Psalmotoxin-1 Docking to Human Acid-sensing Ion Channel-1*

Acid-sensing ion channel-1 (ASIC-1) is a proton-gated ion channel implicated in nociception and neuronal death during ischemia. Recently the first crystal structure of a chicken ASIC was obtained. Expanding upon this work, homology models of the human ASICs were constructed and evaluated. Energy-minimized structures were tested for validity by in silico docking of the models to psalmotoxin-1, which potently inhibits ASIC-1 and not other members of the family. The data are consistent with prior radioligand binding and functional assays while also explaining the selectivity of PcTX-1 for homomeric hASIC-1a. Binding energy calculations suggest that the toxin and channel create a complex that is more stable than the channel alone. The binding is dominated by the coulombic contributions, which account for why the toxin-channel interaction is not observed at low pH. The computational data were experimentally verified with single channel and whole-cell electrophysiological studies. These validated models should allow for the rational design of specific and potent peptidomimetic compounds that may be useful for the treatment of pain or ischemic stroke.

Acid-sensing ion channels are a subfamily of the epithelial sodium channel/degenerin family of proteins (1). The proteins in this family share a general topology; each member has two transmembrane-spanning domains, relatively short intracellular N and C termini, and a large extracellular loop containing multiple cysteine-rich domains. These proteins interact with themselves and other family members to form ion channels with unique properties (1,2). The channels formed are all functionally linked by sensitivity to the small molecule inhibitor amiloride and a general selectivity for conducting sodium despite a sequence identity ranging from 15 to 60% across the family. The epithelial sodium channel/degenerin proteins are important for many physiological and pathophysiological processes. For example, ␣␤␥-epithelial sodium channel channels expressed in the kidney are important in blood pressure homeostasis, whereas homomeric ASIC-1 channels found in neurons are implicated in nociception and neuronal death during ischemia (1,3).
Using fluorescence detection size exclusion chromatography, Jasti et al. (4) found that homomeric chicken acid-sensing ion channel-1 (ASIC-1) 2 could be crystallized and described the structure of a homomeric cASIC-1 channel lacking intracellular domains and in a nonfunctional state. Unfortunately chicken ASICs are very poorly characterized either pharmacologically or functionally. As crystallization of integral membrane proteins is a challenging technique, obtaining crystal structures of the human members of this family may not be feasible. Despite species difference of just a few amino acids, there may be clinically relevant variations between ASIC homologs (5).
This work used homology modeling to deduce structures of the human ASICs. These models were evaluated using structural verification suites and validated using in silico inhibitor docking to recapitulate functional results. Binding predictions were experimentally validated with single channel and wholecell electrophysiology experiments. Together these models and the docked complex create structures that may be used for the rational design or virtual screening of new inhibitors or peptidomimetic compounds.

EXPERIMENTAL PROCEDURES
Template Structures-The work of Gouaux and co-workers (4) described the structure of Gallus gallus ASIC-1 arranged to form a homomeric channel (available as Protein Data Bank code 2QTS). This structure contains six cASIC-1 subunits arranged to form two channels. Chains A, B, and C of the structure were of higher quality and thus were used for modeling (4). Heteroatoms other than the chloride ions were removed.
Target Sequences-The amino acid sequences for full-length chicken ASIC-1, human ASIC-1b, hASIC-2b, hASIC-3a, and hASIC-4a were obtained from the NCBI Protein Database. Pairwise alignments were performed using AlignX in Vec-torNTI Advance 10.3 (Invitrogen). Results of the alignments, as well as accession numbers, are shown in Table 1. Splice variants were chosen to minimize gaps in the alignments. It should be noted that there is considerable confusion in the field regarding human ASIC-1 as compared with the rodent ASIC-1. To clarify, there are two known variants of hASIC-1, which are both very similar to the rodent ASIC-1a. The NCBI recognizes hASIC-1 variant 1 or "isoform a" as the splice variant containing a 46amino acid insertion, whereas "isoform b" is lacking this inser-* This work was supported, in whole or in part, by National Institutes of Health Grant DK37206. This work was also supported by the American Medical Association Foundation. □ S The on-line version of this article (available at http://www.jbc.org) contains supplemental Tables 1-3  tion. However, as isoform b is most identical to rASIC-1a or mASIC-1a, it has often been labeled as hASIC-1a in prior publications. To our knowledge, no experimental work has been done with the actual hASIC-1a, and it is unclear whether it is of biological importance. Homology Modeling-MODELLER 9v2 was used to perform automatic homology modeling of templates from cASIC-1 (6). N and C termini were removed from target sequences as no data are available for them within the crystal structure of cASIC-1. Support for the chloride ions was enabled in MOD-ELLER, and they were considered during the modeling of the channels. The scripts used for modeling, in addition to adding support for the chloride ions, increased the thoroughness of the default optimization protocol. The variable target function method optimization was set to "slow" with the maximum iterations set at 300. The molecular dynamics with simulated annealing optimization was also set to slow, and the entire process was repeated three times to generate one high quality model. Models for homomeric hASIC-1b, hASIC-2b, hASIC-3a, and hASIC-4a channels were created. As an internal control, the template was submitted to MODELLLER as a target for modeling against itself. Following modeling, each structure was parameterized with the GROMOS96 43a1 force field and solvated with a simple point charge water model using genbox with GROMACS 3.3.1 (7). The system was energy-minimized using the steepest descents algorithm with no position restraints until the system converged to machine precision on the Coosa computer cluster at the University of Alabama at Birmingham.
