Marked difference in saxitoxin and tetrodotoxin affinity for the human nociceptive voltage-gated sodium channel (Na_v1.7)

Abstract

Human nociceptive voltage-gated sodium channel (Na_v1.7), a target of significant interest for the development of antinociceptive agents, is blocked by low nanomolar concentrations of (−)-tetrodotoxin(TTX) but not (+)-saxitoxin (STX) and (+)-gonyautoxin-III (GTX-III). These findings question the long-accepted view that the 1.7 isoform is both tetrodotoxin– and saxitoxin-sensitive and identify the outer pore region of the channel as a possible target for the design of Na_v1.7-selective inhibitors. Single- and double-point amino acid mutagenesis studies along with whole-cell electrophysiology recordings establish two domain III residues (T1398 and I1399), which occur as methionine and aspartate in other Na_v isoforms, as critical determinants of STX and gonyautoxin-III binding affinity. An advanced homology model of the Na_v pore region is used to provide a structural rationalization for these surprising results.

The propagation of electrical signals in nerve cells is mediated by an ensemble of proteins, among which voltage-gated sodium ion channels (Na_vs) play a leading role. In mammalian cells, functional Na_vs are expressed as a large pore-forming α-subunit with four nonidentical repeats within a single polypeptide and one or two auxiliary β-proteins (1); 10 gene loci have been identified that encode the 10 unique isoforms of the α-subunit (Na_v1.1–1.9 and Na_x). Mutations of the Na_v1.7 isoform are known to result in disease states, with hyperfunctioning mutants conferring increased pain sensitivity to patients and hypofunctioning mutations resulting in complete insensitivity to pain (2, 3). These findings have generated a tremendous interest in Na_v1.7 as a clinical target for the treatment of pain (4–7). Designing small-molecule modulators of specific Na_v subtypes, however, is challenged by the high sequence homology between the different α-subunits and the lack of detailed structural information for any mammalian Na_v channel (8, 9).

Nature has provided a collection of topologically unique small molecules that bind with high affinity and varying degrees of isoform selectivity to the Na_v family of proteins. The guanidinium toxins, exemplified by the fugu poison (−)-tetrodotoxin (TTX), are one such class of agents that blocks Na⁺ influx by lodging in the outer mouth of the α-subunit defined by the pore loops (site 1) (10). Other guanidine-derived natural products, namely (+)-saxitoxin (STX) (11) and the gonyautoxins (known collectively as paralytic shellfish poisons), are proposed to act analogously (12, 13). Early work to characterize Na_vs was made possible with the availability of TTX and STX and led to the general classification of channels as TTX/STX-sensitive or -insensitive based on the relative potency of the toxin for Na current block (i.e., low nanomolar vs. micromolar IC₅₀, respectively). It is widely accepted that TTX and STX are selective nanomolar inhibitors of six of nine Na_v isoforms (Na_v1.1–1.4, 1.6, and 1.7) expressed in mammalian cells, although electrophysiology measurements of TTX and STX against recombinant Na_v subtypes are more prevalent for the former toxin. In a recent find, 4,9-anhydro-TTX, a naturally occurring derivative of TTX, has been identified as a selective antagonist of Na_v1.6 (14); to date, no structural basis for this remarkable toxin–Na_v affinity profile has been offered.

Herein, we show that the α-subunit of human Na_v1.7 (and other primates) possesses a 2-aa sequence variation compared with all other Na_v isoforms, which results in dramatic differences in the binding affinity for TTX and STX. The results of our study challenge the long-standing belief that TTX and STX have comparable activities against all Na_v channels. These findings, combined with docking studies using a highly advanced homology model of the channel pore, provide a detailed view of the molecular interactions that influence ligand binding. Such information offers the rather provocative possibility that site 1 may be a viable locus around which to engineer a selective Na_v1.7 inhibitor.

