Structural Determinants of Phosphatidylinositol 4,5-Bisphosphate (PIP2) Regulation of BK Channel Activity through the RCK1 Ca2+ Coordination Site*

Background: PIP2 has been reported to enhance Ca2+-driven gating, but the molecular determinants of this interplay are not known. Results: PIP2 interacts with specific basic residues and enhances Ca2+ gating through the αA-KDRDD-αB structural elements. Conclusion: The RCK1 Ca2+-binding site is coupled to PIP2. Significance: PIP2 is a key element in the regulation of BK channel activity. Big or high conductance potassium (BK) channels are activated by voltage and intracellular calcium (Ca2+). Phosphatidylinositol 4,5-bisphosphate (PIP2), a ubiquitous modulator of ion channel activity, has been reported to enhance Ca2+-driven gating of BK channels, but a molecular understanding of this interplay or even of the PIP2 regulation of this channel's activity remains elusive. Here, we identify structural determinants in the KDRDD loop (which follows the αA helix in the RCK1 domain) to be responsible for the coupling between Ca2+ and PIP2 in regulating BK channel activity. In the absence of Ca2+, RCK1 structural elements limit channel activation through a decrease in the channel's PIP2 apparent affinity. This inhibitory influence of BK channel activation can be relieved by mutation of residues that (a) connect either the RCK1 Ca2+ coordination site (Asp367 or its flanking basic residues in the KDRDD loop) to the PIP2-interacting residues (Lys392 and Arg393) found in the αB helix or (b) are involved in hydrophobic interactions between the αA and αB helix of the RCK1 domain. In the presence of Ca2+, the RCK1-inhibitory influence of channel-PIP2 interactions and channel activity is relieved by Ca2+ engaging Asp367. Our results demonstrate that, along with Ca2+ and voltage, PIP2 is a third factor critical to the integral control of BK channel activity.

Big or high conductance potassium (BK) channels are activated by voltage and intracellular calcium (Ca 2؉ ). Phosphatidylinositol 4,5-bisphosphate (PIP 2 ), a ubiquitous modulator of ion channel activity, has been reported to enhance Ca 2؉ -driven gating of BK channels, but a molecular understanding of this interplay or even of the PIP 2 regulation of this channel's activity remains elusive. Here, we identify structural determinants in the KDRDD loop (which follows the ␣A helix in the RCK1 domain) to be responsible for the coupling between Ca 2؉ and PIP 2 in regulating BK channel activity. In the absence of Ca 2؉ , RCK1 structural elements limit channel activation through a decrease in the channel's PIP 2 apparent affinity. This inhibitory influence of BK channel activation can be relieved by mutation of residues that (a) connect either the RCK1 Ca 2؉ coordination site (Asp 367 or its flanking basic residues in the KDRDD loop) to the PIP 2interacting residues (Lys 392 and Arg 393 ) found in the ␣B helix or (b) are involved in hydrophobic interactions between the ␣A and ␣B helix of the RCK1 domain. In the presence of Ca 2؉ , the RCK1-inhibitory influence of channel-PIP 2 interactions and channel activity is relieved by Ca 2؉ engaging Asp 367 . Our results demonstrate that, along with Ca 2؉ and voltage, PIP 2 is a third factor critical to the integral control of BK channel activity.
The high conductance potassium (Maxi K, Big K, or BK) channel is activated by both membrane depolarization and increased intracellular Ca 2ϩ concentrations ([Ca 2ϩ ] i ). Although Ca 2ϩ and voltage are thought to act independently to regulate channel opening, a weak allosteric interaction between them makes the voltage sensor movement much more effective (1)(2)(3)(4). The Slo1 pore-forming subunits of BK channels are composed of seven transmembrane segments (S0 -S6) that assemble into tetramers. The S1-S4 transmembrane region of the Slo1␣ subunit forms a voltage-sensing domain, as in other voltage-gated channels, whereas the large C-terminal intracellular ligand-binding domain is responsible for sensing Ca 2ϩ (5)(6)(7)(8)(9)(10)(11)(12)(13). Association of the Slo1␣ subunit with tissue-specific two-transmembrane ␤1-␤4 subunits modifies its functional characteristics (14). The BK channel is expressed in a wide variety of tissues, most notably in the brain and smooth musclecontaining organs, but also among other tissues in reproductive organs (ovary, testes), in the pancreas, and in adrenal glands (15). In humans, malfunction of the BK channel is known to be important to the pathophysiology of epilepsy (16 -19), hypertension (20 -24), cancer (25)(26)(27)(28), and asthma (29).