Peptide Docking-NMR structures for the peptide psalmotoxin-1 were obtained from Protein Data Bank code 1LMM. As this was a solution-derived NMR structure, the top 20 poses of the toxin were separated into individual files, and the hydrogen atoms were removed for inputting into ZDOCK 3.0.1 (8). The transmembrane portions of the models were blocked as they are known to not participate in the interaction of psalmotoxin-1 (PcTX-1) and ASIC-1 (9). Docking was performed between the channel structures and each of the top 20 NMR solution structures of PcTX-1 (10). Further docking of PcTX-1 to ASIC-1 was performed using the "dense" flag that uses a smaller rotational step to obtain more refined docking results at the cost of computational cycles. These results were reranked using ZRANK to obtain the best docked pose (11). Structures of the docked state at pH 6, 7, and 8 were calculated using PDB2PQR 1.3.0 and PROPKA 2.0 (12)(13)(14).
Structural Visualization-Models were visualized using Visual Molecular Dynamics from the University of Illinois. Figures were rendered using Tachyon (15). Calculations of interaction energies were performed on the best docked pose using the Adaptive Poisson-Boltzmann Solver 1.0.0 (16 -20).
Planar Lipid Bilayer Recordings-Planar lipid bilayer recordings were conducted as described previously (21). Briefly hASIC-1b was expressed in Xenopus oocytes, and vesicles were isolated. These ASIC-1-containing vesicles were fused to a planar lipid bilayer bathed with symmetrical 100 mM NaCl and 10 mM MOPS, pH 6.2. The holding potential was ϩ100 mV referred to the virtually grounded trans chamber. Synthetic toxin was produced and purified by Pneumosite LLC (Shreve-port, LA). Dwell time histograms were constructed following the analyses of events performed using pCLAMP software (Axon Instruments) on single channel recordings of 10 min in duration filtered at 300 Hz with an eight-pole Bessel filter before acquisition at 1 ms per point using pCLAMP software and hardware. The event detection thresholds were 50% in amplitude of the transition between closed and open states and 3 ms in duration. Closed and open time constants shown were determined by fitting the closed and open time histograms to the probability density function g(x) ϭ ⌺ j ϭ 1 k a j g o (x Ϫ s j ) where s j is the logarithm of the jth time constant and a j is the fraction of total events represented by the jth component (22) and using the Simplex least square routine of pSTAT. The number of bins per decade in all histograms was 16. For illustration purposes records shown were digitally filtered at 100 Hz using pCLAMP subsequent to acquisition of the analog signal.
Plasmids and Mutant Construct Generation-The constructs used in this study include the bicistronic pBi-eGFP: hASIC-1b and pBi-eGFP:hASIC-2b plasmids, the parent pBi-eGFP vector, pcDNA3.1-hASIC-2b, and a mutant hASIC-2b (muthASIC-2b) construct containing mutations E344D, G348D, L349F, and A351V rendering residues 340 -355 of hASIC-2b identical to residues 343-358 of hASIC-1b. PAGE-purified oligonucleotides encoding for the mutated domain were purchased from Invitrogen and used in Stratagene's QuikChange II XL Site-Directed Mutagenesis kit according to the manufacturer's protocol. The mutations were verified by sequencing the domain in question at the Genomic Core Facility of University of Alabama, Heflin Center for Human Genetics.