Results

Given the intense interest in human Na_v1.7 as a target for pain (4–7, 15) and growing interest in clinical applications of guanidinium toxins as antinociceptive agents (16–23), we began a series of electrophysiology recordings with TTX, STX, and gonyautoxin-III (GTX-III) against recombinant hNa_v1.7 α-subunit expressed in CHO cells. For comparative purposes, an identical series of measurements was made using rat Na_v1.4 (CHO). These data reveal striking and unexpected differences in affinity for TTX vs. STX and GTX-III against the two isoforms. As shown in Fig. 1 and Table 1, STX blocks rNa_v1.4 with an IC₅₀ value of 2.8 ± 0.1 nM, consistent with reported literature values (12). By contrast, the IC₅₀ value of STX for hNa_v1.7 is 702 ± 53 nM, a 250-fold difference potency. Structurally related GTX-III also displays a significant reduction in potency against hNa_v1.7 compared with rNa_v1.4 (1,513 ± 55 vs. 14.9 ± 2.1 nM, respectively). Analogous experiments performed with TTX confirm that the affinity of this toxin for both channel isoforms is comparable (hNa_v1.7: 18.6 ± 1.0 nM; rNa_v1.4: 17.1 ± 1.2 nM).

Fig. 1.

Whole-cell voltage-clamp electrophysiology measurements of STX and TTX against rNa_v1.4 and hNa_v1.7. (A) Langmuir isotherms of STX and TTX binding to recombinant rNa_v1.4 and hNa_v1.7 expressed in CHO cells; error bars represent SD for n ≥ 3. (B and C) Representative normalized current vs. time plots of hNa_v1.7 varying TTX and STX concentrations, which highlight the reduced potency of STX for inhibition of hNa_v1.7. Additional normalized current vs. time plots for STX binding are provided in Fig. S1.

Table 1.

Measured IC₅₀ values ± SD (nanomolar; n ≥ 3 cells) for STX, GTX-III, and TTX against WT and mutant Na_vs

	hNav1.7	hNav1.7 T1398M-I1399D*	rNav 1.4	rNav1.4 M1240T	rNav1.4 D1241I	rNav1.4 M1240T-D1241I
STX	702 ± 53	2.3 ± 0.2	2.8 ± 0.1	73 ± 2.6	53 ± 4.6	1,153 ± 60
GTX-III	1,513 ± 55	22 ± 2.9	14.9 ± 2.1	228 ± 4.3	38 ± 1.8	1,084 ± 46
TTX	18.6 ± 1.0	5.0 ± 0.9	17.1 ± 1.2	466 ± 42	8.7 ± 0.8	90 ± 4.7

IC₅₀ values were determined by fitting I/I_o vs. log[toxin], where I₀ = normalized current, to a Langmuir isotherm (toxin concentrations ≥ 4). Currents were elicited by 10-ms step depolarizations from a holding potential of −100 to 0 mV.

*T1398 and I1399 in hNa_v1.7 are structurally homologous with M1240 and D1241 in rNa_v1.4.

Comparative analyses of the primary sequences thought to comprise the pore loop regions of rNa_v1.4 and hNa_v1.7 have led to the identification of two amino acid variations found in repeat III of the α-subunit, M1240T and D1241I (rNa_v1.4 numbering). A BLAST search of the National Center for Biotechnology Information database reveals this double variation to be unique to human and other primate Na_v1.7 (Table 2 and Table S1); no other mammalian Na_v isoforms have both threonine and isoleucine at structurally homologous sites in the polypeptide. Accordingly, a series of reciprocal single- and double-point mutants of rNa_v1.4 and hNa_v1.7 was generated to test whether one or both of these amino acids were responsible for altering the binding affinity of STX and GTX-III.

Table 2.

Pore-forming sequence alignment of Na_v1.7 from select primate and nonprimate species

Animal	Domain I	Domain II	Domain III	Domain IV
Primate
Human	RLMTQDYWEN	RVLCGEWIET	VATFKGWTII	ITTSAGWDGL
Chimpanzee	RLMTQDYWEN	RVLCGEWIET	VATFKGWTII	ITTSAGWDGL
Rhesus monkey	RLMTQDYWEN	RVLCGEWIET	VATFKGWTII	ITTSAGWDGL
Nonprimate
Rabbit	RLMTQDYWEN	RVLCGEWIET	VATFKGWMDI	ITTSAGWDGL
Rat	RLMTQDYWEN	RVLCGEWIET	VATFKGWMDI	ITTSAGWDGL
Cow	RLMTQDYWEN	RVLCGEWIET	VATFKGWMDI	ITTSAGWDGL

The domain III M→T, D→I variation is underlined and observed only in primate Na_V1.7. Selectivity filter amino acids, DEKA, are bold. A complete list of Na_v1.7 pore-forming sequence alignments is in Table S1.