The Slo1 cytosolic domain is composed of two RCK (regulator of K ϩ conductance) domains, RCK1 and RCK2 (30). The first Ca 2ϩ -binding site identified was in a region termed the Ca 2ϩ -bowl that contains a series of Asp residues located in the RCK2 domain (6) (also see Fig. 1, B, C (left), and D (bottom)). The second Ca 2ϩ -binding site was identified in the RCK1 domain at position Asp 367 (Fig. 1, A and D (top), shows Asp 367 in the KDRDD loop in yellow) (10). Subsequently, the side chain of Glu 535 (Fig. 1C, right) was reported to be a part of the RCK1 Ca 2ϩ coordination site together with Asp 367 (31). High-resolution crystal structures of the cytosolic domain have been obtained either with the RCK2 Ca 2ϩ -bowl occupied by Ca 2ϩ (PDB 3 code 3MT5) (32) or in the absence of Ca 2ϩ (PDB code 3NAF) (33). Ca 2ϩ has not yet been resolved in the RCK1 site, although the side chains of Asp 367 and Glu 535 are positioned such that they could coordinate Ca 2ϩ (Fig. 1C, right, 3NAF structure). Interestingly, in the presence of Ca 2ϩ in the Ca 2ϩ -bowl, Asp 367 and Glu 535 point away from each other, such that they could not possibly coordinate Ca 2ϩ (Fig. 1C, left, 3MT5 structure). Despite these structural advances in the Ca 2ϩ -sensing cytosolic domains, the lack of a full-length structure that includes the transmembrane domains as well as the lack of the RCK1 Ca 2ϩ -bound state have precluded a structural understanding of how the channel is gated by either voltage or Ca 2ϩ and how the weak allosteric coupling between these two gating mechanisms greatly enhances channel activation.
Phosphatidylinositol 4,5-bisphosphate (PIP 2 ), which has been shown to activate most ion channels and transporters (34,35), has also been reported to directly activate BK channels by a single study (36). PIP 2 was found to enhance Ca 2ϩ -driven gating by increasing mean open time and decreasing mean closed time kinetics. The PIP 2 -induced activation was also found to be potentiated by the ␤1 but not by the ␤4 accessory channel subunits. These PIP 2 effects were found to be relevant in vascular myocytes, possibly contributing to the BK control of vascular tone.
We set out to investigate the structural determinants of the Ca 2ϩ -dependent PIP 2 regulation of Slo1␣ activity. We ex pressed the mSlo1␣ channel in Xenopus oocytes and mainly used inside-out macropatches to study regulation of its activity by PIP 2 . Increases in [Ca 2ϩ ] i enhanced the apparent affinity for PIP 2 through the Ca 2ϩ coordination residue Asp 367 in the KDRDD loop. Furthermore, in the absence of [Ca 2ϩ ] i , it became clear that Asp 367 as well as its two flanking basic residues, Lys 366 and Arg 368 (Fig. 1A), served to inhibit channel activation by decreasing the apparent affinity for PIP 2 . These results suggested that the KDRDD loop exerted an inhibitory effect on channel activation through PIP 2 . Mutagenesis results showed that this coupling proceeded from the KDRDD loop through the ␣A helix to the ␣B helix (Fig. 1D, top). PIP 2 docking simulations with the two available crystal structures and mutagenesis identified two basic residues in the ␣B helix, Lys 392 and Arg 393 , as critical elements in the coordination of PIP 2 (Fig.  1C, right). These results suggest that PIP 2 could serve in the role of allosterically coupling the cytosolic RCK1 structural elements (KDRDD loop, ␣A and ␣B helices) to the membranegating elements of the channel.

EXPERIMENTAL PROCEDURES
Mutagenesis and Channel Expression-Mouse Slo1 (mSlo1) cDNA was a gift from the laboratory of Dr. Christopher Lingle (Washington University, St. Louis, MO), of which the vector pXMX was designed to promote expression or increase RNA stability (10, 37-39). All mutations were generated by Pfu- FIGURE 1. RCK structural elements involved in Ca 2؉ sensing in Slo1 channels; relation to PIP 2 revealed by docking. A, sequence with secondary structure elements of the mouse Slo1␣ channel following the S6 inner helix and through the ␤C strand of the RCK1 domain. The KDRDD loop (Lys 366 -Asp 370 ) is highlighted in yellow, and the Lys 392 /Arg 393 residues, which communicated directly with PIP 2 , are shown in red. B, sequence with secondary structure elements containing the Ca 2ϩ -bowl in the RCK2 domain. The Ca 2ϩ -bowl (Asp 897 -Asp 901 ) is shown in green. C, structural models of Slo1 incorporating crystal structures of cytosolic domains in the presence and absence of Ca 2ϩ . Left, ribbon structures of two subunits of the mSlo1␣ subunit are shown, one in gray and the other in gold, whereas the RCK2 domain of the gray subunit has been highlighted in green (the other two subunits have been removed for clarity). This model was built based on the 3MT5 (cytosolic domain of Slo1 with the Ca 2ϩ -bowl occupied (32)) and 2R9R (Kv1.2-Kv2.1 chimeric channel (56)) coordinates. PIP 2 headgroup (diC1) docking simulations were performed on this model, using Autodock. 100 docking runs were conducted to yield 100 conformations of diC1-channel complex. Each red dot represents the C1 atom of diC1, which indicates the location of diC1 in the complexes. Most PIP 2 headgroups were located nearest the S4-S5 linker, the RCK2 Ca 2ϩ -bowl coordination site is circled (in red), and a Ca 2ϩ ion is shown (in blue). Right, the same process as on the left but using the 3NAF coordinates (cytosolic domain of Slo1 in the absence of Ca 2ϩ (33)). In this figure, all 100 positions of diC1 from the docking simulation result are shown (red dots). Most PIP 2 headgroups were located closest to the ␣B residues Lys 392 and Arg 393 . In addition, the binding energy from Autodock indicated that diC1 showed the most favorable binding energy in the 3NAF model. The putative RCK1 Ca 2ϩ coordination site, showing the critical residues Asp 367 and Glu 535 pointing toward each other, is circled (in red). D, the RCK1 structural elements that harbor the KDRDD loop containing Asp 367 with its two neighboring basic residues (top) and the RCK2 structural elements that comprise the Ca 2ϩ -bowl with five critical Asp residues directly coordinate the Ca 2ϩ ion (bottom).
