Identification of Novel Cholesterol-binding Regions in Kir2 Channels*

Background: Cholesterol modulates inwardly rectifying potassium (Kir) channels. Results: Using a combined computational-experimental approach, we identified two putative cholesterol-binding regions in Kir2.1 that suggest the existence of a novel cholesterol binding motif. Conclusion: Cholesterol binds to nonannular surfaces in the transmembrane domain of Kir2.1. Significance: These findings provide new insights into the mechanisms underlying lipid regulation of ion channels. Inwardly rectifying potassium (Kir) channels play an important role in setting the resting membrane potential and modulating membrane excitability. We have recently shown that cholesterol regulates representative members of the Kir family and that in the majority of the cases, cholesterol suppresses channel function. Furthermore, recent data indicate that cholesterol regulates Kir channels by specific sterol-protein interactions, yet the location of the cholesterol binding site in Kir channels is unknown. Using a combined computational-experimental approach, we show that cholesterol may bind to two nonanular hydrophobic regions in the transmembrane domain of Kir2.1 located between adjacent subunits of the channel. The location of the binding regions suggests that cholesterol modulates channel function by affecting the hinging motion at the center of the pore-lining transmembrane helix that underlies channel gating either directly or through the interface between the N and C termini of the channel.

Inwardly rectifying potassium (Kir) channels play an important role in setting the resting membrane potential and modulating membrane excitability. We have recently shown that cholesterol regulates representative members of the Kir family and that in the majority of the cases, cholesterol suppresses channel function. Furthermore, recent data indicate that cholesterol regulates Kir channels by specific sterol-protein interactions, yet the location of the cholesterol binding site in Kir channels is unknown. Using a combined computational-experimental approach, we show that cholesterol may bind to two nonanular hydrophobic regions in the transmembrane domain of Kir2.1 located between adjacent subunits of the channel. The location of the binding regions suggests that cholesterol modulates channel function by affecting the hinging motion at the center of the pore-lining transmembrane helix that underlies channel gating either directly or through the interface between the N and C termini of the channel.
In recent years, a growing number of studies have implicated cholesterol as a major regulator of ion channel function. An increase in membrane cholesterol has been shown to regulate different types of K ϩ channels, Ca 2ϩ channels, Na ϩ channels, and Cl Ϫ channels, as described in detail in several recent reviews (1)(2)(3)(4). The most common effect of cholesterol on ion channels is a decrease in channel activity that may be due to a decrease in the open probability, in the unitary conductance, and/or in the number of active channels in the membrane.
Here we study the mechanism underlying cholesterol regulation of Kir2 channels, which we have shown earlier to be suppressed by increased levels of membrane cholesterol (5)(6)(7)(8). The members of this subfamily of constitutively active, strongly inwardly rectifying K ϩ (Kir) 4 channels are critically involved in regulation of the excitability and contraction in a variety of cell types and in maintaining the membrane potential in several types of non-excitable cells (9 -11).
Earlier studies have suggested that cholesterol regulates ion channel function by altering the physical properties of the lipid bilayer. Accumulating evidence indicates, however, that cholesterol may also regulate ion channel function via direct binding to the channel proteins. Several earlier studies demonstrated that specific sterol-protein interactions regulate the function of the nicotinic acetylcholine receptor (12)(13)(14)(15). Our studies demonstrated that specific sterol-protein interactions are responsible for the suppression of Kir channels. First, we showed that cholesterol and its optical stereoisomer, epicholesterol, have opposite effects on endothelial Kir current (5) and that there is no correlation between the effects of different sterols on membrane fluidity and on the activity of a purified bacterial analog of Kir channels, KirBac1.1, incorporated into a pure lipid environment (liposomes) (16). Consistent with these observations, it was shown more recently that ent-cholesterol, an enantiomer of cholesterol (17), has no effect either on KirBac1.1 or on Kir2.1 function (18). Differential effects of cholesterol analogues on channel function were also demonstrated in several other types of ion channels, including the nicotinic acetylcholine receptor (19), BK channels (20), and TRPV1 channels (21). Most recently, our laboratory has shown that cholesterol binds to purified KirBac1.1 channels and that cholesterol-KirBac1.1 binding is essential for the inhibitory effect of cholesterol on channel activity (22).
Our earlier studies have identified a series of residues in the C and N termini of the channel that are crucial for the sensitivity of Kir2 channels to cholesterol (23)(24)(25). These residues form a distinct cytosolic structure, a cholesterol sensitivity belt that surrounds the cytosolic pore of the channel in proximity to the transmembrane (TM) domain (24). In addition, we also identified functional links between a central residue of the cholesterol sensitivity belt and residues in the N terminus and distant C terminus (26,27). However, based on computational analysis, none of these residues form a cholesterol-binding region in Kir channels.
