Structure–Activity Relationship Evaluation of Wasp Toxin β-PMTX Leads to Analogs with Superior Activity for Human Neuronal Sodium Channels

Abstract

Beta-pompilidotoxin (β-PMTX) is a 13-amino acid wasp venom peptide that activates human neuronal sodium channel Na_V1.1 with weak activity (40% activation at 3.3 μM of β-PMTX). Through rational design of β-PMTX analogs, we have identified peptides with significantly improved activity on human Na_V1.1 (1170% activation at 3.3 μM of peptide 18). The underlying structure–activity relationship suggests importance of charge interactions (from residue Lys-3) and lipophilic interactions (from residue Phe-7 and Ser-11). Three top-ranked analogs showed parallel activity improvement for other neuronal sodium channels (human Na_V1.2/1.3/1.6/1.7) but not muscular subtypes (Na_V1.4/1.5). Finally, we found that analog 16 could partially rescue the pharmacological block imposed by Na_V1.1/1.3 selective inhibitor ICA-121431 in cultured mouse cortical GABAergic neurons, demonstrating an activating effect of this peptide on native neuronal sodium channels and its potential utility as a neuropharmacological tool.

Many venom peptides act as neuroactive substances by inhibiting or activating neuronal ion channels.¹ The natural diversity of venom peptides represents a rich repertoire of molecular interactions with ion channels, the understanding of which may inform rational discovery of ion channel modulators as therapeutics.¹⁻³ In particular, certain venom peptides display selectivity for specific subtypes of ion channels within a phylogenetic family.^3,4 An intriguing case is the recent discovery of Hm1a and Hm1b as selective activators for Na_V1.1 relative to other subtypes of voltage-gated sodium channels (Figure 1A).^5,6 The binding site of Hm1a and Hm1b was mapped to specific loops and helices of the fourth voltage-sensing domain (VSD4) in Na_V1.1 (aka site 3).⁵ While this work encourages elucidation of the structural determinants in Hm1a/1b for binding, their nature as knottin peptides presents significant challenges with synthesis, mainly because their signature inhibitor cysteine knot scaffold requires correct disulfide folding.^5b Here we explored the potential of engineering Pompilidotoxins (PMTXs), a different class of sodium channel activating venom peptides, also proposed to bind at site 3, to probe structural determinants for modulation of Na_V1.1.⁷⁻⁹ The PMTX scaffold contains only 13 amino acids and no cysteine knot, thus offering a synthetic advantage over the knottin scaffold (Figure 1B). Moreover, naturally existing PMTXs have selectivity for neuronal sodium channel subtypes (Na_V1.1/1.2/1.3/1.6/1.7) over skeletal muscle and cardiac channel subtypes (Na_V1.4 and Na_V1.5, respectively);⁸ thus, improved analogs may serve as useful tools in neuropharmacology.¹⁰⁻¹³

Figure 1.

(A) Structure of Hm1a and Hm1b. Sequence of amino acids indicated by one-letter code, lines represent disulfide linkage between residues. (B) Chemical structure of β-PMTX. (C) Modeled conformations of β-PMTX. Two conformations from a simulated ensemble are shown in green and yellow tubes. Positively charged residues are shown in blue. A locally kinked conformation was more populated around Gly-5, while no specific conformations were suggested in the C-terminal region.

For this study, we synthesized analogs of β-PMTX to understand the structure–activity relationship (SAR) with respect to activity for hNa_V1.1 and selectivity over other isoforms of human Na-channels such as hNa_V1.2–1.7. Peptide modifications were designed based on hypotheses generated from an analysis of (1) the SAR of previously reported PMTX analogs,^8,14 (2) key features of sodium channel binding site 3 (a proposed binding site for β-PMTX⁹), (3) learnings from binding of knottin toxins at site 3,¹⁵ and (4) conformational restriction of a simulated conformation.

As β-PMTX is a flexible linear peptide, we hypothesized that conformational restriction of the β-PMTX backbone to its pharmacologically active confirmation(s) may improve its activity. Accordingly, molecular dynamics simulation was performed to identify potential pharmacologically active conformation of β-PMTX. A 3D-modeled conformation of β-PMTX suggested the following: (1) β-PMTX may adopt a kinked conformation assisted by Gly-5 residue (plus importance of Gly-5 for activity has been reported for close analog, α-PMTX¹⁴), and (2) this conformation could be further stabilized by electrostatic interaction between spatially proximal negatively charged residue Asp-8 and positively charged residues Arg-1 or Lys-3 (Figure 1C). Based on these observations, we hypothesized that a β-turn-like conformation pivoting around Gly-5 could be a key feature of the pharmacologically active (i.e., sodium channel bound) conformation. To test this, Gly-5 was replaced by other turn-favoring residues such as d-Pro (peptide 2, Table 1) and d-Ala (peptide 3); however, these changes were not tolerated. Next, to probe whether positions Arg-1, Lys-3, and Asp-8 on the modeled beta-hairpin are critical for binding to sodium channels or for formation of secondary structure via charge interactions between them, we designed peptides 4 and 5 in which these residues were interchanged. Both of these peptides were also inactive, suggesting that correct positioning of these residues is critical for activity. We also attempted to rigidify the proposed beta-hairpin by creating disulfide-link containing peptides 6 and 7. These peptides were inactive as well. Although the beta-hairpin model is plausible, it appeared to be very sensitive to structural perturbation and did not lead to analogs with improved activity.

