Synthetic analogs of the snail toxin 6-bromo-2-mercaptotryptamine dimer (BrMT) reveal that lipid bilayer perturbation does not underlie its modulation of voltage-gated potassium channels

Abstract

Drugs do not act solely by canonical ligand-receptor binding interactions. Amphiphilic drugs partition into membranes thereby perturbing bulk lipid bilayer properties and possibly altering the function of membrane proteins. Distinguishing membrane perturbation from more direct protein-ligand interactions is an ongoing challenge in chemical biology. Herein, we present one strategy for doing so, using dimeric 6-bromo-2-mercaptotryptamine (BrMT) and synthetic analogs. BrMT is a chemically unstable marine snail toxin that has unique effects on voltage-gated K⁺ channel proteins, making it an attractive medicinal chemistry lead. BrMT is amphiphilic and perturbs lipid bilayers, raising the question of whether its action against K⁺ channels is merely a manifestation of membrane perturbation. To determine whether medicinal chemistry approaches to improve BrMT might be viable, we synthesized BrMT and 11 analogs and determined their activities in parallel assays measuring K⁺ channel activity and lipid bilayer properties. Structure-activity relationships were determined for modulation of the Kv1.4 channel, bilayer partitioning, and bilayer perturbation. Neither membrane partitioning, nor bilayer perturbation, correlate with K⁺ channel modulation. We conclude that BrMT’s membrane interactions are not critical for its inhibition of Kv1.4 activation. Further, we found that alkyl or ether linkages can replace the chemically labile disulfide bond in the BrMT pharmacophore, and we identified additional regions of the scaffold that are amenable to chemical modification. Our work demonstrates a strategy for determining if drugs act by specific interactions or bilayer-dependent mechanisms, and chemically stable modulators of Kv1 channels are reported.

INTRODUCTION

Biological membranes are composites of lipid bilayers and embedded proteins. It has long been known that membrane protein function is sensitive to the composition of the host bilayer.(1-4) Commonly, drugs that modulate membrane proteins are presumed to target proteins, while in fact many act by changing the bulk properties of the host bilayer, thereby altering membrane protein conformational equilibria.(5-8) Modulators that act by bilayer perturbation promiscuously modulate a broad spectrum of unrelated membrane proteins.(6, 7, 9-15) When interpreting the mechanisms underlying the physiological actions of a drug, it thus becomes crucial to determine whether the action of an amphiphilic modulator may involve bulk bilayer perturbation, in addition to more specific interactions.

A prominent example of the importance of understanding drug mechanism involves capsaicin, a natural product of chili peppers that stimulates mammalian peripheral neurons to evoke a sensation of burning heat. Capsaicin perturbs bilayers and modulates a wide variety of membrane proteins, including Na⁺, K⁺, and TRP channels. Capsaicin modulates voltage gated Na⁺ and K⁺ channels via lipid bilayer perturbation,(5, 7) but capsaicin also has a specific receptor site on TRPV1.(16) Medicinal chemistry approaches have been successful in generating analogs of capsaicin that are selective TRPV1 inhibitors.(17) Similar efforts to selectively modulate Na⁺ or K⁺ channels with capsaicin analogs would be foolhardy, however, because modulation of membrane proteins via bilayer perturbation is fundamentally promiscuous. Thus, determining if lipid bilayer perturbation underlies modulation of a target is critical for the prediction of undesired effects on other membrane proteins.

Because lipophilic and amphiphilic drugs, by their chemical nature, partition into membranes and perturb the function of transmembrane proteins, it is a significant challenge to determine whether bilayer perturbation is the relevant mechanism underlying modulation of any particular target protein. To identify whether drugs operate by a bilayer mechanism, we previously developed a method of testing modulators for promiscuous activity against multiple unrelated classes of membrane proteins.(7) Although this method is effective, it requires significant resources and expertise with many membrane protein preparations. Herein we report a greatly simplified strategy for using structure-activity relationships (SARs) of a modulator against a single target of interest, in combination with synthetic membrane assays, to dissect the effects of bulk bilayer perturbation from those of direct protein binding.

