Mutations in Nature Conferred a High Affinity Phosphatidylinositol 4,5-Bisphosphate-binding Site in Vertebrate Inwardly Rectifying Potassium Channels*

Background: A native Kir channel in a distant relative of vertebrates interacts weakly with PIP2. Results: Mutagenesis restores the vertebrate channel sensitivity to PIP2 in sponge channels. Conclusion: A basic residue in the tether helix of Kir is required for high affinity PIP2 regulation. Significance: Evolution conferred a high affinity interaction of vertebrate Kir channels with PIP2, which is lacking in a distant relative. All vertebrate inwardly rectifying potassium (Kir) channels are activated by phosphatidylinositol 4,5-bisphosphate (PIP2) (Logothetis, D. E., Petrou, V. I., Zhang, M., Mahajan, R., Meng, X. Y., Adney, S. K., Cui, M., and Baki, L. (2015) Annu. Rev. Physiol. 77, 81–104; Fürst, O., Mondou, B., and D'Avanzo, N. (2014) Front. Physiol. 4, 404–404). Structural components of a PIP2-binding site are conserved in vertebrate Kir channels but not in distantly related animals such as sponges and sea anemones. To expand our understanding of the structure-function relationships of PIP2 regulation of Kir channels, we studied AqKir, which was cloned from the marine sponge Amphimedon queenslandica, an animal that represents the phylogenetically oldest metazoans. A requirement for PIP2 in the maintenance of AqKir activity was examined in intact oocytes by activation of a co-expressed voltage-sensing phosphatase, application of wortmannin (at micromolar concentrations), and activation of a co-expressed muscarinic acetylcholine receptor. All three mechanisms to reduce the availability of PIP2 resulted in inhibition of AqKir current. However, time-dependent rundown of AqKir currents in inside-out patches could not be re-activated by direct application to the inside membrane surface of water-soluble dioctanoyl PIP2, and the current was incompletely re-activated by the more hydrophobic arachidonyl stearyl PIP2. When we introduced mutations to AqKir to restore two positive charges within the vertebrate PIP2-binding site, both forms of PIP2 strongly re-activated the mutant sponge channels in inside-out patches. Molecular dynamics simulations validate the additional hydrogen bonding potential of the sponge channel mutants. Thus, nature's mutations conferred a high affinity activation of vertebrate Kir channels by PIP2, and this is a more recent evolutionary development than the structures that explain ion channel selectivity and inward rectification.

Inwardly rectifying potassium (Kir) 2 channels represent a superfamily of one pore domain and two transmembrane domain channels. Kir channels are present in organisms from bacteria to humans and are found throughout the animal kingdom. In animal cells, they play an important role in regulating the resting membrane potential, K ϩ homeostasis, and membrane excitability. An ancestral precursor to all Kir channels is predicted to have originated in prokaryotes followed by changes in channel structure and function at the prokaryoticeukaryotic transition and, most likely, throughout the time course of animal evolution (3,4).
Like many ion channels and transporters, the activity of Kir channels from vertebrates, including mammalian and nonmammalian species, is dependent on a membrane lipid, phosphatidylinositol 4,5-bisphosphate (PIP 2 ) (1, 2, 5-7). PIP 2 is found primarily in the inner leaflet of the plasma membrane (8), and its availability is dynamically controlled by neurotransmitter, hormone, and growth factor receptors that couple to phospholipase C (9,10). Signaling that activates phospholipase C is expected to lower the availability of PIP 2 , which reduces or prevents the activation of vertebrate Kir channels. Understand-ing how channels are regulated by PIP 2 is of fundamental physiological importance to cellular signaling and K ϩ homeostasis in eukaryotic cells.
