Molecular determinants of α-conotoxin potency for inhibition of human and rat α6β4 nicotinic acetylcholine receptors

Nicotinic acetylcholine receptors (nAChRs) containing α6 and β4 subunits are expressed by dorsal root ganglion neurons and have been implicated in neuropathic pain. Rodent models are often used to evaluate the efficacy of analgesic compounds, but species differences may affect the activity of some nAChR ligands. A previous candidate α-conotoxin-based therapeutic yielded promising results in rodent models, but failed in human clinical trials, emphasizing the importance of understanding species differences in ligand activity. Here, we show that human and rat α6/α3β4 nAChRs expressed in Xenopus laevis oocytes exhibit differential sensitivity to α-conotoxins. Sequence homology comparisons of human and rat α6β4 nAChR subunits indicated that α6 residues forming the ligand-binding pocket are highly conserved between the two species, but several residues of β4 differed, including a Leu–Gln difference at position 119. X-ray crystallography of α-conotoxin PeIA complexed with the Aplysia californica acetylcholine-binding protein (AChBP) revealed that binding of PeIA orients Pro13 in close proximity to residue 119 of the AChBP complementary subunit. Site-directed mutagenesis studies revealed that Leu119 of human β4 contributes to higher sensitivity of human α6/α3β4 nAChRs to α-conotoxins, and structure–activity studies indicated that PeIA Pro13 is critical for high potency. Human and rat α6/α3β4 nAChRs displayed differential sensitivities to perturbations of the interaction between PeIA Pro13 and residue 119 of the β4 subunit. These results highlight the potential significance of species differences in α6β4 nAChR pharmacology that should be taken into consideration when evaluating the activity of candidate human therapeutics in rodent models.

Nicotinic acetylcholine receptors (nAChRs) containing ␣6 and ␤4 subunits are expressed by dorsal root ganglion neurons and have been implicated in neuropathic pain. Rodent models are often used to evaluate the efficacy of analgesic compounds, but species differences may affect the activity of some nAChR ligands. A previous candidate ␣-conotoxin-based therapeutic yielded promising results in rodent models, but failed in human clinical trials, emphasizing the importance of understanding species differences in ligand activity. Here, we show that human and rat ␣6/␣3␤4 nAChRs expressed in Xenopus laevis oocytes exhibit differential sensitivity to ␣-conotoxins. Sequence homology comparisons of human and rat ␣6␤4 nAChR subunits indicated that ␣6 residues forming the ligand-binding pocket are highly conserved between the two species, but several residues of ␤4 differed, including a Leu-Gln difference at position 119. X-ray crystallography of ␣-conotoxin PeIA complexed with the Aplysia californica acetylcholine-binding protein (AChBP) revealed that binding of PeIA orients Pro 13 in close proximity to residue 119 of the AChBP complementary subunit. Sitedirected mutagenesis studies revealed that Leu 119 of human ␤4 contributes to higher sensitivity of human ␣6/␣3␤4 nAChRs to ␣-conotoxins, and structure-activity studies indicated that PeIA Pro 13 is critical for high potency. Human and rat ␣6/␣3␤4 nAChRs displayed differential sensitivities to perturbations of the interaction between PeIA Pro 13 and residue 119 of the ␤4 subunit. These results highlight the potential significance of species differences in ␣6␤4 nAChR pharmacology that should be taken into consideration when evaluating the activity of candidate human therapeutics in rodent models.
Native nAChRs containing the ␣6 subunit are broadly classified into two subtype categories: those that contain the ␤2 subunit and those that contain the ␤4 subunit. The ␣6␤2* subtype (the asterisk denotes the potential presence of additional subunits in native receptors) has a limited distribution profile in the nervous system but is abundantly expressed in certain regions of the brain and spinal cord (13)(14)(15)(16)(17)(18)(19). The ␣6␤4* subtype probably has an even more restricted expression profile. Functional evidence for ␣6␤4* nAChR expression in sensory neurons of rat and mouse dorsal root ganglia (DRG) has been demonstrated (20,21), although their functional role in these cells is mostly unknown.
DRG contain neurons that perform a wide range of sensory functions, including proprioception and the detection of harmful or painful stimuli. Recent evidence suggests that ␣6␤4 nAChRs represent a new molecular target for the treatment of neuropathic pain. Development of neuropathic pain-like symptoms has been shown to be inversely correlated with CHRNA6 expression levels in the DRG of mice (21). Strains with high levels of CHRNA6 expression show lower levels of mechanical allodynia in several neuropathic and inflammatory pain models, and those with low levels of CHRNA6 expression are more susceptible to developing neuropathic pain. Intriguingly, CHRNA6 null mice show no analgesic responses to nicotine, whereas mice with a gain-of-function mutation show increased analgesic responses. In addition, a history of chronic pain syndromes in humans correlates with levels of CHRNA6 expression (21).
Rodent models of neuropathic pain are often used to study mechanisms of nociception as well as to evaluate potential therapeutics that can modulate the transmission and perception of pain. However, rodent models can be compromised by a number of species-related factors that complicate translation of results obtained in rodent studies to human clinical trials. One of these factors is the difference in ligand sensitivity between human and rodent receptors and ion channels.
Conotoxins are small peptides found in the venom of carnivorous marine snails of the Conus genus and are used by these mollusks to capture prey. ␣-Conotoxins (␣-Ctxs) belong to a subclass of conotoxins and are antagonists of nAChRs. Some ␣-Ctxs are capable of distinguishing among the various nAChR subtypes (22). ␣-Ctx Vc1.1 (23,24) is an example of a ligand developed as a treatment for neuropathic pain that showed promising results in rodent models but failed to produce similar levels of analgesia in human clinical trials. It was later demonstrated that decreased sensitivity of human versus rat nAChRs to Vc1.1 might have contributed to this outcome. It is therefore important to determine how species differences influence receptor sensitivity to ligands. To this end, we have examined the interaction of human and rat ␣6␤4 nAChRs with ␣-Ctxs to elucidate molecular determinants of ligand potency for this receptor subtype. Through structure-activity studies of PeIA, site-directed mutagenesis of ␣6 and ␤4 subunits, X-ray crystallography, and NMR studies of PeIA, we have identified important molecular determinants of ␣-Ctx potency for human and rat ␣6␤4 nAChRs. The information obtained from these studies offers important insights into the pharmacology of ␣6␤4 nAChRs and might facilitate the development of selective ligands that interact with this potential neuropathic pain target.