Recent studies have shown that a well known cholesterol binding motif, the cholesterol recognition amino acid consensus (CRAC) motif (28,29), is found in TRPV1 channels (21) and in BK channels (30), representing putative cholesterol binding sites. CRAC motifs are also found in several types of AChR, but it has been suggested that it is not the CRAC itself but rather its inverted sequence, CARC, that may be responsible for cholesterol interactions with AChR (31). However, sequence analysis shows that Kir2.1 channels have no CRAC motifs with an exception of a cytoplasmic segment that is highly energetically unfavorable for cholesterol binding. Kir2.1 channels also have no cholesterol consensus motif (CCM), another well established cholesterol-binding motif (32). The only known cholesterol-binding motif that can be found in Kir2.1 sequence is CARC. That does not preclude, however, the possibility that cholesterol may interact with Kir2.1 channels through other previously unidentified interaction motifs. In this study, therefore, we employ an alternative strategy to identify putative cholesterol-binding regions in Kir2.1 channels, using a combined computational-experimental approach, which is independent of any known cholesterol binding motifs. Our results identify a novel nonannular cholesterol-binding region/surface in Kir2.1 channels that has no correspondence to any of the established cholesterol-binding motifs.

EXPERIMENTAL PROCEDURES
Expression of Recombinant Channels in Xenopus Oocytes-Point mutations were generated using the QuikChange sitedirected mutagenesis kit (Stratagene, La Jolla, CA). cRNAs were transcribed in vitro using the "Message Machine" kit (Ambion, Austin, TX). Oocytes were isolated and microinjected as described previously (24,33). Expression of channel proteins in Xenopus oocytes was accomplished by injection of the desired amount of cRNA. Oocytes were injected with 0.5 ng of cRNA of the channel. All oocytes were maintained at 17°C. Two-electrode voltage clamp recordings were performed 1 day following injection.
Cholesterol Enrichment of Xenopus Oocytes-Treatment of Xenopus oocytes with a mixture of cholesterol and lipids has been shown to increase the cholesterol/phospholipid molar ratio of the plasma membrane of the oocytes (34). Thus, in order to enrich the oocytes with cholesterol we used a 1:1:1 (w/w/w) mixture containing cholesterol, porcine brain L-␣phosphatidylethanolamine, and 1-palmitoyl-2-oleoyl-sn-glycero-3-phospho-L-serine (Avanti Polar Lipids, Birmingham, AL). The mixture was evaporated to dryness under a stream of nitrogen. The resultant pellet was suspended in a buffered solution consisting of 150 mM KCl and 10 mM Tris/HEPES at pH 7.4 and sonicated at 80 kHz in a bath sonicator (Laboratory Supplies, Hicksville, NY). Xenopus oocytes were treated with cholesterol for 1 h.
Two-electrode Voltage Clamp Recording and Analysis-Whole-cell currents were measured by conventional twomicroelectrode voltage clamp with a GeneClamp 500 amplifier (Axon Instruments, Union City, CA), as reported previously (24,33). A high potassium solution was used to superfuse oocytes (96 mM KCl, 1 mM NaCl, 1 mM MgCl 2 , 5 mM KOH/ HEPES (pH 7.4)). Basal currents represent the difference of inward currents obtained (at Ϫ80 mV) in the presence of 3 mM BaCl 2 in high potassium solution from those in the absence of Ba 2ϩ . A minimum of two batches of oocytes were tested for each normalized recording shown. Recordings from different batches of oocytes were normalized to the mean of whole-cell basal currents obtained from control untreated oocytes. The mean of each batch of control untreated oocytes was normalized to 1. Statistics (i.e. mean and S.E.) of each construct were calculated from all of the normalized data recorded from different batches of oocytes.
Molecular Docking of Cholesterol to Kir2.1-Molecular docking was carried out using ArgusLab (35,36). The model of the channel used (KDB database ID H011) 5 (38) was based on the chimera between the cytosolic domain of Kir3.1 and the TM domain of KirBac1.3 (Protein Data Bank code 2QKS; resolution 2.2 Å) (39). The Gly 177 residues that are located one helical turn below the helix bundle crossing residues in the four subunits of the channel were grouped in order to define the center of the binding site box. The size of the binding site box was set around the Gly 177 binding center at its maximal allowable value of 60 ϫ 60 ϫ 60 Å 3 . Possible binding sites of cholesterol were searched within the binding site box. An exhaustive search was carried out using grids at a resolution of 0.4 Å, and a flexible ligand docking mode was employed. The maximal number of poses was set at 150. In order to achieve high accuracy (36), all 138 possible poses that were identified by the program were included in the analysis. Clustering analysis was carried out by using a k-medoid algorithm, a classic partitioning technique that groups n objects into k clusters. Each pose was associated with the closest center pose of a cluster as determined by a Euclidean distance metric.