based mutagenesis using the QuikChange TM kit and verified by sequencing. cRNA was transcribed using the MessageMachine kit SP6 (Ambion) and injected into Xenopus laevis oocytes (0.3-5 ng/oocyte), depending on the expression level of the given channel protein. X. laevis oocytes were harvested and used for cRNA injection as described previously (40 -42). Currents were normally recorded within ϳ2 weeks.
Electrophysiology-Macroscopic currents were recorded from standard excised inside-out patches with an A-M 2400 patch clamp amplifier (A-M Systems, Inc.). pClamp (Molecular Devices) was used to drive stimulus protocols and digitize currents. The signals were low pass-filtered at 10  ] i , currents were elicited by voltage pulses from Ϫ180 to 200 mV (20 ms) at 10-mV increments, whereas the voltages before and after the pulses were held at Ϫ120 mV. In 0 [Ca 2ϩ ] i , currents were elicited by voltage pulses from Ϫ100 to 300 mV or to 380 mV for some mutants (8 ms) at 10-mV increments, unless otherwise mentioned, whereas the voltages before and after these pulses were held at Ϫ100 mV.
BK single-channel currents were recorded from oocytes under the standard inside-out patch configuration. The solutions in the pipette and bath were the same as used in macroscopic current recordings except that 2 mM MgCl 2 and[Ca 2ϩ ] i concentrations were changed as indicated. Activity rundown in different intracellular [Ca 2ϩ ] i was measured 20 -30 min following excision at the indicated voltage. 10 M PIP 2 was perfused from the intracellular side (bath solution), and its effect was measured 5 min later when the BK channel activity reached steady state. Preparation of different [Ca 2ϩ ] solutions was as described previously (37,43).
Whole-cell currents in Xenopus oocytes were recorded by conventional two-electrode voltage clamp as described previously (40). Recordings were performed with a GeneClamp500 amplifier (Axon Instruments) 3-5 days after cRNA injection. Electrodes were filled with 1.5% (w/v) agarose in 3 M KCl. The bath was perfused with the same solution as that used in the pipette solution for inside-out patches. Microelectrodes had a resistance of 0.3-1.0 megaohms. Wortmannin (Wtmn) treatment involved incubation of oocytes for 2-2.5 h before recording. In experiments with intact oocytes, intracellular Ca 2ϩ levels were controlled by application of 2 nM ionomycin in the bath that contained different Ca 2ϩ concentrations. Data acquisition and analysis were carried out using pClamp9 (Molecular Devices) and Origin (Microcal) software.
Data Analysis-The relative conductance was determined by measuring the steady-state current amplitudes at the indicated voltages. The G-V curves were fitted with the Boltzmann function, where G/G max is the ratio of conductance to maximal conductance, z is the number of equivalent charges, e is the elementary charge, V is membrane potential, and V1 ⁄ 2 is the voltage where G/G max reaches half of the maximum. k is Boltzmann's constant, and T is the absolute temperature. Data in all figures are expressed as mean Ϯ S.E. Statistical significance was evaluated by Student's t test, and p Ͻ 0.05 was considered significant.
Homology Modeling-The crystal structures of Kv1.2/2.1 (PDB code 2R9R), hSlo1 (PDB code 3MT5 for the Ca 2ϩ -bowloccupied model, and 3NAF for the Ca 2ϩ -bowl-free model) were used as templates to develop homology models for the mSlo1 channel. In order to build the homology model of mSlo1, we first constructed a hybrid model template composed of the Kv1.2/Kv2.1 structure and of the hSlo1 3NAF structure. The structure of the Kv1.2/Kv2.1 S1-S6 was docked onto the hSlo1 structure based on the orientation of the four BK linkers. We then used the fused crystal structure templates (Kv1.2/2.1 transmembrane domains) and mSlo1 channel (GI: 347144) for sequence alignment using the ClustalW server (44), followed by minor manual adjustments in non-homologous regions.