Cell Culture and Transfection-CHO-K1 cells were maintained in 1:1 Dulbecco's modified Eagle's medium/F-12 (Hyclone) supplemented with 10% fetal bovine serum (Hyclone) and 1% penicillin/streptomycin (Invitrogen). Cells were transfected with plasmids using Lipofectamine 2000 (Invitrogen) according to the manufacturer's protocol using 4 g of total DNA and 12 l of lipid. Cells were replated onto glass coverslips and used 24 -96 h later. For homomeric hASIC-1b or hASIC-2b channels, the bicistronic constructs were transfected individually allowing for patching of visually green cells expressing channel protein. For patching homomeric muthASIC-2b channels, a ratio of 1 g of pBi-eGFP parent vector to 3 g of hASIC-2b chimera DNA was cotransfected. For heteromeric channels, a bicistronic construct with a fluorescent reporter was cotransfected in a 1:3 ratio with another plasmid. The 1:3 ratio ensures that cells expressing the fluorescent tag are very likely to be expressing the cotransfected plasmid as well.
Patch Clamp-Micropipettes with an electrical resistance of 3-5 megaohms were prepared using a Narashigi PP-83 twostage micropipette puller and filled with 120 mM KCl, 5 mM NaCl, 10 mM HEPES, 0.4 mM CaCl 2 , 2 mM MgCl 2 , 1 mM EGTA, and 2 mM MgATP (pH 7.4). The whole-cell configuration was achieved by abutting the pipette tip with a cell, applying suction, forming a Ͼ1-gigaohm seal, and rupturing the membrane to achieve cytoplasmic access. The barrels of a VC-77MCS perfusion system (Warner Instruments) were moved adjacent to the cell for rapid, local perfusion. Signals were recorded with pCLAMP 9 using an Axopatch 200B patch clamp amplifier and a DigiData 1320 digitizer (Molecular Devices). The signal was sampled at 5 kHz and low pass filtered at 5 kHz with the fourpole Bessel filter of the Axopatch 200B patch clamp amplifier. Currents were recorded by holding the membrane voltage at Ϫ60 mV and perfusing with a modified Krebs buffer (130 mM NaCl, 2 mM CaCl 2 , 10 mM D-glucose, 10 mM HEPES, and 10 mM MES, pH 7.4 with HCl). For calcium permeability measurements, NaCl was replaced with N-methyl-D-glucamine. Acid pulses of 10-s duration were applied every 20 s, and little to no desensitization was seen in the absence of access resistance changes. Thus, data were normalized to the pulse immediately prior to initiation of the experimental protocol. Although data suggest that PcTX-1 binding is decreased at pH 5, hASIC-2 containing channels are not significantly activated until pH 5 (2).
Statistics-Data were analyzed using Clampfit (Molecular Devices), Excel 2007 (Microsoft), and SAS 9.1 (SAS Institute Inc.). Data are presented as averages Ϯ 95% confidence intervals (CIs) as calculated by Excel 2007. One-way ANOVAs performed in SAS 9.1 with ␣ ϭ 0.05 were used to assay for differences between groups. Scheffe's post hoc test was used to define different groups as it is better suited for unequal sample sizes and is a very conservative measure.

Computational Results
Homology Modeling-MODELLER uses a peptide sequence alignment to create a model structure from a template. Pairwise alignments of the human ASIC proteins against the crystallized portion of cASIC-1 show that there is significant identity between the human and chicken ASIC proteins with greater than 50% identity and 60% similarity at the peptide level ( Table  1). Studies of homology modeling of membrane proteins have shown that MODELLER is able to produce valid results with as little as 25% identity; thus it is expected that valid models can be obtained for the ASIC proteins (23). Furthermore the functional similarities between these proteins, as they all conduct cations and are sensitive to the small molecule amiloride, favors a conserved structure.
An initial visual inspection, shown in Fig. 1, of the energyminimized structures shows a general conservation of the template structure and allows for visualization of the residues that are altered as compared with the template. Using the NIH Structure Analysis and Verification Server and Matching Molecular Models Obtained from Theory-mult server, struc-tures were compared against theoretical and observed parameters for other protein structures as well as the template structure (supplemental Tables 1 and 2) (24 -28). Based solely on structure, the hASIC-1b and hASIC-2b structures are most similar to the template cASIC-1, whereas the hASIC-3a and hASIC-4a diverge more from the template as expected based on the sequence alignment. All the models scored similarly to the template as would be expected for valid structures.