Molecular Dynamics (MD) Simulations-The complex structure was embedded into explicit lipid bilayer and solvated in 150 mM KCl with the CHARMM-GUI interface of Jo et al. (40). CHARMM-36 parameters were used for DMPC lipid bilayer and bound cholesterol molecule (41)(42)(43). The TIP3P water model and recently developed ion parameters were used for all molecular dynamics simulations (44). Three K ϩ ions were added into the selectivity filter of the TM domain of the Kir channel to occupy sites S0:S2:S4, and the cavity was hydrated by running grand-canonical Monte-Carlo simulations (45). It was found that average water occupancy at the inner cavity of the channel is 28 Ϯ 3 molecules. The resulting system contained ϳ120,000 atoms total. Each channel-cholesterol complex was equilibrated at 303.15 K with the NPAT ensemble for 10 ns using periodic boundary conditions in a tetragonal box of 93.8 ϫ 93.8 ϫ 132.5 Å 3 . This was followed with simulation runs of 50 ns. The temperature was maintained with a Lowe-Anderson thermostat as implemented in NAMD. All molecular dynamics simulations were performed with a program suite NAMD version 2.9 (46). Subsequent analysis of the system was performed using the CHARMM program suite (35b1r1) (47).
Molecular Mechanics/Poisson-Boltzmann Surface Area Computations of Binding Free Energies-Enthalpies of cholesterol binding to the Kir channel were computed as averages from an ensemble of structures (5000 for each binding pose) sampled from evenly distributed points over 50-ns runs. Electrostatic contributions to binding free energy were obtained by solving a linearized version of the Poisson-Boltzmann equation as implemented in PBEQ module of CHARMM (48,49). A dielectric constant of 2 was assigned to the protein, and the protein-solvent surface was defined using the set of optimized atomic radii from Nina et al. (50), and both ions in the filter and water molecules in the inner cavity of the channel were kept during PBEQ computations. Following the numerical recipe of Chandra et al. (51), the membrane was represented as a 24-Å slab of low dielectric 2. To better describe lipid dynamics, we extracted the positions of the DMPC lipid's heavy atom in the bilayer and used them to represent the neutral slab around the protein and bound cholesterol. The focusing method was used to solve the Poisson-Boltzmann equation with a coarse grid spacing of 1.5 and fine grid spacing of 0.5 Å, respectively. The protein-ligand complex occupied ϳ80% of the grid volume.
The non-electrostatic components of binding enthalpies were represented with scaled Lennard-Jones and solvent-accessible surface area (SASA) terms, as described by Swanson et al. (52). The SASA term was parameterized to represent Gibbs free energies of solvation of small non-polar solutes and thus contains both enthalpic and entropic contributions related to the formation of hydrophobic surfaces and to solvent expulsion from the protein pockets (52). Following Swanson and McCammom (52), we assessed the entropy contribution to the free energy of binding using the quasiharmonic approximation. Assuming isotropic rotation, it is possible to evaluate the translational, rotational, and vibrational contributions to the association entropies. The massweighted average coordinates from the last 50 ns of the MD simulations were used as a reference coordinate for computations of the binding entropy. Uncertainties were determined from block-averaging analysis with 5 blocks in total.
Contact analysis between the channel and the lipids/cholesterol was carried out as described previously (53,54) within a heavy-heavy atom cut-off distance of 3.7 Å. Contact analysis between the channel and the phospholipids was done for subunits that did not have a bound cholesterol molecule. The information from the contact analysis was used to define the residues that form the cholesterol binding pocket. These residues were selected for per-residue interaction energy evaluations.

Identification of Possible TM Cholesterol-binding Regions by Molecular Dynamics
Simulations-In order to identify putative cholesterol binding sites, we used a combination of molecular docking followed by MD simulations. In our recent study, we performed molecular docking of cholesterol to the cytosolic domain and its interface with the TM domain of Kir2.1 and found, based on this analysis, that all potential cholesterol binding sites are located not in the cytosolic domain itself but at its interface with the TM domain with one exception in the cytosolic domain that is highly energetically unfavorable for cholesterol binding (24). In this study, we extend this analysis to the TM domain, also including its interface with the cytosolic domain. To achieve this goal, molecular docking of cholesterol to Kir2.1 was performed within a box that includes the TM domain and its interface with the cytosolic domain, as depicted in supplemental Fig. 1A. We then grouped the resulting poses in six clusters based on the similarity between each two poses, as plotted in supplemental Fig. 1B. Four of these clusters were located in the TM domain between helices of adjacent subunits, and two clusters were located at the boundary between the TM helices and the N-terminal slide helix (supplemental Fig. 1C). The clusters identified here at the interface of the TM and cytosolic domains are found in the same regions as the clusters that were found in our previous study of the cytosolic domain described above (24). The centers of each cluster provide the starting points for the MD simulations.