Three mutant forms of rNa_v1.4 and one mutant of hNa_v1.7 have been prepared and were successfully expressed in CHO cells and showed current densities similar to WT. Electrophysiology recordings give compelling evidence that M→T and D→I substitutions are important modulators of STX/GTX-III affinity to Na_v. As noted in Table 1, 1-aa mutations in rNa_v1.4 result in only moderate increases in measured IC₅₀ values for both STX and GTX-III. The potency of STX and GTX-III to the double mutant, rNa_v1.4 M1240T-D1241I, however, is greatly reduced (1,153 ± 60 nM and 1,084 ± 46 nM, respectively). To further examine the combined effect of M1240 and D1241 on STX/GTX-III binding, commensurate mutations were made to T1398 and I1399 in hNa_v1.7. Against hNa_v1.7 T1398M-I1399D, the IC₅₀ values for STX and GTX-III were determined to be 2.3 ± 0.2 and 22 ± 2.9 nM, respectively. The measured affinities for TTX against all three rNa_v1.4 mutants differ considerably from the other two toxins. A marked increase in TTX IC₅₀ is observed only for the single-point mutant M1240T, which was previously noted in the work by Jost et al. (24). Compared with TTX binding against WT rNa_v1.4, the measured IC₅₀ value for toxin block of rNa_v1.4 D1241I is slightly decreased (8.7 ± 0.8 nM), whereas the IC₅₀ value for rNa_v1.4 M1240T-D1241I is increased by fivefold (90 ± 4.7 nM). As indicated by these data, changes to these repeat III residues in the channel pore do not largely alter TTX binding. This conclusion is affirmed by experiments that measured TTX block of hNa_v1.7 T1398M-I1399D, for which the TTX IC₅₀ value is 5.0 ± 0.9 nM.

In an attempt to rationalize the observed differences in potency between STX, GTX-III, and TTX against both WT and mutant Na_vs, we performed a series of docking studies with homology models of the pore structure. For the purpose of this analysis, two reported Na_v homology models and one model that we developed were evaluated (25, 26). All three models use a prokaryotic K⁺ channel (KcsA, MthK, or K_vAP) as a starting template. Our model is comprised of the pore helix, the P-loop, and the S6 helix from each domain (Fig. 2 A and B). The putative toxin binding site is formed by residues 400–404, 755–758, 1,239–1,242, and 1,530–1,533. Docking models of STX and TTX were created against available amino acid mutagenesis data (27), which show E403, E755, E758, M1240, and D1532 to be critical for toxin block (Table S2). The model developed by Zhorov and our model are both able to recapitulate the strong binding interactions between the toxins and these select residues (Fig. S2). The two models have similar positioning of backbone atoms (rmsd = 1.5 Å for backbone atoms only); however, significant conformational differences are apparent in amino acid side chain atoms (rmsd = 3.2 Å for nonbackbone atoms). An additional constraint is present in the Zhorov structure that fixes an interaction between the 7,8,9 guanidinium of STX and D400. Mutagenesis data (D400A) show that this aspartate residue does not contribute appreciably to STX binding (Table S2). For the purpose of introducing a minimum number of constraints to our model, specific contact between the guanidium toxin and domain I Y401 was not included during model construction. Docking poses for TTX, STX, and GTX-III, however, show the aromatic ring of the tyrosine residue positioned within 3.5 Å of the respective guanidinium moiety. Such findings, particularly with TTX, are consistent with experimental data that posit a guanidinium cation-π interaction with Y401 (28, 29).

Fig. 2.