We also built an mSlo1 model based on the 3MT5 crystal structure of hSlo1. Because 3MT5 was crystallized as a monomer and did not contain the BK linker, we used the 2R9R-3NAF structure to construct the hybrid template of 2R9R-3MT5. The 3MT5-based homology model of mSlo1 was then built based on the hybrid template of 2R9R-3MT5. Homology models of the mSlo1 channel were generated using the MODELLER program (45). PIP 2 and Slo1 Docking-We used the AUTODOCK program (46) to dock the PIP 2 headgroup into the mSlo1 model structures. The grid-based potential maps that were generated for the mSlo1 channel, using CHNOP (i.e. carbon, hydrogen, nitrogen, oxygen, and phosphorus) elements, sampled on a uniform grid containing 100 ϫ 70 ϫ 100 points, were 0.375 Å apart for the free energy calculations. The grid box was centered at the side chain of residue Arg 393 of mSlo1, which was found by our functional studies to be important for PIP 2 sensitivity. The Lamarckian genetic algorithm was used to identify the docking conformations of the PIP 2 headgroup. 100 docking simulations were performed. The final docked PIP 2 headgroup configurations were selected based on docked binding energies and cluster analysis. Two potential binding sites of mSlo1 channel for PIP 2 were identified by docking simulations, formed by positively charged residues Lys 392 and Arg 393 .
Chemicals-diC8 PIP 2 and PIP 2 antibody (PIP 2 Ab) were purchased from Avanti Polar Lipids. Other chemicals, such as Wtmn, ionomycin, Mg-ATP, and polylysine (poly-K ϩ ) were purchased from Sigma-Aldrich. Stocks and working solutions were prepared using protocols according to the manufacturer's instructions.

RESULTS
Slo1 Channels Are PIP 2 -sensitive-Consistent with the conclusions of Vaithianathan et al. (36), we also found that Slo1 channels expressed in Xenopus oocytes are PIP 2 -sensitive. Scavenging of endogenous PIP 2 with a combination of poly-K ϩ and PIP 2 Ab in excised patches or treatment with micromolar concentrations of Wtmn in intact cells caused significant inhibition of Slo1 currents ( Fig. 2A). Wtmn is known to block the activity of most phosphatidylinositol 3-kinases at nanomolar concentrations. 100 nM Wtmn showed no effect on BK currents (Fig. 2, Aa and Ac). In contrast, 25 M wortmannin, which also blocks phosphatidylinositol 4-kinases, thus reducing resynthe-sis of PIP 2 to the plasma membrane, showed strong inhibition of BK currents (Fig. 2, Ab and Ac). Moreover, direct application of PIP 2 , in excised patches that had been previously treated with a combination of polylysine and PIP 2 antibody to deplete endogenous PIP 2 , showed robust current reactivation (Fig. 2B). PIP 2 altered the voltage-dependent activation kinetics of Slo1 currents. 300 M PIP 2 showed faster activation kinetics than 10 M PIP 2 , especially at depolarizations to less positive potentials (Fig. 2, Ca and Cb).
Rundown of Slo1 Unitary Currents and Reactivation by PIP 2 -Single channel recordings in the inside-out mode of the patch clamp technique held at ϩ40 mV in [Ca] i ϭ 100 M FIGURE 2. Slo1 channels are sensitive to PIP 2 . Aa, effect of a 2-h preincubation with 100 nM Wtmn on Slo1 currents. Current traces were evoked by a voltage ramp protocol (top) ranging from Ϫ100 to ϩ160 mV (1 s), using whole-cell (two-electrode voltage clamp) recordings in oocytes. Ab, same as in Aa but preincubated with 25 M Wtmn. Ac, summary bar graphs from Aa and Ab. Ba, time course recording of Slo1 current amplitude of an inside-out macropatch recording from Xenopus oocytes injected with mSlo1 channels. Slo1 current in 100 M [Ca 2ϩ ] i was inhibited by PIP 2 scavengers PIP 2 Ab (1-2:1000) ϩ poly-K ϩ (300 g/ml) but was reactivated by exogenous application of 30 M PIP 2 . Bb, representative traces for Slo1 currents recorded at the time points indicated by numbers (1-3, color-coded) in Ba. Current traces were evoked by the voltage step protocol shown above. Inset, tail currents expanded. Bc, normalized G-V curves for Slo1 in 30 M [Ca 2ϩ ] i showing the reactivation of Slo1 channels by PIP 2 following current inhibition by PIP 2 Ab (1-2:1000) ϩ 300 g/ml poly-K ϩ . Ca, the actual macroscopic current traces were recorded in the presence of 10 M (black) or 300 M PIP 2 (purple). Voltage was stepped from a holding potential of Ϫ100 mV to ϩ150 mV and then back to Ϫ100 mV. The current trace in 10 M PIP 2 was rescaled to have the same peak amplitude with that in 300 M PIP 2 (purple). Inset, tail currents expanded. [Ca 2ϩ ] i ϭ 0 M. Cb, activation time constants by 10 M (Ⅺ) or 100 M (E) PIP 2 following depletion of endogenous PIP 2 by PIP 2 scavengers (PIP 2 Ab (1-2:1000) ϩ poly-K ϩ (300 g/ml). Error bars indicate mean Ϯ S.E.