The rationale for following molecular docking analysis with MD simulations is that although molecular docking is an efficient approach for finding potential binding sites, it lacks two major factors in determining the interactions between the channel and cholesterol. First, it does not include the lipids that surround the channel that may play a critical role because cholesterol is expected to interact with both the channel protein and the membrane lipids. Second, in typical molecular docking applications, the channel is not free to adjust its conformation in the presence of the cholesterol molecule, which prevents accurate determination of putative binding sites. As a result, molecular docking might not be able to predict cholesterol binding sites of the channels accurately. It is, however, an efficient approach for generating starting points for MD simulations.
We proceeded, therefore, to run 50-ns full-membrane allatom MD simulations using the six centers obtained from the docking analysis as starting points. The significant adjustment of the initial predictions obtained from the docking analysis underscores the importance of combining the two methods. Specifically, during the MD simulations, the cluster centers obtained from the molecular docking moved and rearranged in two distinct cholesterol-binding regions between adjacent subunits (see Fig. 1, A and B): 1) three clusters that have been embedded between the TM helices (poses 1-3) converged into a region at the center of the TM domain, and 2) two clusters that have been identified at the interface of the TM domain with the cytosolic domain (poses 4 and 5) converged into a second region, which is closer to the cytosolic domain. All five clusters are located between neighboring subunits of the chan-nel. In region 1, the cholesterol molecule is located between the outer TM helix of one subunit and the slide and TM helices of the adjacent subunit. In region 2, cholesterol is situated between the slide helix of the N terminus of one subunit and the interface between the TM inner helix and the C terminus of its neighboring subunit. The sixth cluster that was identified by the docking analysis at the TM domain closest to the extracellular domain (pose 6) appears to be a false positive of the docking analysis, because the cholesterol molecule unbound and diffused away from the channel during the simulation. The location of the five poses after 50 ns is shown in Fig. 1, A and B.
Analysis of the trajectories of the MD simulations shows that the flexible cholesterol molecule continuously explores a considerable conformational space in each of the two binding regions shown in Fig. 1, A and B. This is evident in supplemental Fig. 2, which depicts the maps of the weighted densities of the center of mass of the cholesterol molecule that demonstrate that in each pose, the center of mass of the cholesterol molecule moves primarily within the red colored region and its immediate surroundings.
To gain further insights into the energetics of cholesterol binding to the two putative cholesterol-binding regions, we calculated the binding enthalpy (⌬H bind ), a quantitative indicator of changes in the binding energy. As evident from Table 1, the binding energy of the cholesterol molecule to the channel depends on two key factors, van der Waals interactions between cholesterol and the channel residues (⌬E vdw ) and surface area effects due to expulsion of solvent molecules from hydrophobic surfaces (⌬E SASA ) (55). Our calculations show that the average binding energy (⌬H bind ) of cholesterol at region 1 is slightly stronger than the average binding energy to the cholesterol molecule at region 2 (2 kcal/mol average difference between the poses of regions 1 and 2, respectively). However, because (similarly to previous studies using the MM/ PBSA approach (52)) also in our calculations the S.E. values in ⌬H bind were within 2 kcal/mol, it is impossible to determine whether there is a significant difference between cholesterol binding energies at the two regions.
Moreover, to assess further the binding affinity and stability of cholesterol to the channel, we also calculated the equilibrium free energy of the process, ⌬G bind , which is composed of two thermodynamic components, the binding enthalpy (⌬H bind ) described above and the binding entropy (ϪT⌬S bind ). Calculation of the entropic component of the free energy (ϪT⌬S bind ) indicates that, following cholesterol binding to the channel, there is an entropic penalty of 3-5 kcal/mol, which accounts for loss of translational and vibrational degrees of freedom. The entropic penalty appears to be similar for all poses with a very slight increase in region 2 (1 kcal/mol average difference between the poses of regions 1 and 2, respectively; see Table 1). As a result, the binding free energy calculated for poses 1-3 (region 1) is negative (favorable), whereas the binding free energy calculated for poses 4 -5 (region 2) is positive (unfavorable). The absolute value of ⌬G bind is small and corresponds to weak binding affinity of cholesterol to region 1. Also, despite the unfavorable equilibrium free energy of cholesterol binding to region 2, because of the small absolute value of ⌬G bind and given a possible error, we cannot rule out the possibility of weak cholesterol binding to region 2.