The evolutionary origin of PIP 2 -mediated regulation of Kir channels is unknown. Prokaryotic Kir channels are inhibited by PIP 2 (11,12), whereas Kir channels from all eukaryotic cells studied to date are activated by PIP 2 (2,6,10,13). This is an interesting difference, but PIP 2 is not present in the prokaryotic cell membrane (2,6), and thus its inhibitory effect on bacterial Kir channels lacks physiological relevance. Among eukaryotes, the structural basis for Kir channel activation by PIP 2 has been clarified by many studies (1,2), culminating in the co-crystallization of PIP 2 with chicken Kir2.2 (7). Among the critical components of the protein-lipid interaction are several basic residues within the channel that provide an electrostatically favorable interaction with the negative charges on the phosphate groups of PIP 2 . However, despite the tremendous insight gained by many years of physiology and structural biology experiments, the functional priority of the basic residues in the Kir channel and whether they are sufficient to explain high affinity interactions of the lipid and channel remain untested outside the confines of vertebrate species such as rat, mouse, human, and chicken. Because there is a 400 -600 million year evolutionary gap between humans and basal invertebrates such as sponges (14), there is a good possibility that evolutionary pressures resulted in structural changes to regulatory sites on ion channels (15).
To explore the impact of evolutionary pressure on structural changes in ion channel proteins, we studied PIP 2 regulation of AqKir, a K ϩ channel with strong inward rectification, which was previously cloned from the marine sponge Amphimedon queenslandica (16). Sponges are an important comparative group for understanding the relationships between extant unicellular protists and the first multicellular organisms from which all animals evolved (14,(17)(18)(19)(20). By studying the physiological and structural impact of "nature's mutations," we provide a window into the functional consequences of evolution that is represented in distantly related animal species. A Kir channel cloned from a marine sponge (AqKir) lacks two positive charges that are important for high affinity interaction with PIP 2 in vertebrate Kir2 channels (7). We report that, following time-dependent rundown, AqKir channels are insensitive to re-activation by PIP 2 added to the internal surface of an excised membrane patch, despite additional evidence from intact oocytes that suggests that PIP 2 may be required for this channel's function. In addition, AqKir channels, which we engineered to possess the two positive charges present in the PIP 2binding site of vertebrate Kir2 channels, showed strong activation by diC 8 PIP 2 , with a sensitivity that was even 3-fold greater than that for vertebrate Kir2.1 channels. Our results confirm that certain positively charged residues (corresponding to residues 80 and 189 in cKir2.2) are absolutely necessary to confer high sensitivity to PIP 2 and represent evolutionary changes that are more recent than the evolution of structures that confer K ϩ selectivity and inward rectification, which are conserved from sponge to human (16). Finally, we propose that other structures present in distant relatives of vertebrate Kir channels must facilitate the activation properties conferred by PIP 2 such that restoration of even a single positive charge (N180K in AqKir) recapitulates the high sensitivity to PIP 2 present in vertebrate Kir channels.
For inside-out patch clamp recordings, we used an A-M Systems Model 2400 or an Axon Instruments 200B patch clamp amplifier. Macropatches were excised from oocytes following manual removal of the vitelline membrane (24 -26). Electrodes were pulled from borosilicate glass using a Sutter P-97 puller and gave a tip diameter of 5-15 m and a resistance of 0.5-2 megohms when filled with a solution containing (in mM) 96 KMeSO 3 , 1 MgCl 2 , and 10 HEPES, pH 7.4. The standard bath (internally facing) high K ϩ ("HK") solution contained 96 KCl, 5 EGTA, and 10 HEPES, pH 7.4, unless otherwise mentioned.
Currents were elicited with a voltage ramp protocol from Ϫ100 to ϩ100 mV over 1 s. Signals were low pass-filtered at 10 kHz and digitized at 20-s intervals.
Electrophysiological data were recorded on computers equipped with Digidata 1320A (Axon Instruments) A/D hardware. Axon's Clampex acquisition and Clampfit analysis software (versions 7-9) were used.
Data Analysis and Curve Fitting-Data were also transferred to Microsoft Excel and Microcal Origin (Northampton, MA) for additional analysis, curve-fitting, and the production of figures. Results were calculated from cells with negligible background currents and for TEVC, only from cells with membrane potentials not more depolarized than Ϫ30 mV in 5 mM external K ϩ . Results are expressed as mean Ϯ S.E. with n ϭ number of oocytes tested with a minimum of three batches of oocytes per experiment. The EC 50 value of PIP 2 effects for each channel was obtained by fitting concentration-response data to a Hill Equation: where I/I max is the normalized current; [X] i is the PIP 2 concentration applied from the intracellular side; n is the Hill coefficient, and EC 50 is the concentration required to reach half of the maximum activation.