Coupling of Ca 2؉ and PIP 2 Sites Controls BK Channel Activity
showed high open probability (P o ϭ 0.88; Fig. 3, A and D) immediately after excision. Under these conditions, no significant rundown of activity was seen, and 10 M PIP 2 did not enhance further channel activity. In contrast, if the patch was held at Ϫ40 mV, activity showed significant rundown, and diC8-PIP 2 stimulated activity significantly in a reversible manner (Fig. 3, B and E). This enhanced sensitivity to PIP 2 of channel activity in more depolarized membranes has been described for other channels (e.g. TRPM8 (47)). In 10 M [Ca 2ϩ ] i concentrations, rundown and reversible reactivation by PIP 2 were highly significant (Fig. 3, C and F). At 1 M [Ca 2ϩ ] i , we observed even stronger rundown both at ϩ40 and ϩ80 mV (data not shown). Interestingly, 10 M diC8-PIP 2 was not enough to reactivate this channel at either ϩ40 or ϩ80 mV, suggesting again a further decrease in PIP 2 sensitivity with a decrease in Ca 2ϩ concentration. However, a higher diC8-PIP 2 concentration (40 M) could partially reactivate BK channel activity at ϩ80 mV (data not shown). Collectively, the experiments in Figs. 2 and 3 demonstrate the PIP 2 dependence of Slo1 current activation.
Ca 2ϩ Binding to the Asp 367 Site Enhances PIP 2 Affinity-We first compared the apparent affinity of the Slo1-WT channel to PIP 2 in solutions containing no added Ca 2ϩ (assumed to be ϳ0.5 nM and referred to as 0 Ca 2ϩ ) and in the presence of 100 M [Ca 2ϩ ] i . In 0 Ca 2ϩ , the V1 ⁄ 2 of Slo1 was 173 mV (n Ͼ 10). In contrast, in 100 M Ca 2ϩ , the V1 ⁄ 2 was shifted to ϳϪ7.8 mV (n Ͼ 6) (Fig. 4A). G-V relationships were constructed at different concentrations of diC8-PIP 2 after inhibition with poly-K ϩ and PIP 2 Ab (just as shown in Fig. 2B), and the relative conductance values at ϩ170 mV (for 0 Ca 2ϩ ) (Fig. 4B) or at Ϫ10 mV (for 100 M Ca 2ϩ ) (Fig. 4C) were plotted as a function of the PIP 2 concentration tested (Fig. 4D). In the presence of 100 M [Ca 2ϩ ] i , the Slo1 channel's apparent affinity to PIP 2 increased ϳ2-fold relative to 0 M [Ca 2ϩ ] i (Fig. 4, B-D and G). Similarly, in the presence of 100 M [Ca 2ϩ ] i , the apparent affinity to PIP 2 of an epilepsy-dyskinesia D369G mutant increased to a similar extent as the Slo1-WT compared with that in the absence of [Ca 2ϩ ] i (ϳ3-fold; Fig. 4, E and G). D369G has been shown previously to increase channel activity by decreasing the flexibility of the KDRDD loop without influencing Ca 2ϩ binding itself (48). Interestingly, compared with the wild type Slo1, the D369G mutant showed a significant enhancement in its PIP 2 apparent affinity both in the absence and in the presence of Ca 2ϩ (Fig. 4G). Interestingly, the D367G mutant that disrupts the Ca 2ϩ binding in the RCK1 domain also increased the apparent affinity for PIP 2 (Fig.  4, F and G). As expected, increasing [Ca 2ϩ ] i could not further enhance the PIP 2 affinity of the D367G mutant (Fig. 4, F and G) due to the absence of the local conformational change induced by Ca 2ϩ binding. A similar result to the PIP 2 effect on the D367G mutant was obtained from the D367A mutant. (Fig. 5, C and H).
To test whether neighboring residues to Asp 367 affected the PIP 2 apparent affinity of the channel, additional mutants in the KDRDD loop were tested (Fig. 5, A-H). Results showed similar increases of PIP 2 affinity by the K366N and R368N but not by the D370N and D379N mutants (Fig. 5, D-G). These results suggested that the Asp 367 residue that coordinates Ca 2ϩ and its two flanking basic residues (Lys 366 and Arg 368 ) are coupled to PIP 2 regulation of channel activity.