Sequences of human and rat ␣6 subunits are highly conserved, but several residues of ␤4 that form the ligand-binding pocket differ
A sequence alignment of human and rat ␣6 subunits indicates that the extracellular domains are similar, and in fact residues that form the canonical ligand-binding pocket are strictly conserved (Fig. S1). The closest residue to the ligand-binding pocket that differs in the ␣6 sequences is located at position 177. In humans, this residue is Ile, but in rat, a Val is present. Analysis of the ␤4 subunits revealed more significant sequence differences occurring at positions 110, 118, and 119. In human ␤4, these residues are Leu, Val, and Leu, and in rat, they are Val, Ile, and Gln. These three nonconserved residues are all located in the ligand-binding pocket and might contribute to the species differences in ␣-Ctxs potencies for inhibition of human and rat ␣6/␣3␤4 nAChRs.
The ␤4 subunit is a major determinant of the species differences in PeIA, PnIA, and TxIB potencies for human versus rat ␣6/␣3␤4 nAChRs A series of experiments was performed to determine whether differences in the amino acid sequences of human and rat ␤4 subunits are important for ␣-Ctx potency for ␣6/␣3␤4 nAChRs. In the first experiment, one group of oocytes was injected with cRNAs encoding human ␣6/␣3 and rat ␤4 subunits, and another group was injected with cRNAs encoding rat ␣6/␣3 and human ␤4 subunits. The potencies for PeIA, PnIA, and TxIB were then reassessed on these hybrid combinations with the hypothesis that the IC 50 values for receptors composed of human (H) ␣6/␣3 and rat (R) ␤4 subunits would be similar to R␣6/␣3R␤4 nAChRs. Likewise, the IC 50 values for inhibition of R␣6/␣3H␤4 and H␣6/␣3H␤4 nAChRs would also be similar. Consistent with these predictions, the IC 50 curves for inhibition of H␣6/␣3R␤4 by PeIA, PnIA, and TxIB are right-shifted toward those for R␣6/␣3R␤4 nAChRs (Fig. 3 (A-C) and Table  2). Conversely, the IC 50 curves for inhibition of R␣6/␣3H␤4 are left-shifted toward H␣6/␣3H␤4 nAChRs (Fig. 3 (A-C) and Table 2). These results indicate that the ␤4 subunit is an important determinant of the species differences in ␣-Ctx potency for inhibition of ␣6/␣3␤4 nAChRs.
As additional evidence in support of the importance of the ␤4 subunit for conferring differential ␣-Ctx potencies, we tested the same three ␣-Ctxs on human and rat ␣6/␣3␤2␤3 nAChRs. Conotoxins are classified into various subclasses based on the number of cysteine residues present in the sequence and the number of amino acid residues between them (63). Peptides of the ␣-Ctx subclass contain two disulfideconnected pairs of cysteines. Thus, for example, ␣-Ctxs with four residues between Cys 2 and Cys 3 and seven between Cys 8 and Cys 16 belong to the 4/7 framework subclass. Residues in black are variable, and those in red are conserved among this ␣-Ctx set. The disulfide bonds between Cys residues are depicted with lines. *, C-terminal amidation.

