study, no Mg²⁺ was added, whereas Kubo et al. 1993 used 3 mM Mg²⁺. 5 mM Mg²⁺ decreased single-channel conductance of IRK1 to 13 pS (Sabirov et al. 1997).

When the NH₂ terminus of ROMK was replaced with the corresponding region of IRK1, the current amplitude of the chimera (Chm1) was not changed significantly. However, when the ECL (consisting of MP, P, and PM regions) of ROMK was replaced, the chimera (Chm107) had larger currents (−6.3 vs. −3.7 pA at −100 mV). Replacement of the COOH terminus of ROMK2 with that of IRK1 (Chm13) decreased the current amplitude to −2.2 pA. These results are consistent with those of a previous study involving ROMK1–IRK1 chimeras (Taglialatela et al. 1994).

When the M1 or M2 region of IRK1 was replaced with that of ROMK2 (Chm4 and Chm12, respectively), there was no significant change (Fig. 2). However, when the ECL of IRK1 was converted to that of ROMK2 (Chm108), the current decreased from −2.8 to −2.0 pA.

In a summary of inward conductances (Fig. 3), the COOH terminus and the ECL stand out as major determinants of this conductance. As indicated in Fig. 3, the wild-type and chimeric channels can be classified into four groups, depending on the origin of these two components. Group 1 has both the ECL and the COOH terminus from ROMK2. Group 2 has both regions from IRK1. Group 3 has the ECL from ROMK2 and the COOH terminus from IRK1. Finally, Group 4, which shows the largest conductance (63 pS) has the ECL from IRK1 and the COOH terminus from ROMK2. M2 may also contribute somewhat because, in Group 2, the chimeras that have M2 from ROMK2 have slightly higher conductances than those with the M2 from IRK1. However, in the subsequent analysis, we will focus only on the ECL and the COOH terminus because these regions have the largest effects.

The simplest model that can explain these results is one in which the conduction pathway has two regions that impede ion flow (resistors) arranged in series (see discussion and Fig. 8). The “outer” resistor is formed by the extracellular region, and the “inner” resistor by the COOH terminus. In this model, the inner resistance of IRK1 was larger than that of ROMK2, while the outer resistance of ROMK2 was larger than that of IRK1.

Chimeras were constructed to evaluate which region of the ECL is the primary determinant of conductance. In our scheme (Choe et al. 1999), the ECL is divided into three parts: (a) the most conserved, pore forming, P region, (b) the MP region, between M1 and P, and (c) the PM segment, between P and M2 (Fig. 1). These chimeras showed that exchanging the P region of ROMK2 with that of IRK1 increased the conductance to an even greater extent than exchanging the entire ECL (Fig. 4). The difference is attributable to MP, since chimeras where this segment was derived from IRK1 had lower conductances.

The P regions of ROMK2 and IRK1 have only two amino acid differences. One is at the −1P position and the other is at the 3P position. Therefore, we studied mutants in which either the −1P or 3P position in ROMK2 was changed to the corresponding amino acid of IRK1. The L117I mutant, involving the −1P position, had the same conductance as wild type, whereas the V121T mutant, involving the 3P position, had a conductance similar to that of Chm107 (Fig. 4), although slightly smaller than that of Chm7. Therefore, the 3P position seems to be responsible for most of the differences in conductances arising from different ECLs.

We also studied NH₄⁺, Rb⁺, and Tl⁺ selectivities of single channels in the cell-attached mode. Fig. 5 shows current traces of ROMK2 and IRK1 for K⁺, NH₄⁺, Rb⁺ and Tl⁺. Replacing pipette K⁺ with different ions alters the reversal potential in the hyperpolarizing direction for NH₄⁺ and Rb⁺, and in the depolarizing direction for Tl⁺. Bi-ionic potentials from linear-regression analysis of the data and permeability ratios are shown in Table. The overall permeability sequence for both ROMK2 and IRK1 is Tl⁺ > K⁺ > Rb⁺ > NH₄⁺. Permeability ratios of NH₄⁺, Rb⁺, Tl⁺, versus K⁺ for both ROMK2 and IRK1 are similar.