KDR Mutants of the KDRDD Loop Increase Slo1 Channel
Activation in the Absence of Ca 2ϩ -A hallmark of BK channel function is that intracellular Ca 2ϩ binding can allosterically couple to the voltage sensor movement and enhance channel activity (1). In 300 M Ca 2ϩ , the V1 ⁄ 2 of Slo1 shifted by as much as 190 mV (Fig. 6A) compared with 0 [Ca 2ϩ ] i . As mentioned above, Ca 2ϩ sensitivity in Slo1 channels is mainly conferred by two sites, the RCK2 Ca 2ϩ -bowl (five consecutive Asp residues, 897-901 in mSlo1) and the RCK1 Asp 367 /Glu 535 Ca 2ϩ coordination site (see Fig. 1). Both Ca 2ϩ -bowl 5D5N mutant (D897N/ D901N) and Asp 367 /Glu 535 mutants significantly decreased the Ca 2ϩ -induced shift in V1 ⁄ 2 (Fig. 6, B-D and I). However, although the 5D5N (Fig. 6B) or Glu 535 mutants (e.g. E535G or E535A) did not change the V1 ⁄ 2 in the absence of Ca 2ϩ (Fig. 6C), Asp 367 mutants (e.g. D367G or D367A) induced significant leftward shifts of V1 ⁄ 2 in the absence of Ca 2ϩ (Fig. 6D). Mutation of the two basic residues flanking Asp 367 (i.e. K366N and R368N; see Figs. 5A and 6, top right) caused similar left shifts of the V1 ⁄ 2 in the absence of Ca 2ϩ , without affecting the Ca 2ϩ -induced shift in V1 ⁄ 2 (Fig. 6, E and F). In contrast, mutants of the remaining two Asp residues of the KDRDD loop (D369N and D370N) did not show a significant effect (Fig. 6, G and H). Summarized data for ⌬V1 ⁄ 2 compared with Slo1 WT in the presence of 300 and 0 M Ca 2ϩ are shown in Fig. 6, I and J. These results indicated that the KDR mutants of the KDRDD loop increased PIP 2 affinity (Fig. 5) and left-shifted activation of the channel in the absence of Ca 2ϩ (Fig. 6, D-F and J), suggesting that these mutants increase Slo1 activity by increasing the channel's PIP 2 affinity.
Two Basic Residues in the ␣B Helix Involved in Direct Channel-PIP 2 Interactions-To gain insight into how PIP 2 interacts with Slo1, we performed 100 docking simulations of the PIP 2  JULY 4, 2014 • VOLUME 289 • NUMBER 27

JOURNAL OF BIOLOGICAL CHEMISTRY 18865
headgroup (diC1) with either of two Slo1 models. These two models included a common homology model of the transmembrane domain of Slo1, using the Kv1.2/2.1 chimera (PDB code 2R9R) as a template together with each of the two available crystal structures of the cytosolic domains of this channel in the presence and absence of Ca 2ϩ (PDB code 3MT5 and 3NAF, respectively). The C1 atom of the diC1 in each of the 100 conformations obtained is represented by dots in Fig. 1C. In the absence of Ca 2ϩ , most of the diC1 molecules aggregated around the ␣B helix, involving the two basic residues, Lys 392 and Arg 393 (Figs. 1C (right) and 7A), in marked contrast to the docking simulations in the presence of Ca 2ϩ bound to the Ca 2ϩ -bowl (Fig. 1C, left). Next, we tested experimentally whether these two residues are involved in PIP 2 sensitivity. First, electrophysiological data showed that although both K392N and R393N mutants inhibited activation of Slo1 in the absence of Ca 2ϩ (Fig.  7, B, C, and E), only the R393N mutant showed a parallel shift of V1 ⁄ 2 in the presence of Ca 2ϩ (Fig. 7, C and E). Furthermore, both neutralization mutations of Lys 392 and Arg 393 right-shifted the diC8-PIP 2 dose-response relationships, causing a 5-6-fold increase in the PIP 2 EC 50 (Fig. 7, F and G). These results could be explained by hypothesizing that the Arg 393 interaction with PIP 2 couples the Ca 2ϩ -induced conformational change, whereas the Lys 392 residue interaction with PIP 2 is independent of Ca 2ϩ binding. The crystal structure and docking simulation results support this idea because Lys 392 points away from the ␣B helix in the absence of Ca 2ϩ , whereas Arg 393 points toward the ␣B helix (Fig. 1C, right). Accordingly, Ca 2ϩ binding induces a helical conformational turn of ␣B to facilitate the Arg 393 interaction with PIP 2 (nearly a 90°turn) but has no effect on the Lys 392 orientation.
Hydrophobic Coupling of the ␣A and ␣B Helices Plays a Critical Role in the Activation of Slo1 Channels-How could changes in the conformation of the KDRDD loop be communicated to the PIP 2 -interacting residues in the ␣B helix? Comparison of the 3MT5 (Asp 367 /Glu 535 pointing away from each other; Fig. 1C, left) and 3NAF (Asp 367 /Glu 535 pointing toward each other; Fig. 1C, right) structures reveals that in the 3MT5 structure, several residues between the ␣A and ␣B helices (␣A, Val 356 , Leu 360 , Lys 361 , and Leu 364 ; ␣B, Phe 391 and Phe 395 ) form predominantly hydrophobic interactions ( Figs. 1C and 8, top, right inset). Mutants of these residues were tested for their involvement in (a) the [Ca 2ϩ ] iinduced shift in V1 ⁄2 and (b) the effect on channel activation in the absence of [Ca 2ϩ ] i . Whereas the V356A mutant showed no significant changes on either of the two effects (Fig. 8, A,  G, and H), the L360A mutant significantly affected only the Ca 2ϩ -induced shift in V1 ⁄2 (Fig. 8, B, G, and H). In contrast, the K361N significantly left-shifted only the channel's activation in the absence of [Ca 2ϩ ] i (Fig. 8, C, G, and H). Ala mutations of the remaining three hydrophobic residues, Leu 364 , Phe 391 , and Phe 395 , all significantly affected both effects (Fig. 8, D-H). Interestingly, F395A, unlike the other mutants, exhibited a greater inhibition on the channel's activation in the absence of [Ca 2ϩ ] i (Fig. 8, F and H). Thus, mutation of hydrophobic residues whose side chains point toward the crevice between ␣A and ␣B enabled the Ca 2ϩ -dependent effect.