Determinants of ␣-Ctx potency for ␣6␤4 nAChRs
In the ␣6␤2␤3 nAChR, the canonical ␣-Ctx ligand-binding site is located at the interface between the ␣6 and ␤2 subunits. We aligned the sequences of human and rat ␤2 subunits and found that residues known to interact with ␣-Ctxs are strictly conserved (29,30), and in fact only three conservative differences were noted in the entire extracellular ligand-binding domain sequence (Fig. S2). When we tested PeIA on human and rat ␣6/␣3␤2␤3 nAChRs, we found that the IC 50 values for the two Determinants of ␣-Ctx potency for ␣6␤4 nAChRs species were nearly identical ( Fig. 4A and Table 3). Similar results were obtained for PnIA and TxIB ( Fig. 4 (B and C) and Table 3). In all comparisons, a Ͻ2-fold difference in the IC 50 values was found.
Site-directed mutagenesis of human ␤4 implicates residue 119 as a critical determinant of high-potency PeIA binding to human ␣6/␣3␤4 nAChRs The high sequence similarity of human and rat ␣6 and ␤4 subunits allowed us to focus on select residues that might contribute to the species differences in ␣-Ctx potency for human and rat ␣6/␣3␤4 nAChRs. We used site-directed mutagenesis to generate receptor mutants where nonconserved residues of ␣6 and ␤4 subunits were switched to the residues found in the homologous positions of the other species and then assessed the effects of these mutations on ␣6/␣3␤4 nAChR sensitivity to PeIA. In the ␣6 subunit, there is an Ile-Val difference between human and rat sequences, respectively, at position 177. We found that when Ile 177 of human ␣6 was switched to Val, the IC 50 value of PeIA was similar to the value obtained for human ␣6/␣3␤4 nAChRs ( Fig. 5A; Table 4), indicating that this residue contributes very little to the species difference in PeIA potency. Next, residues Leu 110 , Val 118 , and Leu 119 of human ␤4 were individually switched to Val, Ile, and Gln, respectively, and expressed with human ␣6/␣3 subunits. Similar to the ␣6 I177V mutation, L110V and V118I mutations in the ␤4 subunit had very little impact on the IC 50 value of PeIA ( Fig. 5A and Table 4). However, when the ␣6/␣3 subunit was expressed with the ␤4 L119Q mutant subunit, the IC 50 value increased by ϳ10-fold, and the curve shifted to the right toward that of rat ␣6/␣3␤4 nAChRs ( Fig. 5A and Table 4). To determine whether residues 118 and 119 might play a combined role in PeIA binding, we made a double V118I,L119Q mutant human ␤4 subunit and reassessed the potency of PeIA. The sensitivity of the ␣6/␣3␤4 V118I,L119Q mutant to PeIA was similar to that of the ␣6/␣3␤4 L119Q single mutant, indicating no added effect with the combined mutations ( Fig. 5A and Table 4). To further assess the influence of position 119 on PeIA potency, we made an additional human ␤4 mutant where Leu 119 was changed to Phe, the residue found in the homologous positions of human and rat ␤2 subunits (Fig. S2). The IC 50 value for inhibition of ␣6/␣3␤4 L119F mutant nAChRs by PeIA was similar (Ͻ2-fold difference) to the value for receptors with native ␤4 subunits ( Fig. 5A and Table 4). Mutation experiments were also performed for rat ␣6/␣3␤4 nAChRs and, similar to the results found for human receptor ␣6 V177I , ␤4 V110L , ␤4 I118V , and ␤4 Q119F mutations, had very little effect on the sensitivity of rat  Table 3. Error bars, S.D. from at least four oocytes for each IC 50 determination.  Determinants of ␣-Ctx potency for ␣6␤4 nAChRs ␣6/␣3␤4 nAChRs to PeIA relative to nonmutated receptors ( Fig. 5B and Table 4). However, in contrast to the results obtained with human ␣6/␣3␤4 L119Q and ␣6/␣3␤4 V118I,L119Q nAChRs, mutation of the homologous residues of rat ␤4 had no effect on the sensitivity of rat ␣6/␣3␤4 nAChRs to PeIA ( Fig. 5B and Table 4). The results of these receptor mutation experiments indicate that position 119 of human ␤4 plays an important role in the interaction between PeIA and human ␣6/␣3␤4 nAChRs, but other factors appear to be involved in determining the lower potency of PeIA for rat ␣6/␣3␤4 nAChRs.

X-ray crystallography of PeIA complexed with Aplysia californica AChBP reveals close spatial proximity of residue Pro 13 of PeIA to Met 119 of the AChBP complementary subunit
The marine mollusk Aplysia californica expresses a watersoluble protein called AChBP that functions to modulate AChmediated synaptic transmission in this organism (31). X-ray crystallography studies of the AChBP have been used to elucidate binding interactions between ligands and nAChRs (32)(33)(34)(35). Several high-resolution crystal structures have been reported for the AChBP complexed with ␣-Ctxs (36 -39) but not with PeIA. To gain further insights into the interactions between PeIA and human ␣6␤4 nAChRs, we performed X-ray crystallography of the AChBP complexed with PeIA. The resulting 2.34 Å resolution structure of the AChBP-PeIA complex is shown in Fig. 6 (A-C). An examination of the complex reveals that PeIA residue Pro 13 is oriented toward the complementary subunit and in close proximity to Met 119 of the AChBP (Fig. 6C). In the ␣6␤4 receptor complex, the complementary subunit corresponds to the ␤4 subunit. A sequence alignment of the human ␤4 ligand-binding domain with the AChBP indicates that Met 119 of the AChBP is homologous with human ␤4 Leu-119 (Fig. S3). The proximity of Met 119 of the AChBP to Pro 13 of PeIA suggests that an interaction might occur between these two residues, and functional analysis of human ␣6/␣3␤4 L119M mutant nAChRs expressed in oocytes indicates that a Met in position 119 influences PeIA activity (Fig. S3).