In cases with multiple comparisons to a single set of control values, a single factor ANOVA with a post-hoc Dunnett's test was used with p Ͻ 0.05 considered to be significant. Other statistical evaluations included paired or unpaired Student's two-tailed t-tests, as appropriate, with p Ͻ 0.05 considered to be significant.
Chemicals-AASt ("long chain") PIP 2 was purchased from Avanti Polar Lipids (Alabaster, AL). diC 8 PIP 2 was purchased from Avanti Polar Lipids or Echelon Biosciences (Salt Lake City, UT). The lipids were prepared as described previously (23,25,27). All other chemicals were purchased from Sigma or VWR.
Modeling of AqKir/PIP 2 Channel Complexes-The AqKir channel model was constructed based on the chicken Kir2.2 x-ray crystal structure template (PDB code 3SPI (7)). The sequence alignment between Kir2.2 and AqKir channels was generated by ClustalW2. The sequence identity between Kir2.2 and AqKir is ϳ40%. We used the program MODELLER 9 version 12 (28) to generate 10 initial homology models. The model assessment and selection were based on the internal objective function, 1,2-dioleoyl-sn-glycero-3-phosphoethanolamine, and GA341 assessment scores in the MODELLER program to select models with the lowest free energy potentials. The full-length PIP 2 molecules were added into the AqKir model by superimposition with the short-chain derivative of PIP 2 in the Kir2.2 crystal structure (7). Models for AqKir channel mutants were constructed by a mutation function using Discovery Studio version 3.5 molecular modeling program (Accelrys Software Inc., San Diego).
Molecular Dynamics Simulations-The PIP 2 -AqKir channel complexes were immersed in an explicit palmitoyloleoylphosphatidylcholine bilayer generated from the VMD membrane package (29). After being solvated with SPC water molecules, neutralized by KCl (150 mM) as the counter ions and including K ϩ located in the selectivity filter as obtained from the crystal structures, each system involved ϳ140,000 atoms in the MD simulations. GROMACS version 4.5.4 (30) was used to conduct the simulation with the GROMOS96 53a6 force field (31). The parameters for PIP 2 and lipid were generated as described previously (32). Long range electrostatics were calculated using the particle mesh Ewald (PME) method (33) with a 12 Å cutoff. van der Waal interactions were modeled using Lennard-Jones 6 -12 potentials with 14-Å cutoff. All simulations were conducted at a constant temperature of 300 K using the Berendsen thermostat. The system pressure was coupled at isotropic (X ϩ Y, Z) directions referenced to 1 bar using the Berendsen method (34). All bonds were constrained with the LINCS algorithm (35). The time step was 2 fs, and the neighboring list was updated every 10 time steps.
Prior to production runs, energy minimization of 3000 steps of the steepest descent were carried out on each system followed by a 0.5-ns two-step equilibration process. In the first 0.2 ns, channels, K ϩ ions, and PIP 2 were position-restrained using a constant force of 1000 kJ/mol/nm 2 , allowing lipid and water molecules to move freely. The restraint was weakened to 10 kJ/mol/nm 2 in the following 0.3-ns equilibration. An electrical field of 0.02857 V/nm was applied in this step as well as the production run, along the z axis of the box to maintain the lower potentials in the intracellular side. A 40-ns production run was conducted on each system, and coordinates were saved every 10 ps for analysis.
Interaction Energy Calculations-The Discovery Studio 3.5 molecular modeling program (Accelrys Software Inc.) was used to conduct energy minimizations and calculate the interaction energy between the PIP 2 and the AqKir channels. All structures were subjected to energy minimization using Smart Minimizer algorithm (2000 steps) and Generalized Born implicit solvent model using the CHARMm force field. Interaction energies between the PIP 2 and the channel were calculated using a distance-dependent dielectric constant (⑀ ϭ 5r) implicit solvent model. The decomposed interaction energy contribution of residues to the total interaction energy was analyzed to identify the critical interacting residues in the protein complexes. A residue was considered to be critical if it bound the ligand with an interaction energy lower than Ϫ1 kcal/mol.