Inhibitor Docking-The examinations of the models showed that the hASIC models appear valid; however, they have yet to be tested. Ideally one would be able to crystallize the proteins in question and compare the model against the experimental structure. However, crystallization of integral membrane proteins is an arduous task. Fortunately one can still examine whether these models can recapitulate functional data that are readily observable and have already been well studied with the ASICs. For example, homomeric ASIC-1 channels are known to be inhibited by PcTX-1, a peptide toxin found in the venom of the Trinidad chevron tarantula (9,29,30). The NMR solution structure for this 40-amino acid peptide has been solved (10). The peptide appears to be specific for ASIC-1, showing no functional effect on other ASICs (31), and does not appear to bind to the other ASIC proteins (9). Domains involved in binding (9,30) as well as a putative functional surface on the toxin (10) have been established. This allows for a clear cut in silico docking test: will PcTX-1 dock in silico to the hASIC-1b model and not to the other hASIC models? If so, is the docking site consistent with the published literature?
Using ZDOCK 3.0.1 at low resolution, the top 20 NMR structures of PcTX-1 were docked to the human ASIC models. The low resolution docking is much less computationally intensive allowing relatively rapid screening of all the models. Transmembrane domains were blocked from docking as these domains have been shown to have no effect on PcTX-1 binding (9). ZDOCK uses shape complementarity and electrostatics to perform rigid body docking (32). In this case, by using 20 NMR structures of PcTX-1, some biologically observed flexibility was introduced into the ligand. ZDOCK outputs the top 2000 docked poses for each NMR structure and grades the quality of the docking with a larger score signifying a stronger interaction. Fig. 2, panel A, shows the average scoring of the toxin docking FIGURE 1. Structures of the template cASIC-1 and models shown as ribbon diagrams colored by subunit. One subunit from each template has overlays of the van der Waals surfaces of residues that are conserved (yellow), nonconserved (purple), or insertions (green) as compared with the template. Identical residues are not overlaid, allowing one to visually appreciate the conservation in the peptide sequence across these proteins. Not shown are deletions: there 16 amino acids missing in hASIC-3a and 5 amino acids missing in hASIC-4a as compared with the template. PcTX-1 Docking to hASIC-1 with the models with docking to hASIC-1b scoring higher on average than the other models. Using denser docking results and reranking with ZRANK produced a best docked pose for PcTX-1 and hASIC-1b as shown in Fig. 3. The orientation of PcTX-1 docking with the hASIC-1b channel is consistent with the predictions for an inhibitor cysteine knot fold toxin as noted by Escoubas et al. (10). The residues involved in the docking appear to electrostatically orient the toxin with residues Arg-27 and Arg-28, shown in purple, leading a positively charged region into the docking site created at the interface of the channel subunits (Fig. 3 or supplemental Fig. 1). Arg-26 and Arg-27 of the toxin are capable of hydrogen bonding with Glu-235 of hASIC-1b, and Phe-30 can hydrogen bond with His-173. The docking site in the channel also agrees with the domains noted by Salinas et al. (9) in the mapping of the PcTX-1 binding site on rat ASIC-1 as well as the smaller high affinity domain defined by Chen et al. (30). However, unlike the model proffered by Salinas et al. (9), the docking site appears to be created by the interaction of domains on adjacent subunits. A complete list of residues involved within the docking site, defined as within 6 Å of the toxin or channel, is available as supplemental Table 3.
This new insight from the computational models may explain the remarkable selectivity of PcTX-1 for homomeric ASIC-1 channels. To examine this further, the PcTX-1 structure that interacted best with hASIC-1b from the dense docking results was docked to models of heteromeric hASIC-1b and hASIC-2b channels. Models containing two ASIC-1 subunits docked similarly to the homomeric hASIC-1a model, whereas those containing two ASIC-2 subunits scored similarly to homomeric hASIC-2b models (Fig. 2, panel B). This suggests that there may be preferential construction of ASIC heteromers such that they contain only one ASIC-1 subunit as coexpression of ASIC-1 with other ASIC subunits has been shown to abolish inhibition by PcTX-1 (31). A similar phenomenon of preferential heteromerization and assembly has been noted for the related epithelial sodium channel proteins (33,34). Closer computational analysis of the best docked pose also explains lingering questions regarding the binding of PcTX-1 to ASIC-1. Calculations of energies of interactions between the ASIC-1 model and PcTX-1 show that there is a large contribution to the binding energy, ⌬G, by the coulombic interactions of the positively charged toxin with the negatively charged docking pocket created by the channel. This finding partially explains the bellshaped binding curve observed by Salinas et al. (9) where binding peaks at ϳpH 7.0 and falls off to about 25% or less at pH 6.0 and 8.0. By calculating the protonation state and electrostatic potentials of the docked structure at pH 6.0, 7.0, and   (9). The domains shown in cyan, a subset of the domains in blue, are part of the high affinity binding domain that was refined by Chen et al. (30). Shown in purple are residues Arg-27 and Arg-28 of PcTX-1 that appear to lead a positively charged peg into a negatively charged pocket in the channel, consistent with the expectations of Escoubas et al. (10). Panel B shows the surface of the binding pocket colored by residue type: basic, red; acidic, blue; polar, green; and nonpolar, white. Supplemental figure 1 consists of this figure as embedded three-dimensional models.