To quantitatively estimate the inward slope conductance, we measured current amplitudes at voltages of up to 100 mV negative to the reversal potentials: 0 to −100 mV for K⁺, −60 to −160 mV for NH₄⁺, −20 to −120 mV for Rb⁺, and 20 to −80 mV for Tl⁺ (Fig. 6). However, it was difficult to measure unitary events of Rb⁺ current for IRK1 at voltages positive to −120 mV, so we increased the voltage range to −200 mV to obtain a single-channel current of −0.6 pA. Tl⁺ ion currents in IRK1 often opened to a maximal level, and then declined, sometimes gradually, to the main conductance level (Fig. 5). Because we don't know the mechanism underlying this phenomenon, for the analysis of Tl⁺ conductance, we took the main conductance state, although the conclusions reached are the same if the maximum current levels are used. Although inward slope conductances were determined from a linear-regression analysis, the I-V plots in Fig. 6 and Fig. 7 deviate slightly but consistently from linearity. The data can be explained by kinetic models with at least three energy barriers and two energy wells (see Fig. 9, and see also Choe 1999). The main points of this paper, however, can be adequately illustrated with simple linear resistive elements.

The conductance ratios of NH₄⁺, Rb⁺, or Tl⁺ versus K⁺ for ROMK2 and IRK1 are shown in Table. The conductance sequence for ROMK2 is: NH₄⁺ > K⁺ > Tl⁺ > Rb⁺. For IRK1, the conductance sequence is: K⁺ > Tl⁺ > NH₄⁺ >> Rb⁺. Thus, the conductance ratios of ROMK and IRK differ in sequence order. Also, the conductance and permeability ratios of channels are different. The most striking examples are for the NH₄⁺ and Tl⁺ selectivity of ROMK2: γ_NH4⁺/γ_K⁺ is 1.62, whereas P_NH4⁺/P_K⁺ is 0.09; γ_Tl⁺/γ_K⁺ is 0.53, whereas P_Tl⁺/P_K⁺ is 2.30.

To identify the regions responsible for the selectivity differences between ROMK2 and IRK1, we studied the selectivities of two representative chimeras, Chm25 (having the NH₂ and COOH termini of IRK1) and Chm107 (having the ECL of IRK1). Fig. 6 shows I-V plots of these chimeras, as well as those of wild-type ROMK2 and IRK1. Table summarizes the selectivities of these chimeras. The permeability sequence of both Chm25 and Chm107 is: Tl⁺ > K⁺ > Rb⁺ > NH₄⁺, which is similar to that of ROMK2 and IRK1. However, the conductance ratio sequences for Chm25 and Chm107 are: NH₄⁺ > K⁺ > Tl⁺ > Rb⁺ and NH₄⁺ ≈ K⁺ > Tl⁺ > Rb⁺, respectively. The latter differs from those of both ROMK2 and IRK1.

In both cases, the ratios γ_NH4⁺/γ_K⁺ and γ_Rb⁺/γ_K⁺ of the chimeras are closer to those of the wild-type channel from which the ECL of the chimera was derived. However, the Tl⁺ versus K⁺ conductance ratio appeared to be determined by a different region. The γ_Tl⁺/γ_K⁺ of Chm25 and Chm107 more closely resemble those of the wild-type channel from which the cytoplasmic termini were derived. Thus γ_Tl⁺/γ_K⁺ for Chm107 is 0.37, closer to the value for ROMK2 (0.52). Conversely, γ_Tl⁺/γ_K⁺ for Chm25 is 0.82, closer to the value for IRK1 (0.83).

To identify the terminus responsible for Tl⁺ ion selectivity, we tested the Tl⁺ ion conductance of Chm1 and Chm13. These chimeras are, respectively: ROMK2 with the NH₂ terminus of IRK1 and ROMK2 with the COOH terminus of IRK1 (Table). The Tl⁺ and K⁺ conductances of Chm1 are similar to those of ROMK2. However, the conductances of Chm13 are closer to those of Chm25. These results, together with the effect of COOH terminus on K⁺ conductance, suggest that the Tl⁺ versus K⁺ selectivity is controlled mainly by the COOH terminus.

