Single Amino Acid Substitutions in κ-Conotoxin PVIIA Disrupt Interaction with the Shaker K+ Channel*

κ-Conotoxin PVIIA (κ-PVIIA), a 27-amino acid peptide with three disulfide cross-links, isolated from the venom ofConus purpurascens, is the first conopeptide shown to inhibit the Shaker K+ channel (Terlau, H., Shon, K., Grilley, M., Stocker, M., Stühmer, W., and Olivera, B. M. (1996) Nature 381, 148–151). Recently, two groups independently determined the solution structure for κ-PVIIA using NMR; although the structures reported were similar, two mutually exclusive models for the interaction of the peptide with theShaker channel were proposed. We carried out a structure/function analysis of κ-PVIIA, with alanine substitutions for all amino acids postulated to be key residues by both groups. Our data are consistent with the critical dyad model developed by Ménez and co-workers (Dauplais, M., Lecoq, A., Song, J., Cotton, J., Jamin, N., Gilquin, B., Roumestand, C., Vita, C., de Medeiros, C., Rowan, E. G., Harvey, A. L., and Ménez, A. (1997)J. Biol. Chem. 272, 4802–4809) for polypeptide antagonists of K+ channels. In the case of κ-PVIIA, Lys7 and Phe9 are essential for activity as predicted by Savarin et al. (Savarin, P., Guenneugues, M., Gilquin, B., Lamthanh, H., Gasparini, S., Zinn-Justin, S., and Ménez, A. (1998) Biochemistry 37, 5407–5416); these workers also correctly predicted an important role for Lys25. Thus, although κ-conotoxin PVIIA has no obvious sequence homology to polypeptide toxins from other venomous animals that interact with voltage-gated K+ channels, there may be convergent functional features in diverse K+ channel polypeptide antagonists.

The venomous cone snails have evolved a neuropharmacological strategy largely based on highly sophisticated peptide ligands, the conotoxins (for reviews see Refs. 3 and 4). Particular molecular forms of both voltage-gated and ligand-gated ion channel families are targets of peptides belonging to specific conotoxin families. Among the ligand-gated ion channels that Conus peptides are known to inhibit are nicotinic acetylcholine receptors (5), N-methyl-D-aspartate receptors (3,6,7) and 5HT 3 receptors (8). In the voltage-gated ion channel su-perfamily, Conus peptides specific for various subtypes of calcium channels (9) and sodium channels (10) have been described. -Conotoxin PVIIA from the venom of the Eastern Pacific fish-hunting species Conus purpurascens was the first conopeptide described that targets voltage-gated potassium channels (1,11).
The role of -conotoxin PVIIA in prey capture has been established (1); this peptide is key to the very rapid immobilization of fish prey by C. purpurascens. This cone species uses the following two groups of toxins to affect different physiological programs in the prey after venom injection: a very rapid excitotoxic shock and a slower onset but irreversible neuromuscular block. -Conotoxin PVIIA plays a role in the initial excitotoxic shock response to venom injection; in combination with ␦-conotoxin PVIA (which delays inactivation of voltage-gated sodium channels), an immediate tetanic-like immobilization is elicited upon injection into fish.
The voltage-gated potassium channel Shaker, cloned from Drosophila melanogaster, is sensitive to -conotoxin PVIIA (1). Electrophysiological characterization of the interaction of -PVIIA 1 with chimeras between the Shaker channel and a -conotoxin-insensitive potassium channel from rat brain (Kv-1.1) demonstrated that the binding determinants on Shaker were in the pore region of the channel, presumably at the extracellular face (11).