The medicinal chemistry target in this study is the natural product ion channel modulator dimeric 6-bromo-2-mercaptotryptamine (BrMT, 1a). A component of the defensive mucus of the marine snail Calliostoma canaliculatum, it inhibits voltage-gated K⁺ channels of the Kv1 and Kv4 subfamilies.(18) BrMT is an allosteric modulator that inhibits channels by slowing the voltage activation steps that precede pore opening, without blocking the central channel pore.(19, 20) Allosteric modulators of Kv channels are valuable not only as research tools but also potentially as therapeutics.(21, 22) BrMT itself has limited utility because it contains a chemically labile disulfide bond that is degraded by light and reducing conditions.(19) BrMT is thus an attractive target for medicinal chemistry efforts to improve its stability.

Several observations suggest that the activity of BrMT against Kv channels may be affected by non-specific membrane partitioning. First, high concentrations of BrMT applied to outside-out membrane patches disrupt the patch-clamp seal.(19) Second, a series of chimeras between the BrMT-sensitive Shaker Kv channel and the insensitive Kv2.1 channel suggest that the region imparting sensitivity is in the S1, S2, and/or S3 transmembrane regions of sensitive channels.(23) Third, the wash-in and wash-out kinetics of BrMT are multiphasic, suggesting slow accumulation of BrMT in the cell membrane during prolonged exposures.(23, 24) Together, these effects are consistent with BrMT partitioning in and out of cell membranes and acting through the membrane to alter channel function. Similar to many other amphiphilic molecules that act by bilayer perturbation, the biological effect of BrMT, with its two aminoethyl groups, depends on the side of the membrane to which it is applied.(25-28) BrMT slows Kv channel voltage activation only when applied from the extracellular side of the membrane,(19) suggesting that its two positive charges may prevent it from crossing the membrane entirely. Certain Kv modulator peptides from animal venoms partition into, but do not cross, the outer leaflet of the plasma membrane bilayer. Many of these peptides bind to the transmembrane voltage sensor domains of the channels.(29-31) However, other closely related venom peptides modulate ion channels via bilayer perturbation.(32) It remains unclear whether BrMT modulates K⁺ channels by direct channel binding, or in an indirect manner, by perturbing the bilayer, or a combination of both.(33) To elucidate the mode of action of BrMT and potentially improve its properties as a lead compound for future mechanistic or therapeutic studies, we synthesized a series of analogs, including several with stable disulfide replacements. The resulting SARs were assessed separately in membrane partitioning, perturbation, and ion channel assays to test whether specific or non-specific interactions drive K⁺ channel activities with these bis-indole compounds.

MATERIALS AND METHODS

The Supporting Information contains detailed descriptions of the synthesis of all BrMT analogs, cell culture, electrophysiology, gramicidin-based fluorescence quench assay, and isothermal titration calorimetry.

RESULTS

Synthesis of novel BrMT analogs

To determine the minimal structural features required for modulation of Kv channels with BrMT (1a), we designed a flexible synthesis that would enable facile modification of the tryptamine scaffold as well as the disulfide linker (Scheme 1). Our synthetic route is similar to that reported by Gallin and Hall.(34) 6-Bromotryptamine (4) was prepared from 6-bromoindole according to the sequence reported by Davidson.(35) Protonation of 4 with trichloroacetic acid, followed by reaction with freshly distilled S₂Cl₂,(36, 37) yielded a mixture of mono-, di-, and trisulfides 1a–c that was characterized by LC-MS. Using a protocol reported by Showalter for the preparation of bisindole diselenides as tyrosine kinase inhibitors,(38) we increased the yield of the desired disulfide product 1a by treating the mixture with sodium borohydride to reduce the di- and trisulfides. Extraction of the non-polar monosulfide 1c with ether from the basic aqueous solution of the resulting indole-2-thiolate, followed by oxidation of the thiolate with hydrogen peroxide gave the disulfide 1a, which was purified by semi-preparative HPLC and treated with HCl in dioxane to yield the bis-hydrochloride salt. Five different tryptamines were prepared via variations of literature protocols (see Supporting Information for details), and these were transformed into the analogous bistryptamine-disulfides 5–8 (Table 1) according to the sequence of reactions in Scheme 1.

Table 1.

