Conformational changes at cytoplasmic intersubunit interactions control Kir channel gating

The defining structural feature of inward-rectifier potassium (Kir) channels is the unique Kir cytoplasmic domain. Recently we showed that salt bridges located at the cytoplasmic domain subunit interfaces (CD-Is) of eukaryotic Kir channels control channel gating via stability of a novel inactivated closed state. The cytoplasmic domains of prokaryotic and eukaryotic Kir channels show similar conformational rearrangements to the common gating ligand, phosphatidylinositol bisphosphate (PIP2), although these exhibit opposite coupling to opening and closing transitions. In Kir2.1, mutation of one of these CD-I salt bridge residues (R204A) reduces apparent PIP2 sensitivity of channel activity, and here we show that Ala or Cys substitutions of the functionally equivalent residue (Arg-165) in the prokaryotic Kir channel KirBac1.1 also significantly decrease sensitivity of the channel to PIP2 (by 5–30-fold). To further understand the structural basis of CD-I control of Kir channel gating, we examined the effect of the R165A mutation on PIP2-induced changes in channel function and conformation. Single-channel analyses indicated that the R165A mutation disrupts the characteristic long interburst closed state of reconstituted KirBac1.1 in giant liposomes, resulting in a higher open probability due to more frequent opening bursts. Intramolecular FRET measurements indicate that, relative to wild-type channels, the R165A mutation results in splaying of the cytoplasmic domains away from the central axis and that PIP2 essentially induces opposite motions of the major β-sheet in this channel mutant. We conclude that the removal of stabilizing CD-I salt bridges results in a collapsed state of the Kir domain.