Recently, two different groups published structures for -PVIIA, determined on the basis of multidimensional NMR techniques (12,13); the structures were generally similar. However, the interpretation of likely mechanisms of the molecular orientation of -conotoxin PVIIA in the channel mouth suggested by the two groups, based on the structural information, differed. Scanlon et al. (12) hypothesized that -PVIIA interaction with the Shaker potassium channel might be strongly analogous to the charybdotoxin-Shaker channel interaction (for a review of the charybdotoxin family, see Refs. 14 and 15). Although the two toxins share little primary sequence homology, it was suggested that specific amino acids in the -conotoxin (His 11 , Arg 18 , Lys 19 , and Arg 22 ) might correspond to four key residues in charybdotoxin (Lys 11 , Arg 25 , Lys 27 , and Arg 31 ). In contrast, Savarin et al. (13) suggested that a dyad motif of lysine and an aromatic residue (Lys 7 and Phe 9 or Phe 23 ) shared by a variety of unrelated toxins that inhibit voltage-gated potassium channels, including charybdotoxin from scorpion and BgK from sea anemone (2), would be a key determinant for -conotoxin PVIIA binding to the Shaker channel.
We have carried out a mutant cycle analysis of -conotoxin PVIIA; the results identify a binding surface on the toxin that is clearly critical for high affinity K ϩ channel blocking and provide functional evidence for evaluating the two hypotheses that were presented on the basis of structural information.

EXPERIMENTAL PROCEDURES
Solid-phase Peptide Synthesis-The protected peptide resin was built using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry as described previously (11) with side chains protected as follows: pentamethylchromasulfonyl (Arg), t-butyl (Hyp, Asp, and Ser), S-trityl (Asn, Gln, His, and Cys), and t-butoxycarbonyl (Lys). Protocols for the removal of peptide from resin, HPLC purification, and disulfide bridge formation have been detailed before (11). Under standard glutathione oxidation conditions, native -PVIIA folds into the biologically active form with Ͼ90% efficiency. Although some analogs folded with much lower efficiency, the correctly folded isomer was often well separated from other isomers during HPLC purification, making identification of the correctly folded form straightforward. Electrospray mass spectroscopy was carried out on all peptide analogs to confirm correct synthesis.
Verification by NMR and Circular Dichroism (CD)-One isoform, Asn 21 -Ala, could not be properly folded, and two other isoforms, Hyp 4 -Ala and Asn 5 -Ala, showed much reduced block but were obtained in yields that were insufficient for complete testing. Multiple substitutions for Arg 2 , Lys 7 , and Phe 9 were synthesized and folded (Arg 2 -Ala, Arg 2 -Gln, Arg 2 -Lys, Lys 7 -Ala, Lys 7 -Arg, Phe 9 -Ala, Phe 9 -Met, and Phe 9 -Tyr). The pattern of mixed folded isomers obtained for all analogs was very similar to wild-type -PVIIA. The earliest eluting major peak was collected and purified, and in each case, presumed to have the same disulfide connectivity as native PVIIA; one substitution for each of the three loci (Arg 2 -Ala, Lys 7 -Ala, and Phe 9 -Met) was analyzed by NMR and CD to confirm directly that the disulfide connectivity was native. However, Hyp 4 -Ala, Asn 5 -Ala, Leu 12 -Ala, and Asp 14 -Ala were difficult to purify in high yield, and their structures have not been confirmed. The Leu 12 -Ala analog in particular showed an unusual pattern of isomers following folding and needs to be further investigated.
Electrophysiological Measurements-The coding region of Shaker H4 employed in this study was cloned into Bluescript vector. Point mutations in the pore region were introduced by polymerase chain reaction mutagenesis, and the amplified sequences were confirmed by sequencing. Capped cRNA was synthesized in vitro after linearizing the plasmid with HindIII, and the transcription with T7 RNA polymerase was performed by a standard protocol (16).
Oocytes from Xenopus laevis were prepared as described previously (17,18). Frogs were anesthetized with 0.2% tricaine in ice water during surgery. cRNA coding for Shaker H4 was injected into oocytes, and the cells were incubated 1-7 days to allow expression of the protein. Prior to the electrophysiological measurements the vitelline membranes of the oocytes were removed mechanically with fine forceps. Whole cell currents were recorded under two-electrode voltage clamp control using a Turbo-Tec amplifier (npi electronic, Tamm, Germany). The intracellular electrodes were filled with 2 M KCl and had a resistance between 0.6 and 1 M⍀. Current records were low-pass filtered at 1 kHz (Ϫ3 db) and sampled at 4 kHz. The bath solution was normal frog Ringer's containing (in mM) the following: 115 NaCl, 2.5 KCl, 1.8 CaCl 2 , 10 Hepes, pH 7.2 (NaOH).