Compilation of properties of BrMT derivatives

Structure	IC50 (μM)a	Hill coeff.a	ClogPb	KPW→Lc	gA2 (μM)d
	1.1 ± 0.1 N = 10	1.7 ± 0.3 N = 10	5.32	560 ± 5 N = 3	40 ± 5 N = 3
	0.8 ± 0.2 N = 10	1.4 ± 0.3 N = 10	4.99	1000 ± 200 N = 5	13 ± 1 N = 3
	26 ± 9 N = 9	1.1 ± 0.5 N = 9	4.06	800 ± 200 N = 5	28 ± 1 N = 3
	2.1 ± 0.2 N = 21	1.8 ± 0.2 N = 21	4.81	450 ± 80 N = 4	12 ± 1 N = 3
	2.1 ± 0.2 N = 7	1.3 ± 0.2 N = 7	5.32	1900 ± 600 N = 4	inactivee (50 μΜ) N = 3
	inactivee (100 μΜ) N = 8	–	5.15	600 ± 200 N = 4	inactivee (50 μΜ) N = 3
	2.7 ± 0.3 N = 23	1.9 ± 0.3 N = 23	3.98	3100 ± 700 N = 3	2.9 ± 0.2 N = 3
	70 ± 30 N = 10	0.7 ± 0.2 N = 10	3.62	3900 ± 500 N = 4	1.9 ± 0.1 N = 3
	2.3 ± 0.3 N = 6	1.8 ± 0.2 N = 6	4.72	3300 ± 800 N = 3	12.8 ± 0.7 N = 3
	35 ± 7 N = 8	1.3 ± 0.3 N = 8	5.77	670 ± 90 N = 4	5.1 ± 0.3 N = 3
	inactivee (25 μΜ) N = 6	–	4.82	1400 ± 200 N = 4	inactivee (50 μΜ) N = 3
	inactivee (25 μΜ) N = 8	–	4.21	3500 ± 1100 N = 3	32 ± 5 N = 3

^aCalculated as best fit for inhibition of Kv1.4 ± standard deviation of parameter fit.

^bCalculated using ChemAxon MarvinSketch v. 17.4.3.0.

^cK_P^W→L is the equilibrium constant for partitioning from water to lipid. Calculated as the geometric mean ± positive standard error.

^dGramicidin A (gA) channel activity, as measured by the rate of quenching of intravesicular ANTS fluorescence by Tl⁺. gA₂ fits shown in Supp. Figure S2.

^eIndicates IC₅₀ or gA₂ was not measured (highest concentration tested in parentheses).

The relative instability of bistryptamine-disulfides, and their potential for disulfide exchange reactions in vivo (39), inspired us to explore the use of alternative linkers between the indole moieties. Several two-carbon indole linkers have been reported in the literature,(40, 41) but we expected that these would be too rigid and/or short to be effective disulfide replacements. Accordingly, we pursued a convergent synthesis of symmetrical bis-indoles by reacting suitable aniline derivatives with bis-alkynes. Dipropargyl ether was selected as our first choice, as it could provide a three-atom, ether-based linker between the two indole rings that would provide a distance between indole moieties comparable to that of the disulfide linker.

The optimized synthesis of the ether-linked compounds is given in Scheme 2. Commercially available 4-bromo-2-nitroaniline was subjected to a Sandmeyer reaction(42) to yield aryliodide 9d in 94% yield. The nitro group of 9d was reduced to aniline 9a using SnCl₂ and conc. HCl, followed by N-acylation with acetic anhydride to provide the amide 9c. Amide 9c was subjected to the double Sonogashira cross coupling conditions (Table S1, entry 7) to yield the bis-alkyne 10c in 90% yield. Treatment of 10c with aqueous TBAF, followed by amide hydrolysis gave 11 (see Supporting Information for details). Alkylation of the bis-indole 11 with Eschenmosher’s salt (43) gave the bis-gramine 12 in excellent yield. Owing to the potential instability of 12, it was converted without purification to the bis-nitrile 13 upon treatment with excess iodomethane and sodium cyanide in DMF. Reduction of the nitrile groups in 13 with LiAlH₄ was problematic and led to partial reduction of the aryl bromide, but reduction of 13 with alane(44) proceeded smoothly to give the bis-amine 14 (termed BrET) in 45% yield. Acetylation of 14 with acetic anhydride furnished the bis-amide 15 in 60% yield. The related BrMT analogs 16–19 (Table 1) were prepared by a sequence of reactions similar to those depicted in Scheme 2. Initial attempts to reduce 19 to the corresponding diamine were unsuccessful in generating compound of acceptable purity.