Results
Sequence Comparisons-An amino acid sequence comparison between sponge and vertebrate Kir channels shows that AqKir is missing two canonical positively charged residues at positions 78 and 180, within the outer and tether helices, respectively (Fig. 1A), of the PI(4,5)P 2 -binding site in the crystal structure of chicken Kir2.2 (7) and is represented in a homology model of AqKir (Fig. 1B). Both positive charges are present in members of all but one of the seven vertebrate Kir channel subfamilies (Fig. 1A). Thus, AqKir is the only known functional Kir channel that lacks multiple residues previously shown to be important for regulation by PIP 2 (27,36). A BLAST search for additional distant relatives of the vertebrate Kir channels found three putative Kir channels in the genome of the sea anemone Nematostella vectensis (Nv). All three lack a positive charge within the tether helix region of the PIP 2 -binding site in cKir2.2, although the positive charge in the outer helix region, which is missing in the sponge channel, is preserved in the three sequences from sea anemone (Nv1, His; Nv2-3, Arg; Fig. 1A). As both sponges and sea anemones are modern day representatives of more primitive metazoans, these structural similarities provide evidence for the impact of evolution on the function of vertebrate Kir channels.
In several experiments, we tested a catalytically inactive mutant of CiVSP (C363S (21) and observed no change in Kir current amplitudes upon membrane depolarization (data not shown). As a second test for PIP 2 dependence, the amplitude of resting Kir currents was measured after preincubation with 20 M wortmannin, a concentration that inhibits phosphatidylinositol 4-kinase activity and thus lowers endogenous levels of PI(4,5)P 2 (40). Compared with time-matched, vehicle-treated (DMSO) controls using oocytes from the same frogs, wortmannin pretreatment resulted in a significant reduction in the resting AqKir and Kir2.3 currents but not Kir2.1 currents (Fig. 2C). Notably, Kir2.1 has a higher affinity for PIP 2 interaction than does Kir2.3 (41). Third, when muscarinic (M1) acetylcholine receptors were co-expressed with AqKir and activated by acetylcholine, the inward AqKir K ϩ currents were inhibited (Fig. 2, D and E), whereas Kir2.1 currents were not inhibited (Fig. 2E). The muscarinic inhibition of AqKir mimics the established effect on Kir2.3 currents by M1 receptor activation of a G protein-coupled phospholipase C (41). Together, these three results from intact oocytes suggest that, despite the lack of pos-itively charged residues in key locations of the vertebrate Kir2 channel PIP 2 -binding site (Fig. 1), AqKir demonstrates known characteristics of PIP 2 -dependent channel regulation. However, the effects of CiVSP and wortmannin suggest that the AqKir-PIP 2 interaction is weaker than that for the Kir2.1-PIP 2 interaction.
Loss of Activity in Excised Patches-To further test the prediction of PIP 2 dependence of AqKir channels, we examined the impact of membrane patch excision on current activity of AqKir channels. In oocytes expressing AqKir currents, patch excision first resulted in a transient increase in current amplitude, which was quickly followed by decay of the current such that channel activity was lost within 2 min (Fig. 3A). In 90% of the patches, the "rundown" of AqKir current demonstrated two phases, as observed in Fig. 3A (black circles). Overall, the loss of AqKir current was faster than the well known, time-dependent decay of vertebrate Kir channels, cKir2.2, and particularly Kir2.1, measured under the same ionic conditions. We found that the time to decay to 50% of the initial on-cell AqKir current was 12-fold faster than for Kir2.1 and 4-fold faster than for cKir2.2 currents (Fig. 3B). As rundown of Kir channels is commonly explained by the loss of PIP 2 from the excised patch due to phosphatase-mediated dephosphorylation of PIP 2 (42), the faster rate of AqKir current decay suggests that a higher concentration of PIP 2 is needed to sustain the sponge Kir channels than the vertebrate Kir channels. On-cell Current Magnitude-On-cell patch recordings obtained before patch excision showed that the average magnitude of the AqKir channel currents was 24-fold lower than that found for Kir2.1 (Figs. 3C and 4C). Likewise, the wholecell current amplitudes measured in TEVC showed a markedly lower current level for AqKir (Ϫ1.9 Ϯ 0.39 A) when compared with cKir2.2 (Ϫ4.6 Ϯ 1.0 〈) or Kir2.1 (Ϫ60 Ϯ 4.8 A; recordings in HK solution). However, a double mutant I78R/N180K restored large AqKir current amplitudes in TEVC (Ϫ65 Ϯ 10 A; n ϭ 7-17) to values comparable with the Kir2.1 channels. The reduced magnitude of the wild-type AqKir channel activity in the on-cell patch and in whole-cell recordings is consistent with the hypothesis that wild-type AqKir channel activity is limited by weaker interactions of PIP 2 with the wild-type AqKir channel than with mutant AqKir channels.