Mutants of all four residues that significantly altered Slo1 activation in the absence of Ca 2ϩ (␣A, K361N and L364A; ␣B, F391A and F395A) also enhanced the apparent affinity for PIP 2 (Fig. 9, A-F). The enhancement of the PIP 2 apparent affinity was smallest for F395A (Fig. 9, D-F). This residue was the only one that stood out from the otherwise perfect correlation between inhibitory effects of residues on the channel's activation in the absence of Ca 2ϩ and their inhibitory effects on PIP 2 apparent affinity. Examination of our modeled structure of the full-length Slo1 channel that incorporated the 3NAF (Ca 2ϩfree) crystal structure suggests that Phe 395 may come in close proximity to the Tyr 336 residue in the C-linker that immediately follows the S6 helix (Fig. 1C). Thus, it is possible that the F395A mutation affected the stability of the channel's open state in a manner that extended beyond its effect on PIP 2 sensitivity. These results revealed that mostly hydrophobic interactions between the ␣A and ␣B helices decreased PIP 2 affinity, suggesting that the coupling between the KDRDD loop and the PIP 2 interaction residues is mediated through specific interactions in these two helices.
Mutants with Decreased Apparent Affinity to PIP 2 Exhibit Increased Current Rundown-Membrane patch excision of channels in ATP-free solutions frequently results in current rundown that can be reversed by application of PIP 2 (e.g. (49). However, for channels with high affinity for PIP 2 , current rundown can be minimal or none. Slo1 excised inside-out patches in 100 M [Ca 2ϩ ] i display minimal current rundown (see single channel data shown in Fig. 3), probably due to their high affinity for PIP 2 (apparent affinity for diC8-PIP 2 is ϳ6 M; see Fig. 4G). In contrast, the K392N and R393N mutants that decreased the channel's apparent affinity for PIP 2 ϳ6 -7-fold caused significant current rundown (Fig. 10, A-H), whereas channel activation and deactivation rates were significantly slowed down (data not shown). ] i , the current traces for the G-V curves were obtained from voltage pulses with a 10-mV increment from Ϫ180 to ϩ180 mV; the holding and repolarizing voltages were Ϫ120 mV. In the absence of [Ca 2ϩ ] i , voltage was stepped from Ϫ100 to ϩ300 mV, and the holding and repolarizing current was at Ϫ100 mV. For clarity, the currents shown in the insets are the traces elicited at Ϫ20, ϩ40, ϩ80, ϩ120, and

DISCUSSION
In this study, we examined the molecular determinants of the coupled relationship between Ca 2ϩ and PIP 2 in enhancing Slo1 activity. We found that Ca 2ϩ relieved a KDRDD loop inhibitory influence on channel activation by increasing the apparent affinity to PIP 2 . Neutralization mutations of three KDRDD loop residues, Lys 366 , Asp 367 , or Arg 368 , also relieved this inhibition in the absence of Ca 2ϩ by enhancing the channel's apparent affinity to PIP 2 . But where did PIP 2 act on the channel, and how did the KDRDD communicate with PIP 2 ? Docking simulations of PIP 2 with models of mouse Slo1 channels (based on crystal structures of the human Slo1 cytosolic domain and the rat Kv1.2/2.1 chimera transmembrane domain) identified Lys 392 and Arg 393 in the ␣B helix as putative PIP 2 -interacting residues. Neutralization mutations of these two residues decreased PIP 2 sensitivity and also the channel activation in the absence of Ca 2ϩ . Interestingly, Arg 393 , which points toward the ␣A helix, also decreased channel activation in the presence of 300 M Ca 2ϩ , whereas Lys 392 , which points toward the membrane, did not (Fig. 7). This result prompted us further to examine resi-dues that enabled communication between the ␣A and ␣B helices. The two available structures of the cytosolic domains reveal a large conformational change in the structural element of ␣A-KDRDD-␣B from the RCK2 Ca 2ϩ -occupied site (3MT5) to the RCK1 site that could potentially be occupied by Ca 2ϩ (3NAF). Mutation of four residues, two from the ␣A helix (K361N and L364A) and two from the ␣B helix (F391A and F395A), all affected channel activation in the absence of Ca 2ϩ with a concomitant effect on the apparent affinity to PIP 2 . With the exception of F395A, all other mutants decreased channel activation in the absence of Ca 2ϩ (Fig. 8H). Phe 395 seems to come close to Tyr 336 (located immediately following the S6 gate), a potential interaction that may influence the response beyond the effects of this residue in its communication with the ␣A helix. The Hill slopes of the PIP 2 dose-response curves were typically less than 1. For Kir channels, the Hill slopes are around 1-2, suggesting that at least 1-2 PIP 2 molecules are required for channel opening. Values less than 1 could signify negative cooperativity among subunits for channel opening. Alternatively, they may signify more than one PIP 2 interaction site with similar affinities. The latter interpretation is consistent with our docking simulations of PIP 2 to the two available crystal structures, which suggest state-dependent interaction modes for PIP 2 .