BrMT derivatization alters potency of ion channel modulation

We assessed the modulatory effects of BrMT analogs on currents through voltage-gated potassium channels. To measure the dose–response relations of many compounds in parallel, we developed an automated whole-cell voltage clamp assay against a BrMT-sensitive ion channel. BrMT inhibits Kv1 (Shaker-type) channels from invertebrates and vertebrates.(18) Among members of the Kv1 family, Kv1.4 was chosen because it is a potential target for chronic pain, and lacks high-affinity modulators.(45-48) Moreover, Kv1.4 traffics efficiently and reliably to the plasma membrane in mammalian cell lines, making it optimal for automated electrophysiology.(49, 50) We therefore created a CHO-K1 cell line with tetracycline-inducible expression of Kv1.4 to provide a scalable cell culture with consistent current levels amenable to automated patch clamp methods.

Amphiphilic drugs often have variable potencies in different experimental preparations. This preparation-dependent variability may stem from membrane partitioning that is sensitive to the exact composition of the cell membrane, solution flow, and other factors.(24, 51, 52) Not surprisingly, BrMT has different potencies against Shaker Kv1 channels in patches and whole cell voltage clamp of HEK cells, CHO cells, and Xenopus oocytes.(23) To serially compare the effects of BrMT derivatives with minimal variability, the assays were conducted with a commercially available automated patch clamp system that applies consistent solution flow for different experiments.

When BrMT was applied to voltage-clamped cells expressing Kv1.4, it inhibited the currents by slowing gating kinetics and reducing peak currents (Figure 1A). The effects of BrMT on kinetics were quantified by fitting a double exponential to the rise and decay of current after a depolarizing voltage step (Figure 1B). Increasing concentrations of BrMT progressively slowed the gating kinetics (Figure 1C). These effects are similar to those seen with Shaker Kv1 channels.(19) BrMT has two phenomenological effects on Kv1 channels: a slowing of activation kinetics and a diminishment of peak current amplitude after a voltage step.(19) Peak current amplitudes of Kv1.4 are affected more dramatically than the rate of channel activation by BrMT (Figures 1, 2A). This greater sensitivity of the peak amplitudes may be due to stronger coupling between activation and inactivation in Kv1.4 as compared to Shaker, channels.(53) For comparison among the BrMT analogs, we therefore used the IC₅₀ values determined from the changes in peak current because their fitting was better constrained than the results obtained from the more complex curve fitting procedures needed to analyze the activation kinetics. Modulation of Kv1.4 currents is apparent over a range of concentrations (Figure S1), and IC₅₀ estimates were successfully obtained for most BrMT analogs (Figure 2). The Hill slopes deduced from these fits are between 1 and 2 for most analogs. Hill slopes of greater than 1 could arise from cooperative interactions of the BrMT analog with multiple subunits of Kv channels.(20) One analog, 15, has a lower Hill slope, which could indicate that this weakly modulating analog acts by a different mechanism, but its low potency precludes any significant interpretations.

Figure 1. BrMT inhibits Kv1.4 channel activation.

(A) Kv1.4 current responses to indicated BrMT concentrations during voltage steps to from −100 mV to 0 mV. (B) Kv1.4 current responses. Colored lines correspond to same BrMT concentrations as panel A. Stippled lines are fits of Equation 2, in Supporting Information. Voltage step to from −100 mV to 0 mV. Vehicle τ_act = 1.72 ± 0.02, τ_inact = 67.3 ± 0.2; 1.5 μM τ_act = 2.54 ± 0.02, τ_inact = 136.6 ± 0.6; 3 μM τ_act = 3.17 ± 0.03, τ_inact = 181 ± 2; 6 μM τ_act = 4.2 ± 0.4, τ_inact = 300 ± 20. (C) Time constants of activation (filled circles) and inactivation (empty circles). Error bars indicate SEM. n = 9. Stippled lines are fits for Kd as in ref. 19. K_d from τ_act = 2.9 ± 1.0 μM, K_d from τ_inact = 1.6 ± 0.7 μM.

