The K⁺ channel signature sequence of murine Kir2.1: mutations that affect microscopic gating but not ionic selectivity

Abstract

1. We have studied the effects on ionic selectivity and gating of Kir2.1 of replacing Tyr (Y) in the GYG signature sequence with Phe (Y145F), Leu (Y145L), Met (Y145M), Ala (Y145A) or Val (Y145V).

2. The mutant Y145F showed no changes in ionic selectivity (as indicated by the permeability coefficient ratios P_Na/P_K or P_Rb/P_K), indicating that a hydrogen bond between Tyr and other residues is not essential for K⁺ selectivity. Y145L, Y145M, Y145A and Y145V did not express as monomers.

3. None of the channels made from covalently linked tandem dimers with wild-type and mutant subunits (WT-mutant) had altered ionic selectivity (P_Na/P_K or P_Rb/P_K), indicating that 4-fold symmetry is not required.

4. Macroscopic currents activated under hyperpolarization and the time constants for activation were reduced e-fold per 23 mV hyperpolarization in wild-type. This gating, believed to be due to the release of polyamines from the pore, was little affected by mutation of Y145. There was similarly little effect on the relationship between chord conductance (g_K) and membrane potential.

5. Unitary conductance (140 mm[K⁺]_o) was also little affected by mutation and was reduced only in channels formed from WT-Y145M, from 22.7 ± 0.4 pS (n= 5) in wild-type to 17.1 ± 0.5 pS (n= 4) in WT-Y145M.

6. Steady-state recording of unitary currents showed that channel open times were affected by the residue that replaced Tyr in GYG. Channel openings were particularly brief in WT-Y145V, where the mean open time was reduced from 102 ms at −120 mV in wild-type to 6 ms in WT-Y145V.

7. Thus in Kir2.1, GFG can act as a K⁺ selectivity filter, as can G(L/M/A/V)G, at least in dimers also containing GYG. Channel open time duration depended on the residue at position 145, consistent with the H5 region helping to determine the dwell time of the channel in the open state.

The principal selectivity filter of potassium channels is found in the pore (H5 or P) region, which contains a consensus sequence of amino acid residues, GYG, thought to be crucial for the regulation of K⁺ selectivity (Heginbotham et al. 1994). Certain mutations in voltage-gated K⁺ channels - deletion of YG (Heginbotham et al. 1992) or replacement of Y by certain other residues (Heginbotham et al. 1994) - result in the loss of ionic permeation or make channels non-selective among monovalent cations. Mutations of residues within H5 outside this immediate consensus may also alter selectivity (Yool & Schwarz, 1991; Taglialatela et al. 1993). In G-protein-activated inward rectifier K⁺ channels (Kir3.0), mutation of a Gly in the conserved GYG motif also results in the loss of K⁺ selectivity (Slesinger et al. 1996; Kofuji et al. 1996; Navarro et al. 1996).

Models of the selectivity filter, based on scanning cysteine accessibility mutagenesis, suggested that the aromatic side-chain of Tyr in GYG faced the pore (Lü & Miller, 1995; Dart et al. 1998b), offering the possibility that K⁺ selectivity is conferred by cation-π interactions (Kumpf & Dougherty, 1993). In a structural model of Kir2.1 (Dart et al. 1998b), selectivity was proposed as being conferred by two aromatic residues (Y145 and F147) via cation-π interactions and also by backbone carbonyl groups provided by T142 and G144.

However, the report by Doyle et al. (1998) of the crystal structure of a K⁺ channel (KcsA) from the bacterium Streptomyces lividans makes it unlikely that cation-π interactions play a significant part in K⁺ channel selectivity. The elegant results of Doyle et al. (1998) indicate that a group of residues, Thr-Val-Gly-Tyr-Gly, that includes the signature sequence orient their side-chains away from, rather than into, the pore, so that main-chain carbonyl oxygen atoms line the selectivity filter. These carbonyl oxygens must be held at precisely the distance required optimally to co-ordinate a K⁺ ion (see also Bezanilla & Armstrong, 1972). In KcsA, this precision is achieved partly through each Tyr residue in GYG forming a hydrogen (H-) bond with a Trp residue in a pore helix that is formed from the N-terminal end of H5. It is also partly achieved through van der Waals forces between residues of the selectivity filter on the one hand and of the pore helix and elsewhere on the other. An implication of this result is that 4-fold symmetry may be a requirement for K⁺ channel selectivity.

In the recently discovered family of tandem domain K⁺ channels (Kt), however, one of the two GYG sequences in the channel subunit may be replaced by GLG (e.g. TWIK-1; Lesage et al. 1996b) or GFG (e.g. TASK-1; Duprat et al. 1997). Other differences also exist between P1 and P2, the two H5 regions in each subunit. Thus, 4-fold symmetry is not required in these channels, nor is GYG required to be present in every H5 region for K⁺ selectivity.