Structure-activity studies identify Pro 13 as important for PeIA potency on human and rat ␣6/␣3␤4 nAChRs
The crystal structure of PeIA complexed with the AChBP suggests that PeIA residue Pro 13 probably interacts with the ␤4 subunit. Therefore, we used analogs of PeIA where Pro 13 was substituted with different amino acids to probe the interaction between this residue and position 119 of the ␤4 subunit.  Table 5). However, substitution of Pro with Ala or Arg resulted in more substantial decreases in potency by ϳ7and ϳ20-fold, respectively ( Fig. 7A and Table 5). The rank order potency of these analogs is PeIA. The IC 50 values for these analogs were then determined for rat ␣6/␣3␤4 nAChRs. Although all analogs showed reduced potency, relative to native PeIA, there were differences in the magnitude of the changes for rat compared with human ␣6/␣3␤4 nAChRs. The potency of  Table 4. Data for inhibition of human (dashed red) and rat (dashed green) ␣6/␣3␤4 nAChRs by PeIA were previously presented and shown for ease of visual comparison. Error bars, S.D. from at least four oocytes for each IC 50 determination.    Table 5. Error bars, S.D. from at least four individual oocytes for each IC 50 determination. Data for inhibition of human ␣6/␣3␤4 (dashed red) and rat ␣6/␣3␤4 nAChRs (dashed green) by PeIA were presented previously and shown for ease of visual comparison.

Determinants of ␣-Ctx potency for ␣6␤4 nAChRs NMR analysis of PeIA demonstrates that the backbone structure and side-chain positions are unchanged with Ala substitution of Pro 13
One possible effect of substituting PeIA Pro 13 with other amino acids is that changes in the tertiary structure of the peptide might occur. To determine whether large changes in the backbone structure or repositioning of the side chains could account for the loss of potency observed with the P13A substitution, we performed NMR analysis on [P13A]PeIA and compared its structure with that of native PeIA. A pattern of negative secondary shifts across residues Ala 7 -Val 10 indicated the presence of a short helix (Fig. 8A) in agreement with the NMR solution structure of PeIA determined previously (41). Analysis of the secondary H␣ shifts, as calculated by subtracting random coil H␣ shifts from the peptide H␣ shifts, suggested a slight difference in [P13A]PeIA across the C-terminal section of the peptide from residue Asn 11 to Cys 16 . Therefore, we calculated the three-dimensional solution NMR structure of [P13A]PeIA to determine whether there were changes in the overall struc-ture of the peptide. A total of 98 distance restraints were determined from NOESY data collected in aqueous solution at 290 K, along with 14 dihedral angle restraints. Restraints for two hydrogen bonds (Val 10 HN-Pro 6 CO and His 12 HN-Cys 8 CO) were also added based upon preliminary structures and amide chemical shift/temperature coefficients (42). The ensemble of the 20 lowest-energy structures overlay well (root mean square deviation 0.60 Ϯ 0.16 Å for the backbone atoms) (Fig. 8B). A comparison with native PeIA revealed no apparent effect of the P13A substitution on either peptide backbone (root mean square deviation 1.02 Å, backbone atoms) or side-chain orientation (Fig. 8C). These structural data suggest that the losses of potency observed with the P13A-substituted analog are not due to large changes in tertiary structure of the peptide. 13 (25)). An ␣-helix is present from residue 7 to 10, and the side chains of residue 13 are shown as sticks.  (Fig. 9 (B and C) and Table  6). In each case, the -fold increase in the IC 50 value for each double peptide-receptor mutant combination was larger than the sum of the each individual mutant IC 50. These data are consistent with a direct interaction between residue 13 of PeIA and residue 119 of human ␤4.

Position 119 of human and rat ␤4 and human ␤2 subunits interacts differentially with residue 13 of ␣-Ctxs to influence binding
The results shown in Fig. 5 indicate that mutating residue 119 of ␤4 to Phe has no impact on the potency of PeIA for human or rat ␣6/␣3␤4 nAChRs. However, we note that there are some differences in the potencies of the Pro 13 -substituted PeIA analogs on human ␣6/␣3␤4 and those previously found for rat ␣6/␣3␤2␤3 nAChRs. PeIA shows a ϳ7-fold loss of potency on human ␣6/␣3␤4 nAChRs with the P13A substitution ( Fig. 7A and Table 5), yet rat ␣6/␣3␤2␤3 nAChRs are equally sensitive to PeIA and the P13A analog (40). Furthermore, PnIA, which also has a Pro in position 13, is significantly more potent on human ␣6/␣3␤2␤3 than ␣6/␣3␤4 nAChRs, in contrast to PeIA, which is equally potent on both subtypes (Fig.  3 (A and B) and Table 3). These differences in potency suggest that the binding interactions between Pro 13 of ␣-Ctxs and position 119 of the ␤2 and ␤4 subunits may not be equivalent. Therefore, we tested [P13A]PeIA for its ability to inhibit the ␣6/␣3␤4 L119F mutant and compared the IC 50 value with those for human ␣6/␣3␤4 and ␣6/␣3␤2␤3 nAChRs. The Leu to Phe mutation in position 119 of ␤4 had no effect on the loss of PeIA potency for ␣6/␣3␤4 nAChRs caused by the P13A substitution ( Fig. 10A and Table 6). Similar results were observed with the [P13R]PeIA analog ( Fig. 10A and Table 6). In each case, the higher IC 50 Table 6. Error bars, S.D. from at least four individual oocytes for each IC 50 determination. Dashed lines, data previously presented and shown for ease of visual comparison.