Table 1. SAR around β-PMTX Analogs^a.

^aSequence of amino acids indicated by one-letter code. Letters shown in red represent replaced amino acid residue. C* represents disulfide linkage between residues in peptides 6 and 7. Abbreviations: 1-Nal, 3-(1-naphthyl)-l-alanine; 2-Nal, 3-(2-naphthyl)-l-alanine; Phe(4-Cl), l-4-chlorophenylalanine. A dose–response curve was obtained on the Q-Patch using recombinant HEK-hNav1.1 cells. The values represent the area under the curve (AUC) and its standard error. The AUC ratio was calculated as (peptide AUC/baseline AUC).

Next, we decided to focus on basic residues of β-PMTX owing to the following findings: (1) acidic residue Glu-1616 of rNa_V1.2 has been reported to be important for β-PMTX activity,⁹ (2) acidic residue Asp-1376 at the site 3 (a proposed binding site for β-PMTX) in rat Na_V1.4 is essential for function of Hm1a,⁵ and (3) removal of basic residue Lys-3 from α-PMTX was not tolerated, whereas removal of acid residue Asp-8 via D8A mutation was tolerated.¹⁴ Based on these observations, we designed peptide 9 to investigate whether replacement of Lys-3 with a more basic Arg residue would increase activity. Encouragingly, peptide 9 showed improved activity. Furthermore, Kawai and co-workers have reported that S11A mutation in α-PMTX is tolerated without significant changes in activity.¹⁴ Therefore, replacement of Ser-11 with lipophilic residues was explored. Encouragingly, combining K3R and S11L mutations in a single peptide 10 led to a further boost in activity.

Encouraged by this result, we explored substitution of Ser-11 with other lipophilic substituents in peptides 11, 12, 13, and 14. These substitutions were tolerated, but they did not lead to improve activity compared to peptide 10.

Previously, lipophilic residues such as Phe-15 present in the scorpion toxin Lqh-II have been postulated to be among key residues that bind at site 3 of neuronal sodium channels, thus contributing to the activity of Lqh-II.¹⁵ In addition, changing Phe-7 in a truncated PMTX-analog to a non-natural amino acid modestly improved potency (4-fold) of this analog as a Na-channel blocker.¹⁶ These findings suggest an essential role of endogenous lipophilic residues in mediating the interaction between venom peptides and sodium channels. Therefore, we hypothesized that modifying Phe-7 to bear different levels of lipophilicity might also influence the activity of β-PMTX. Replacing Phe-7 with more lipophilic amino acids such as 1-Nal (peptide 16) and 2-Nal (peptide 17) resulted in a substantial boost in activity compared to β-PMTX. By contrast, a swap with a polar tryptophan residue (peptide 15) led to complete loss of activity. Encouraged by these data, we synthesized peptide 18, which incorporated the three individual amino acid changes that had led to improvements in activity (K3R, F7-2Nal, and S11L). Excitingly, peptide 18 is significantly more active than β-PMTX (compare 1170% activation of Na_V1.1 channel observed for peptide 18 with 40% for β-PMTX at 3.3 μM)¹⁷ and the most active PMTX analog reported so far (Figure 2).

Figure 2.

(A) Modulation of Na_v1.1 currents by 3.3 μM of peptide 18 (left) and β-PMTX (right). (B) Dose-dependent effects of peptide 18 (left, 123 nM to 3.3 μM) and β-PMTX (right, 123 nM to 30 μM) on the rate of inactivation of seven sodium channel subtypes, measured as the area under the curve (AUC) for current traces in Q-patch assays. The AUC ratio was calculated as (peptide AUC/baseline AUC).

Having significantly improved Na_V1.1 activity, we then characterized peptides 16–18 for sodium channel subtype selectivity. All three peptides retained a preference for neuronal Na-channel subtypes (Na_V1.1/1.2/1.3/1.6/1.7) over skeletal muscle and cardiac subtypes (Na_V1.4 and 1.5, respectively) (Table 2). Regarding selectivity among the five neuronal channel subtypes, the peptides retained a similar selectivity pattern observed for β-PMTX, but the separation between channels was greater with new analogs (e.g., peptide 17). The SAR identified for Na_V1.1 activity did not translate as effectively to channel subtype selectivity; thus, it remains to be seen which alternative structural features in the scaffold may be required for selectivity.