The defining structural feature of inward-rectifier potassium (Kir) channels is the unique Kir cytoplasmic domain. Recently we showed that salt bridges located at the cytoplasmic domain subunit interfaces (CD-Is) of eukaryotic Kir channels control channel gating via stability of a novel inactivated closed state. The cytoplasmic domains of prokaryotic and eukaryotic Kir channels show similar conformational rearrangements to the common gating ligand, phosphatidylinositol bisphosphate (PIP 2 ), although these exhibit opposite coupling to opening and closing transitions. In Kir2.1, mutation of one of these CD-I salt bridge residues (R204A) reduces apparent PIP 2 sensitivity of channel activity, and here we show that Ala or Cys substitutions of the functionally equivalent residue (Arg-165) in the prokaryotic Kir channel KirBac1.1 also significantly decrease sensitivity of the channel to PIP 2 (by 5-30-fold). To further understand the structural basis of CD-I control of Kir channel gating, we examined the effect of the R165A mutation on PIP 2 -induced changes in channel function and conformation. Single-channel analyses indicated that the R165A mutation disrupts the characteristic long interburst closed state of reconstituted KirBac1.1 in giant liposomes, resulting in a higher open probability due to more frequent opening bursts. Intramolecular FRET measurements indicate that, relative to wild-type channels, the R165A mutation results in splaying of the cytoplasmic domains away from the central axis and that PIP 2 essentially induces opposite motions of the major ␤-sheet in this channel mutant. We conclude that the removal of stabilizing CD-I salt bridges results in a collapsed state of the Kir domain.
Eukaryotic inward-rectifier potassium (Kir) 3 channels are expressed in many tissues, including heart, skeletal muscle, kid-ney, cochlea, glia, neurons, retina, pancreas, small intestine, and stomach, and are essential for regulation of excitability and maintenance of cellular potassium homeostasis (1,2). Mutations of Kir channels have been associated with multiple diseases, ranging from cardiac failure to renal, ocular, pancreatic, and neurological abnormalities (3)(4)(5). All eukaryotic Kir channels require phosphatidylinositol 4,5-bisphosphate (PIP 2 ), a direct regulator of many other ion channels (6,7), for activation. Crystal structures of eukaryotic Kir channels in complex with PIP 2 have revealed the PIP 2 binding pockets (8,9) and have also indicated structural rearrangements associated with gating, including rigid-body motions of the cytoplasmic domain coupled to TM2 bending and/or formation of a tethering helix that pulls the cytoplasmic domain toward membrane surface (8,9).
Many disease-causing mutations are located in the cytoplasmic domain of Kir channels where multiple regulators, such as protons, sodium, ATP, and G-protein ␤␥ subunits bind (1, 2, 10 -14). However, molecular mechanisms that couple ligand binding to conformational changes and intersubunit interactions of the cytoplasmic domain are still unclear. As a bacterial homologue (15)(16)(17), KirBac1.1 is missing three critical residues in each of the linking strands between the cytoplasmic domain and the transmembrane domain (8,18,19). These strands are directly involved in PIP 2 binding in eukaryotic Kir channels (8,18,19). Hence KirBac1.1 has a distinct PIP 2 -binding pocket structure and differential gating responses to PIP 2 in comparison with eukaryotic Kir channels (20): eukaryotic Kir channels are opened by PIP 2 , but KirBac1.1 is closed in the presence of PIP 2 . Utilizing intramolecular FRET approaches, our previous work has revealed motions of the KirBac1.1 cytoplasmic domain associated with PIP 2 -dependent channel gating (12,22). This leads to a proposed gating mechanism that is consistent with models of channel gating based on KirBac3.1 and Kir3.2 crystal structures but with opposite gating consequences upon PIP 2 binding between the prokaryotic and eukaryotic channels (9,22,23).
We recently showed that conserved salt bridges in the cytoplasmic domain interface (CD-I) stabilize a PIP 2 -accessible conformation of eukaryotic Kir channels and that breaking these salt bridges results in stabilization of a PIP 2 -inaccessible "inactivated" closed state (24). In the present work we show that breaking of the CD-I salt bridge by the R165A mutation stabilizes open KirBac1.1 channels also by reducing apparent PIP 2 sensitivity. We examined the structural changes that accom-pany this mutation by intramolecular FRET measurements. The data lead us to conclude that CD-I stabilization by the Arg-165-Glu-202 salt bridge (or equivalent salt bridge in eukaryotic Kir channels) stabilizes a tight conformation of the tetrameric Kir channel domains that is PIP 2 -accessible, which results in higher open probability of eukaryotic Kir channels and lower activity of KirBac channels.

An arginine residue at the cytoplasmic intersubunit interface modified PIP 2 sensitivity of Kir2.1
As shown in crystal structures of Kir2.2 and Kir3.2 (8,9,25), Kir2.1 residue Arg-204, located in the interface between cytoplasmic domains (CD-I) forms a salt bridge with Glu-241 of the adjacent subunit. These charged residues are highly conserved throughout both prokaryotic and eukaryotic Kir channels (Fig.  1, A and B), and disruption of this salt bridge reduces the open probability (P o ) of Kir2.1 and Kir6.2 channels by induction of a PIP 2 -inaccessible inactivated closed state (26). At the ambient PIP 2 levels in membranes from cell lines such as COSm6 (27,28), recombinant Kir2.1 channels exhibit a high P o (ϳ0.95) that is only slightly enhanced by additional cytoplasmic PIP 2 (Fig.   1C). Kir2.1[R204A] mutant channels exhibit lower channel activity than wild type (WT) in excised patches but are then significantly more activated by similar exposure to PIP 2 ( Fig. 1, D and E), indicating a reduced apparent PIP 2 sensitivity and reflecting the stabilization of the inactivated state by this mutation (24).