PcTX-1 Docking to hASIC-1 8.0, one can appreciate that the coulombic contribution drops as the pocket in hASIC-1b becomes more positively charged at the lower pH (Fig. 4). Furthermore based on the electrostatic and solvation components of the binding energy, there appears to be a stabilization of the toxin-channel complex as evidenced by the highly negative ⌬G.

Experimental Tests of the Model
Planar Lipid Bilayer Studies-There is evidence in the literature that PcTX-1 binds to and stabilizes the desensitized state of ASIC-1; this is consistent with the computational data. However, these data are limited to extrapolations from whole-cell currents (5,29,30). In the present study, the effect of the synthetic PcTX-1 on hASIC-1b channels was assayed using the planar lipid bilayer technique. We have previously described the effects of Ca 2ϩ and protein kinase C on hASIC-1b using this technique (35,36). With this technique, hASIC-1b channel transitions can be observed continuously at pH 6.2 in nominally calcium-free solutions with the channel spending as much as 90% of the time in an open state with an apparent unitary conductance of ϳ19 picosiemens (Fig. 5, panel A). Addition of 25 nM PcTX-1 to the external solution resulted in an obvious change of the single channel behavior with a relatively long lived closed state becoming more evident and a decrease in the duration of time spent by hASIC-1b in the open state (Fig. 5,  panel A). At this concentration, the open probability of the channel decreased from 0.91 to 0.33. No change in unitary conductance was apparent. Also PcTX-1 was only effective when added to the presumptive extracellular face of hASIC-1b as expected by the binding site defined by the work above and previously in the literature (9,30). Fig. 5, panel B, shows that increasing the PcTX-1 concentration resulted in a dosedependent decrease of open probability of the hASIC-1b.
Examining the kinetics more closely in the absence and in the presence of PcTX-1, a double exponential function described the closed time distributions (Fig. 6, panel A): the presence of PcTX-1 caused no change in a short lived closed state ( CЈ ), but the time spent by channel in its relatively long lived closed state ( CЉ ) was lengthened. A single exponential function described fairly well the open time distributions of hASIC-1b in the absence and in the presence of PcTX-1 (Fig. 6, panel A). The presence of PcTX-1 decreased the duration of time spent by hASIC-1b in the open state ( O ). This is also illustrated in the reciprocal plots of open and closed states of hASIC-1b as a function of PcTX-1 concentration (Fig. 6, panel B). These findings demonstrate that O of hASIC-1b is inversely proportional to PcTX-1 concentration. The short lived closed state, CЈ , appeared to be independent whereas the relatively long lived closed state, CЉ , was linearly proportional to PcTX-1 concentration. It should be noted that because of our sampling and filtering rates the presence of a change in the short lived closed state may not have been observed. Thus, the pattern observed is likely that of a toxin that reduces the open time and increases the time As can be visually appreciated, the pocket on the surface of the channel shifts from being negatively charged to being neutral/positive as the pH drops. Not shown is the electrostatic potential of the toxin that is relatively unaltered with the pH drop. This is recapitulated in the binding energy calculations that show that the coulombic contribution becomes more positive as the pH is reduced, making the interaction less favorable.

PcTX-1 Docking to hASIC-1
spent in the long lived closed state while not altering the conductance of the channel. This is consistent with an allosteric modulator where the channel is being stabilized in the desensitized state as proposed by Chen et al. (29,30). These data also reinforce the binding energy calculations that show a stabilization of the toxin-channel complex as compared with the toxin and channel alone.