Our results have provided compelling evidence that when Ca 2ϩ is bound to the Ca 2ϩ -bowl and Ca 2ϩ coordination at the RCK1 site is absent, the ␣A and ␣B helices are tightly coupled to exert an inhibitory effect on channel PIP 2 interactions. Ca 2ϩ binding to the RCK1 site (or mutations that serve to "uncouple" the two helices) seems to relieve the RCK1-mediated decrease in PIP 2 affinity and to enhance channel activation. Consideration of the two available structures (3NAF and 3MT5) suggests that simultaneous Ca 2ϩ binding to both RCK1 and RCK2 sites may not be possible. Our data suggested that ablation of either the Ca 2ϩ -bowl site (5D5N mutant) or the RCK1 E535 coordination site, either of which decreases Ca 2ϩ sensitivity, did not alter the RCK1 site inhibitory effect on activation of the Slo1␣ channel in the absence of Ca 2ϩ (Fig. 6, B, C, and J). Only mutations in the ␣A-KDRDD-␣B structural elements removed the inhibitory effect on channel activity and enhanced PIP 2 sensi-  tivity and activation of the Slo1 channel in the absence of Ca 2ϩ . These results strongly argue that it is not Ca 2ϩ binding per se but rather the KDR residue conformations within the KDRDD loop that control channel activation in the absence of Ca 2ϩ by decreasing PIP 2 sensitivity. Ca 2ϩ binding to Asp 367 serves to relieve this inhibitory effect.
The ␤1 but not the ␤4 accessory channel subunits have been reported to potentiate PIP 2 -induced activation of BK channels (36). Using double mutant cycle analysis, the ␤2 subunit was found to enhance the Ca 2ϩ sensitivity of the Slo1␣ pore-forming subunit by directly coupling its Glu 44 and Asp 45 residues, located just before the first ␤2 transmembrane domain, with the Lys 392 and Arg 393 residues of the ␣B helix of Slo1 (50). Because we found that Lys 392 and Arg 393 are critical residues for PIP 2 sensitivity, it remains to be examined whether the ␤2-mediated enhancement of Slo1 currents is a reflection of altering channel-PIP 2 interactions. Similarly, whether the ␤1 potentiation of PIP 2 -induced activation involves the same ␣B residues remains to be tested.
Our study focused on the relationship of Ca 2ϩ and PIP 2 sensitivity for Slo1 channel activation. We did not investigate channel residues that may also affect sensitivity to PIP 2 but are not part of the ␣B helix. Could such residues be specifically coupled to gating by voltage? Recent work from different laboratories, including ours, has shown that Kv1.2 channels utilize the S4-S5 linker and the N terminus to couple the movement of the voltage sensor to PIP 2 (51,52). Thus, the relationship of PIP 2 and voltage-dependent gating in Slo1␣ channels remains an open question to pursue.
A recent report examining PIP 2 sensitivity of Kv channels in intact cells showed that voltage-gated channels other than Kv7 channels would not respond to a number of manipulations that decreased PIP 2 (53). However, experiments from other studies from excised patches have shown that some of the same channels are sensitive to PIP 2 (see Ref. 53 for discussion). Unlike most Kv channels tested, Kv1.2 and Shaker channels have been shown to be PIP 2 -sensitive in both intact cells and excised patches (51)(52)(53)(54). Several reasons for the differences in PIP 2 sensitivity seen between intact cells and excised patches have been considered (53), yet the relevant question is what is the physiological purpose of a high affinity interaction of a given channel with PIP 2 , if it is not that PIP 2 depletion serves as a signal to inhibit channel activity? Slo1 channels are highly sensitive to PIP 2 . Our studies using dose-response curves with the soluble diC8-PIP 2 following endogenous PIP 2 depletion by scavengers suggest a Slo1a EC 50 (Fig. 4). The Slo3 channel has shown even higher PIP 2 sensitivity (EC 50 ϳ2.5 M) (40). As has been shown for certain channels (55), the length of the acyl chain may also contribute to the apparent affinity of Slo1 channels to PIP 2 , making diC8 PIP 2 assessments of apparent affinity less meaningful (36). Regardless of what the apparent affinity of Slo1 channels is to the native PIP 2 , it is clear that upon patch excision, the currents do not run down as they would for Kir channels with comparable apparent affinity for PIP 2 (e.g. Kir2.1 with a diC8-PIP 2 EC 50 of ϳ2-3 M). Our study has shown that the strong apparent affinity of the Slo1␣ channel to PIP 2 can be utilized in gating the channel by coupling the structural elements ␣A-KDRDD-␣B and decreasing the apparent affinity of the channel for PIP 2 (EC 50 from ϳ2 M in the D367G mutant to ϳ14 M in the WT).