Taken together, our MD simulations suggest that the two regions may form cholesterolophilic surfaces with a preference for binding region 1, which is located in the center of the TM domain. Furthermore, our analysis suggests that within each of these two regions, the cholesterol molecule exhibits significant flexibility, interacting with a different set of hydrophobic residues.
Identification of Cholesterol Recognition Residues in the Transmembrane Cholesterol Binding Regions-To examine which of the putative cholesterol-binding regions identified by the MD simulations has residues that play a role in the sensitivity of the channel to cholesterol, we mutated multiple residues that are predicted to interact with the cholesterol molecule in  OCTOBER 25, 2013 • VOLUME 288 • NUMBER 43 JOURNAL OF BIOLOGICAL CHEMISTRY 31157 each of the regions described above (9 -16 residues have been mutated for each of the five poses described above). In order to decrease the likelihood for loss of function due to the mutagenesis as well as nonspecific global effects on the channel structure, we mutated each residue to a residue of a similar type that is only slightly different from a structural point of view. More specifically, the three hydrophobic residues Ile, Leu, and Val were mutated among themselves; Ala was mutated to Val, the smallest of the above three hydrophobic residues, Ser was mutated to Thr or Val, and Met was mutated to Ile, which is the closest to Met in structure among the hydrophobic residues. As predicted, almost all of these mutations resulted in functional channels with only a few exceptions (see supplemental Table 1).

Cholesterol Regulation of Kir Channels
As Fig. 2 demonstrates, all five sites include residues whose mutation significantly affects the sensitivity of the channel to cholesterol (Fig. 2, A-C). The majority of these residues (9 of 10) belong to the non-polar aliphatic group of amino acids. These include residues in both the N-terminal slide helix (Leu 69 , Ala 70 , and Val 77 ) as well as in the outer (Leu 85 , Val 93 , and Ser 95 ) and inner (Ile 166 , Val 167 , Ile 175 , and Met 183 ) TM helices. Notably, in each of the sites, 3-6 of these residues surround the cholesterol molecule, forming with it multiple van der Waals interactions (Fig. 2, D-H). We thus suggest that these residues form cholesterol recognition regions that bind cholesterol molecules. Because the poses within the two cholesterolophilic regions overlap, several of the residues that abrogate cholesterol sensitivity appear in more than one site; Leu 85 and Val 167 are common to the three sites of region 1, and Leu 69 , Ala 70 , Val 77 , and Met 183 are common to the two sites of region 2. Moreover, two of the residues, Leu 69 and Leu 85 , are common to both regions 1 and 2, suggesting that they can interact with more than one cholesterol molecule, one from each of the two cholesterol-binding regions at the same time.
With the majority of the residues that are involved in the formation of the cholesterol-binding regions located at the membrane interface, we examined whether phospholipids could bind to these putative sites as well. Interestingly, whereas cholesterol has at least 1.7 contacts/residue, the average number of contacts between phospholipids and the majority of the residues in the putative cholesterol binding sites is less than 0.4, suggesting that these are nonannular surfaces that are occluded from phospholipid binding. Further evidence for the preference of the different poses to cholesterol as compared with phospholipids is shown in supplemental Fig. 3, which depicts the perresidue decomposition of the interaction energies for each of the poses for cholesterol and the DMPC lipids that form the membrane in the simulations. Furthermore, the coordination mode described above, which depends on interactions with Leu 69 , Leu 85 , Ser 95 , and Leu 167 , would require the phospholipid molecule to be in an off-plane position, further decreasing the likelihood of phospholipid binding at the putative cholesterol binding sites.
Our results show, however, that only a subset of residues that surround the cholesterol molecule in each of the putative cholesterol-binding regions affect cholesterol sensitivity of the channels, whereas mutations of other residues in each of the poses did not have an effect (Fig. 3 and supplemental Table 1). The latter include a subgroup of non-polar aliphatic residues, similar to those that are found to interact with cholesterol (9 of 20) as well as six aromatic residues and four positively charged residues. No aromatic or positively charged residues were found to affect cholesterol sensitivity of the channels. These observations suggest that only some of the residues that are located at each of the five poses are critical for cholesterol recognition. However, we cannot exclude the possibility that the lack of an effect may be due to the fact that mild mutations may not sufficiently affect the interaction of the channel with the cholesterol molecule.