To identify the segments of the ECL responsible for K⁺:Rb⁺ selectivity, we tested the macroscopic Rb⁺ versus K⁺ conductance ratio of chimeras using the two-electrode voltage clamp technique. When the bath solution was switched from 5 K⁺ to 5 Rb⁺ solution, the chimeras having the P region of ROMK2 (Chm8 and 9) had G_Rb⁺/G_K⁺ = 0.45 ∼ 0.51, similar to that (0.44) of ROMK2 wild type; whereas the chimeras having the P region of IRK1 (Chm6, 7, 10, 37, and 45) had G_Rb⁺/G_K⁺ = 0.05 ∼ 0.17, similar to that (0.05) of IRK1 wild type. These results suggest that the P region is important for the Rb⁺ versus K⁺ selectivity.

Therefore, we evaluated only Chm8 and the two point mutants, L117I and V121T, to narrow down the determinant of the single-channel conductance ratio of Rb⁺ and NH4 versus K⁺. These results are shown in Fig. 7 and Table. Both Chm8 (which is IRK1 except for the P region of ROMK2) and the point mutant L117I have a single-channel conductance sequence of NH₄⁺ > K⁺ > Rb⁺, similar to that of ROMK2. The conductance sequence of the V121T mutant was K⁺ > NH₄⁺ >> Rb⁺, a sequence identical to that of IRK1. Therefore, we conclude that the 3P position of IRK and ROMK is responsible for the selectivity differences among the cations: K⁺, Rb⁺, and NH₄⁺.

Since the 3P position is very important for determining the conductances of K⁺, NH₄⁺, and Rb⁺ of the channels, we substituted the Val residue at this position in ROMK2 with 18 other amino acids to further understand the molecular mechanism of conductance and selectivity.

We observed no significant current from 10 mutants in which the V121 residue (3P position) was replaced with one of the following: Asp, Glu, Lys, Arg, Tyr, Pro, Trp, Phe, Gly, or Leu. The Leu mutant had a resting membrane potential of −70 mV at 5 mM K⁺ bath solution, consistent with K⁺ channel expression, but had <1 μA of current at a voltage of −120 mV. This is puzzling since Ile, a similar amino acid, formed a fully functional channel. In the other nine mutants, the potential was −30 to −50 mV even 3 d after cRNA injection, which is similar to that of noninjected oocytes. Since none of the charged amino acids showed significant currents except for histidine, we expect that the histidine exists in the deprotonated form at least when the channel is conducting. In the cases of Tyr, Trp, and Phe, the large size of the side chains might occlude the pore. The absence of current of the Pro mutant may result from a structural modification since this side chain changes the direction of peptide backbone abruptly.

The remaining nine mutants, together with ROMK2, expressed currents that were large enough for single-channel analysis. The conductances and selectivities of these channels are presented in Table. The first five rows summarize results with the five nonpolar side chains (Val, His, Met, Ile, and Ala). These, together with two of the polar side chain mutants (Gln and Cys) shown in the next two rows, had an ROMK2-type conductance sequence: NH₄⁺ > K⁺ > Rb⁺. On the other hand, two of the mutants with polar side chains (Thr and Ser) at position 121 had an IRK1-type conductance sequence: K⁺ > NH₄⁺ > Rb⁺. Finally, the mutant with Asn at this position showed a novel selectivity sequence: NH₄⁺ > Rb⁺ > K⁺.