Homology model-derived images highlighting differences between rNa_v1.4 and hNa_v1.7. STX bound in site 1 of the α-subunit of the homology model of rNa_v1.4 is viewed from (A) above and (B) the side of the pore. The four domains of the pore are shown in cartoon representations and colored green (domain I), cyan (domain II), magenta (domain III), and yellow (domain IV). In B, residues comprising the DEKA selectivity filter are displayed as space-filling models; atoms are colored red (oxygen), blue (nitrogen), white (hydrogen), yellow (sulfur), and green (STX carbons) or white (protein carbons). In C–H, STX is docked into rNa_v1.4 (C) and the identical pose in hNa_v1.7 (D) to highlight the steric differences of the M1240T and D1241I variations. In C and D, STX is colored similarly as in A and B. The molecular surface of the protein is in white, with M1240 (C) or T1398 (D) shown in yellow and D1241 (C) or I1399 (D) in red. Electrostatic potential surfaces of rNa_v1.4 (E), hNa_v1.7 T1398M-I1399D (F), rNa_v1.4 M1240T-D1241I (G), hNa_v1.7 (H); depicted range is from −20 (red) to +5 kT (blue). In each image, residues 1,240 (1,398) and 1,241 (1,399) are displayed as stick figures.

Given recent reports of X-ray crystal structures for bacterial voltage-gated sodium channels Na_vAb (8) and Na_vRh (9), we have examined one of these channels (Na_vAb) as a template for the eukaryotic sodium channel. Alignment of the primary sequences of Na_vAb and Na_v1.4 was performed to mark the indices on Na_vAb corresponding to the P-loop region and toxin binding site (Table S2). Analysis of the Na_vAb X-ray structure reveals that outer vestibule loop amino acids E403, E758, M1240, and D1532 (Na_v1.4 numbering) do not face the pore lining and thus, would be precluded from interacting with bound toxins. A more thorough evaluation of Na_vAb as a template for eukaryotic Na_v1.4 has been recently reported (30). Sequence homology between the prokaryotic and eukaryotic channels in the pore loop region is low, and the former is not inhibited by TTX (31). Adjustment of the sequence alignment by inserting an additional amino acid in the P-loop of Na_v1.4 is needed to produce a model that faithfully recapitulates TTX binding data. From this analysis, the conclusions of which are reinforced by our computational experiments, we conclude that the structure of Na_vAb is no more accurate a starting template for construction of a Na_v1.4 homology model than KcsA or K_vAP.

Comparing structures of homology models of hNa_v1.7, rNa_v1.4, and the four mutant proteins reveals the influence of mutations on the binding pocket size, shape, and electrostatic surface at site 1. The site 1 binding pocket in the outer vestibule of hNa_v1.7 is larger than the binding pocket of rNa_v1.4, owing primarily to the replacement of the methionine residue with the sterically smaller threonine (Fig. 2 C and D). In Fig. 2 E–H, the electrostatic surface potentials predicted for rNa_v1.4, hNa_v1.7 T1398M-I1399D, rNa_v1.4 M1240T-D1241I, and hNa_v1.7, respectively, are shown. In the region of residues 1,240 and 1,241, the surface electrostatic potential of both rNa_v1.4 and hNa_v1.7 T1398M-I1399D varies in the region between −11 and −16 kT, whereas the surface potential of hNa_v1.7 and rNa_v1.4 M1240T-D1241I is a more positive −3 to −8 kT. Analysis of the single-point mutants rNa_v1.4 M1240T and D1241I reveals that the effect of these amino acid substitutions is cumulative, with both threonine and isoleucine contributing to a larger, less negative toxin binding site (compared with WT rNa_v1.4) (Fig. S3).

Analyses of the steric volume and electrostatic surface potential for each toxin highlight marked differences between the three natural products (Fig. 3). The molecular volumes for both STX and TTX are similar (251 and 252 ų, respectively), despite their disparate molecular shapes; GTX-III is 54 ų larger as a result of the C11 sulfate moiety. At physiological pH, TTX and GTX-III are both monocations, whereas STX exists entirely in its diprotonated form. Accordingly, large differences between the three toxins are noted in the calculated electrostatic surface potentials. Although the guanidinium group of TTX is positive at the molecular surface, the rest of the molecule is largely neutral with small regions of charge density. This finding is in contrast to STX, where the electrostatic potential is above 4 kT across the entire surface of the molecule, except for the carbamate side chain. The addition of the sulfate on GTX-III partially neutralizes the surface potential of the guanidinium groups and creates two negative regions surrounding the C11 sulfate and C13 carbamate.