In this paper, we attempted to test a number of hypotheses concerning the selectivity of Kir channels. Firstly, because Kir channels lack the Trp residues found in the pore helix of Kv (and KcsA; see Fig. 1A), we asked what interactions might stabilize the selectivity filter in Kir at its appropriate diameter and whether H-bonding is essential. We formed the hypothesis that H-bonding is essential for K⁺ selectivity by considering a homology model based on KcsA. We then tested the effects of replacing Tyr by Phe, which lacks the hydroxyl group required for the formation of a H-bond, on selectivity against Na⁺ and Rb⁺. In Kir6.0 (K_ATP; Inagaki et al. 1995a,b) and in K⁺ channels of the eag family (Warmke & Ganetsky, 1994), Phe replaces Tyr in the GYG triplet of residues. Our results show that H-bonding is not essential for ionic selectivity.

Figure 1. Structures of the selectivity filter of KcsA and Kir2.1.

A, sequences of the H5 region in the bacterial channel KcsA and in murine Kir2.1. The sequences of the pore helices are shown in italics and the K⁺ channel signature sequence (TxxTxGYG) in bold. Tyr and the residues with which it forms H-bonds are shown in red. B, the selectivity filter of KcsA (PDB accession code 1BL8; Doyle et al. 1998), illustrating the H-bonds involving the side-chain hydroxyl group of Y78. The carbon atoms of the four chains comprising the tetramer are coloured differently. H-bonds are shown as cyan lines - the continuous line denotes a H-bond that is probably conserved in Kir2.1 (see C) and the dotted line a H-bond that is not present in Kir2.1. C, the selectivity filter in our model of Kir2.1, illustrating the H-bond involving the side-chain hydroxyl group of Y145. The carbon atoms of the four chains comprising the tetramer are coloured differently. The probably conserved H-bond - involving the side-chain of T139, the first residue of the TxxTxGYG motif - is shown as a continuous cyan line.

Secondly, we asked whether 4-fold symmetry is a requirement for K⁺ channel selectivity in Kir. We tested the hypothesis that 4-fold symmetry is essential by using tandem dimeric channel constructs made as described previously (Dart et al. 1998a). We then replaced one of the two Tyr residues in GYG by Phe (Y145F), by Leu (Y145L, as in certain forms of Kt), by Met (Y145M, as in the vanilloid receptor VR1; Caterina et al. 1997), by Ala (Y145A) and by Val (Y145V). The results suggest that 4-fold symmetry is not required at the level of the GYG triplet. As a corollary, however, we found that channels in which every subunit had its GYG replaced by GLG, GMG, GAG or GVG appeared not to conduct K⁺.

Thirdly, since Kir channels are channels in which permeant K⁺ activates through displacing gating polyamine molecules from the channel pore (Lopatin et al. 1994), we asked whether these mutations affected such channel gating, as measured in macroscopic currents. The results showed little evidence of altered gating as a result of these mutations in H5.

Finally we asked whether unitary conductance and microscopic kinetics are altered by mutations in H5. We found that certain of our mutants produced striking reductions in channel mean open time.

A preliminary account of some of these results has been given (So et al. 1999).

METHODS

Computational chemistry

Structural models of Kir2.1 were constructed by comparative modelling (MODELLER; Sali & Blundell, 1993) using the crystal structure of KcsA (Doyle et al. 1998; Protein Data Bank (PDB) accession code 1BL8) as a template and the sequence alignment proposed by Minor et al. (1999). In addition to the restraints derived automatically by MODELLER, restraints were also applied to represent: 4-fold (wild-type and Y145F) or 2-fold (Y145L, Y145M, Y145A and Y145V) symmetry, and a salt bridge between E138 and R148 in adjacent subunits (Yang et al. 1997).

Molecular biological methods

The coding region of Kir2.1 cDNA (Stanfield et al. 1994) was cloned into the expression vector pcDNA3 as an EcoR I-Xho I fragment. Substitution mutations were generated by oligonucleotide-directed in vitro mutagenesis either by the method developed by Taylor et al. (1985) and supplied in kit form by Amersham Pharmacia Biotech, or by the method of Kunkel (1985). Mutations were verified by sequencing of the entire Kir2.1 cDNA. Ion channels were transiently expressed in Chinese hamster ovary cells (CHO-K1/SF, European Collection of Animal Cell Cultures) using the pFx-8 cationic lipid transfection reagent (Invitrogen) according to the manufacturer’s instructions. As a marker of transient transfection in CHO cells, plasmid DNA containing the cDNA for green fluorescent protein (Molecular Probes) was co-transfected with the Kir2.1 cDNA.

