Direct modulation of Kir channel gating by membrane phosphatidylinositol 4,5-bisphosphate.

Multiple ion channels have now been shown to be regulated by phosphatidylinositol 4,5-bisphosphate (PIP2) at the cytoplasmic face of the membrane. However, direct evidence for a specific interaction between phosphoinositides and ion channels is critically lacking. We reconstituted pure KirBac1.1 and KcsA protein into liposomes of defined composition (3:1 phosphatidylethanolamine:phosphatidylglycerol) and examined channel activity using a 86Rb+ uptake assay. We demonstrate direct modulation by PIP2 of KirBac1.1 but not KcsA activity. In marked contrast to activation of eukaryotic Kir channels by PIP2, KirBac1.1 is inhibited by PIP2 incorporated in the membrane (K(1/2) = 0.3 mol %). The dependence of inhibition on the number of phosphate groups and requirement for a lipid tail matches that for activation of eukaryotic Kir channels, suggesting a fundamentally similar interaction mechanism. The data exclude the possibility of indirect modulation via cytoskeletal or other intermediary elements and establish a direct interaction of the channel with PIP2 in the membrane.

Multiple ion channels have now been shown to be regulated by phosphatidylinositol 4,5-bisphosphate (PIP 2 ) at the cytoplasmic face of the membrane. However, direct evidence for a specific interaction between phosphoinositides and ion channels is critically lacking. We reconstituted pure KirBac1.1 and KcsA protein into liposomes of defined composition (3:1 phosphatidylethanolamine:phosphatidylglycerol) and examined channel activity using a 86 Rb ؉ uptake assay. We demonstrate direct modulation by PIP 2 of KirBac1.1 but not KcsA activity. In marked contrast to activation of eukaryotic Kir channels by PIP 2 , KirBac1.1 is inhibited by PIP 2 incorporated in the membrane (K1 ⁄ 2 ‫؍‬ 0.3 mol %). The dependence of inhibition on the number of phosphate groups and requirement for a lipid tail matches that for activation of eukaryotic Kir channels, suggesting a fundamentally similar interaction mechanism. The data exclude the possibility of indirect modulation via cytoskeletal or other intermediary elements and establish a direct interaction of the channel with PIP 2 in the membrane.
Phosphoinosotides constitute a major group of signaling molecules in eukaryotic membranes (1,2) and modulate an ever growing list of ion channels, whether by application of exogenous phospholipids to the cytoplasmic membrane surface or by manipulation of endogenous phospholipids (3)(4)(5)(6)(7)(8)(9)(10)(11)(12). However, the nature of the phosphoinositide-channel interaction remains elusive. For one extensively studied group, the inward rectifier K (Kir) channels, there is an emerging consensus that a direct interaction occurs between cytoplasmic domains of the channel and inositol headgroups, based on electrophysiological analysis (5,(13)(14)(15)(16) and biochemical analysis of isolated channel domains (5,17,18). Direct interaction of functional channels with phospholipids in the membrane has been difficult to demonstrate unequivocally, and without this, quantification of the dose-response relationships for channel modulation by phospholipids is obviated, and further mechanistic understanding is limited (19,20).
The recent cloning and crystallization (21), as well as functional analysis of KirBac1.1 channels reconstituted in lipid membranes (22), provides the opportunity to examine channel activity using a highly purified protein preparation in membranes of defined composition and permits direct test of the nature of the channel-phosphoinositide interaction.

EXPERIMENTAL PROCEDURES
Methods are essentially as described previously (22). KcsA and KirBac1.1 in pQE60 vector were expressed in BL21* (DE3) cells after induction with isopropyl ␤-D-thiogalactopyranoside. Bacteria were lysed by sonication, incubated 2-4 h with 30 mM decylmaltoside (Anatrace), then centrifuged at 30,000 ϫ g for 30 min, and the supernatant was applied to a cobalt affinity column. The column was washed with 20 -30 volumes of wash buffer (50 mM Tris-HCl, pH 8.0, 150 mM KCl, 10 mM imidazole, and 5 mM decylmaltoside) and eluted with 1-2 ml of wash buffer containing 500 mM imidazole. Protein was concentrated using 30-kDa centrifugal filters (Millipore) and stored in wash buffer with 20 mM imidazole.