Discussion
Compounds expected to have analgesic properties are often tested in rodent models of pain with the expectation (and hope) that the results will be translatable to humans. However, several factors can influence whether a drug that is effective in rodents will also produce the same efficacy in humans. Some of these factors include bioavailability, metabolism, and, importantly, sensitivity of the molecular target to the compound of interest. Species differences in the amino acid sequences of target molecules can have substantial effects on ligand potency. For example, ␣-Ctx Vc1.1 was an effective analgesic in rodent models of neuropathic pain but failed to show similar efficacy in human clinical trials (23)(24)(25). A single amino acid difference between human and rat ␣9 nAChR subunits was later shown to confer higher potency for rat over human ␣9␣10 nAChRs, a potential analgesic target of Vc1.1 (43).
There are several factors that might contribute to the higher ␣-Ctx potency for human ␣6/␣3␤4 nAChRs. First, there are Figure 10. ␣-Ctx potency for human ␣6/␣3␤4 nAChRs is not affected by mutation of Leu 119 to Phe. X. laevis oocytes expressing human nAChRs were subjected to TEVC electrophysiology as described under "Experimental procedures," and the IC 50  Determinants of ␣-Ctx potency for ␣6␤4 nAChRs three nonconserved residues of human and rat ␤4 subunits that form the ligand-binding pocket and might directly interact with ligands potentially affecting ligand affinity. Nonconserved residues outside the ligand-binding pocket might also contribute, albeit indirectly, to ligand affinity by altering the tertiary structure of the subunit. This, in turn, might affect the way the subunits associate with each other to create the ligand-binding surfaces. Here, we have examined the effect of substituting nonconserved residues of human ␣6 and ␤4 subunits with residues found in the homologous positions of the respective rat subunits. These residues include Ile 177 of human ␣6 and Leu 110 , Val 118 , and Leu 119 of human ␤4 (Fig. S1). We observed that mutation of Ile 177 to Val in the ␣6 sequence has no effect on PeIA potency, but substantial effects were found with mutations of human ␤4. When Leu 119 was mutated to Gln, the potency of PeIA for ␣6/␣3␤4 nAChRs was reduced by ϳ10fold, but when Leu 119 was mutated to Phe, the potency was unchanged (Fig. 5A and Table 4). PeIA potency was also reduced by ϳ4-fold when Leu 119 was mutated to Met (Fig. S3). Additional mutations of human ␤4 were made to determine whether other nonconserved residues contributed to species differences in PeIA potency. We found that the potency of PeIA  5% (n ϭ 4)). The responses of ␣6/␣3␤4 Q119L and ␣6/␣3␤4 Q119F nAChRs after exposure to 10 M [P13R]PeIA were also significantly larger than those obtained with ␣6/␣3␤4 nAChRs under the same conditions (98 Ϯ 2% (n ϭ 4) and 86 Ϯ 4% (n ϭ 4), respectively, of control responses compared with 58 Ϯ 3% (n ϭ 4) for ␣6/␣3␤4 nAChRs). Statistical significance was determined using an analysis of variance and Fisher's least significant difference test (***, p Յ 0.001; ****, p Յ 0.0001). Error bars, S.D. for the indicated number of individual replicates. D, representative current traces for inhibition of mutant ␣6/␣3␤4 Q119L and ␣6/␣3␤4 Q119F nAChRs by the indicated PeIA analogs. Traces in black are control responses, and those in red are responses in the presence of the ␣-Ctxs.
Determinants of ␣-Ctx potency for ␣6␤4 nAChRs was not affected by either L110V or V118I mutations. A double mutant combining the V118I and L119Q mutations was then made to determine whether an interaction between Ile and Gln could potentially influence PeIA potency, but no further decrease in PeIAs IC 50 value, compared with that for the ␣6/␣3␤4 L119Q single mutant, was observed. It should be noted that mutation of Leu to Val and Val to Ile are conservative changes that might not result in easily observable effects on PeIA potency. Nevertheless, these experiments demonstrate that of the three nonconserved ligand-binding pocket residues between human and rat ␤4, only the Leu-Gln difference at position 119 contributes to the higher affinity of PeIA for human ␣6/␣3␤4 nAChRs. Similar experiments were conducted for rat ␣6/␣3␤4 nAChRs, but strikingly, we found that none of the mutations in rat subunits substantially changed the potency of PeIA, including the ␤4 Q119L mutation ( Fig. 5B and Table 4). Therefore, we took a different approach to further examine the interaction of PeIA with ␣6/␣3␤4 nAChRs.