Fig. 3.

Electrostatic potential surfaces from −4 (red) to 4 kT (blue) of the toxins (A) TTX, (B) STX, and (C) GTX-III highlighting the differences in charge distributions between the molecules.

Discussion

Ion conduction in six of nine mammalian isoforms of Na_v is blocked by TTX at low nanomolar concentrations (i.e., TTX-sensitive). The three so-called TTX-resistant isoforms differ by a 1-aa substitution in site 1, a structural modification that is responsible for the dramatic reduction in TTX affinity as shown by preparation of reciprocal mutants (32–34). Additionally, organisms that accumulate TTX or STX have evolved similar 1-aa variations in the outer vestibule of the channel that decrease toxin affinity—in some cases, by more than 1,000-fold (24, 35, 36). Studies that compare TTX and STX block of single-channel WT or mutant isoforms reveal parallel trends between the toxins (27, 37). Previous investigations of Na_v1.7 from rabbit (38), cow (39), and rat (40) have shown that this isoform is both TTX- and STX-sensitive. In contrast to these reports, we have found that a naturally occurring double-point variation of two repeat III residues in hNa_v1.7 results in selective destabilization of STX and GTX-III, but has minimal perturbation on TTX binding. Intriguingly, this natural variation of adjacent amino acids in repeat III appears to be unique to primate Na_v1.7 (Table 2 and Table S1).

The observed trends in toxin binding affinities noted in this study can be rationalized by differences in electrostatic and steric properties between the three toxins and six protein models. The rNa_v1.4 D1241I mutant, for example, removes a negative charge from the putative toxin binding site. The binding of STX, which has the most positive surface electrostatic potential and thus would form the strongest electrostatic interaction with the native channel, is reduced by 19-fold to this mutant. Additionally, the rNa_v1.4 M1240T mutation increases the volume of the binding site and reduces shape complementarity to the bound toxins. IC₅₀ values for STX and TTX binding are both decreased 26-fold with the inclusion of this mutation, whereas binding of the larger GTX-III to the more sterically open site is reduced by only 15-fold. In comparing toxin block of the single and double mutants of rNa_v1.4, the effects of both amino acid changes seem to be additive: all three toxins exhibit decreased affinity to rNa_v1.4 M1240T-D1241I. A similar rationale can be used to explain toxin binding affinity to the double mutant hNa_v1.7 T1398M-I1399D compared with hNa_v1.7.

We have identified a naturally occurring variation in 2 aa that line the outer pore region of Na_v and significantly alters guanidinium toxin binding profiles. These findings refute conventional wisdom, which regards Na_v1.7 as both a TTX- and STX-sensitive channel isoform. Homology modeling of the pore helix, P-loop, and S6 regions of the channel along with ligand docking studies provide a molecular context for understanding differences in toxin affinities between different WT and mutant Na_v isoforms. Our findings offer the rather exciting possibility that selective inhibitors of hNa_v1.7 could be designed around site 1 because of the uniqueness of its outer pore structure vis-à-vis other sodium channel isoforms. In the absence of crystallographic data for mammalian Na_vs, the rational design of Na_v1.7 pore blockers should be enabled with our homology model.

Materials and Methods

Electrophysiology.

Electrophysiology experiments were performed on CHO cells transfected with an expression vector containing the full-length cDNA coding for the appropriate WT or mutant Na_v sodium channel α-subunit. The preparation of plasmids containing cDNA encoding for WT rNa_v1.4 and hNa_v1.7 has been described previously (41, 42). Cells were transfected using the method of calcium phosphate precipitation; cotransfection with eGFP was used as a marker of transfection efficiency.