Patch Clamp Studies-As the model for PcTX-1 interaction previously suggested that the docking site was self-contained within one subunit and our computational findings disagreed with this, we tested the finding using a chimeric hASIC-2b subunit. As Salinas et al. (9) and Chen et al. (30) focused on showing a direct involvement of residues identical to 157-185 of hASIC-1b, our work focused on the less well defined domain also implicated in binding of PcTX-1 to ASIC-1 defined as domain 5 by Salinas et al. (9) that contains residues 271-371 of hASIC-1b. Using the best docked pose of PcTX-1 with hASIC-1b, a small section of hASIC-2b that was similar to domain 5 was targeted for mutagenesis to render it identical to hASIC-1b, narrowing the 100-amino acid domain down to just 10 residues. Only four point mutations were required to make residues 340 -355 of hASIC-2b identical to residues 343-358 of hASIC-1b; however, it was postulated that alteration of these residues would create a PcTX-1 docking site within heteromeric hASIC-1b/muthASIC-2b channels.
CHO-K1 cells were first transfected with hASIC-1b, hASIC-2b, or muthASIC-2b alone or in combinations thereof. To establish that heteromeric transfected cells expressed heteromers, the well described calcium permeability of homomeric ASIC-1 channels was exploited (37)(38)(39)(40). In our hands, when external sodium was replaced with the impermeant cation N-methyl-D-glucamine and the only major cation gradients were for calcium influx and potassium efflux, homomeric hASIC-1b-transfected cells showed a significant acid-induced inward current of ϳ63.3% of the peak current as compared with conditions with sodium at pH 5.0 (Fig. 7). Conversely hASIC-2b-or muthASIC-2b-transfected cells showed on average an inward current of 4.7 or 6.5%, respectively, showing significantly decreased permeability to calcium relative to hASIC-1b (Fig. 7). Moreover as the calcium permeability is reported to be half that of the potassium permeability for hASIC-2b, after a short influx of calcium, an outward current was detected, signifying the efflux of potassium (Fig. 7, inset) (40). Combined with a fluorescent reporter and hASIC-1b in a bicistronic vector, this allowed us to differentiate cells that had both hASIC-1b and hASIC-2b or muthASIC-2b from those with solely hASIC-1b based on a significantly decreased calcium influx and the presence of an acid-induced outward current.
When synthetic toxin was applied to these cells at 25 nM in both pH 7.4 and the acid pulse, there was a significant decrease in acid-induced currents after 2 pulses or ϳ60 s of exposure to PcTX-1 in cells transfected solely with hASIC-1b and no effect in cells transfected with hASIC-2b or muthASIC-2b (Fig. 8,  panel A). In cells expressing heteromers of hASIC-1b and hASIC-2b or muthASIC-2b, there was no significant effect of 25 nM PcTX-1 (Fig. 8, panel B). However, a consistent but slight increase in acid-induced currents was observed in cells transfected with hASIC-1b and muthASIC-2b. Using 100 nM PcTX-1, a significant increase in acid-induced current was elu-

PcTX-1 Docking to hASIC-1
cidated from cells expressing the hASIC-1b/muthASIC-2b heteromers, whereas cells expressing hASIC-1b and wild type hASIC-2b showed no effect (Fig. 8, panel B). Although an increase in current was not the initial expectation, similar PcTX-1-induced increases in current for chimeric channels have been shown by Salinas et al. (9) and Chen et al. (30). For example, Chen et al. (30) show that PcTX-1 activates rASIC-1b while inhibiting rASIC-1a. Although the N terminus of rASIC-1b is very different from that of rASIC-1a, the rest of the protein is nearly identical, and much like our muthASIC-2b it contains half of the PcTX-1 docking site we defined, residues 343-358 of hASIC-1b. This result strongly suggests that the muthASIC-2b is able to interact with hASIC-1b to create an interaction site for PcTX-1, affirming the hypothesis that the PcTX-1 docking site is created at subunit interfaces.