X-ray crystallography studies of PeIA complexed with the AChBP showed that PeIA Pro-13 is in close proximity to Met-119 of the complementary subunit. Based on this observation, we synthesized analogs of PeIA where Pro 13 was substituted with Hyp, Gln, Ala, or Arg to determine whether Pro 13 and position 119 of the ␤4 subunit interacted (Fig. 7 (A and B) and Table 5). Substitution of Pro 13 with Hyp or Gln reduced the potency of PeIA for human ␣6/␣3␤4 nAChRs but by only ϳ3-fold. More substantial decreases in potency resulted when Pro 13 was substituted with Ala (ϳ7-fold) or Arg (ϳ20-fold). Although all four PeIA analogs showed reduced potencies for rat ␣6/␣3␤4 nAChRs, they did not show the same magnitude of change found for human ␣6/␣3␤4 nAChRs. The most substantial differences were found with the P13Q and P13R substitutions (Fig. 7 (A and B) and Table 5).
Additional support for an interaction between PeIA Pro 13 and position 119 of the ␤4 subunit was obtained by determining the potencies of the PeIA analogs on the human ␣6/␣3␤4 L119Q (Fig. 8 (A-C) and Table 6) and rat ␣6/␣3␤4 Q119L mutants ( Fig.  11B and Table 7). These double ligand-receptor mutant combinations resulted in additional reductions in PeIA potency for both species, lending support for a direct interaction between PeIA Pro 13 and position 119 of the ␤4 subunit.
To ensure that the reduced potencies observed with the PeIA analogs were not the result of changes in the overall structure of the peptide due to substitution of rigid Pro, we examined the structure of [P13A]PeIA and compared it with that of the native peptide using NMR. Although some differences in the ␣H shifts of residues 10 -15 were observed, the overall backbone structure appeared unchanged relative to the native peptide (Fig. 8). Furthermore, the orientations of the Pro 13 and Ala 13 side chains were similar in the native peptide and the analog, respectively, suggesting that the losses in potencies observed with the PeIA analogs are not due to repositioning of the side chains. Last, the PeIA potency differences observed for human ␣6/␣3␤4 and ␣6/␣3␤4 L119Q nAChRs are not due to changes in the potency or efficacy of the agonist acetylcholine (Fig. S4).
Zhangsun et al. showed that mutation of Phe 119 to Gln in the rat ␤2 subunit results in increased potency of LvIA for ␣3␤2 nAChRs (46). By contrast, we found that for human ␣6/␣3␤4 nAChRs, mutation of ␤4 Leu-119 to Gln decreased the potency of PeIA by ϳ10-fold, and mutation to Phe had no effect ( Fig. 5A and Table 4). These results initially suggested that Phe can substitute for Leu in position 119 with respect to PeIA binding and might offer a "protective" effect against the reduction in potencies observed with the [P13A]PeIA and [P13R]PeIA analogs because the IC 50 values for [P13A]PeIA and [P13R]PeIA on human ␣6/␣3␤2␤3 nAChRs are unchanged relative to native PeIA ( Fig. 10B and Table 6). However, when [P13A]PeIA and [P13R]PeIA were tested on the human ␣6/␣3␤4 L119F mutant, the loss of PeIA potency was similar to that found for nonmutated receptors (Fig. 10A and Table 6). Additionally, the rat ␣6/␣3␤4 Q119F mutant as well as the ␣6/␣3␤4 Q119L mutant also showed lower sensitivity to [P13R]PeIA and [P13Q]PeIA (Fig.  11 (B and C) and Table 7). These results suggest that position 119 of human ␤2 and ␤4 and rat ␤4 are not equivalent with respect to ligand binding even when the same amino acid is present (by mutation) in the receptor. Additional evidence for the nonequivalency of position 119 for ligand binding was obtained with PnIA. There was no difference in PnIA potency between human ␣6/␣3␤4 and the ␣6/␣3␤4 L119F mutant and, consequently, no change in the ability of PnIA to discriminate between human ␣6/␣3␤2␤3 and ␣6/␣3␤4 nAChRs (Fig. 10C  Determinants of ␣-Ctx potency for ␣6␤4 nAChRs and Table 6). However, PnIA showed substantially increased potency on rat ␣6/␣3␤4 Q119F and ␣6/␣3␤4 Q119L mutants ( Fig.  11A and Table 7). In summary, the results of these receptor and ligand mutation studies argue for species-and subtype-specific ligand-receptor interactions that are context-dependent. Therefore, we urge caution directly comparing ligand-receptor interactions among species as well as generalizing results obtained for different nAChR subtypes of the same species. Furthermore, the fact that both PeIA and PnIA have a Pro residue in position 13 but only PnIA shows differential potencies for mutant rat, but not human, ␣6/␣3␤4 nAChRs also argues for caution when generalizing interactions between receptor residues and residues of different ␣-Ctxs.
The ␣6␤4 subtype is an emerging, novel target for the treatment of neuropathic pain, but very little information is available concerning the interaction of ligands with this subtype at the molecular level. Ligands selective for ␣6␤4 nAChRs might be critical to avoid off-target effects that can occur due to interactions with closely related subtypes, particularly ␣3␤4. In this study, we have identified ␤4 Leu-119 as an important residue of the human ␤4 subunit that interacts with ␣-Ctxs. To our knowledge, this is the first report to functionally identify this key interaction. The information contained in this study might ultimately guide the design of ligands that target ␣6␤4 nAChRs for the treatment of neuropathic pain.