For Rb ϩ flux assay, purified protein was added to 3:1 phosphatidylethanolamine:phosphatidylglycerol (POPE 2 :POPG, Avanti) at 2-3 g/ml lipid and incubated 30 min. Disposable polystyrene columns (Pierce catalog number 29920) were packed with Sephadex G-50 (fine) beads (1 ml), swollen overnight in buffer A or B (buffer A: 450 mM KCl, 10 mM HEPES, 4 mM NMG, pH 7; buffer B: 450 mM sorbitol, 10 mM HEPES, 4 mM NMG, 50 M KCl, pH 7.0). Liposomes were formed by placing 100 l of detergent-solubilized lipid/protein mixture directly on top of the gel bed of a prespun column A and collected into glass tubes by spinning at 1000 ϫ g (2500 rpm; Beckman TJ6). Extraliposomal solution was exchanged for buffer B by centrifugation through a prespun column B. 50-l aliquots of the radioactive mixture were taken at time points indicated, and extraliposomal 86 Rb ϩ removed by passage over a 0.5-ml Dowex cation exchange column in the NMGH ϩ form. Samples were mixed with scintillation fluid and counted in a liquid scintillation counter. Valinomycin was used to assay maximal 86 Rb ϩ uptake.
In all figures, values indicated are mean Ϯ S.E.

RESULTS AND DISCUSSION
KirBac1.1 with a His 6 tag was purified by affinity on a Co 2ϩ column. As indicated in Fig. 1A the eluted protein is highly purified. While we cannot exclude minor additional contaminants, associated proteins, if they exist, are not present in stoichiometric amounts. Fig. 1B shows that the purified protein generates robust channel activity when incorporated into pure POPE:POPG membranes. Fig. 1B also shows that incorporation of PtdIns(4,5)P 2 into the liposomes at less than 1% mole fraction ( Fig. 1B) does indeed significantly modulate KirBac1.1 activity, while KcsA activity is unaffected (Fig. 1C). However, contrary to expectations from eukaryotic Kir channel experiments, Kir-Bac1.1 channels are actually inhibited by PIP 2 , being half-maximally inhibited at ϳ0.3% mole fraction (Fig. 1C). KcsA is an important control, indicating that incorporation of PtdIns(4,5)P 2 in the liposomal membrane does not non-specifically abolish ion channel activity. Interestingly, the half-maximal inhibitory concentration of PtdIns(4,5)P 2 is similar to the expected PtdIns(4,5)P 2 content of mammalian cell membranes (23,24).
This first experiment defines a direct interaction of PtdIns(4,5)P 2 with the channel protein and excludes any possibility of intermediary components, such as cytoskeletal proteins (25,26). It is surprising that KirBac1.1 is inhibited by PtdIns(4,5)P 2 in the liposomal membrane, whereas application of PtdIns(4,5)P 2 to eukaryotic Kir channels in native membranes always causes activation (9). Nevertheless, we hypothesize that the nature of the PtdIns(4,5)P 2 -channel interaction is physically similar in both cases and that potentially straightforward explanations for activation versus inhibition may exist, as discussed below. However, any such explanation first requires that essential structural requirements for phosphoinositide modulation of Kir channels are replicated in the two cases. In eukaryotic Kir channels, it is well established that modulation requires a negatively charged head group and an acyl chain and is proportional to the number of phosphates on the head group (9,13). As shown in Fig. 2, these requirements are also met for modulation of KirBac1.1 activity. Without the inositol headgroup, diacylglycerol incorporation or addition of IP 3 both fail to affect channel activity (Fig. 2, A and B). Unphosphorylated PtdIns fails to inhibit channels, and singly phosphorylated PtdIns(4)P is ϳ10-fold less effective than PtdIns(4,5)P 2 or PtdIns(3,4,5)P 3 ( Fig. 2A).
Among eukaryotic Kir channels, there is varying apparent specificity for the phosphate position on the inositol headgroup (9,27), with preferential activation by phosphoinositides when the phosphate is in position 4 or 5 for most Kir family members but decreased or no preference for Kir3.x and Kir6.2 channels.