To estimate the IC 50 value of -conotoxin PVIIA and the corresponding analogs, whole cell currents of oocytes expressing either wild-type Shaker H4 K ϩ channel or mutant isoforms were recorded, and the peak currents, at a test voltage of 0 mV, were measured. IC 50 values were calculated according to IC 50 ϭ fc/(1 Ϫ fc) ⅐ [Tx], where fc is the fractional current, and [Tx] is the toxin concentration applied. This is an approximation for calculating the affinity of the toxin for the closed state of the ion channel protein (19). Toxin concentrations were determined by assuming that the integrated absorbance (determined by HPLC) was equivalent for all analogs (1 mol ϭ 33 A 220 units) except for Ala substitution for aromatic residues, where this value was reduced by 10%. The maximal toxin concentration tested for analogues of -conotoxin PVIIA that showed a low affinity to the Shaker K ϩ channel was usually 50 M. All electrophysiological experiments were performed at room temperature (19 -22°C).
Mutant Cycle Analysis-This method provides a means of revealing interactions between given pairs of toxin and channel residues as FIG. 1. Mutation of the residues His 11 , Arg 18 , Lys 19 , and Arg 22 to alanine results in isoforms of -conotoxin PVIIA that block the Shaker K ؉ channel with a potency close to the unsubstituted toxin. Whole cell currents recorded from oocytes expressing Shaker H4 K ϩ channels upon depolarization to 0, 20, and 40 mV are shown before (Control) and after addition of 1 M of the -conotoxin isoform indicated. The holding potential was Ϫ100 mV. Currents in the presence of toxin exhibit an apparent slowing of the activation/inactivation kinetics and an increased steady state current. This is because of the state-dependent unblocking during depolarization (see Ref. 19). a These analogs gave very low yields upon folding, insufficient for an NMR analysis. The L12A analog in particular had an aberrant elution profile. All of these analogs were inactive (for O4A and N5A, IC 50 Ͼ 10,000 nM; for L12A and D14A, IC 50 Ͼ 50,000 nM). Although the analogs were tested in the electrophysiological assay, because we were not able to obtain independent experimental evidence that these are properly folded, the results should be interpreted with caution. described earlier (20,21). Values for the coupling coefficient ⍀ were obtained by using the formula ⍀ ϭ (IC 50 WTchannel:WTtoxin ⅐IC 50 mutchannel:muttoxin)/(IC 50 WTchannel:muttoxin ⅐ IC 50 mutchannel: WTtoxin). Unity reflects no interaction, whereas deviation from unity indicates progressively stronger interactions. For ⍀ Ͻ 1 the reciprocal is given for ease of comparison of their deviation from unity (20). The change in coupling energy, ⌬⌬G, for a given pair of toxin-channel residues was calculated according to ⌬⌬G ϭ RT ln ⍀ (with R ϭ 8.314 Jmol Ϫ1 K Ϫ1 and T ϭ 295 K) (20,22). The studies of Schreiber and Fersht (21) suggest that a pair of residues showing a change in ⌬⌬G Ͼ 2.1 kJ/mol Ϫ1 may be expected to lie within 5 Å of each other (see also Ref. 23).