Covalently linked tandem dimers of wild-type and mutant Kir2.1 cDNAs (WT-mutant) were made as described by Dart et al. (1998a).

Electrophysiological methods

Membrane currents were measured by whole-cell or single channel recording using an Axopatch 200A amplifier (Axon Instruments). For whole-cell recording, records were filtered at 5 kHz (-3 dB, 8-pole Bessel), digitized at 10 kHz using either a Digidata 1200 or a TL-1 interface (Axon Instruments) and analysed on a 486 computer. Patch pipettes were filled with a solution containing 10 mm Hepes and 10 mm EDTA; pH adjusted to 7.2 with KOH; KCl added to bring [K⁺] to 140 mm. The extracellular solution contained (mm): KCl, 70; NaCl, 70; CaCl₂, 2; MgCl₂, 2; and Hepes buffer (pH 7.2), 10. NaCl replaced KCl when the K⁺ concentration was reduced in experiments to measure P_Na/P_K. In experiments to determine P_Rb/P_K, the extracellular solution contained (mm): KCl or RbCl, 70; N-methyl-d-glucamine (NMDG), 70; CaCl₂, 2; MgCl₂, 2; and Hepes, 10; titrated to pH 7.2 with HCl. The calculated junction potential between pipette (140 mm K⁺) and bath (70 mm K⁺) solutions used during sealing in experiments to obtain P_Na/P_K was 4.1 mV (pipette negative; using the program JPCalc, P. H. Barry, University of New South Wales). No correction was applied for this junction potential, since only changes in, rather than absolute values of, reversal potential were measured. In whole-cell recording we used 75 % compensation for series resistance errors. In measuring macroscopic kinetics results were used only from experiments where inward currents were less than 3 nA in amplitude at −132 mV. Thus the voltage error was less than 4 % at −132 mV with a series resistance of less than 6 MΩ, the electrode resistance being less than 4 MΩ.

For single channel recording using inside-out membrane patches, records were filtered at 2 kHz and digitized at 10 kHz. We used the intracellular, 140 mm K⁺ solution as the bath solution, while the patch pipette contained 140 mm K⁺ (mm: KCl, 140; CaCl₂, 2; MgCl₂, 2; and Hepes buffer (pH 7.2), 10).

Unitary current amplitudes were measured directly from the traces or by forming amplitude histograms of selected sections of recording containing clear single open and closed current levels. Gaussian distributions were fitted to the amplitude histograms to obtain the unitary current. Open probability, open time histograms and mean open times were analysed with a suite of programs developed using the AxoBASIC library (Axon Instruments). The open probability (P_o) was calculated either from amplitude histograms or using cursors set at 50 % of the open level according to the equation:

In this equation, t_j is the time when j channels are simultaneously open during a recording that lasts for time T and N is the total number of channels in the patch.

Experiments were carried out at room temperature (20-23 °C). Averaged results throughout this paper are given as means ±s.e.m. Student’s unpaired t test was performed, and P values of less than 0.05 were regarded as significant.

RESULTS

Possible H-bonding in the selectivity filter of Kir2.1

Figure 1B shows the selectivity filter of KcsA, with the two H-bonds in which Tyr in GYG (Y78 in KcsA, equivalent to Y145 in Kir2.1, Fig. 1A) participates. Y78 H-bonds with a Trp residue in the pore helix (W68) and with T72, the first residue in the signature sequence TxxTxGYG of an adjacent channel subunit (Fig. 1A and B). Since the Trp in KcsA is replaced by Phe (F135) in Kir2.1 (Fig. 1A), this second H-bond is likely to be the only one available to stabilize the selectivity filter of Kir2.1. Our model-building of Kir2.1 (Fig. 1C) confirms this prediction. To test whether this H-bond is required for selectivity in Kir2.1, we investigated five mutants, Y145F, Y145L, Y145M, Y145A and Y145L, where the residue at position 145 cannot form H-bonds.

Selectivity is not altered by replacement of Tyr in GYG by Phe

Of the mutants we used, only Y145F gave currents after expression of monomeric channel subunits. These Y145F channels showed inward rectification (Fig. 2A), with currents activating rapidly under hyperpolarizing voltage pulses. Currents also inactivated slowly from their maximum value, and this inactivation process took many seconds to reach completion. Selectivity against Na⁺ was tested by measuring the change in the reversal potential (V_rev) when [K⁺]_o was altered, being replaced by Na⁺ (Fig. 2B). In wild-type channels, V_rev changed 56.8 ± 3.7 mV (n= 4) per 10-fold change in [K⁺]_o. It changed 56.9 ± 2.0 mV (n= 3) for Y145F. The absence of any alteration indicates that Na⁺ is excluded with equal efficacy in wild-type and in the Y145F mutant.

Figure 2. Channels with GFG have similar properties to those with GYG.