Application of PIP 2 to Inside-out Patches-To further test this hypothesis, we used direct application of two forms of PIP 2 to the inside-facing surface of inside-out membrane patches. The lipids were prepared fresh from frozen stocks and applied in the HK bath solution using a gravity-driven perfusion system. Long chain (AASt) PIP 2 contains the 20-carbon polyun-saturated fatty acid, arachidonic acid (AA) and the 18-carbon saturated stearic acid (St) in its two acyl tails. The hydrophobic nature of AASt-PIP 2 limits its use to single dose applications to membranes or bilayers. In contrast, water-soluble diC 8 PIP 2 contains two 8-carbon unsaturated fatty acid tails and can be applied reversibly, multiple times; it is most useful in quantifying the concentration dependence of PIP 2 and comparing the relative sensitivities for the lipids in different channel constructs.
The experiment in Fig. 4 tests the ability of direct application of AASt PIP 2 to re-activate Kir currents following rundown due to patch excision (i/o arrow). At a concentration of 10 M, AASt PIP 2 completely restored the Kir2.1 current (Fig. 4A) but even 25 M incompletely re-activated AqKir (Fig. 4B). In these recordings, the on-cell current magnitude was 24-fold higher for Kir2.1 when compared with AqKir (Fig. 4C), similar to previous results shown in Fig. 3C. To negate the impact of the initial (on-cell) current value on the strength of the re-activation by PIP 2 in each construct, we compared the magnitude of the recovered current after AASt PIP 2 (Fig. 4D, black bars) with the on-cell current amplitude before rundown (Fig. 4D, gray bars). AASt-PIP 2 (10 M) reactivated all of the Kir2.1 current but only about 25% of the AqKir WT current (Fig. 4D). This incomplete reactivation of the sponge Kir channel activity is consistent with results from our TEVC experiments suggesting that PIP 2 interactions with AqKir are weaker than for Kir2 channels.
Direct application of diC 8 PIP 2 to the intracellular face of the membrane patches expressing AqKir also did not restore channel activity (Fig. 5). The lowest tested concentrations of diC 8 PIP 2 (0.1 and 1 M) showed only very small increases in AqKir currents after rundown (Fig. 5A), and higher concentrations showed no activity. Clearly, diC 8 PIP 2 could not restore the sponge channel activity. On average, the current activity with diC 8 PIP 2 was so low that AqKir WT current amplitudes reached, at most, 10% of the on-cell patch current. Thus, a concentration-response relationship for channel activation could not be determined for the wild-type sponge Kir channel (see Fig. 6C, black circles), although Kir2.1 showed characteristic re-activation by diC 8 PIP 2 using the same recording conditions and same preparations of the lipid (Fig. 5, B and D). Currentvoltage relationships for the two constructs showed that the strongly rectifying inward AqKir currents were not recovered by even 100 M diC 8 PIP 2 (Fig. 5C), whereas Kir2.1 was maximally activated by 10 M diC 8 PIP 2 (Fig. 5D).