RESULTS AND DISCUSSION
Analogs of -PVIIA containing alanine substitutions in each amino acid position of the peptide were synthesized and folded (see "Experimental Procedures"), and the affinity of these isoforms was functionally assayed by electrophysiological measurements using the Xenopus oocyte expression system. Fig. 1 shows whole cell currents from oocytes expressing Shaker H4 K ϩ channels measured at different test potentials. Addition of 1 M of four isoforms, His 11 -Ala, Arg 18 -Ala, Arg 19 -Ala, and Arg 22 -Ala, of -conotoxin PVIIA leads to a block of the evoked currents of about 60 to 80%. Alanine substitutions of Ile 3 , Gln 10 , Asp 13 , and Ser 17 gave similar results. For these eight isoforms, the IC 50 values were all within 5-fold of wild-type -conotoxin PVIIA (see Table I). We note that some substitutions (Ile 3 -Ala and Gln 10 -Ala, for example) cause a bigger change in the IC 50 value than do others (such as Asp 13 -Ala, Ser 17 -Ala, and Lys 19 -Ala, which were all within 2-fold of wildtype PVIIA). Thus, residues such as Ile 3 or Gln 10 could well make some contribution to the interaction of PVIIA with the Shaker channel, but this would be relatively minor compared with the residues discussed below.
In contrast to the mutations of the eight amino acid residues of -conotoxin PVIIA described above, mutations of other residues resulted in a major reduction of the affinity for Shaker K ϩ channels; the IC 50 values of the different isoforms are summarized in Table I. Examples of -conotoxin PVIIA analogs that dramatically diminished activity are presented in Fig. 2. Because the alanine mutation at position 2 (Arg 2 -Ala) resulted in an isoform of -conotoxin PVIIA that did not block the evoked Shaker K ϩ currents, this residue was also mutated to a lysine (Arg 2 -Lys), and the affinity of this isoform was investigated. The upper panel of Fig. 2 shows that even 20 M of the isoform Arg 2 -Lys of -conotoxin PVIIA has almost no effect on the evoked Shaker K ϩ currents. This demonstrates that even a charge-conserving substitution at residue 2 of -conotoxin PVIIA results in a profound reduction in toxin affinity, suggesting that the arginine at this position is critically important for the interaction of the toxin with the channel protein. Similar results were obtained with the alanine mutation of the lysine at position 7. The middle panel of Fig. 2 shows that even at a concentration of 50 M the Lys 7 -Ala isoform had almost no blocking effect on the evoked currents. The same results were obtained with mutations of the phenylalanine at position 9 ( Fig. 2, lower panel). Interestingly, even a conservative substitution of phenylalanine to a tyrosine reduced the block by an order of magnitude (see Table I).
FIG. 2. Mutation of residues Arg 2 , Lys 7 , and Phe 9 results in isoforms of -conotoxin PVIIA that do not block Shaker K ؉ channels. Whole cell currents recorded from oocytes expressing Shaker His 4 K ϩ channels are shown before (Control) and after addition of 20 or 50 M of the -conotoxin PVIIA isoform indicated. The pulse protocol was as in Fig. 1.   FIG. 3. A model of the interaction surface of -conotoxin PVIIA. Red indicates that the IC 50 value following alanine substitution in this position was Ͼ10 M, yellow labels residues where alanine substitution resulted in an intermediate IC 50 value (500 nM to 10 M), and green labels residues where analogs had an IC 50 value within 10-fold of the native peptide (Ͻ500 nM). White residues were not substituted, could not be properly folded, or the relevant analogs were not available in sufficient quantity to verify proper folding (see the legend to Table I). Cysteines are shown in gray. The coordinates for the structure shown in the figure were based on the NMR structure of -conotoxin PVIIA (Protein Data Bank number 1KCP) (13).
Alanine mutations of the amino acid residues Gln 6 , Phe 23 , Asn 24 , and Lys 25 exhibited an intermediate behavior in that the IC 50 values of these isoforms were higher than 500 nM but less than or near 10 M. A summary of the IC 50 values of all isoforms tested is given in Table I. Thus, the alanine walk mutagenesis of -conotoxin PVIIA has identified several residues critical for functional block of the Shaker channel. A diagrammatic representation of the most important residues is shown in Fig. 3. The critical residues are colored red, and it is immediately apparent that these are adjacent to each other in the three-dimensional structure, though not in the primary amino acid sequence. The toxin surface, which presumably interacts with the outer vestibule of the Shaker pore to cause a channel block, is strongly suggested by our study to contain Arg 2 , Lys 7 , and Phe 9 as key amino acids.