Secondary Cholesterol-sensitive Residues in the Cholesterolophilic Surfaces of the Channel-Besides the residues that affect cholesterol sensitivity of the channels within the putative cholesterol-binding regions, adjacent residues may also play a role in the regulation of the channel by cholesterol in an indirect way. We thus screened residues predicted to be located away from the cholesterol molecule but found within 4 Å of the "cholesterol recognition" residues that engulf the cholesterol molecule at each of the five sites (Fig. 2). As can be seen in Fig. 4, five additional residues affected the sensitivity of the channel to cholesterol. These include three residues in the slide helix of the N terminus (Tyr 68 , Cys 76 , and Ile 79 ) and two residues in the inner helix (Phe 159 and Ser 165 ).
Along with the cholesterolophilic surfaces of the channel that extend over a substantial part of the TM domain, the analysis also predicts a smaller membrane-facing region close to the extracellular domain of the channel to which cholesterol does not bind (see supplemental Fig. 4, A and B). This region includes outer (Cys 101 , Trp 104 , and Leu 105 ) and pore helix (Phe 129 ) residues. We have previously shown that mutations of these outer helix residues to both alanine and leucine do not affect the sensitivity of the channel to cholesterol (23). Supplemental Fig. 4C shows that also mutation of Phe 129 to a leucine does not alter the channel's cholesterol sensitivity, confirming that, indeed, this outlying region forms a cholesterol insensitive section. Moreover, within each subunit, this region connects to an elevated chain of cholesterol-insensitive membrane-facing residues of the membrane-spanning outer helix TM1 (Arg 82 , Met 84 , Val 86 , Ile 87 , Leu 90 , Ala 91 , Leu 94 , Leu 97 , and Phe 98 ) (see Fig. 5A). Notably, elevated chains from two adjacent subunits border the groove defined by poses 1, 2, and 3 (cholesterolophilic region 1) between the corresponding channel subunits.

DISCUSSION
With multiple lines of evidence indicating that cholesterol may regulate ion channels by specific sterol-protein interactions, the next fundamental question is to identify the specific regions of the channel proteins that are responsible for cholesterol binding. Two recent studies identified well known CRAC motifs in TRPV1 (21) and in BK channels (30), showing that mutations of residues included in CRAC motifs affect the sensitivity of these channels to cholesterol. It is important to note, however, that although three cholesterol-binding motifs are currently identified (28, 29, 31, 32), they may not account for all of the possible cholesterol-protein interactions. We chose, therefore, to employ a combination of all-atom full-membrane molecular dynamics simulations with an experimental func-tional assay to identify the location of a cholesterol binding site in Kir2.1 channels that is not based on the known cholesterolbinding motifs. Our studies identified two putative cholesterol-  Fig. 2 are shown in red. Residues whose mutation did not affect the sensitivity of the channel to cholesterol are shown in yellow (Fig. 3, B-D), and residues whose mutation resulted in non-functional channels are shown in   F133I and I137V of the pore helix (C); and F159I, V162L, and S165V of the inner helix (D) (n ϭ 10 -90). Significant difference is indicated by an asterisk (*, p Յ 0.05). Also shown are representative traces of the whole-cell basal currents recorded at Ϫ80 mV/ϩ80 mV for the WT Kir2.1 channel and each of the mutants in A-D. The color of the asterisks is in accordance with the description in the legend of Fig. 2. E, ribbon representation of the TM and extracellular domains of two adjacent subunits of Kir2.1 (in gray and pink) depicting the locations of secondary cholesterolsensitive residues (in orange) that are located within 4 Å of the cholesterol recognition sites depicted in Fig. 2 (in red). Also shown are the locations of the cholesterol molecules at sites 1-5 (cyan sticks and surface representations). Error bars, S.E.
binding regions in the TM domain in nonannular surfaces that are located in between transmembrane ␣-helices of two adjacent subunits.
We have previously identified a series of residues that form a structured belt that surrounds the cytosolic pore of the channel close to its interface with the TM domain and modulate the cholesterol sensitivity of Kir channels (8,(23)(24)(25). However, based on the crystal structure of the cytosolic domain of Kir2.1 (56), the residues of the cholesterol sensitivity belt neither interact directly with the lipid bilayer nor form a hydrophobic pocket that might harbor a cholesterol molecule. Indeed, there was no correspondence between the location of these residues and any of the possible cytosolic cholesterol binding sites that were obtained in a docking analysis of the cytosolic domain of the channels, suggesting that the cholesterol sensitivity belt does not form a cholesterol binding site. Instead, based on correlation between these residues and residues in the cytosolic gate of the channel, we proposed that these residues are critical to channel gating (24). Notably, we have also shown earlier that mutations of the lipid-facing residues of the outer TM helix that have been identified in previous studies (57) had no significant effect on cholesterol sensitivity of the channels, suggesting that cholesterol does not bind to a boundary site of Kir2 channels (23), a site that is located on the TM surface of the protein. In this study, therefore, we tested the possibility that cholesterol might bind to nonannular surfaces that occlude membrane phospholipids and are located between ␣-helices in the TM domain (58).