Recapitulating Mutations in Nature in AqKir-Channel proteins that associate with PIP 2 are known to engage in electrostatic interactions with the negatively charged headgroup of PIP 2 and positively charged residues in the protein (2,7). To test the critical nature of the two basic residues in vertebrate Kir channel-PIP 2 interactions (Fig. 1), we used site-directed mutagenesis to add the positive charges to AqKir, individually and together. Although wild-type whole cell and on-cell AqKir currents were small and failed to be restored by diC 8 applied to excised patches (Fig. 5A, 6A), even a single mutation, N180K, restored the capability of re-activation by diC 8 PIP 2 (Fig. 6, A  and C). The effects of diC 8 on N180K were dose-dependent, reversible, and saturated by 30 M (Fig. 6, A and C); this mim- The dotted line indicates zero current. B, time to 50% of maximal rundown (t 50 ) was measured for each construct (n ϭ 6 -15). C, magnitude of the on-cell current prior to patch excision is compared by normalizing to the average magnitude for Kir2.1 currents (n ϭ 6 -15). The cRNAs for Kir2.1 and cKir2.2 were injected into oocytes at the same concentration, but AqKir was injected at ϳ4-fold higher concentration, as needed to obtain measurable currents.
icked the effect of diC 8 on vertebrate channels such as Kir2.1 (Fig. 6C). We also tested a double mutant, I78R/N180K, which restored the two missing positive charges to AqKir and resulted in large Kir currents in whole-cell recordings (Fig. 6B, red bar). The double mutant was exquisitely sensitive to diC 8 PIP 2 ; the on-cell current amplitude was completely recovered with less than 10 M of the lipid (Fig. 6, A and C). Fitted concentrationresponse relationships (Fig. 6D) showed that the single mutant AqKir N180K has an apparent affinity similar to the vertebrate Kir2.1 channel, whereas the double mutant I78R/N180K has a higher sensitivity to diC 8  An additive effect on diC 8 sensitivity of I78R in the double mutant (Fig. 6C, red circles) points to an important role of this arginine residue although, in macropatches, we could not reliably measure currents from the single mutant, I78R, following diC 8 application. However, the mutant I78R produced measurable currents for TEVC recordings (Figs. 6B and 7) and was capable of re-activation by AASt PIP 2 in inside-out patch recordings (Fig. 8, C and D).
We also examined the AqKir mutant channels for changes in CiVSP-mediated inhibition of whole oocyte current. The previously documented fast inhibition of whole-cell wild-type AqKir currents by depolarization-mediated CiVSP activation ( Fig. 2A) was slowed by single or double mutations at residues Ile-78 and Asn-180 (Fig. 7, A and C). The time course of recovery with hyperpolarization to de-activate CiVSP showed the inverse effect; recovery was slow in WT AqKir but faster with mutation of Ile-78 or Asn-180 or both (Fig. 7, B and D). Indeed, quantitative comparisons of the time to 50% inhibition (t 50 ) and the t 50 for recovery following inhibition point to an important role for each positive charge alone and an additive effect of the two mutations (Fig. 7, C and D). Overall, the data suggest that the loss of PIP 2 due to phosphatase activity has a stronger impact on WT than on mutant AqKir channels, which further demonstrates that these critical positive charges support the high sensitivity of PIP 2 interaction with the channel. Furthermore, our results suggest that PIP 2 concentrations in the membrane of intact cells (frog oocytes) falls to very low levels upon CiVSP activation, less than that required to sustain the activity of the double mutant AqKir channels for which 1 M diC 8 PIP 2 maximally activated them.