Oocyte two-electrode voltage-clamp electrophysiology
Xenopus laevis frogs were obtained from Xenopus Express (RRID:SCR_016373). Protocols for the isolation of oocytes from the frogs were approved by the University of Utah institutional animal care and use committee. Methods describing the preparation of cRNA encoding human and rat nAChR subunits for expression of nAChRs in Xenopus oocytes have been described previously (44). The human and rat ␣6/␣3 constructs were generated by replacing the extracellular ligand-binding domain of the ␣3 subunit with that of the ␣6 subunit as described previously (47,48). These constructs were used because injection of oocytes with cRNAs encoding human ␣6 and ␤4 subunits resulted in no functionally expressed receptors across multiple donors (data not shown). The rat ␣6/␣3 construct was used for comparison and has been previously shown to display similar sensitivities to ␣-Ctxs compared with nonmutated ␣6 (49). Preliminary experiments varying the ratio of rat ␣6/␣3 to ␤4 subunit cRNAs by 10:1 or 1:10 to favor the formation of receptors with different stoichiometries had no effect on PeIA potency (data not shown). In all subsequent experiments, the oocytes were injected with equal ratios of cRNA for all subunit combinations.

Data analysis
Concentration-response data were obtained from a minimum of four oocytes. ␣-Ctx concentrations of Յ1 M were applied to the oocyte by continuous perfusion, and the ACh responses in the presence of the ␣-Ctxs were normalized to the average of at least three control responses. ␣-Ctxs were applied only after the ACh response-to-response variation was Ͻ10%. The variance of the responses is provided as the ϮS.D. and shown with error bars. ␣-Ctxs were routinely retested on oocytes from different donors to ensure data reproducibility. To estimate the IC 50 value for inhibition of the ACh responses by a given ␣-Ctx, the normalized data were analyzed by nonlinear regression and fit using a four-parameter logistic equation in Prism (RRID:SCR_002798) (GraphPad Software Inc., La Jolla, CA). The IC 50 values are presented with corresponding 95% confidence intervals to evaluate the precision of the IC 50 estimate. Although in many cases, the confidence intervals are nonoverlapping, for the purposes of this study, the difference between two IC 50 values is considered significant if Ն3-fold. For ␣-Ctx-receptor combinations where inhibition by the maximal ␣-Ctx concentration tested was less than ϳ40%, the response after a 5-min static bath exposure to 10 M ␣-Ctx was compared with control response, and the means were compared using a one-way analysis of variance and Fisher's least significant difference to determine significance. Significance was determined at the 95% level (p Ͻ .05). Concentrationresponse curves for activation of nAChRs were obtained according to the following procedures. ACh was applied in ascending concentrations, and the current amplitudes for each individual oocyte were analyzed by nonlinear regression and fit using a four-parameter logistic equation in Prism to obtain the calculated plateau value for activation. All ACh-evoked responses were then normalized to the calculated plateau value to obtain a percentage response and then analyzed with the same four-parameter logistic equation. Data for the ACh curves were collected using oocytes from three different donors to ensure reproducibility. Acetylcholine chloride (catalog no. A6625), potassium chloride (catalog no. P3911), and BSA (catalog no. A2153) were purchased from Sigma-Aldrich. Sodium chloride (catalog no. S271), calcium chloride dihydrate (catalog no. C79), magnesium chloride hexahydrate (catalog no. M33), sodium hydroxide (catalog no. S313), and HEPES (catalog no. BP310) were purchased from Fisher Scientific.
Site-directed mutagenesis of ␣6/␣3 and ␤4 subunits cDNAs for human ␣6/␣3 and human and rat ␤4 subunits in the pGEMHE vector were used as starting templates. cDNAs for rat ␣6/␣3 were in the pT7TS vector. Oligonucleotide primers were designed to individually span positions 177 of the ␣6/␣3 subunit and 110, 118, and 119 of the ␤4 subunit and included nucleotide substitutions to mutate each respective amino acid residue. The double mutant constructs were made using a single primer pair that mutated positions 118 and 119 in the same PCR. All oligonucleotides were synthesized by the DNA/Peptide Facility, part of the Health Sciences Center Cores at the University of Utah. Pfu Turbo (catalog no. 600250) DNA polymerase (Agilent Technologies, Santa Clara, CA) was used to extend the primers. The PCR conditions were as follows: 95°C for 120 s for denaturation followed by 30 cycles of 95°C for 30 s, 65-78°C (depending on the primers used) for 7 min, 72°C for 60 s, and a final extension step for 10 min at 72°C. The reaction was then digested with DpnI (catalog no. R0176S) to remove template cDNA (New England Biolabs, Ipswich, MA). Chemically competent DH5␣ (New England Biolabs, catalog no. C2987I) or 10-␤ (New England Biolabs, catalog no. C3019I) cells were used for transformation. The cells were grown at Determinants of ␣-Ctx potency for ␣6␤4 nAChRs 37°C for 1 h after transformation and then plated on agar plates containing ampicillin and maintained at 37°C overnight. Several colonies were selected from each plate and individually grown overnight at 37°C in Luria-Bertani medium containing ampicillin (Fisher Scientific, catalog no. BP1760). cDNA was isolated from the cells using a Qiaprep Spin Miniprep Kit (catalog no. 27104) (Qiagen, Valencia, CA) followed by sequencing at the University of Utah DNA Sequencing Core Facility to verify incorporation of the mutations. cDNAs for human ␣6/␣3 and human and rat ␤4 were linearized overnight at 37°C using the restriction enzyme NheI (New England Biolabs, catalog no. R0131S), and rat ␣6/␣3 cDNA was linearized using SalI (New England Biolabs, catalog no. R0138T) using the same protocol. Following linearization, the cDNAs were purified and eluted with water using the Qiaquick PCR purification kit (Qiagen, catalog no. 28104). cRNA was then prepared from the linearized cDNA using Ambion's mMESSAGE mMACHINE T7 Transcription Kit (Fisher Scientific, catalog no. AM1344) and purified using a Qiagen RNeasy Mini Kit (Qiagen, catalog no. 74104).