The equivalent Arg-165 of prokaryotic KirBac1.1 also reduced sensitivity to PIP 2
In contrast to the activating effect of PIP 2 on Kir2.1, prokaryotic KirBac1.1 is strongly inhibited by PIP 2 , assessed using Rb efflux from liposomes with or without PIP 2 ( Fig. 2A) (19). The R165C mutation significantly diminishes KirBac1.1 channel inhibition by PIP 2 ( Fig. 2A), although sensitivity can be restored by modifying the cysteine with MTS reagents carrying positive, but not negative charged groups ( Fig. 2A). We examined the effect of additional substitutions at residue 165. Although introduction of the charge-maintaining R165K mutation did not affect PIP 2 sensitivity, neutralization to either R165A or R165C significantly reduced PIP 2 sensitivity (Fig. 2B). Although not fully resolved in the available KirBac1.1 crystal structure (16), structural alignments indicate that the side chain of Arg-  (16). A double cysteine (R165C/ E202C) mutant was constructed and purified to test potential for physical connection between residues Arg-165 and Glu-202 in KirBac1.1. Interestingly, the KirBac1.1 E202C mutant does not express a protein that could be purified, but expression is rescued by introducing the additional R165C mutation. In addition, WT KirBac1.1 runs almost completely as a monomer in SDS-PAGE, but the R165C/ E202C double mutant showed a clear dimeric band in oxidizing conditions (Fig. 2B) indicating formation of an intersubunit R165C-E202C disulfide bridge, reflecting close proximity of the two residues in the intact tetramer. Moreover, these multimer bands were more prominent in the presence of diC8-PIP 2 ( Fig. 2. B and C), implying that the Arg-165-Glu-202 bridge may be stabilized in the PIP 2driven closed state of the KirBac1.1 channel. Functional and structural conservation of the Arg-165-Glu-202 and equivalent salt bridges in both prokaryotic and eukaryotic Kir channels (24) allows us to investigate the structural role of intersubunit interactions in controlling Kir channel gating generally using KirBac1.1 as a model.

Stoichiometric effect of Arg-165-Glu-202 bridges on PIP 2 sensitivity of KirBac1.1
Kir channels are tetrameric proteins. To gain insight to the number of Arg-165-Glu-202 bridges within a tetramer that are required to maintain PIP 2 sensitivity, we generated tandem KirBac1.1 constructs, which permits introduction of different numbers of R165A mutations in different configurations within the functional channel. Gel filtration profiles show that all such concatemeric proteins exist primarily as tetramers (Fig. 3A). Tandem dimeric or tetrameric KirBac1.1 appear as dimer or tetramer bands, respectively, in SDS-PAGE (Fig. 3B) together with some lower oligomer bands presumably resulting from linker degradation. The tandem dimer KirBac1.1 channel, with two R165A mutations at diagonally opposed subunits within the tetramer, shows even higher PIP 2 sensitivity than WT, implying that a conformation with 2-fold symmetry, as is present in KirBac3.1 crystal structures, may favor channel closure (29). By contrast, a tandem tetramer construct with two adjacent R165A mutations became markedly less PIP 2 sensitive, and KirBac1.1 tetramers with either three or four R165A mutations (Fig. 3B) were equally PIP 2 insensitive (Fig. 3C). These results suggest that release of even one protomer from the Arg-165-mediated intersubunit interaction impaired PIP 2 sensitivity and release of two or more impaired sensitivity even more so.

The Arg-165-mediated intersubunit contacts stabilized the KirBac1.1 long-closed state
To uncover the kinetic basis of altered PIP 2 sensitivity in R165A mutants, we performed single-channel recordings on purified KirBac1.1 reconstituted into giant liposomes. Consistent with previous reports (15,22), single WT KirBac1.1 channels exhibited very heterogeneous gating behaviors, with multiple conductance levels (Fig. 4A). Openings occur in bursts, each characterized by rapid flickering transitions between multiple conductance levels and separated by prolonged interburst closures (Fig. 4A). Although the singlechannel conductance profile within the burst did not differ markedly from that of WT, R165A mutant channels exhibited shorter prolonged closures (Fig. 4, A and B). Detailed kinetic analysis indicates that the R165A mutation did not alter the mean closed time significantly (Fig. 4F), but it markedly destabilized the longest closed state (Fig. 4C), reducing the interburst duration Ͼ8-fold (Fig. 4, D and G).
Consequently, the R165A mutant demonstrated more frequent burst-openings than WT and, therefore, higher P o (Fig. 4E).