And indeed, our MD simulations support the hypothesis that the putative cholesterol-binding regions that we have identified are nonannular surfaces and prefer cholesterol to phospholipids. As described under "Results," additional interaction energy decomposition indicates that phospholipids only transiently interact with the residues that form cholesterol binding pockets. For every pose in which cholesterol resides stably for the entire length of the MD simulations, the bound cholesterol molecule interacts more favorably than a single lipid targeting the same residues. The rigid ring structure of the cholesterol molecule provides an advantage in interactions with the aromatic residues Phe 73 and Trp 81 . At the same time, the hydrophobic tail of the bound cholesterol is stabilized by multiple interactions with the Leu 69 , Ala 70 , and Leu 85 region as well as with a hydrophobic pocket in an adjacent subunit centered on Ile 166 and Val 167 . The lipids reside only transiently in this region. According to a contact analysis, there are an average of 35 lipid binding/unbinding events to the putative cholesterolspecific binding pockets proposed in our work. Therefore, we conclude that pocket specificity depends on matching hydrophobic and aromatic moieties in the channel.
Our functional studies show that, as predicted by the computational analysis, residues whose mutation significantly affects the sensitivity of Kir2.1 to cholesterol engulf the cholesterol molecule. This correspondence between the functional assay and the results of the molecular modeling supports the notion that cholesterol binds to the two regions of the channel, the TM region and the TM-cytosolic domain interface region.
Analysis of the binding enthalpy and free energy, which are quantitative indicators of the binding energy and affinity of cholesterol to the channel, suggests that cholesterol may bind more strongly to the TM cholesterol-binding region and with a higher affinity than to the TM-cytosolic domain interface region. Furthermore, because mutations of both regions abrogate the sensitivity of the channels to cholesterol, we propose that region 2 represents a transient site with weak and possibly short lived cholesterol binding, which is nevertheless necessary for cholesterol to have an effect on channel function. It is possible that transient binding to the interface region is necessary for cholesterol to access a more stable binding region in the TM domain. showing the TM residues whose mutation affects the sensitivity of the channel to cholesterol (in red balls) relative to the location of the five cholesterol sites (in cyan sticks and surface representation). Also shown are the continuous chains of residues that border the cholesterol binding groove in the channel (in blue balls). B, schematic model illustrating the location of the two cholesterolbinding regions along with labeling of the channel regions. Note that for purposes of clarity, the model shows the cholesterol molecules next to one of the two adjacent channel subunits with which they are predicted to interact.
With the critical residues that affect cholesterol sensitivity being primarily non-polar aliphatic, the cholesterol-binding regions that we have identified differ from both the CRAC and the CCM cholesterol binding motifs, which include an aromatic residue and a positively charged residue in addition to a non-polar aliphatic residue. Specifically, the CRAC motif is -(L/V)X 1-5 YX 1-5 (R/K)-, where X 1-5 represents 1-5 residues of any amino acid (28,29). The CCM cholesterol binding motif includes sites on adjacent helices: (W/Y)(I/V/L)(K/R) on one helix and (F/Y/R) on the second helix (32). In contrast, in the case of Kir2, each cholesterol-binding region includes multiple non-polar aliphatic residues whose mutations significantly affected and in most cases abrogated the sensitivity of the channel to cholesterol. Interestingly, however, there is some overlap between the TM putative cholesterol-binding region that we identify in this study and a CRAC motif-based putative cholesterol binding site in TRPV1 channels (21). Specifically, alignment of Ser 5 of TRPV1 with TM1 of Kir2.1 shows that Leu 585 in TRPV1 channels, which was shown to be important for cholesterol sensitivity of TRPV1, is located at the equivalent position to Val 93 of Kir2.1, which, as we show here, is important for cholesterol sensitivity of Kir2.1. However, mutations of the Kir2.1 residues that are located at the equivalent positions to the other two residues of the CRAC motif in TRPV1, Arg 579 and Phe 582 , did not affect the sensitivity of Kir2.1 to cholesterol. This difference between the two channels may be attributed to the lack of a positively charged residue in this region of Kir2.1, which, as noted above, is a necessary component of the CRAC motif. Furthermore, the corresponding two residues in Kir2.1, Ile 87 and Leu 90 , are a part of the elevated chain of cholesterolinsensitive membrane-facing residues of the membrane-spanning outer helix (Fig. 5A). We conclude, therefore, that the cholesterol-binding regions in Kir2.1 channels differ significantly from the CRAC motif-based putative region identified earlier in TRPV1 channels.