Figure 2. Structure-activity relationship of Kv1.4 inhibition.

Average inhibition of peak Kv1.4 currents as function of concentration. The shapes and colors of markers correspond to compounds with modifications indicated by shapes in left column. Error bars indicate SEM. Lines are fits of Equation 1, in Supporting Ingormation, with parameter values indicated in Table 1. Color and pattern of lines correspond to compounds with modifications indicated by shapes in left column. (A) BrMT and analogs with modifications at the 6-positions on the indole rings. (B) Analogs with modifications of the disulfide linker and amines. (C) Ineffective analogs without aminoethyl groups.

Factors affecting channel inhibition emerge from analysis of SAR data. Replacement of the 6-bromo moiety of BrMT with a chloro or methyl group as in 5 and 7 had little effect on inhibitor potency, whereas substitution with the smaller and more electronegative fluoro group as in 6 led to a decrease in potency by an order of magnitude (IC₅₀ = 26 μM). Prior measurements with a BrMT analog containing only a hydride at position 6 indicated that it was also an order of magnitude less potent than BrMT against Shaker K⁺ channels.(54) Moving the bromo group to the 5-position on the indole ring as in 8 had only a minor effect on potency. These results suggest that variable indole substituents are tolerated.

BrMT loses its potency against Kv channels when the disulfide is reduced to form monomeric compounds.(19) On the other hand, the disulfide linkage between the indole groups is remarkably tolerant to replacement. For example, activity is retained when the disulfide moiety in BrMT is replaced with the ether linkage in 14, or when the disulfide in the 6-chloro-substituted analog 5 is replaced with the trimethylene linker in 16. That alterations in the linker between the indoles have only mild effects on Kv potency was also reported by Gallin and Hall,(34) who found that Kv1 channel inhibition was maintained for 1a–c, wherein the number of sulfur atoms bridging the indole rings was varied from one to three.

The discovery that Kv inhibition is retained with an ether or alkyl linkage between the indole rings permitted additional SAR studies using these more chemically stable scaffolds. We found that compounds 11, 18, and 19, which each lack the aminoethyl side chain, are inactive at the highest concentrations tested. Moreover, the N-acetyl derivative 15 and the N,N-dimethylaminoethyl analog 17 are less potent than the parent compound BrET (14) by more than an order of magnitude. Collectively, these results indicate that modification at the 2- and 6-positions of the indole rings of BrMT are well-tolerated, but the aminoethyl groups at the 3-positions are critical for channel activity: compounds lacking basic sidechains at the 3-position were inactive (e.g., 11, 18, 19).

BrMT derivatization alters bilayer partitioning, but partitioning does not predict ion channel modulation

We measured the degree to which the BrMT analogs partition into membranes using isothermal titration calorimetry (ITC).(55) These experiments were conducted with suspensions of large unilamellar phospholipid vesicles (LUVs) (see Supporting Information) as a surrogate for the bilayer component of cell membranes. All of the analogs partitioned into the bilayer with water→bilayer partition coefficients (K_P^W→L) ranging between 450 and 3900 (Table 1); the calculated log ([n-octanol]/[water]) (ClogP) values are also listed for comparison. K_P^W→L increased when the disulfide linkage was replaced with a 3-atom alkyl or an ether linkage. For example, replacing the disulfide linkage of BrMT (1a) (K_P^W→L = 560) with an ether linkage as in 14 (K_P^W→L = 3100) led to an increase in K_P^W→LP, as did replacing the disulfide bridge of the chloro analog 5 (K_P^W→L = 1000) with a trimethylene linkage in 16 (K_P^W→L = 3300). If channel modulation results from compounds partitioning into the membrane bilayer, we would expect that the Kv1.4 IC₅₀ values would correlate with the K_P^W→L values, but this is not the case (Figure 3). Examination of small structural changes that influence partitioning and/or channel modulation also indicate that the two properties are independent. Replacing the disulfide linkage of BrMT (1a) with the ether moiety in 14, increases K_P^W→L but has little impact on IC₅₀. The acetylated, non-basic compound 15 has an increased fraction in membranes versus its basic parent 14 (K_P^W→L = 3900 vs. 3100), but its potency against Kv1.4 drops substantially (IC₅₀ = 70 vs. 2.7 M) (Figure 3, brown arrows). Replacing the disulfide linkage of 5 with a trimethylene linkage in 16 also increases K_P^W→L with little impact on IC₅₀. Comparison of 16 to the related N,N-dimethylaminomethyl compound 17 is a rare example of a relatively minor structural change that simultaneously decreases both the K_P^W→L and inhibitory potency (Figure 3, green arrows). Overall, increases in the partition coefficient were not coupled to an increase of inhibitory potency. Fitting a linear regression to K_P^W→L and IC₅₀ values did not yield significance (r = 0.34, p = 0.37). The fraction of a BrMT analog in the bilayer does not correlate with its potency, which suggests that bilayer partitioning is not sufficient for Kv channel modulation.