The R165A mutation induces structural changes that promote channel opening
We examined the conformational changes in channel structure that are induced in the presence and absence of PIP 2 on the R165A background using a macroscopic FRET approach (12). We introduced cysteine residues at different locations throughout the cytoplasmic domain on the KirBac1.1 R165A background and measured the FRET efficiencies between the fluorophore pair EDANS and DABCYL-plus covalently attached to purified channel proteins reconstituted into liposomes (Fig. 5, Table 1). Eight of 17 fluorophore-labeled cysteine mutants on the R165A background retained high channel activity (Fig. 5A), and all exhibited reduced PIP 2 inhibition compared with WT (Fig. 5, C  and D). We previously demonstrated that, on the WT background, the FRET-predicted intersubunit distances of labeled residues significantly correlated with those from the KirBac1.1 crystal structure (R ϭ 0.71, p Ͻ 0.001; Fig. 5C) (12), but a weaker correlation was revealed in the KirBac1.1[R165A] mutant (R ϭ 0.063, p Ͻ 0.804; Fig. 5C), with predicted C␣-C␣ distances being uniformly greater than on the WT background (Fig. 5D).
Previously we mapped PIP 2 -dependent structural arrangements of WT KirBac1.1 channel in liposomes (12). The results showed that PIP 2 induces rigid body motions of the cytoplasmic domain secondary structures, in particular a decrease in the tilt of the major ␤-sheet to align closer with the membrane normal, effectively opening the cytoplasmic vestibule (Fig. 6, C and F). However, in the R165A mutant channel background, the PIP 2 -induced conformational changes reported by FRET exhibited very different patterns (Fig. 6, A and D, Table 1). The major ␤-sheet still underwent what appears to be a rigid motion, but all residues in the major ␤-sheet now moved outward from the central axis in the presence of PIP 2 (Fig. 6, B and E), potentially underlying the failure of PIP 2 binding to promote channel closure in the mutant channel (see "Discussion").

Conserved CD-I salt-bridge roles in pro-and eukaryotic Kir channels
All Kir channels share highly conserved structure overall and are all highly sensitive to PIP 2 (7,19). All eukaryotic Kir chan-nels are activated by PIP 2 , whereas KirBac1.1 channels are inhibited. Previous studies (24,30) show that highly conserved salt-bridge pairs in the interface between cytoplasmic domains (CD-I) reduced apparent PIP 2 sensitivity of eukaryotic Kir channels by stabilizing a PIP 2 -inaccessible, inactivated closed state (24). In the present study we show that breaking this intersubunit interaction also reduces apparent sensitivity to PIP 2 in prokaryotic KirBac1.1, resulting in stabilization of a PIP 2 -inaccessible open state.