Recently, a CARC motif has been shown to be more consistent in predicting cholesterol recognition motifs in integral membrane proteins (58). As hinted by its name, the CARC motif, (R/K)X 1-5 (Y/F)X 1-5 (L/V), is an inverted CRAC motif. In our case, we have identified two CARC sequences at the interface between the slide helix and the TM domain. These include 1) Arg 67 -Phe 73 -Val 77 and 2) Arg 82 -Phe 88 -Leu 90 . Our data show that there is a partial overlap between the first CARC motif and the cholesterol-binding region that we have identified at the interface of the TM domain with the cytosolic domain (putative cholesterol-binding region 2). Specifically, as we have shown in Fig. 2, the V77I mutation results in loss of cholesterol sensitivity. The roles of the other two residues (Arg 67 and Phe 73 ), however, could not be tested because their mutations result in non-functional channels (see supplemental Table 1). In contrast, we have shown that mutations of all three residues of the second CARC sequence in Kir2.1 channels do not affect the sensitivity of the channel to cholesterol (see Fig. 3) (23), suggesting that it does not describe a cholesterol binding site in these channels. Furthermore, within the TM domain in the vicinity of region 1, no CARC motif was identified. We thus propose that although the CARC motif may represent a part of the putative cholesterol-binding region that we have identified at the interface of the TM and cytosolic domains of Kir2.1, the cholesterol-binding region at the center of the TM domain of the channel differs from all three of the previously identified cholesterol binding motifs.
In terms of the mechanism, the locations of the binding regions of cholesterol in Kir2.1 suggest that cholesterol may impact channel function by either affecting the hinge region of the channel at the center of the pore-lining inner helix of the TM domain or the interaction between the inner helix of the TM and the slide helix of the N terminus, two regions that have been shown earlier to play central roles in Kir gating (59 -63).
Cholesterol-binding region 1 located at the center of the TM domain that we have identified here overlaps with the hinge region of the inner helix of the TM domain of the channel, as described earlier. Specifically, comparative analysis of the crystal structures of two related bacterial channels, KcsA and MthK, reveals that a highly conserved central glycine of the inner TM helix plays the role of a gating hinge (59). Furthermore, the corresponding central glycine of the homomeric Kir3.4* channel was shown to play a central role in Kir gating, and it was suggested that the flexibility of the glycine at this position is required for ensuring the frequent gating of the helix bundle crossing of the channel (60). Our further analysis of the neighboring residues of this key glycine residue in these channels suggested that hinging occurs not at the glycine itself but at the residue immediately preceding the central glycine of the inner helix (61). Here, we show that the equivalent residue in Kir2.1 where the hinging is predicted to occur, valine at position 167, is one of the residues whose mutation we found to abrogate cholesterol sensitivity in all three sites of region 1 (see Fig. 5B). We propose, therefore, that cholesterol binding may interfere with the hinging motion of the inner helix stabilizing the channel in the closed state.
In region 2 located at the interface of the TM and cytosolic domains, on the other hand, cholesterol connects between the slide helix of the N terminus and the C-linker that connects the C terminus and the inner TM helix. This region has also been implicated in the control of Kir2 gating, and it was proposed that interactions between the N and C termini provide a tangential force that mechanically gates the channel (62). Moreover, it has been recently suggested (63) that rearrangements in the interactions pattern between the N and C termini of Kir channels occur when the channels are activated by phosphatidylinositol 4,5-bisphosphate, a major regulator of channel function, which is required for activation of Kir channels (37,62,64,65). We propose, therefore, that interactions between cholesterol and residues located in region 2 may affect the gating mechanism that leads to the opening of the inner helix gate while bending the pore-lining helix at the central glycine hinge.
These mechanisms may extend to other ion channels. Specifically, it has been previously shown that the glycine in the middle of the inner helix, which, as described above, was suggested to play a role of a central hinge in K ϩ channel gating, is 80% conserved among potassium and cyclic nucleotide-gated channel sequences (60). Therefore, we propose that if cholesterol binds to the region immediately adjacent to this residue, it might interfere with the hinging motion of the inner helix during channel gating in multiple channels.