Sodium currents were measured using the patch-clamp technique in the whole-cell configuration with an Axopatch-200b amplifier (Axon Instruments), which was previously described in the work by Moran et al. (43). Borosilicate glass micropipettes (Sutter Instruments) were fire-polished to a tip diameter yielding a resistance of 1.0–2.0 MΩ in the working solutions. The pipette was filled with 40 mM NaF, 1 mM EDTA, 20 mM Hepes, and 125 mM CsCl, and the pH was adjusted to 7.4 with solid CsOH. The external solution had the following composition: 160 mM NaCl, 2 mM CaCl₂, and 20 mM Hepes; the pH was adjusted to 7.4 with solid CsOH. Current densities were generally between 2 and 4 nA, except for the Na_v1.4 M1240T mutant, which consistently gave higher current densities.

Stock solutions of each of the toxin derivatives (160 mM NaCl, 2 mM CaCl₂, 20 mM Hepes; pH adjusted to 7.4 with solid CsOH) were maintained at 4 °C and diluted with external solution before recording. STX and GTX-III were synthesized according to routes previously published by our laboratory (44, 45). TTX was purchased from Ascent Scientific and used without additional purification. Current measurements were recorded under continuous perfusion and controlled manually by syringe addition. STX, TTX, and GTX-III are potent neurotoxins and should be handled with appropriate care for safety.

The output of the patch-clamp amplifier was filtered with a built-in low-pass, four-pole Bessel filter having a cutoff frequency of 10 kHz and sampled at 100 kHz. The membrane was kept at a holding potential of −100 mV. Pulse stimulation and data acquisition used 16-bit D–A and A–D converters (Digidata 1322A; Axon Instruments) controlled with the PClamp software (Axon Instruments). Leak currents were subtracted using a standard P/4 protocol of the same polarity. Access resistance was always <4 MΩ, and the cell capacitance was between 4 and 20 pF as measured by the compensating circuit of the amplifier. All measurements were done at room temperature (20–22 °C). Recordings were made at least 5 min after establishing the whole-cell and voltage-clamp configurations to allow for stabilization of the voltage-dependent properties of the channels. Currents were elicited by 10-ms step depolarizations from a holding potential of −100 to 0 mV. Data were normalized to control currents, plotted against toxin concentration, and analyzed using custom software developed in the Igor environment (Wavemetrics). Data were fit to Langmuir isotherms to elicit IC₅₀ values and expressed as mean ± SD.

Mutatgenesis.

Primers were ordered from PAN Services.

Double mutations were performed through iterative single mutations:

Mutant Nav	Primer sequence
Nav1.4 M1240T	5′-GCCACATTCAAGGGTTGGACGGATATCATGTATGC
Nav1.4 D1241I	5′-GGGTTGGATGATTATCATGTATGCAGCTGTGGACTCC
Nav1.4 M1240T D1241I	5′-CAAGGGTTGGACGATTATCATGTATGCAGCTGTGGAC
Nav1.7 T1398M	5′-GCAACTTTTAAGGGATGGATGATTATTATGTATGCAGCAGTGG
Nav1.7 T1398M I1399D	5′-GCAACTTTTAAGGGATGGATGGATATTATGTATGCAGCAGTGG

Site-directed mutagenesis was performed using the Quikchange XL Site-Directed Mutagenesis Kit (Stratagene) according to the manufacturer’s protocols. The mutations were confirmed by DNA sequencing of the relevant section of the resulting plasmid.

Molecular Modeling.

Homology models were created using Modeler (46). Computational ligand preparation was achieved using OpenEye tools Omega and Molcharge (OpenEye Scientific). Docked poses were initially generated using FRED (OpenEye Scientific) and minimized in LigandScout (inte:Ligand; Maria Enzersdorf Austria) under a built-in Merck Molecular Force Field. Surflex (Tripos) was used to relax the residues in the binding pocket to the minimized poses of the ligands. Electrostatic potential surface calculations were performed using APBS (47). Specific details of the homology model creation and computational docking can be found in SI Materials and Methods. The homology models generated in this work are provided in Datasets S1, S2, S3, S4, S5, and S6.