While the inward single-channel K⁺ conductances of ROMK2 and IRK1 are similar, this can be viewed as the net result of opposite contributions of two parts of the channels. The ECL of IRK1, including the P region, has a much higher conductance than that of ROMK2, whereas having the COOH terminus of IRK1 lowers the conductance. Exchanging the ECLs and COOH termini of the two channels changes the overall conductance in a way that suggests that these two parts of the channels form independent resistances. The simplest interpretation of these results is that the COOH terminus, perhaps in combination with other elements of the channel, forms the inner part of the conductive pore, which can confer a resistor in series with that of the P region (Fig. 8). Another interpretation of the result is that replacement of COOH terminus affects the structure of selectivity filter through propagation of a conformational change. However, this idea does not predict the observation that replacing the COOH terminus of ROMK with that of IRK increases channel resistance by an amount equal to the decrease observed when the converse exchange is made. It is also difficult to reconcile with the fact that changing the COOH terminus alters the conductance of the channel without affecting the striking selectivity patterns among the permeant cations K⁺, NH₄⁺, and Rb⁺ (Table and Fig. 6).

One prediction made from the series resistors model is that in ROMK2, with a relatively large outer resistor and a relatively small inner resistor, the voltage drop across the outer resistor should be smaller than that across the same region of IRK1. In principle, this voltage drop can be estimated using a blocking ion that binds to a site between the two resistors. Mg²⁺ is one candidate for such a blocker. It reduces the single-channel conductance of IRK1 and ROMK2 when added to the cytoplasmic side of the membrane (Lu and MacKinnon 1994; Wible et al. 1994; Chepilko et al. 1995), and its affinity depends on electrostatic interactions with amino acid side chains in the M2 segment (Fakler et al. 1994; Lu and MacKinnon 1994; Wible et al. 1994), which should lie between the two resistors. Mg²⁺ block has a voltage dependence consistent with a binding site within the electric field. In the cases of ROMK and other weak inward-rectifier K⁺ channels, the apparent fraction of the field sensed at that site (δ_Mg) is 0.28–0.44 of the total measured from the cytoplasmic side (Horie et al. 1987; Lu and MacKinnon 1994; Nichols et al. 1994; Chepilko et al. 1995; Spassova and Lu 1998). For IRK1 and other strong inward-rectifier K⁺ channels, δ is higher, ranging from 0.56 to 1.01 from the cytoplasmic side (Matsuda 1988; Horie and Irisawa 1989; Lopatin et al. 1994; Stanfield et al. 1994; Wible et al. 1994; Yang et al. 1995a), except for one study that reported a value of 0.33 (Fakler et al. 1994). These numbers are in general agreement with the model. If δ_Mg = 0.3 for ROMK2, the ratio of inner to outer resistors (R_i/R_o) would be 0.3/0.7 = 0.43. Since the resistance of a single channel is (38 pS)⁻¹ or 26 GΩ, R_i will equal 8 GΩ and R_o will equal 18 GΩ. For IRK1, the corresponding values can be calculated from the difference in resistance when the ECL or COOH terminus of ROMK2 is replaced with the corresponding part of IRK. This gives R_i = 27 GΩ and R_o = 8 GΩ. This predicts δ_Mg = 8/35 = 0.77, within the range of measured values.

Mg²⁺ might not be the perfect blocker for this analysis. At least part of the voltage dependence of Mg²⁺ block has been attributed to the displacement of K⁺ ions across the outer mouth of the channel (Spassova and Lu 1998). Thus, the apparent electrical distance for block may overestimate the true fraction of the field sensed at the binding site. However, even if the inner resistor of ROMK2 is small, that of IRK would be ≥19 GΩ, a significant fraction of the total. The inner pore is unlikely to be as narrow as the selectivity filter because other studies have indicated that large cations can enter the channel from the inside, suggesting that the opening of the pore to the cytoplasm is relatively wide. Polyamines such as spermine and spermidine block Kir channels in a voltage-dependent fashion from the inside (Lopatin et al. 1994, Lopatin et al. 1995; Taglialatela et al. 1995; Yang et al. 1995a; Lopatin and Nichols 1996; Pearson and Nichols 1998), indicating that the narrowest region of the inner pore must be at least as wide as the diameter of these molecules, which are ∼4-Å wide and 7–18-Å long.