X-ray crystallography
The AChBP from A. californica was expressed and purified as described previously (51,52). Briefly, AChBP was expressed with an N-terminal FLAG epitope tag and secreted from stably transfected human embryonic kidney 293S cells lacking the N-acetyglucosaminyltransferase I (GnTI Ϫ ) gene (53). The protein was purified with FLAG antibody resin and eluted with FLAG peptide (Sigma-Aldrich, catalog no. F3290). Affinity-purified protein was further characterized by size-exclusion chromatography in a Superdex 200 16/60 gel filtration column (GE Healthcare) in 25 mM Tris-HCl (pH 7.4), 150 mM NaCl, and 0.02% NaN 3 (w/v). From this process, the pentameric association could be ascertained, and monomeric subunits and trace contaminants could be removed. Purified AChBP pentamers were then concentrated using a Millipore YM50 Centricon ultrafiltration unit (Fisher Scientific) to a final concentration of ϳ5 mg/ml.

Complex formation and crystallization
The PeIA-AChBP complex was formed by dissolving 10 mol of lyophilized PeIA with 50 l of purified and concentrated protein at a concentration of 5 mg/ml. The PeIA-AChBP complex co-crystals were obtained by the vapor-diffusion hanging-drop method. Concentrated protein complex was mixed in a 1 l:1 l solution consisting of 0.1 M Tris-HCl (pH 8.0), 0.25 M MgCl 2 , 20% (w/v) PEG 4000, incubated at 22°C, and suspended over 500 l of the same solution. Crystals of 0.3 ϫ 0.3 ϫ 0.2 mm final size appeared after a few weeks.

X-ray diffraction data collection
PeIA-AChBP complex co-crystals were transferred to a cryoprotectant solution consisting of 0.1 M Tris-HCl (pH 8.0), 0.25 M MgCl 2 , 12% (w/v) PEG 4000, and 10% (v/v) glycerol and flash-cooled in liquid nitrogen. A full set of X-ray diffraction data were collected at 278°C at beamline 8.2.1 (Advanced Light Source, Berkeley, CA). Diffraction data were processed and scaled using HKL2000 (54). Final data statistics are given in Table S1.

Structure refinement
The PeIA-AChBP complex structure was solved by the molecular replacement method using PHASER (55) using the AChBP/␣-Ctx BuIA structure (PDB entry 4EZ1) as a search model. The electron density maps were manually fitted in COOT (56) with iterative structure refinement done using phenix.refine (RRID:SCR_014224) (57), resulting in a final model with R work and R free of 19.6 and 22.8%, respectively. Refinement statistics are listed in Table S1. Atomic coordinates and structure factors have been deposited in the PDB (entry 5JME). Cartoon representations of the structures were generated using PyMOL (RRID:SCR_000305) (58). It should be noted that there was density for a fifth PeIA in the data obtained. However, although the density clearly showed the position of the peptide backbone, it was not sufficient for accurate positioning of the side chains. Out of an abundance of caution, it was decided to omit the final toxin chain to avoid any possible misinterpretation.

NMR spectroscopy and structure calculations
Peptide samples (1.0 mg) were dissolved in 550 l of 10% D 2 O, 90% H 2 O (pH ϳ3), and spectra were recorded on a Bruker Advance III 600-MHz spectrometer equipped with a cryoprobe. Data were collected at 290 K using TOCSY, NOESY H-N HSQC, and H-C HSQC experiments. Spectra were acquired with mixing times of 80 ms (TOCSY) or 200 ms (NOESY) and 4,096 data points in F2 and 512 in F1. Chemical shifts were referenced to internal 2,2-dimethyl-2-silapentane-5-sulfonate at 0 ppm. Spectra were processed with Topspin version 3.5 (Bruker Biospin) and assigned using the program CcpNmr Analysis (59). Structure calculations of [P13A]PeIA were based upon distance restraints derived from NOESY spectra and on backbone dihedral angle restraints generated using TALOSϩ (60). A family of 20 lowest-energy structures consistent with the experimental restraints was calculated using CYANA (61) and assessed using Molprobity (RRID:SCR_014226) (62). Experimental restraints and stereochemical quality assessment outcomes are provided in Table S2.