DISCUSSION
Relevance-The ASIC family of proteins is an important therapeutic target for pathologies such as pain, cancer, stroke, epilepsy, or anxiety (41). However, finding potent and specific modulators of these channels has required careful screening of venoms and small molecule libraries using time-consuming functional assays (31,(42)(43)(44). Many of the agents found either lack specificity or are clinically difficult to administer, such as amiloride or PcTX-1 (45). A structure of the target permits the rational design or the virtual screening of molecules that will interact specifically with the protein rather than blindly screening drug libraries in vitro. As crystal structures of membrane proteins are still difficult to obtain, this work leverages the finding of Jasti et al. (4) to define structures of the human ASIC proteins. Using a classic homology modeling approach, we produced valid models of the four human ASIC proteins. These models of homomeric ASICs were validated in silico to conform both to observed and theoretical structural parameters as well as to the template cASIC-1 structure as shown in the supplemental data and to recapitulate in silico the interactions of the FIGURE 7. Shown are the average peak inward acid-induced whole-cell currents for CHO-K1 cells transfected with hASIC-1b, hASIC-2b, muthASIC-2b, and combinations thereof. Acid-induced currents in the Na ϩ -free N-methyl-D-glucamine solutions were normalized to the prior acid pulse in Na ϩ -containing solutions. Values for the hASIC-1b homomers were used as the calibrator. Cells transfected with hASIC-2b or muthASIC-2b only showed a significantly lower whole-cell current in the absence of Na ϩ as compared with cells transfected only with hASIC-1b. Cells transfected with hASIC-1b and either hASIC-2b or muthASIC-2b showed currents statistically similar to those of cells expressing only hASIC-2b or muthASIC-2b. Shown above the bars are representative traces of the acid-induced peaks (n Ն 3 per group, Ϯ95% CI, ANOVA, Scheffe's post hoc test, ␣ ϭ 0.05).

A.
B. Currents were normalized to the peak prior to PcTX-1 treatment. Cells transfected with hASIC-1b were significantly inhibited by PcTX-1 application after the second pulse or after ϳ60 s of toxin application. Cells expressing heteromeric hASIC-1b and either hASIC-2b or muthASIC-2b showed no effect of 25 nM PcTX-1 treatment, but a significant increase in peak acid-induced currents was noted in cells expressing hASIC-1b/muthASIC-2b with 100 nM PcTX-1 as compared with cells expressing hASIC-1b/hASIC-2b channels. Together these data strongly suggest that the mutations in hASIC-2b create an interaction site for PcTX-1 at the interface with hASIC-1b (n Ն 3 per group, Ϯ95% CI, ANOVA, Scheffe's post hoc test, ␣ ϭ 0.05).

PcTX-1 Docking to hASIC-1
peptide toxin PcTX-1 as shown in Fig. 2. Moreover these models were tested experimentally in single channel and whole-cell electrophysiological studies.
Computational Observations-The result of docking PcTX-1 to the models confirms that the hASIC-1b binding site is consistent with expectations (9, 10). Both domains 3 and 5, as described by Salinas et al. (9), form the docking site. However, there are additional residues contributed from domains 2, 4, and 6 as well. Similar conclusions have been reached by the computational studies of Pietra (46) who uses similar techniques but limits his study to only hASIC-1b. Of note, the docking site is created by the interaction of these domains within two separate subunits, not in one single subunit as the initial models suggested (9). Moreover our docking studies clarify some confused points in the field.
For example, the crystal structure of cASIC-1 was thought to be in a closed or desensitized state (1). Combined with the functional data from other studies, our ability to dock PcTX-1 to the structure strongly argues that the molecule is in a desensitized state or that there is no significant structural difference between the closed or desensitized states (5,9,29,30). The latter idea that the structure of the desensitized state is indistinct is contradicted by functional and binding data that both show that the interaction of PcTX-1 with the ASIC-1 channel correlates strongly to the desensitization state of the channel and not the closed state (5,9,29,30). For example, Salinas et al. (9) show with radiolabeled PcTX-1 that binding increases as the pH is lowered from 8 to 7, which correlates with the channel becoming more desensitized. Sherwood and Askwith (5) also show that point mutations, distinct from our binding site, appear to affect the desensitization of hASIC-1b allowing it to be inhibited by PcTX-1 at a resting pH 7.4 in the manner of rASIC-1a. An interesting aspect of the binding curve generated by Salinas et al. (9) is a decrease in binding as the pH is lowered from 7 to 5. As discussed above, this can be explained by alterations in the coulombic interactions, which as Escoubas et al. (10) correctly predicted appear to steer the toxin into a pocket on the channel. Furthermore this decrease in the coulombic contributions to binding explains the reduced effectiveness of PcTX-1 when applied concomitantly at acidic pH. The charged pocket that nestles the positively charged toxin in the docked state becomes less inviting as the pocket becomes more positive with a drop in pH. This is also visible in the functional data.
Experimental Observations-The dose-response curve in the bilayer studies (Fig. 5, panel B) was right shifted by approximately a log as compared with prior results (9,29,30). Prior experiments with synthetic toxin have measured the interaction at pH 7-8 with rASIC-1a, whereas these single channel experiments measured the interaction at a pH of 6.2. This right shift is then consistent with the computational data regarding a reduced interaction at lower pH due to a reduction in the coulombic interaction.