A, membrane currents recorded from CHO cells transfected with wild-type or mutant (Y145F) forms of Kir2.1. [K⁺]_o= 70 mm; [K⁺]_i= 140 mm. Currents illustrated were recorded under depolarization and hyperpolarization in 20 mV steps up to 80 mV positive and down to 110 mV negative to the holding potential, −17 mV (equal to the K⁺ equilibrium potential, E_K). Voltage steps lasted 50 ms. B, current-voltage relationships measured in the presence of different [K⁺]_o, Na⁺ replacing K⁺ (ordinate, potassium current; abscissa, membrane potential). K⁺ concentrations were: 140 mm, □; 70 mm, •; and 35 mm, ▵. In both wild-type and Y145F, the reversal potential followed the expectations from Nernst (Table 1). C, current-voltage relationships measured in the presence of 70 mm K⁺ (•) and 70 mm Rb⁺ (□). P_Rb/P_K was unaltered by the mutation (Table 1).

Similarly there was no change in the selectivity against Rb⁺ (Fig. 2C). This selectivity was measured from the change in reversal potential (ΔV_rev) found when Rb⁺ replaced K⁺ in the external solution:

R, T, and F have their usual thermodynamic meanings and P_Rb and P_K are permeability coefficients (see for example Aidley & Stanfield, 1996). With the wild-type channel, the reversal potential changed by -10.2 ± 1.6 mV (n= 4), giving P_Rb/P_K= 0.67 ± 0.04. There was no change in the mutant Y145F, with P_Rb/P_K= 0.67 ± 0.03 (n= 3). Thus although an aromatic residue may be required in position 145 in at least some subunits (see also below), a H-bond between Y145 and T139 cannot be a requirement for K⁺ selectivity. This is in contrast to our modelling studies, which suggest that this H-bond is essential for maintaining the structure of the selectivity filter. Only in the models of wild-type Kir2.1 is the structure of the selectivity filter similar to that in KcsA; the carbonyl oxygens move significantly away from this position in models of all mutants (see Discussion).

is 4-fold symmetry required for k⁺ selectivity?

We have used covalently linked dimers in which only one of the two coding regions contained the mutant cDNA Y145F, Y145L, Y145M, Y145A or Y145V. Channels formed in this way will have Y145 mutated in only two of the four subunits, since we have shown that these dimers fold so that two form each channel (Dart et al. 1998a). All the dimeric constructs used gave currents after expression in CHO, and all showed inward rectification (Fig. 3). In the WT-Y145V mutant, current noise was noticeable at negative voltages (Fig. 3D; see also single channel recording in Fig. 6 and 8).

Figure 3. Currents in covalently linked tandem dimers.

Membrane currents recorded from CHO cells transfected with mutant, tandem dimers. The mutants were WT-Y145M (Y-M, A), WT-Y145F (Y-F, B), WT-Y145L (Y-L, C) and WT-Y145V (Y-V, D). The mutants are illustrated in ascending order of side-chain hydrophobicity at position 145 (Kyte & Doolittle, 1982). We also formed dimers from WT-Y145A (not illustrated). [K⁺]_o= 70 mm; [K⁺]_i= 140 mm. Currents illustrated were recorded under depolarization and hyperpolarization in 10 mV steps up to 80 mV positive and down to 110 mV negative to the holding potential, −17 mV (equal to E_K).

Figure 6. Single channel recording in wild-type and in dimeric constructs.

A, single channel currents were recorded using inside-out patches. [K⁺]_o= 140 mm; [K⁺]_i= 140 mm; membrane potential =−120 mV. The mutants are illustrated in ascending order of side-chain hydrophobicity at position 145. The channel opens for a long time in WT (a), WT-Y145A dimer (b), WT-Y145F dimer (d) and WT-Y145L dimer (e), whereas there are briefer openings in WT-Y145M dimer (c) and quite different channel kinetics in WT-Y145V dimer (f). O, open state(s); C, closed state. B, dwell time histogram for openings at −120 mV. The abscissa gives the dwell time (log scale), the ordinate the occurrence of events (square root scale). The continuous line represents a single exponential best fit (using maximum likelihood) with a time constant of 102 ms in wild-type (a) and 6 ms in WT-Y145V dimer (b).

Figure 8. Substates in Y145 mutants.

The single channel currents were recorded in inside-out patches. [K⁺]_o= 140 mm; [K⁺]_i= 140 mm. O, open state; C, closed state; S, substate. The percentage amplitude of the conductance substate was obtained by dividing the amplitude of the substate by that of the open state. The unitary current amplitudes of the substate were measured directly from the traces. A, WT-Y145F dimer; B, WT-Y145M dimer; C, WT-Y145L dimer; D, WT-Y145M dimer. E, voltage dependence of substates.