Structural consequences of CD-I disruption
Kir channel crystal structures have demonstrated that there are four binding pockets for PIP 2 in each channel (8,9), but stoichiometric requirements of PIP 2 for Kir channel gating and the molecular details of intersubunit interactions in coordinat- ing PIP 2 -induced conformational changes remained unclear.
Our results indicate that the existence of even one subunit with zero Arg-165-mediated interactions significantly attenuated PIP 2 sensitivity of the tetrameric KirBac1.1 channel (Fig. 3C). Because Arg-165 is located far from the location of PIP 2 binding, the change in PIP 2 sensitivity reflects a long-range consequence of structural disruption and may reflect the requirement for coordination between subunits in channel gating. Compared with WT channels, macroscopic FRET measurements (Fig. 5A) reveal a striking increase of apparent intramolecular distances of 10 -20 Å at almost all residues in the KirBac1.1[R165A] cytoplasmic domain, suggesting that the R165A mutation leads to a major dilation of the cytoplasmic domain. This dramatic result could be explained by a breaking apart of the CD-I. In this case the PIP 2 -binding site, located at the CD-I would be broken and would have to reform for PIP 2 to bind. Consistent with this suggestion, at almost every residue in the major ␤-sheet, PIP 2 induced a decrease in the predicted intramolecular distances. This is in marked contrast to the findings in the WT background background (12) (Fig. 6), as illustrated by the schematic in Fig. 7.
There are obvious parallels between the CD-I collapsed mechanism we propose and rupture of the ligand-binding domain dimer interface that mediates glutamate receptor desensitization (31). Recent cryo-EM structures of resting, activated, and desensitized glutamate receptor conformations sug- Mutants with substantially lower rubidium uptake than WT (Ͼ20%, gray) were not analyzed further. B, relative PIP 2 inhibition of E/D-labeled KirBac1.1 cysteine mutants on the R165A background. Relative PIP 2 inhibition was calculated from A and normalized to WT. All mutants were variably less sensitive than WT (green). C, C␣-C␣ distance between adjacent subunits of labeled residues assessed by FRET versus those determined in the KirBac1.1 (PDB code 2WLL) crystal structure. KirBac1.1 cysteine-substituted proteins on the WT or R165A background were purified, labeled, and reconstituted into liposomes (POPE:POPG ϭ 3:1) with or without 1% PIP 2 , apparent FRET efficiencies were measured, and C␣-C␣ distances were calculated using a tetrameric FRET model as that described previously (12). D, schematic illustration of significant structural changes induced by the R165A mutation in KirBac1.1 subunits. The C␣ of residues subjected to FRET measurements are highlighted as spheres, with those that move toward the pore axis colored blue and those that move away colored red upon the addition of PIP 2 or introduction of R165A background mutation. The major ␤-sheet uniformly moves away from the central axis (arrow) in the R165A background.
gest that the associated physical motions of domain rearrangements during desensitization are also extensive and involve loss of interface interactions between the ligand-binding domain as well as the amino-terminal domain dimers (32,33). Two dis-tinct eukaryotic Kir channel cytoplasmic domain conformations are apparent in available crystallographic structures (8,18). In the absence of PIP 2 , the PIP 2 -binding site is unstructured, whereas in the presence of PIP 2 the binding site is highly  structured, and the whole cytoplasmic domain is pulled closer toward the membrane surface. It is conceivable that the CD-I disrupted and compact configurations are equivalent to the apo (PDB code 3JYC) (18) and PIP 2 -bound (3SPI) (8) cKir2.2 structures, respectively. Intersubunit CD-I C␣-C␣ distances are longer in the apo configuration, which may, therefore, be favored by mutations that disrupt the salt bridges (Fig. 7), although further experimentation will be required to systematically assess this idea.

Functional consequences of CD-I disruption
Single channel recordings in KirBac1.1 (Fig. 4) demonstrated that the R165A mutation increased the channel P o by significantly disrupting PIP 2 -driven long closures. In eukaryotic Kir channels, the equivalent mutation resulted in decreased channel P o by inducing a long, PIP 2 -inaccessible, closed state (24). Separation of this state from PIP 2 -accessible closed states is most obvious in ATP-sensitive (Kir6.2) channels in which the mutation R192A, equivalent to R165A in KirBac1.1, leads to an inactivation phenotype: when ATP is removed, channels open but then rapidly and spontaneously inactivate to a closed state (24,26). A consistent kinetic model, in which CD-I disruption results in stabilization or appearance of novel inactivated state accessible from the non-PIP 2 -bound state can account for inactivation in Kir6.2 and reduced P o and enhanced PIP 2 sensitivity in Kir2.1 channels (24). By simply assuming that the gating states are opposite in KirBac channels (i.e. the PIP 2 -stabilized compact CD-I is closed and CD-I disruption stabilizes the open state) the behavior of KirBac1.1 channels are then also readily explained (Fig. 7).