Figure 3. Membrane partition coefficients do not predict channel inhibition.

Numbers by each marker denote the analog identity. Shape and color of markers correspond to Figure 2. Values are from Table 1. Black arrows denote analogs with so low activity that we could not measure it (null response), with the marker indicating highest concentration tested. Brown and green arrows with dashed lines indicate modification of disulfide linker. Brown and green arrows with dotted lines indicate modification of aminoethyl groups.

Membrane perturbation by BrMT derivatives is not correlated with bilayer partitioning

Bilayer partitioning itself did not underlie the variable potencies of Kv channel inhibitors, but it remains a possibility that non-specific bilayer perturbation might affect their inhibition of Kv1.4, because small membrane perturbing amphiphiles are known to inhibit many Kv channels(7). We therefore probed for bilayer-modifying effects using a gramicidin A (gA) based fluorescence assay that monitors changes of gA activity in LUVs.(56, 57) This assay exploits the gramicidin channel-permeant thallous ion (Tl⁺), which quenches the fluorescence of the water-soluble fluorophore 8-aminonaphthalene-1,3,6-trisulfonate (ANTS). gA permits transmembrane Tl⁺ flux only as a dimer, and the equilibrium between monomeric and dimeric gA is altered by lipid bilayer perturbation. ANTS fluorescence is measured after addition of Tl⁺ to a suspension of ANTS-loaded gA-containing LUVs, and the time course of fluorescence quenching (Figure 4A) provides a measure of the changes in intravesicular [Tl⁺]. The initial rate of fluorescence quenching provides a measure of the initial rate of Tl⁺ influx into the LUVs (58-60), which varies with changes in the number of dimeric gA molecules in the LUV membrane. It is thus possible to determine how a drug, or other amphiphile, shifts the gA monomer/dimer equilibrium, which provides a measure of the bilayer-modifying potency of the molecule of interest. The BrMT analog concentration that elicits a doubling of the rate of gA-dependent quenching is denoted as gA₂ and serves as a metric for the bilayer-perturbing potency (Figure S2). Note that a lower gA₂ concentration indicates more potent bilayer perturbation. Some noticeable structural trends were observed (Table 1). Analogs 11 and 18 that lack aminoethyl side chains did not have a measurable gA₂, because they caused no changes in the quench rate at the highest concentration tested. This result indicates that these analogs have minimal effects on bilayer properties. The linkage between the tryptamine monomers also affected bilayer perturbation. For example, the ether-linked derivatives 14 and 15 had the lowest gA₂, and replacing the disulfide linkage of 5 with a trimethylene chain as in 16 did not detectably change gA₂. Other changes to the aminoethyl side chains as shown by 14 vs. 15 and 16 vs. 17 increased the bilayer-modifying potency.

Figure 4. Membrane perturbation is distinct from partitioning and does not predict channel inhibition.

(A) Gramicidin A (gA) channel activity, as measured by the rate of quenching of intravesicular ANTS fluorescence by Tl⁺. For each trace, the gray dot denote the results obtained in the nine individual repeats, whereas the colored dots denote the average value at each time point. (B) Membrane partition–perturbation relationships. Numbers next to the different markers denote the analog’s identity. Shape and color of markers as in Figure 2. Data from Table 1. Black arrows indicate null response, with marker indicating highest concentration tested. Brown and green arrows with dashed lines indicate modification of disulfide linker. Brown and green arrows with dotted lines indicate modification of aminoethyl groups. (C) Membrane perturbation–inhibition relationship (data from Table 1).