The single channel currents were recorded in inside-out patches. [K⁺]_o= 140 mm; [K⁺]_i= 140 mm. O, open state; C, closed state; S, substate. The percentage amplitude of the conductance substate was obtained by dividing the amplitude of the substate by that of the open state. The unitary current amplitudes of the substate were measured directly from the traces. A, WT-Y145F dimer; B, WT-Y145M dimer; C, WT-Y145L dimer; D, WT-Y145M dimer. E, voltage dependence of substates.

As Table 1 and Fig. 4 show, none of the mutants differed from wild-type in the response to changes in [K⁺]_o - all showed a Nernstian response to K⁺ concentration changes when Na⁺ was substituted for K⁺ (Table 1). Thus the removal of 4-fold symmetry in the selectivity filter at the level of GYG does not alter the ability of the channels to select against Na⁺.

Table 1.

Channel properties of Tyr mutants in the GYG signature sequence

	Dependence of Vrev on [K+]o (mV per 10-fold change)	PRb/PK	IRb/IK (-132 mV)	V1/2 (mV)	k (mV)	Dependence of τact on Vm (e-fold) *	τact at −52 mV (ms)	Single channel conductance (pS)
WT (Y145)	56.8 ± 3.7 (4)	0.67 ± 0.04 (5)	0.087 ± 0.04 (5)	−23.4 ± 1.1 (4)	8.9 ± 0.5 (4)	23	0.53 ± 0.09 (3)	22.7 ± 0.4 (5)
Y145F	56.9 ± 2.0 (3)	0.67 ± 0.03 (3)	0.086 ± 0.04 (3)	−19.6 ± 0.4 (3) †	6.2 ± 0.3 (3) †	16	0.28 ± 0.003 (3)	23.6 ± 1.5 (4)
Y–F dimer	56.8 ± 1.9 (4)	0.69 ± 0.03 (3)	0.092 ± 0.02 (3)	−24.1 ± 0.9 (6)	8.8 ± 0.7 (6)	23	0.63 ± 0.02 (6)	23.0 ± 1.1 (3)
Y–L dimer	57.6 ± 3.5 (3)	0.72 ± 0.03 (3)	0.097 ± 0.05 (3)	−24.1 ± 1.3 (6)	7.6 ± 0.4 (6)	22	0.59 ± 0.08 (4)	23.3 ± 0.7 (3)
Y–M dimer	57.7 ± 1.3 (3)	0.68 ± 0.06 (3)	0.088 ± 0.03 (3)	−22.5 ± 0.7 (4)	8.5 ± 0.9 (4)	24	0.43 ± 0.07 (4)	17.1 ± 0.5 (4) †
Y–V dimer	57.2 ± 0.5 (3)	0.68 ± 0.04 (3)	0.083 ± 0.02 (3)	−24.4 ± 0.4 (5)	7.5 ± 0.2 (5) †	20	0.69 ± 0.02 (3)	23.6 ± 0.5 (3)

Averaged values are given as means ± s.e.m. with the number of experiments in parentheses.

^*Values were obtained by fitting the mean value of results from 3–6 cells. V_1/2 and k are defined in eqn (1) of the text: V_1/2 is the voltage at which the K⁺ conductance, g_K, is 0.5 of its maximum value and k is a steepness factor.

^†P< 0.05.

Figure 4. Nernstian response to changes in [K⁺]_o.

A, membrane currents recorded from CHO cells transfected with a mutant (WT-Y145L dimer) form of Kir2.1. [K⁺]_i= 140 mm; [K⁺]_o= 140 mm (a) and 35 mm (b). Currents illustrated were recorded under depolarization and hyperpolarization in 10 mV steps from the holding potential at E_K (0 mV in a;−35 mV in b). B, current-voltage relationships measured in the presence of different [K⁺]_o, Na⁺ replacing K⁺ (ordinate, potassium current; abscissa, membrane potential). Dimeric constructs were used as follows: WT-Y145M (a), WT-Y145F (b), WT-Y145L (c) and WT-Y145V (d). K⁺ concentrations were: 140 mm, □; 70 mm, •; and 35 mm, ▵.

Nor is 4-fold symmetry required for selectivity against Rb⁺, as measured from the shift in V_rev. Replacement of K⁺ by Rb⁺ showed P_Rb/P_K to be unaltered in any of the mutant dimers used (Fig. 5, Table 1). Rb⁺ is known to carry little current through Kir2.1 (Abrams et al. 1996; Thompson et al. 2000). This property depends on residues deeper than GYG (Thompson et al. 2000) and is, as expected, little altered by mutations in GYG (Figs 2C and 5, and Table 1).

Figure 5. Relative permeability to Rb⁺ in dimeric constructs.