Neomycin Inhibition-Neomycin, an amino-glycosidic antibiotic, is a positively charged molecule that binds with high affinity to PIP 2 ; its application to the inside surface of the membrane is expected to inhibit PIP 2 -dependent currents by binding to the anionic headgroup of the lipid (43,44). Thus, we used neomycin to further examine the relative sensitivities to PIP 2 of AqKir mutant channels and the vertebrate Kir2.1 channel. Kir channel currents were allowed to run down in inside-out patch recordings, and long chain AASt-PIP 2 was applied to restore the Kir currents to sustained levels. We then measured the inhibition of Kir2.1 (Fig. 8A) and AqKir mutant channels (Fig. 8B) by increasing concentrations of neomycin in a stepwise fashion. The fractional current inhibition for five constructs was plotted as a function of neomycin concentration, and each data set fitted to a standard Hill equation (Fig. 8C). The fitted K d values for neomycin inhibition quantify the strength of PIP 2 binding (Fig. 8D). The strong neomycin inhibition measured for   AqKirWT was made weaker with each of the single mutants, and the K d value for neomycin inhibition of the lipid-reactivated double mutant channels was even higher than that mea-sured for Kir2.1. These data are consistent with the strengthened diC 8 PIP 2 interactions measured for N180K and the double mutation in AqKir (Fig. 7).  . Mutant AqKir channels reactivated by PIP 2 have a lower sensitivity to inhibition by neomycin than Kir2.1 channels. Inward K ϩ currents recorded in excised patches in HK solution at Ϫ100 mV were briefly exposed to AASt PIP 2 (ϩ 2 mM Mg-ATP) to stably reactivate the currents and then exposed to different concentrations of neomycin. Docking and Energy Analysis-To understand how the mutations affect the AqKir channel and PIP 2 interactions, we developed homology models of AqKir, both wild-type and single and double mutants, in complex with PIP 2 . Fig. 9 depicts a snapshot of AASt PIP 2 bound to AqKir WT (Fig. 9A) and AqKir I78R/ N180K (Fig. 9B). In the double mutant, the positive charges on Lys-180 and Arg-78 orient more closely to the PIP 2 molecule.
We conducted MD simulations on each model system in a membrane environment with explicit solvent and calculated the energy of the PIP 2 -channel interactions. Our results show that PIP 2 docked in AqKir WT, at a site that maps to the Kir2.2 PIP 2 -binding site, has a total interaction energy of Ϫ90.6 kcal/ mol. This interaction became more favorable with the I78R/ N180K mutant, which had a total interaction energy of Ϫ127.4 kcal/mol. Each single mutant also improved the overall favorability of interaction with PIP 2 ; the total interaction energies for I78R and N180K were Ϫ113.4 and Ϫ121.0 kcal/mol, respectively.
To assess quantitatively the importance of specific chemical bonds in the proposed interaction of PIP 2 within the sponge Kir channel, we used energy decomposition analysis ( Table 1). Addition of one basic residue to AqKir increases the total interaction energy ϳ2-fold for I78R (from Ϫ6.77938 to Ϫ11.5338 kcal/mol) and ϳ50-fold for N180K (from Ϫ0.51548 to Ϫ25.1471 kcal/mol). In each case, the electrostatic energy component nearly matched the total interaction energy at that residue. Thus, the more favorable interaction with PIP 2 could be largely explained by improved electrostatic interactions. Comparatively, the mutant N180K, alone or in combination with I78R, had a stronger impact on energetic favorability than the positive charge added at residue 78. This is consistent with the higher priority for N180K in recapitulating the functional reactivation of AqKir by diC 8 PIP 2 (Fig. 6). Fig. 10 provides a visual representation of the favorable changes in electrostatic interactions between PIP 2 and the WT (Fig. 10A) and mutant (Fig. 10, B-D) AqKir channels. The sandwiching of the PIP 2 headgroup between the basic side chains at positions 78 and 180 of the double mutant channel (Fig. 10D) provides additional potential for hydrogen bonding interactions between channel and ligand that are consistent with the enhanced total and electrostatic interaction energy of this mutant. The WT channel may form only one weak (3.2-4.0 Å) hydrogen bond with PIP 2 , whereas the mutants form a mixture of multiple weak and moderate (2.5-3.2 Å) hydrogen bonds with PIP 2 . Overall, each AqKir mutant has stronger hydrogen bonding potential than the WT AqKir channel, which is consistent with an increase in the apparent affinity for PIP 2 in the mutant AqKir channels.