To test the specificity of KirBac1.1 inhibition as a function of phosphate position, we measured channel activity in liposomes with different isomers (Fig. 3). KirBac1.1 shows no preference for headgroup position and as such behaves very similarly to Kir6.x family members (9). Using highly purified channel protein reconstituted into liposomes of defined composition, the above experiments establish a direct interaction between PIP 2 in the membrane and the channel protein: there is no micellar PIP 2 , and interaction must be with PIP 2 incorporated into the liposome membrane. By examining the effects of PIP 2 metabolites, we also unequivocally demonstrate that PIP 2 itself is the critical interactor and, moreover, that lipid structure specificity is the same as that for activation of eukaryotic channels, when applied exogenously. This is consistent with modulation of eukaryotic Kir channels being through a fundamentally similar direct physical interaction with the channel. Why then is KirBac1.1 inhibited by PIP 2 , whereas eukaryotic channels are activated? We suggest that subtle differences in the structure of the two groups of channels can give rise to activation in one case and inhibition in the other. Sequence alignments of eukaryotic and bacterial Kir channels (Fig. 4) indicate significant gap-free homologies in transmembrane regions and conserved N-and C-terminal regions that generate the pore and the Kir cytoplasmic domain (21,28). Notably, however, eukaryotic Kir channels contain two additional 3-amino acid linkers connecting cytoplasmic domains to the slide helix and to the second transmembrane helices. The amino acids in the second linker are generally charged and have repeatedly been invoked as critical interacting charges for PIP 2 activation of eukaryotic Kir channels (5, 13, 29 -31). We hypothesize that these charged linkers in eukaryotic channels are essential for PIP 2 activation; conceivably they appeared only in response to   evolutionary pressure as the otherwise inhibitory effect of PIP 2 caused loss of Kir channel function in early eukaryotic membranes. While the nature and extent of these linkers is clearly important for channel function, further experiments are warranted to examine whether and how their length and nature are involved in transducing PIP 2 interaction to activation or inhibition of channel activity. It is intriguing to note that opposing effects of PIP 2 are seen in the TRP channel family. For some, e.g. TRPM5, TRPM7, and TRPM8, PIP 2 has an activating effect (32)(33)(34). However, PIP 2 inhibits TRPL and TRPV1 channels, and mutation of residues in the same C-terminal "TRP-box" region that is invoked as necessary for PIP 2 activation of TRPM8 channels (34) strengthens or weakens inhibition of TRPV1 by PIP 2 (8).
In studying modulation of channel activity in cell membranes, PIP 2 levels have been indirectly manipulated by hydrolysis (phospholipase) or synthesis (application of MgATP). Such manipulations can cause changes in multiple other membrane constituents and are not quantitative. Although semiquantitation has been cleverly achieved using fluorescent-tagged PIP 2 -associating protein domains (35,36), direct measurement of [PIP 2 ] is also not possible. PIP 2 has also been added to excised membrane patches in micelles. If the supposed consequence (incorporation of PIP 2 into the membrane leaflet) is achievable, it should also be possible to perform essentially this same experiment in the liposome assay. We incubated liposomes in micellar PIP 2 for 5 min prior to Rb ϩ uptake assay (Fig. 5A). KirBac1.1 activity is again inhibited, with no effect on KcsA activity. However, in contrast to the complete inhibition observed when PIP 2 is incorporated into the membrane, inhibition plateaus at ϳ50%, broadly consistent with the expectation that KirBac1.1 channels can insert into the membrane in either of the two possible orientations and that PIP 2 partitions primarily into the outer leaflet of the membrane.
A further approach to reduce [PIP 2 ] in native membranes has been to apply poly-L-lysine, which electrostatically shields the anionic headgroups (13,37). Again, broadly consistent with the hypothesis that PIP 2 headgroups interact with the channel, application of poly-L-lysine reduced the inhibitory effect of membrane PIP 2 on KirBac1.1 fluxes (Fig. 5B).
Multiple ion channel involvements in physiological and disease states, including muscarinic signaling in the brain (10,38), temperature sensing by TRP channels (8,34), inward rectifier currents in cardiac arrhythmias (31), and mechanical regulation of 2P-domain potassium channels (12) are all critically dependent on phosphoinositide signaling. However, almost all studies of channel-lipid interactions to date are inevitably indirect; in many cases it cannot be established whether the PIP 2 is even acting from within the membrane (19,20), and although biochemical studies (e.g. Refs. (18 and 39) have shown that peptide fragments of ion channels can directly interact with PIP 2 , the possibility of intermediates in physiological regulation cannot be excluded. In the present study, we use a novel approach, purified channel protein, reconstituted into lipid vesicles of fully defined composition, to unambiguously establish three critical principles: (i) an action specifically of PIP 2 in the membrane, (ii) direct interaction between the channel and PIP 2 , and (iii) the concentration dependence of this interaction. These are key mechanistic aspects of lipid modulation of channel activity and support the idea that direct interactions underlie eukaryotic Kir channel modulation. This assay provides a model system in FIGURE 5. A, specific inhibition of KirBac1.1 by micellar PtdIns(4,5)P 2 . Liposomes with reconstituted KcsA (filled squares) or KirBac1.1 (open diamonds) channels were incubated in the presence of micellar PtdIns(4,5)P 2 for 5 min prior to initiation of assay. Rb ϩ uptake was measured at 2 min and normalized to uptake in zero [PtdIns(4,5)P 2 ] (n ϭ 3). B, reversal of PtdIns(4,5)P 2 inhibition of KirBac1.1 by polylysine. Liposomes with reconstituted KirBac1.1 were formed with varying membrane [PtdIns(4,5)P 2 ]. Rb ϩ uptake was measured at 2 min plus (filled symbols) or absence (open symbols) of 10 g/ml extraliposomal polylysine (n ϭ 3).