A, membrane currents recorded from CHO cells transfected with a mutant (Y-L dimer) form of Kir2.1. [K⁺]_i= 140 mm; [K⁺]_o= 70 mm (a) or [Rb⁺]_o= 70 mm (b). Currents illustrated were recorded under depolarization up to 63 mV or hyperpolarization down to −127 mV in 10 mV steps from the holding potential of −17 mV. B, current-voltage relationships measured in the presence of 70 mm K⁺ (•) and 70 mm Rb⁺ (□). The dimeric constructs used were: WT-Y145M (a), WT-Y145F (b), WT-Y145L (c) and WT-Y145V (d).

Mutations in GYG have little effect on channel macroscopic kinetics

We investigated channel macroscopic kinetics by measuring the relationships between voltage and both K⁺ chord conductance and rates of activation under hyperpolarization. For each mutant channel, the chord conductance, g_K, was computed from the equation:

where E_K is the potassium equilibrium potential.

g_K was normalized to its maximum value and the normalized conductance, g_K′, was plotted against voltage. The relationship was fitted with a single Boltzmann equation, of the form:

(1)

where V_1/2 is the voltage at which the normalized conductance is 0.5, and k is a steepness factor. The values found for V_1/2 and k are given in Table 1. Of the mutants examined, only Y145F and the WT-Y145V dimer showed a value for k that was slightly reduced from that in wild-type.

Hyperpolarization from E_K of wild-type and mutant channels produced inward currents that increased with time (see Fig. 2A and 3), a process that became more rapid with increasing hyperpolarization. At −52 mV, the time constant for activation (τ_act) was 0.52 ± 0.09 ms (n= 4) in wild-type and rates increased e-fold with a 23 mV hyperpolarization. Little change from these values was found in any of the mutants (Table 1). These results suggest that the H5 region, including GYG, has relatively little effect on channel gating, which we presume to be by polyamines (Lopatin et al. 1994). Previously we have shown that gating is not materially altered in mammalian cells used in expression systems after long periods of whole-cell or outside-out patch recording (Stanfield et al. 1994); apparently polyamines are difficult to wash out of mammalian cells. In the face of so little effect of mutation on channel gating, we have not attempted to measure polyamine affinity directly.

Unitary conductance is little altered by mutations in GYG

If the GYG motif lines the most intimate part of the pore, it may be expected that mutations here will change both selectivity and unitary conductance (γ). We have already shown that none of our mutations altered selectivity. Single channel currents are shown in Fig. 6. Values obtained for γ were 22.7 ± 0.4, 23.6 ± 1.5, 23.0 ± 1.1, 23.3 ± 0.7, 23.6 ± 0.5 and 17.1 ± 0.5 pS for the WT, Y145F, WT-Y145F, WT-Y145L, WT-Y145V and WT-Y145M channels, respectively (Table 1). The WT-Y145M dimer, and only the WT-Y145M dimer, showed a lower conductance compared with wild-type and with the other mutant channels.

Mutations in GYG are associated with changes in unitary kinetics

We did, however, observe an effect of replacement of Tyr in GYG on channel kinetics. In Fig. 6A, traces from single channel recording of wild-type and five mutant channels are shown. Qualitative kinetic differences can be seen between the WT-Y145L and WT-Y145F dimers and the WT-Y145V and WT-Y145M dimers, where channel openings are briefer. We obtained the open time distribution by setting a cursor half-way between the open and closed levels. Conductance substates, which were observed in mutant as well as in the WT channels, were considered as open for the purposes of determining the open time distribution. Mean open times were then calculated by fitting the distribution with a single exponential function (Fig. 6B).

Microscopic kinetics, measured under steady-state conditions in excised, inside-out membrane patches, in the absence of polyamines, are unlikely to reflect the activation of channels that occurs under hyperpolarization owing to the release of these blocking cations. The open time and steady-state P_o fall, rather than increase, as the membrane becomes hyperpolarized, and these changes are measured over a voltage range where channel activation would be complete. Rather, the kinetic behaviour reflects an inactivation process that occurs at negative voltages and may be associated with changes in the conformation of H5 and possibly other regions of the channel (see Choe et al. 1999). Thus for example in wild-type, mean open time was 150 ms at −100 mV but was only 60 ms at −160 mV.