DNA manipulation
KirBac1.1 and human Kir2.1 (Kir2.1) mutants in pQE60 or pcDNA3.1 vector were constructed using a site-directed mutagenesis kit (Agilent Inc.) (17); tandem KirBac1.1 dimeric constructs were made by inserting an additional copy of the cDNA between BamHI and HindIII sites of the WT construct, with an intermediate flexible GGGSGGGS linker; two tandem dimers were subcloned into pET-28a(ϩ) vector between NcoI-SacI and SacI-HindIII to make tandem tetrameric constructs.

Protein expression, purification, and labeling
KirBac1.1 was expressed and purified as previously described (17,34) from BL21-gold(DE3) host cells. For labeling KirBac1.1 cysteine mutants with EDANS c2 maleimide and DABCYLplus c2 maleimide (E/D, Anaspec), tetrameric fractions from size exclusion chromatography were pooled and concentrated to 1.0 mg/ml, then labeled with an E/D mixture at a protein/E/D ratio of 1:10:10 at room temperature for 1 h. Free fluorophores were then removed by gel filtration. Buffer containing 20 mM HEPES, 150 mM KCl, 5 mM decyl-maltoside, pH 7.5, was used for both labeling and gel filtration.

Rubidium flux assays and FRET measurements
Purified KirBac1.1 proteins were reconstituted into POPE: POPG (3:1) liposomes with or without PIP 2 at a protein/lipid ratio of 1:100. Rubidium flux assays and FRET measurements were performed on proteoliposome samples as described previously (12). For KirBac1.1, relative rubidium uptake of all mutants was normalized against WT protein in liposomes without PIP 2 . K i (PIP 2 concentration inhibiting 50% rubidium uptake) was obtained by fitting with a Hill equation using the Solver tool in Microsoft Excel. Free energy perturbations induced by different mutations on PIP 2 sensitivity of KirBac1.1 (⌬⌬G) were calculated using the equation, where R is the gas constant, T is temperature, and K i WT and K i R165X are the half-inhibitory PIP 2 concentrations for Kir-Bac1.1 WT and mutant, respectively.

Electrophysiology and kinetic analyses
Kir2.1 subunit cDNAs in pcDNA3.1 vector were transiently transfected into COSm6 cells, and excised voltage clamp recordings were performed as described previously (35,36). Single channel recordings of KirBac1.1 were performed on reconstituted channels in giant liposomes as described (15). Excised inside-out patch clamp recordings were performed with electrodes of 2-5 megaohms and pipette/bath solutions being KMOPS (158 mM KCl, 10 mM MOPS, pH 7.4). Singlechannel current responses were measured at Ϯ100 mV, and recordings were low-pass-filtered at 1 kHz, sampled at 10 kHz, and converted into digital files in Clampex7 for offline analysis using QuB kinetic software (QuB software). Recordings were digitally filtered to 0.4 kHz, and single-channel amplitudes were first estimated using the Baum-Welch re-estimation algorithm. Recordings were idealized using a 50% threshold method based on the corresponding single-channel amplitudes, which typically consisted of between 2 and 4 subconductance levels, with no imposed dead time. The corresponding idealization was evaluated for each recording by comparison to that of the unfiltered single-channel activity. State modeling and kinetic analyses were applied to idealized data using a maximum interval log-likelihood (MIL) algorithm (37,38). A simple linear kinetic model was constructed with the number of closed states based on maximum log-likelihood criteria (10 LL units/state) with an imposed dead time of 900 s determined by the digital filter frequency to correct for missed events (39). Interburst durations were determined based on the crit (critical time) value used to isolate bursts that consisted of Ͼ5 opening events and were terminated by closure lasting between 30 and 75 ms as previously described (21).

Data analysis
For correlation analysis, the C␣-C␣ distances between labeled residues at adjacent subunits were calculated from measured apparent FRET efficiencies using a tetrameric FRET model described previously (12).