A pore resistance, R_pore, is determined by several factors ():

where ρ, l, and a represent resistivity, the length, and the radius of the pore, respectively. The high inner resistance presumably reflects several possibilities: a long extension of the pore into the cytoplasm, and/or a hydrophobic inner pore wall (which will increase resistivity). The possibility of hydrophobic wall of the inner pore is suggested by the finding that cations with longer hydrophobic elements can be more effective K⁺ channel blockers (Armstrong 1969; Armstrong 1971). Indeed, the KcsA structure shows hydrophobic walls in the pore lined by the second transmembrane domain (Doyle et al. 1998). The wall of the inner pore, which was not determined in the KcsA structure, might be hydrophobic as well.

Another possibility is that the diameter of the inner pore fluctuates. By analogy, the x-ray structure of myoglobin reveals no path for O₂ to reach the heme group (Perutz and Mathews 1966; Case and Karplus 1979), but O₂ and CO bind to the heme group at rates that are close to the diffusion-controlled limit (Hasinoff 1977), suggesting thermal fluctuations of the protein (Elber and Karplus 1987). This mechanism would allow passage of a relatively large molecule through a normally narrow pore.

Table lists the permeation and the conductance ratios of various native and cloned K⁺ channels. All of these K⁺ channels have remarkably similar permeability ratios (P_X/P_K), where Tl⁺ > K⁺ > Rb⁺ > NH₄⁺. Conductance ratios show a greater variation with channel type than do the permeability ratios. This is partly because conductance ratios depend on experimental conditions such as the voltage range over which they are measured. Nevertheless it is clear that different K⁺ channels can have distinct selectivity sequences based on conductance ratios.

While the signature GYG sequence is thought to be the site conferring selectivity for K⁺ ions over Na⁺, the selectivity sequence for permeant ions may have different determinants because ROMK2 and IRK1 have the identical GYG sequence, but yet different selectivity based on the single-channel conductance ratio. Our results indicate that most of the conductance selectivity among the permeant cations (K⁺, Rb⁺, and NH₄⁺) is conferred by the outer resistance. The one exception is that the Tl⁺ versus K⁺ single-channel conductance ratio follows the origin of the two cytoplasmic termini (Table). If both termini originate from ROMK, the Tl⁺/K⁺ conductance ratio resembles that of ROMK2; if both termini originate from IRK, then the ratio resembles that of IRK1.

How can Tl⁺/K⁺ conductance ratio be altered by the relatively wide inner pore, whereas similar sized ions, Rb⁺ and NH₄⁺, are not affected? One possibility is that Tl⁺ may bind to specific sites within the inner pore, such as sulfhydryl groups, and affect the Tl⁺ conductance. A Tl⁺ binding-site hypothesis was first proposed by Ashcroft and Stanfield 1983 to explain inactivation of Tl⁺ current in frog skeletal muscle.

We identified the 3P position (located 3 residues NH₂ terminal to the GYG signature sequence), as a modifying site for ion permeation through inward rectifiers. This is similar to what is observed with voltage-gated channels. For example, Shaker and mslo (Ca²⁺-activated K⁺) channels display significant differences in both γ_NH4⁺/γ_K⁺:0.75/0.18 and γ_Rb⁺/γ_K⁺:0.5/0.07 (Table). These differences can be correlated with differences at the 3P position, where Shaker has a Thr residue and mslo has a Ser residue. In fact, the Shaker mutation T441S (at the 3P position) changed both the conductance and the bi-ionic permeability ratios for Rb⁺ and NH₄⁺ (Yool and Schwarz 1991). This mutant also exhibited an anomalous mole fraction effect in P_Rb⁺/P_K⁺ (Yool and Schwarz 1996). In our experiments, V121T and V121S showed no significant difference in selectivity. The 3P position is also responsible for the 10-fold difference in Ba²⁺ affinity between ROMK2 and IRK1 (Zhou et al. 1996). The homologous position in the Shaker K⁺ channel was also shown to be important for the efficacy of block by extracellular Ba²⁺ (Harris et al. 1998).