Two other positions, Leu 12 and Asp 14 , may also play a critical role in -PVIIA-Shaker interaction. As shown in Table I, the Leu 12 -Ala and Asp 14 -Ala analogs were inactive; however, the yields upon folding these analogs were very poor, and we were unable to obtain sufficient amounts to do an NMR analysis. For this reason, we have been conservative in our interpretation and have not colored these residues red in the model shown in Fig. 3. We note however that these are adjacent to the three confirmed critical residues on the toxin surface.
The differing predictions of Savarin et al. (13) and Scanlon et al. (12) can be compared using these results. Scanlon et al. (12) suggested that four positively charged residues in -PVIIA were analogous to certain residues in charybdotoxin for interaction with the Shaker channel. A key residue in charybdotoxin is Lys 27 , which is postulated to directly occlude the pore of the channel (24); Scanlon et al. (12) suggested that the analogous position would be Lys 19 of -conotoxin PVIIA. Other charged residues that have been shown to be important in the charybdotoxin-Shaker interaction are Arg 25 , Lys 11 , and Lys 31 (25); the suggested analogous residues were Arg 18 , His 11 , and Arg 22 of -conotoxin.
The results of the alanine walk mutagenesis study described above do not support the postulated charybdotoxin analogy. In particular, the Lys 19 -Ala analog has an IC 50 value for the Shaker channel close to that of the wild-type toxin. None of the other proposed residues show much change in IC 50 when substituted by alanine; all gave values within a factor of 5 of the wild-type value and are among the residues represented in green in Fig. 3. The analogous placement of positive charges at the three-dimensional level between charybdotoxin and -conotoxin PVIIA appears not to be functionally significant.
There were prior indications from electrophysiological work that some of the specific interactions of -conotoxin PVIIA residues with outer vestibule amino acids of the Shaker channel pore were probably significantly different from charybdotoxin-Shaker interactions. Most notably, the mutation Phe 425 -Gly on Shaker increases the affinity of charybdotoxin by ϳ2,000-fold (26) but abolishes the binding of -PVIIA at the concentrations tested (11). The conservative mutation Phe 425 -Tyr has close to wild-type -PVIIA affinity, suggesting an important role for an aromatic residue at this position in the PVIIA-Shaker interaction.
Savarin et al. (13) postulated that, like other polypeptide toxins that are able to inhibit K ϩ channels, there may have been convergent evolution in one key feature of otherwise structurally unrelated toxins, the presence of a dyad consisting of a lysine residue (postulated to directly interact with the pore) and a hydrophobic amino acid (2). Upon considering the various choices in -conotoxin PVIIA, Savarin et al. (13) predicted that for -conotoxin PVIIA the likely key residues were Lys 7 and either Phe 9 or Phe 23 of -PVIIA. Our results confirm that Lys 7 and Phe 9 are indeed important for binding of -conotoxin to the Shaker potassium channel (and to a much lesser extent, so is Phe 23 ). Another notable feature of the predictions of Savarin et al. (13) is a potential role for Lys 25 . This residue makes Van der Waals contact with Lys 7 (mean distance of ␣C atoms derived from NMR structure is 4.4 Å), and an arrangement of two proximal lysine residues was compared with a similar motif in dendrotoxin that was found to have functional significance (27). We find that the Lys 25 mutant has the most reduced affinity of the analogs in the intermediate group (those residues colored yellow in Fig. 3), with an IC 50 value ϳ100-fold higher than the wild-type peptide.
A number of pore mutations of the Shaker channel were generated, and affinities of PVIIA and PVIIA mutants were measured. The effects of these single amino acid substitutions on the affinity of wild-type PVIIA are shown in Fig. 4. The mutations Lys 427 -Asn and Lys 427 -Asp resulted in an increased affinity of PVIIA of a factor of ϳ2. Similar effects have been reported for Lys 427 -Glu (12). Several other mutations (Phe 425 -Gly, Thr 449 -Lys, Thr 449 -Tyr, Thr 449 -Gln (11), and Val 451 -Lys) were not sensitive to PVIIA and therefore were not useful for the cycling analysis. All other channel mutations led to a slight reduction of PVIIA affinity or had no measurable effect (see Fig. 4, lower panel).