Bilayer partitioning did not predict the bilayer-perturbing potency, as there was no obvious correlation between K_P^W→L and gA₂ (Figure 4B) (r = −0.27, p = 0.47). Divergent effects of structure on K_P^W→L and gA₂ can also be seen when comparing the related analogs 1a vs. 14 and 15, as well as 5 vs. 16 and 17. For compounds 1a, 14, and 15, a decreasing gA₂ was consistent with an increasing K_P^W→L, as would be expected if an increase in partition coefficient decreased the aqueous concentration needed to reach a mole-fraction in the bilayer that caused the perturbation. Substituting an ether linkage into 1a giving 14, for example, increased bilayer partitioning and perturbation. When the aminoethyl groups in 14 are acetylated to give 15, there is a further slight increase of bilayer partitioning and perturbation (Figure 4B, brown arrows). However, in a sequence of modifications at the same positions of the Cl-substituted disulfide in the series 5, 16, and 17, gA₂ appeared insensitive to K_P^W→L (5→16, Figure 4B, stippled green arrows), and then anti-correlated (16→17). A particularly striking example of the divergence between gA₂ and K_P^W→L is when the bromo group at C6 in 1a is moved to C5 in 8 resulting in a moderate increase in K_P^W→L, yet complete inactivity in the gA assay (Figure 4A, right panel). Overall, BrMT analogs did not show a consistent dependence of bilayer perturbation on partitioning.

Potency of channel modulation is not dictated by bilayer perturbation

Bilayer perturbation was altered by BrMT analogs in a fashion distinct from partitioning, so we then assessed whether perturbation might be the mechanism underlying Kv modulation. When Kv1.4 IC₅₀s are plotted vs. gA₂ as a measure of bilayer perturbation, there is no obvious correlation (r = −0.37, p = 0.36) (Figure 4C). In fact, more potent bilayer-perturbing compounds are generally weaker inhibitors of Kv1.4. Substitution of the native disulfide linkage of 1a for the ether linker in 14 led to a 10-fold decrease in gA₂ (10-fold increase in bilayer-perturbing potency) with a slight decrease in potency of channel inhibition (IC₅₀ = 2.7 μM for 14 vs 1.1 μM for 1a). In contrast, when the disulfide linkage in the 6-chloro analog 5 is replaced with the trimethylene linker in 16, there is little change in gA₂, but a three-fold increase in IC₅₀. Changes to the aminoethyl side chain (e.g., 14→15 or 16→17) decrease gA₂ but increase IC₅₀ (Figure 4C). Notably, analog 8, which does not perturb bilayers at the highest concentration tested, is a strong modulator of Kv1.4, whereas the most bilayer-perturbing analog 15 is a weak modulator of Kv1.4 channels. Inspection of the current traces recorded in the presence of 15 reveals that inactivation is accelerated. This action is consistent with prior findings that bilayer-perturbing detergents modulate inactivation of Kv1 channels,(61) and suggests Kv1.4 modulation by 15 may be via a different mechanism than the less bilayer-perturbing BrMT analogs.

DISCUSSION

Overall, these relationships show that Kv1.4 inhibition does not correlate with either bilayer partitioning or bilayer perturbation. We conclude that though BrMT and its analogs that are active against ion channels all partition into bilayer and usually perturb bilayer properties, the partitioning and alterations in bilayer properties do not drive potassium channel modulation. We conclude that specific interactions between Kv1.4 and “strong” modulators such as BrMT and BrET are critical for activity.

Limitations

Our results show that bilayer partitioning and perturbation are not critical determinants of modulation of Kv channels by BrMT or its analogs. These conclusions rely on interpretations that have caveats:

1. The bilayer partitioning and perturbation experiments were conducted with bilayers formed by synthetic phosphocholine lipids. Results in multi-component bilayers,(62-64) and whole cells,(10) have been found to be similar to those obtained in single-component bilayers. On the basis of these previous validations, we expect the impact of compounds on single-component bilayers will extend to living cells more generally. However, our experiments do not exclude the possibility that BrMT derivatives partition differently into CHO cells and/or interact in a binding pocket with specialized lipids.