The mean open times in Y145F and in the WT-Y145F and WT-Y145L dimers were greater than in WT in the voltage range −120 to −160 mV (Figs 6 and 7A). On the other hand, the mean open times in the WT-Y145M and WT-Y145V dimers were shorter than that in WT over the voltage range explored (-100 to −160 mV; Fig. 7A). This reduction was most marked with WT-Y145V, where the open time was 6 ms at −120 mV, compared with 102 ms in wild-type (Figs 6B and 7A). But the reduction in mean open time did not correlate with the size of the side-chain of the residue involved; in the Y145A dimer (Fig. 6Ab), mean open times were similar to those seen in wild-type. Nor did it appear to follow side-chain hydrophobicity, which is greatest for Val and least for Met and Ala among the substitutions made. We analysed closed time distributions in two of our single channel patches. In one with wild-type channels, at −100 mV, there were three populations of closures, with mean closed durations (areas of closed time probability density function (p.d.f.)) of 0.2 ms (0.40), 16.2 ms (0.44) and 715 ms (0.16). With the Y145V dimer, the mean closed time durations were: 0.2 ms (0.33), 1.56 ms (0.66) and 608 ms (0.01).

Figure 7. Voltage dependence of mean open time and open state probability.

A, voltage dependence of mean open time in Y145 mutants (n = 3-8 experiments). Following Choe et al. (1999), the relationships were fitted with the expression:

where α and β are constants; α is the value of τ at 0 mV and β is a fraction determining the voltage dependence of τ. Lines show the best fit with α= 670, 273 and 37 ms and β= 0.4 for WT, WT-Y145M dimer and WT-Y145V dimer, respectively. B, voltage dependence of open probability in Y145 mutants (n = 3-6 experiments). Following Choe et al. (1999), the relationships were fitted with the expression:

where V′ is the half-point of the Boltzmann expression and k is a steepness factor. The lines show the best fit with P_max= 1, P_min= 0.2, 0.6 and 0.2, V′= -128, −126 and −106 mV, and k= 28, 39 and 49 mV for WT, WT-Y145M dimer and WT-Y145V dimer, respectively.

We considered briefly whether the changes found were due to changes in affinity for blocking ions, such as Ba²⁺, present at trace concentrations in the extracellular solutions. This explanation has been given by Choe et al. (1998, 1999) for the reduction in steady-state P_o at negative voltages, though the presence of traces of Ba²⁺ affected primarily the occupancy of a long-lived closed state, rather than mean open time. We therefore carried out a few measurements using solutions without divalent cations and with EDTA (5 mm) to remove traces of divalent cations such as Ba²⁺. While a slight increase in mean open time was found in the presence of EDTA, substantially lower mean open times were still found in WT-Y145V under these conditions (results not illustrated). Choe et al. (1998) have shown that the principal effect of EDTA in the external solution was to remove the longest population of closures.

We have also calculated steady-state P_o (Fig. 7B). In wild-type, P_o decreased from 0.86 at −60 mV to 0.35 at −160 mV (Choe et al. 1999). P_o showed a similar voltage dependence in the WT-Y145V dimer, decreasing from 0.78 at −40 mV to 0.37 at −160 mV. Other mutants showed less voltage dependence than WT and the WT-Y145V dimer. The reduction in P_o with negative voltage was largely removed by the presence of EDTA, as described by Choe et al. (1998, 1999).

Finally, we have examined the conductance substates recorded in the WT-Y145F, WT-Y145M and WT-Y145L dimers in greater detail (Fig. 8). Conductance substates have previously been recorded in wild-type Kir2.1, and these appear to have levels 1/3 and 2/3 of the fully open state and were associated with the action of pore blockers such as Mg²⁺ (Matsuda, 1988; see also Oishi et al. 1998). In our mutant channels in the absence of internal Mg²⁺, the level of the most prominent substate found was closer to the 2/3 level, but varied both with the voltage and with the residue that was substituted for Tyr (Fig. 8E). The amplitude of the substate in wild-type showed voltage dependence; the amplitude of the substate relative to that of the fully open state (%) increased from 58 % at −60 mV to 83 % at −160 mV and the difference was significant. The relative amplitude of the substate showed a similar voltage dependence in the WT-Y145F dimer. In other mutants, there was less voltage dependence of the relative substate amplitude. The pattern of voltage dependence of the amplitude of the substates in the mutant channels was therefore complex (Fig. 8). However, the variation with voltage in the substate levels in both the wild-type and mutant channels and the lack of frequent occupancy of a substate at around 1/3 of fully open rules out the triple barrel hypothesis of Matsuda (1988) as an explanation for the substates we observed. The variation with voltage also suggests that the origin of the conductance substates described here may be more complex than in the model proposed by Oishi et al. (1998). The presence or absence of their substates was determined only by the presence or absence of a negatively charged residue (D172 in wild-type) in M2 (Oishi et al. 1998).

DISCUSSION

In this study we carried out experiments on channels in which the Tyr residue in the GYG motif was replaced by one of a number of residues: F, L, M, A and V. We found that functional channels could be formed from monomers when Phe was substituted for Tyr. Since the only putative H-bond stabilizing the selectivity filter of Kir2.1 is between Y145 and T139 (Fig. 1) and is absent from the Y145F mutant, which lacks the side-chain hydroxyl group, a H-bond cannot be essential for maintaining the diameter of the selectivity filter. Nor can the hydroxyl group in Tyr be of major importance in co-ordinating K⁺ as Ranatunga et al. (1998) have proposed.