The results of the mutant cycling analysis are shown in Table II and Fig. 5. The mutant PVIIA peptides (with single amino acid substitutions) were tested against twelve different mutant Shaker channel isoforms. In most cases, strong interactions were not detected (the ⍀ values were close to 1, and the coupling energy (⌬⌬G) was close to 0). However, some ⍀ values obtained provide at least suggestive data for the orientation of the peptide in the postulated outer vestibule of the channel. The ⍀ value for the Lys 25 -Ala (toxin)-Met 448 -Lys (channel) pair of 7.3, which is equivalent to a ⌬⌬G value of 4.86 kJ/mol Ϫ1 , indicates an interaction between these two residues. Lys 25 is close to Phe 9 on the toxin molecule (distance of the ␣C atoms is ϳ8 Å). The interaction between Lys 25 and Met 448 makes the relatively modest ⍀ value of 2.7 for the Phe 9 -Tyr (toxin)-Thr 449 -Ser (channel) pair potentially significant. Any other substitution for Phe 9 , except Tyr, decreased the IC 50 value below detectable limits, and we were therefore constrained to using this conservative substitution (even the Phe 9 -Met substitution abolished detectable activity). This is suggestive evidence that the key Thr 449 residue, known to play a role in charybdotoxin binding, may be in close proximity to Phe 9 of PVIIA. Thus, Phe 9 of -conotoxin PVIIA may be at a homologous locus in the vestibule to Tyr 36 in charybdotoxin, i.e. the aromatic residue of the dyad postulated by Ménez and coworkers (2). This would therefore also make Phe 9 analogous to Phe 25 in KTX/AgTX and Tyr 23 in Shk and Tyr 26 in BgK (2,22). The mean distance between the ␣C atom of Lys 7 and the center of the benzene ring of Phe 9 derived from the NMR structure (13) is 8.3 Å (range, 7.4 -10.1 Å). There also seems to be some interaction of Lys 25 with Asp 431 (⍀ 3.7).
The ⍀ values obtained for the Arg 22 -Ala (toxin)-Asn 423 -Lys (channel) pair of 3.6 and the Ile 3 -Ala (toxin)-Ser 424 -Lys (channel) pair of 3.0 might indicate that there is also some interaction of these residues. Unfortunately, because no substitution for Arg 2 and Lys 7 has been found that gives measurable affinity, we were unable to carry out a mutant cycling analysis for these important loci. By using the spatial constraints that can be derived from the mutant cycle experiments, a docking of PVIIA into the KcsA crystal structure reveals that the docking configuration may fit with the mutant cycling data. An alignment of the residues that make up the pore regions of Shaker and KcsA is shown in the upper panel of Fig. 4. By assuming that Lys 25 of -PVIIA is in the vicinity of Leu 81 (Met 448 in Shaker), Phe 9 of the toxin can be positioned close to Tyr 82 (Thr 449 ) of the adjacent subunit of the channel pore vestibule (not shown). In this orientation, Lys 7 of -PVIIA is facing toward the ion channel pore. Furthermore, Arg 22 of -PVIIA is close to Gly 56 (Asn 423 ) and Ala 57 (Ser 424 ) of the same channel subunit with which Lys 25 interacts, whereas Ile 3 of -PVIIA is in the vicinity of Ala 57 (Ser 424 ) of the channel opposite to that with which Phe 9 interacts.
In summary, the data in this report do not support the predictions of Scanlon et al. (12) that there are positive residue homologies between PVIIA and charybdotoxin but are consistent with the critical dyad hypothesis developed by Ménez and co-workers (2) for polypeptide antagonists of voltage-gated K ϩ channels.  0.02 (4) 1.17 (4) 0.14 (2) 0.90 (1) (1)