The mechanism underlying the effect of changing the 3P amino acid on selectivity is unknown. Our analysis suggests that the outer resistor of the IRK1 channel conducts K⁺ more easily than that of ROMK2, but that both Rb and NH₄ conductances are smaller. This could be explained by a small decrease in the pore diameter of the selectivity filter, which would favor the conductance of the smaller K⁺ ion over that of the larger Rb and NH₄ ions. It is also possible that electrostatic effects are involved. According to the x-ray crystal model of the related KcsA K⁺ channel (Doyle et al. 1998), the 3P position is located at the COOH terminus of the pore helix, the end which points toward the pore (Fig. 8). Because an α helix has an associated macro dipole (Hol 1985), changes at the 3P position may change the dipole and thereby affect the electric field within the pore. A stronger field might lower a critical energy barrier for K⁺ more than that for the larger cations.

The selectivity of ROMK2 for NH₄⁺ versus K⁺ is remarkably different for permeability and conductance ratios. NH₄ permeability was 10-fold smaller than that of K⁺, but the conductance was 1.6-fold larger (Fig. 6 and Table). To explain these contradictory results, we examined kinetic models of ion permeation. The purpose of this exercise was not to precisely determine energy profiles, but rather to check whether the simple model can account for the observed results. We used three-site four-barrier models because the x-ray diffraction of a bacterial K⁺ channel crystal showed three discrete electron densities along the pore in single-file manner (Doyle et al. 1998). Also, the three-site four-barrier model gives a better fit for NH₄⁺ I-V data than the two-site three-barrier model (although two-site three-barrier model gives the same general conclusion).

First, we obtained depths and heights of energy wells and peaks for the K⁺ ion energy profile that predicted the I-V relationship for the K⁺ ion permeation. For simplicity, we used equally spaced wells and barriers. This is consistent with the resistance divider ratio for ROMK discussed above if the outer two wells and three barriers, which should cover two thirds of the electric field across the pore, correspond to the outer resistor. To reduce the number of free parameters, the depths of outer and middle wells were set to be equal (Fig. 9 A). The theoretical I-V curve is based on an energy profile that accurately, but not uniquely, simulated the experimental data. Next, we fixed the parameters for K⁺ and changed the energy profile parameters for NH₄⁺ to fit the I-V curve for the bi-ionic condition of NH₄⁺ outside and K⁺ inside the cell (Fig. 9 B). Again, the depths of outer and middle wells for the NH₄⁺ energy profile were set to be equal. The best-fit energy profile suggests that NH₄⁺ has significantly lower affinities than does K⁺ for the binding sites within the pore, as indicated by the shallower energy wells for NH₄⁺ than for K⁺ in the diagram. The higher affinity for K⁺, particularly of the energy well closest to the cytoplasmic side, mainly accounts for the higher K⁺ permeability since this site will be predominantly occupied by K⁺ at membrane potentials positive to −60 mV (Fig. 7 B), preventing NH₄⁺ flow through the channel. On the other hand, the hyperpolarized potentials associated with inward current would prevent occupancy of the channel by K⁺ and favor occupancy by NH₄⁺. Under these conditions, the inward NH₄⁺ conductance is large because the middle energy barriers to NH₄⁺ movement are lower than those to K⁺, increasing its mobility within the pore. We conclude that simple energy barrier models can explain the unusual selectivity pattern of K⁺ and NH₄⁺.

In summary, the differences in K⁺ conductance among ROMK, IRK, and various chimeras can be roughly explained by a model with two resistors in series, composed of the extracellular region and the COOH terminus. The NH₄⁺ and Rb⁺ selectivity differences between ROMK2 and IRK1 depend on the extracellular loop region, and specifically on the amino acid at the 3P position. On the other hand, the cytoplasmic termini seem to be at least partly the locus of Tl⁺ selectivity. The unusual selectivity pattern of K⁺ and NH₄⁺ can be explained by a simple energy barrier model.

This work was supported by National Institutes of Health grants DK27847 (L.G. Palmer) and DK46950 (H. Sackin).

Henry Lester served as guest editor.