Table 2. Effect of β-PMTX Analogs Across Sodium Channel Isoforms at 3.3 μM^a.

	AUC ratio for hNaV1.X activation at 3.3 μM
Compounds	NaV1.1	NaV1.2	NaV1.3	NaV1.4	NaV1.5	NaV1.6	NaV1.7
β-PMTX	1.4 ± 0.05	0.9 ± 0.05	1.2 ± 0.1	0.8 ± 0.07	1.0 ± 0.1	1.3 ± 0.1	1.1 ± 0.06
Peptide16	3.1 ± 0.3	1.8 ± 0.3	2.0 ± 0.3	1.4 ± 0.2	1.7 ± 0.5	3.1 ± 0.3	2.9 ± 0.3
Peptide17	10.7 ± 1.7	2.4 ± 0.2	1.9 ± 0.1	1.6 ± 0.4	4.0 ± 1.3	6.7 ± 0.9	4.0 ± 0.3
Peptide18	12.7 ± 1.2	7.1 ± 0.5	4.4 ± 1.6	1.6 ± 0.08	1.3 ± 0.04	8.1 ± 0.8	9.6 ± 0.5

^aThe values represent the area under the curve (AUC) and its standard error. The AUC ratio was calculated as (peptide AUC/baseline AUC). For complete dose–response curves, see Supporting Information.

Lastly, we tested peptide 16 in cultured mouse cortical GABAergic neurons to examine its effects on native neuronal sodium channels. Cortical GABAergic neurons are known to express Na_V1.1 at high levels (Figure 3A).¹⁸ Consistently, we found that the action potential output of these neurons was sensitive to the Na_V1.1/Na_V1.3 subtype selective inhibitor ICA-121431 (Figure 3B, middle panels).¹⁹ When peptide 16 (300 nM) was applied on top of ICA-121431 (100 nM), action potential firing was partially rescued (Figure 3B, right panels). This is reminiscent of the results others have obtained for Hm1a in mouse Na_V1.1-haploinsufficiency GABAergic neurons,⁵ although peptide 16 is not selective for Na_V1.1. Its rescuing effect could thus reflect combined potentiation of multiple other sodium channel subtypes that coexpress with Na_V1.1 in these neurons. Nevertheless, this is an encouraging demonstration of utility for newly designed peptides as potential tools in neuropharmacology and drug discovery.

Figure 3.

Effects of peptide 16 on the excitability of cultured parvalbumin-expressing cortical GABAergic neurons. (A) Primary cortical neuronal culture from PV-Cre, Ai9-tdTomato double transgenic mice grown for 28 days in vitro. The fluorescent protein tdTomato labels a single parvalbumin-expressing cortical interneuron (arrowhead) in this field of view. Three adjacent neurons do not express tdTomato (arrows). The fluorescent image (left) was overlaid onto a bright field image (right). Scale bar, 200 μm. (B) Input–output trace examples from three representative cultured parvalbumin (PV) positive interneurons. The data demonstrates examples of the maximum firing rate of three individual PV+ interneurons (left panels). Subsequent application of the Na_V1.1/Na_V1.3 inhibitor, ICA-121431, diminished neuronal activity in all interneurons tested (middle panels). Coapplication of ICA-121431 (100 nM) and 16 (300 nM) could partially rescue the baseline interneuron firing activity (right panels), demonstrating the positive modulation of these peptides on interneuron firing.

In summary, we have identified significantly more active analogs of β-PMTX through rational design based on pharmacophore knowledge of scorpion alpha-toxin Lqh-II binding at site 3 of the sodium channels. These peptides retain the selectivity preference of β-PMTX for neuronal sodium-channels over skeletal muscle and cardiac subtypes similar to β-PMTX. Finally, we have demonstrated that peptide 16 can partially rescue the pharmacological block imposed by a Na_V1.1/1.3 selective inhibitor in cultured mouse cortical GABAergic neurons.

Supporting Information Available

The Supporting Information is available free of charge on the ACS Publications website at DOI: 10.1021/acsmedchemlett.9b00415.

• Detailed synthesis procedures for peptides and descriptions of the in vitro assay (PDF)

Author Present Address

^§ Department of Chemistry, Stanford University, 337 Campus Drive, Lokey Building, Stanford, California 94305, United States.

Author Contributions

The manuscript was written through contributions of all authors. C.E.G. and T.P. synthesized peptides, W.G. conducted electrophysiology experiments, M.K. supported computational work, T.P.P. and M.K. designed peptides, T.T. conducted neuronal studies, all authors contributed to the data analysis, and T.P.P. and Y.S. led the project and wrote the manuscript.

The authors declare no competing financial interest.