2. Our interpretation assumes that the BrMT derivatives alter Tl⁺ flux through gA channels by perturbing lipid membranes, not by interacting directly with the gA channel itself. Our experiments cannot completely exclude such direct effects; results with many different structurally diverse amphiphiles show that they have similar effects on right- and left-handed channels, which effectively excludes direct interactions.(6, 8, 65) The achirality of the BrMT derivatives here precludes such validation.

Implications

By using a series of analogs of the snail defensive toxin BrMT, we identified certain molecular features that determine its inhibition of potassium channels as well as its partitioning into and perturbation of membranes. Our findings are consistent with BrMT acting through a drug binding pocket at the lipid bilayer–potassium channel interface.(31, 66, 67) The positive charges of other amphiphilic, membrane partitioning compounds are crucial for their inhibition of Kv1 channels,(68) which is consistent with our observations for BrMT and analogs. Compounds 1a, 5, 7, 8, 14, and 16 each strongly inhibit Kv channels, but of these only the native compound, BrMT (1a) has a minimal impact on membranes. The BrMT analog 7 is the only strong Kv1.4 inhibitor that shows a relative decrease in bilayer partitioning and is a more potent bilayer perturber. On the other hand, 8 is the only strong Kv1.4 inhibitor that is a less potent bilayer perturber, and that has a higher partition coefficient than BrMT. Thus, all of our strongly Kv1.4 inhibiting analogs either partitioned into or perturbed the bilayer more than BrMT. We speculate that membrane partitioning and perturbation may have played a role in the natural selection of BrMT to be the defensive toxin of the sea snail, possibly because of its minimal negative impacts on snail membranes. We also conclude that bilayer interactions are not correlated with inhibition of Kv1.4.

BrMT analogs as research tools

Determining what conformations ion channels adopt is an ongoing challenge of molecular physiology. BrMT is a toxin that selectively binds to resting Kv conformations(19), making BrMT a powerful tool to explore conformational changes of potassium channels. Conformation-selective ligands are of increasing importance for associating functional states of proteins with structurally-defined conformations. For example, conformation-selective toxins were the enabling factors for reconstructing TRPV1 ion channels in an open state(16, 69-71), and co-crystals of ASIC ion channels with peptide toxin gating modifiers led to identifying structures of new channel conformations(72-74). In functional studies, tagged conformation-selective toxins allow imaging of Kv2 channels adopting specific conformations in live cells.(75, 76) BrMT-like compounds similarly can stabilize the resting conformation of Kv1 channel voltage sensors, and they have the potential to be useful tools for studying channel conformations in cells and live tissue. The ether-linked analog BrET (14) identified in this study is comparable in potency to BrMT, but lacks the unstable disulfide moiety of the natural product. With the information that modifications to the indole ring and the linker between them are well-tolerated, a path is revealed for generating additional conformation-selective ligands targeting resting conformations of Kv1 ion channel voltage sensors.

Scheme 1.

Total synthesis of BrMT

Scheme 2.

Synthesis of ether-linker analogs 14 (BrET) and 15

ABBREVIATIONS

ACN

acetonitrile

ANTS

8-aminonaphthalene-1,3,6-trisulfonate (di-sodium salt)

ASIC

acid-sensing ion channel

4-AP

4-aminopyridine

BrET, BrMT analog

2-[2-(<3-(2-aminoethyl)-6-bromo-1>H-indol-2-yl]methoxy}methyl)-6-bromo-1H-indol-3-yl]ethan-1-amine

BrMT

snail toxin 6-bromo-2-mercaptotryptamine dimer

ClogP

calculated log of the partition coefficient ([n-octanol]/[water])

gA

gramicidin A

gA₂

concentration of compound that elicits a doubling of the rate of gA-dependent quenching

H_P^W→L

enthalpy of partitioning from water into lipid

IC₅₀

half-maximal inhibitory concentration

ITC

isothermal calorimetry

K_P^W→L

equilibrium constant for partitioning from water into lipid

Kv

voltage-gated potassium channel

LUV

large unilamellar vesicles

SAR

structure-activity relationship

TRPV1

transient receptor potential vanilloid 1 channel