From the bilayer data, PcTX-1 appears to cause a reduction in the time spent in the open state and an increase in time spent in the long closed state of hASIC-1b. This is compatible with the model proposed by Gründer and co-workers (29,30) that suggests that the toxin binds to the open or desensitized state of rASIC-1a and inhibits the channel by shifting the pH activation curve toward more alkaline values, leading it to become desensitized in the presence of toxin. However, their paradigm is based on extrapolations from whole-cell currents in the Xenopus oocyte system. Our single channel bilayer data showed a similar effect where the desensitized or long closed state is stabilized whereas the open state is destabilized for rASIC-1a or hASIC-1b.
To verify that the PcTX-1 binding site was located at the subunit interfaces, an ASIC-1/ASIC-2 chimera was created where half of the predicted docking site was created in ASIC-2. This would be predicted to lead to an interaction of ASIC-1/ mutASIC-2 heteromers with PcTX-1, whereas wild type ASIC-1/ASIC-2 heteromers would be unaffected as is reported in the literature. This model prediction was borne out in our patch clamp experiments (Fig. 8). However, although an interaction of PcTX-1 was observed in the heteromeric ASIC-1/mutA-SIC-2 and not in the wild type ASIC-1/ASIC-2, PcTX-1 increased the peak acid-induced current rather than inhibiting the channel. Although this does confirm the docking site and is similar to observations made with PcTX-1 and other chimeric channels (9,30), it also reinforces that PcTX-1 does not necessarily just inhibit ASIC-1.
PcTX-1 as a Modulator-Although PcTX-1 is considered a highly potent inhibitor of ASIC-1, there are multiple reports of PcTX-1 potentiating or activating ASIC-1 in various situations (29,30,47). For example, applying 30 nM PcTX-1 concomitantly with a pH pulse of 7.1 or in low calcium activates rASIC-1a overexpressed in oocytes, whereas 30 nM PcTX-1 alone can activate cASIC-1 expressed in COS-7 cells or endogenous cASIC-1 in chicken dorsal root ganglion cells. Whether this ability to potentiate ASIC-1 could be detrimental in clinically relevant scenarios has yet to be closely studied.
Extrapolations to Gating-Although the present study was limited by the rigidity of the structures and may have missed a binding site present when the channel was in a closed or open state, it suggests that the stabilization of the open or desensitized state of ASIC-1 by PcTX-1 occurs at the same domains. This is also noted by Salinas et al. (9) who found that placing domains 1, 2, 3, and 5 of rASIC-1a into rASIC-2a created a chimeric construct that was significantly activated by PcTX-1. However, when domain 4 or 6 of rASIC-1a was also present, no activation of the chimeric rASIC-2a was observed (9). If both domains 4 and 6 were present, inhibition similar to that of rASIC-1a was found (9). Salinas et al. (9) concluded that although domains 3 and 5 were the main mediators of the interactions domains 1, 2, 4, and 6 were needed for positioning the docking site or for transmitting the interaction into inhibition or activation.
Our results showed that residues in domains 2, 4, and 6 could also play a role in the docking of hASIC-1b with the toxin and may mediate structural changes upon binding. One could postulate that perhaps the main docking site domains 3 and 5 are part of a hinge location that is relatively unchanged between the open and desensitized state. Interactions of the toxin or the channel domains 3 and 5 with domains 2, 4, and 6 could then lead to either opening of the channel or desensitization. This is similar to the model suggested by Jasti et al. (4) with the toxin PcTX-1 Docking to hASIC-1 interacting at or near the pH sensor between the finger and thumb. However, as PcTX-1 interacts at this location to cause either activation or inactivation, this model would suggest that the interaction site would be relatively static whereas binding would cause conformational changes elsewhere in the structure, such as in domains 2, 4, and 6, which roughly correspond to the finger, ball, and palm regions, rather than a slight flick through the wrist domain as suggested by Jasti et al. (4).
The availability of these validated structures and the docking site of hASIC-1b with PcTX-1 will spur the isolation and design of novel therapeutics aimed at this family. Future work should leverage molecular dynamics studies to assay for conformational changes in the structural models in the presence and absence of toxin and as a function of pH as well as expand to APETx2, the sea anemone toxin that inhibits ASIC-3-containing channels (43,48,49).