The modelling of Kir2.1 had suggested that the H-bond between the side-chain hydroxyl group on Y145 and T139 was essential to constrain the structure of the selectivity filter in the correct geometry. However, the experimental results for Y145F suggest that van der Waals forces are in fact sufficient to maintain this geometry. This difference probably arises because the low sequence homology between KcsA and Kir2.1 means that these van der Waals forces are not predicted with sufficient accuracy to retain the correct structure.

While van der Waals interactions are probably sufficient to stabilize the selectivity filter in the Y145F mutant, it seems likely that they are no longer strong enough to retain the wild-type-like structure in the other mutants (Y145L/M/A/V) that apparently did not function as ion channels. This in turn suggests a possible reason why dimeric wild-type-mutant constructs were essential for function with mutants that lack an aromatic side-chain.

Similar results have recently been found in KcsA channels (Splitt et al. 2000). Here too, channels formed with GFG have characteristics indistinguishable from those of wild-type KcsA, while channels formed from subunits containing GAG did not show K⁺ channel activity (and indeed did not form a tetramer). In the Shaker channel, a homotetramer with GVG replacing GYG, significant K⁺ selectivity is retained (Heginbotham et al. 1994).

We tested whether 4-fold symmetry is required in the selectivity filter at the level of GYG for K⁺ channel selectivity. We found that this was not the case, since functional channels could be formed from dimers where one Tyr in GYG had been replaced. All replacements we attempted produced such functional channels; the substitutions we used were GLG, GMG, GAG and GVG. The lack of a requirement for Tyr in all positions also makes it unlikely that cation-π interactions determine selectivity as Kumpf & Dougherty (1993) suggested (see also Lü & Miller, 1995; Dart et al. 1998b). The difference in energy level associated with an interaction with the faces of only two as opposed to four aromatic side-chains is likely to be reflected in a significant change in selectivity if this interaction were its principal determinant. These results are, therefore, more consistent with the mechanism of selectivity proposed by Bezanilla & Armstrong (1972) and demonstrated for KcsA by Doyle et al. (1998).

Our results with dimers are reminiscent of the situation found in the emerging family of tandem pore K⁺ channels. These channels contain two P-regions per subunit instead of the single such region found in voltage-gated and inward rectifier K⁺ channel subunits (Lesage et al. 1996a,b; Fink et al. 1996, 1998; Duprat et al. 1997; Reyes et al. 1998; Chavez et al. 1999; Salinas et al. 1999). Often only one P-region has the conserved GYG triplet, whereas the other has instead GFG or GLG. TREK-1 has two GFGs (Fink et al. 1996).

It is noteworthy that Kir3.0 channels, which naturally form as heteromers, have asymmetrical selectivity filters. For example, although both contain the GYG triplet, Kir3.1 and 3.4 have different amino acid sequences in parts of H5. Surprisingly, this asymmetry is necessary for high K⁺ selectivity (Silverman et al. 1998), since channels formed from Kir3.4 alone have measurable permeability to Na⁺ and Ca²⁺.

Changes to the structure of H5 produce only small changes in macroscopic kinetics. Such changes as occur are likely to be the consequence of minor, allosteric changes in channel structure.

As perhaps expected from the lack of change in selectivity in our various dimeric constructs, most of the mutants we studied failed to show a change in unitary conductance. This is consistent with Navaratnam et al. (1995) and Repunte et al. (1999), who have shown that residues in the M1-H5 and H5-M2 extracellular linkers (outside the GYG motif) are important in determining the unitary conductance of inward rectifiers.

However, microscopic or single channel kinetics was altered by replacement of Tyr in GYG, and this alteration was in some cases substantial. In particular, channel open times were reduced significantly in the WT-Y145M and WT-Y145V dimers. Although single channel kinetics are influenced by the presence of trace amounts of Ba²⁺ (Choe et al. 1999), the change in mean open time found here does not appear to be associated with a change in affinity for Ba²⁺, since it persisted in the presence of 5 mm EDTA.

Using Kir2.1-ROMK2 chimeric channels, Choe et al. (1999) have shown that the extracellular segment located between M1 and M2 is an important determinant of channel kinetics. The short openings seen in ROMK2 are conferred onto Kir2.1 by the swapping of this extracellular loop (including the H5 region). Choe et al. (1999) have argued that a conformational change in this region (and possibly in M2) underlies the gating seen in the steady state under single channel recording. Our results are consistent with the occurrence of such a conformational change - a change that is, perhaps, somewhat akin to the C-type inactivation found in Kv channels (Lopez-Barneo et al. 1993).