Discussion
To better understand the impact of evolutionary changes in ion channel structure and function, we studied the PIP 2 regulation of a sponge Kir channel using mutagenesis, electrophysiology, and MD simulations using a homology model. PIP 2 is a known regulator of many ion channel and transport molecules (1,2). The solution of two crystal structures of vertebrate Kir channels complexed with PIP 2 highlights the interaction of positively charged residues between the transmembrane domain and the C terminus with the negatively charged headgroup phosphates (7,45). The sponge Kir channel, AqKir, a distant relative of vertebrate Kir channels, lacks two basic residues in the PIP 2 -binding site. This structural difference predicts low sensitivity to PIP 2 regulation in the sponge Kir channel.  In whole-cell (TEVC) recordings, blocking PIP 2 synthesis or stimulating PIP 2 hydrolysis or dephosphorylation reduced AqKir currents, effects typically associated with a requirement for regulation by PIP 2. Direct application of PIP 2 to inside-out membrane patches failed to restore AqKirWT channel activity after rundown, consistent with an even lower apparent affinity for PIP 2 following excision, as could occur due to loss of a posttranslational modification that is not maintained under excised patch conditions (46). Consistent with this interpretation of our results, in TEVC and excised patches, AqKirWT currents were typically smaller than the AqKir mutant channels that displayed much higher affinity interactions with PIP 2 .
The partial reactivation of wild-type AqKir by long chain AASt PIP 2 , but not diC 8 PIP 2 , suggests that hydrophobic interactions with the acyl chain of the lipid could be important in the mechanism of PIP 2 activation of AqKir current, but alone are insufficient. It is also possible that other phospholipids maintain the WT current due to species-dependent differences in lipid specificity (47)(48)(49), which has not yet been examined in this channel.
Multiple positively charged residues located in the N and C termini of vertebrate Kir channels are important in coordinating with the negatively charged headgroup phosphates of PIP 2 (1,2,7,45). Do the various basic residues play equivalent roles? We have clarified the molecular significance by studying a channel that natively lacks these positively charged residues. Vertebrate levels of PIP 2 sensitivity could be restored by mutational insertion of just one of the basic residues in AqKir, N180K (represented in 4-fold symmetry by the tetrameric channel). These results signify a particularly important role for this residue. MD simulations provided an explanation (see Fig.  10) in that N180K had the greatest impact on strengthening electrostatic interactions. Interestingly, a search for homologs of the sponge Kir channel in other distantly related metazoan species discovered putative Kir channels in sea anemones, which also lack the critical positive charge in the tether helix region. We suggest that mutation at this residue (Lys-K189 in cKir2.2) over the course of vertebrate evolution was perhaps the most critical final change that brought about the PIP 2 dependence of vertebrate Kir channels.
Of course, measurements of channel activity in the excised patches following PIP 2 application cannot determine directly whether the mutations affect PIP 2 binding or channel gating. To this end, the MD simulations provide an explanation that highlights the more energetically favorable molecular interactions following recapitulation of one or both of the positive charges ( Fig. 10 and Table 1).
Whether the sponge Kir channel is physiologically regulated by changes in PIP 2 concentrations in its native environment is unknown, but the possibility surely exists. We ran a KEGG analysis (50) of the sponge genome and confirmed that all of the required components for inositol phosphate metabolism are present, and molecular cloning has previously reported that G q , phospholipase C, and protein kinase C exist in sponges (51). Thus, the capability for PIP 2 regulation of Kir channels or other proteins via phospholipase C signaling pathways is evolutionarily old, with origins at the earliest branch point of extant animal phyla.

Conclusion
Nature's mutations over the course of animal evolution have resulted in a high affinity PIP 2 regulatory site within vertebrate Kir channels. However, critical features that confer high sensitivity to PIP 2 in the vertebrate channels are missing in Kir channels from basal metazoans such as the channel from A. queenslandica studied here. The greatly enhanced PIP 2 sensitivity of AqKir mutants containing two positive charges confirms the absolute necessity of these basic residues for PIP 2 regulation, which is consistent with our homology model and MD simulations of AqKir complexed with PIP 2 . Furthermore, our results support a functional priority for a single positive charge (residue 180 in AqKir) in recapitulating the high sensitivity to PIP 2 .
Changes in the binding site over time, in response to forces that drive evolution, could include changes in the bioavailability of PIP 2 (15) or molecules involved in cell-cell signaling. Conservation of Kir channel selectivity and inward rectification mechanisms (16) have been retained over millions of years of evolution, but acquisition of a new function, PIP 2 regulation by high affinity interaction, is more recent, occurring after the prokaryotic-eukaryotic split.