Substitutions at two distant sites along the thumb-palm domain subunit interface are determinants of differences in Ma-3 activity between rat ASIC1a to ASIC1b

We set out to investigate the molecular basis for the different pharmacology of Ma-3 between rat (r) ASIC1a and rASIC1b, which only differ in the first third of the protein (residues 1–186 in rASIC1a) and share an identical thumb domain containing the principal binding site for mambalgins. First, we confirm that the mambalgin variants Ma-1 and Ma-3 (single T23I substitution) show no significant difference in ASIC pharmacology (Supplementary Fig. 2 and Supplementary Table 1). Ma-3 inhibits rASIC1a ~ 10-fold more potently than rASIC1b under the same pH conditions, with IC50 values of 3.9 nM and 38.3 nM, respectively (conditioning pH 7.45 and stimulating pH 6; Fig. 1, statistics in Supplementary Table 2). Furthermore, whereas rASIC1a is completely inhibited at saturating Ma-3 concentrations, the inhibition of rASIC1b is incomplete, reaching only ~ 60% inhibition (Fig. 1c and d). To elucidate the molecular basis of this subtype-dependent activity, we targeted non-conserved residues at the interface between the thumb domain and the palm domain of the adjacent subunit, away from the core mambalgin-binding residues on the thumb, changing the residue in rASIC1a to the corresponding residue from rASIC1b (Fig. 1a–e and Supplementary Table 2). Ma-3 showed equipotent inhibition of the single mutations A82T, H173S, F174Y, and A178P compared to wild-type rASIC1a. However, mutants S83T, Q84E, R175C, and E177G were slightly less sensitive to inhibition by Ma-3. These four residues can be spatially grouped with Ser83 and Gln84 located in the lower palm region that is involved in regulating channel desensitisation kinetics [19], and Arg175 and Glu177 in the upper palm region within the acidic pocket. Notably, S83T and Q84E showed incomplete inhibition at maximal concentrations of Ma-3 (~ 90% rather than full inhibition). Nevertheless, no individual mutation could fully explain the subtype-dependent potency and efficacy of Ma-3. Therefore, we sequentially combined mutations until we generated the quadruple mutant (rASIC1a S83T/Q84E/R175C/E177G, or SQRE) (Fig. 1d and e). With the incorporation of each additional mutation, the Ma-3 concentration-response curve progressively resembled that of rASIC1b, with the pharmacological effects of Ma-3 on the quadruple mutant also closely matching its effects on rASIC1b.

We also investigated whether differences in kinetics of Ma-3 activity and the functional state of channels during peptide application could contribute to the observed pharmacological difference of Ma-3 on rASIC1a and rASIC1b. We evaluated the rate of onset of Ma-3 inhibition by applying peptide at ~ 10 times the IC50 to channels at pH 7.45 for different durations. We then measured inhibition by activating channels at pH 6. Maximal inhibition was achieved within 45 s of Ma-3 application at both subtypes, and the time course of inhibition could be fit with a single exponential function to give t1/2 values of 6.5 s for rASIC1a and 3.8 s for rASIC1b (Supplementary Fig. 3a and b). Notably, complete inhibition of rASIC1b currents was not observed even after 120 s of exposure to Ma-3. We next assessed the recovery of currents from inhibition by Ma-3 at the same concentrations, using pH 7.45 during the conditioning period. This was comparable between the two subtypes with > 75% of current recovered after 60 s of washout of both channels (rASIC1a t1/2 = 21.6 s; rASIC1b t1/2 = 12.8 s; Supplementary Fig. 3a and b). This result agrees with the idea that variations in the activity profile of Ma-3 between rASIC1a and rASIC1b are not due to differences in the kinetics of inhibition or washout under the conditions used for collecting the data in Fig. 1.

Ma-3 is a gating modifier of ASIC1, and the different pharmacology between rASIC1a and rASIC1b has been tested using the same conditioning and stimulating pH for both channels. However, the pH-dependence of activation and steady-state desensitisation differs between rASIC1a and rASIC1b, with rASIC1b having a lower proton sensitivity by ~ 0.25 pH units (Supplementary Fig. 3c and d). The conformational state of ASICs depends on the pH, and can exist in any of the closed, desensitised, or open states, with the steady-state transitions studied via pH curves. We reevaluated the activity of Ma-3 at a single concentration for inhibition of rASIC1a and rASIC1b at different pH values. This is to compare the equivalent relative conditioning and stimulating pH for each subtype. Changing the conditioning pH from 7.45 to 7.75 did not significantly alter the degree of inhibition observed for rASIC1a (Supplementary Fig. 3e). For rASIC1b, a greater degree of current inhibition was only observed at a conditioning pH where the population of channels being studied started to undergo steady-state desensitisation (Supplementary Fig. 3f). Under the equivalent conditioning pH (pH50 SSD + 0.45 pH units) and activating rat ASIC1a and ASIC1b each with their respective half-maximal activating pH, we still observed a significant difference in the efficacy of inhibition between subtypes (Supplementary Fig. 3g). Therefore, we conclude that the observed pharmacological differences of Ma-3 between rASIC1a and rASIC1b is unlikely to be from the pH conditions used for testing alone. Instead, we propose a more complex mechanism where the predominant binding of Ma-3 to rASIC1a or rASIC1b occurs to a different channel conformation, and once bound the peptide stabilises a different conformation. This is consistent with our findings that residues on the channel in regions known to be important for proton gating underlying the subtype-dependent pharmacology of Ma-3.

To gain further insights into the mechanistic basis for the difference in Ma-3 activity at rASIC1a, rASIC1b, and the rASIC1a SQRE mutant, we analysed the rise and decay times of currents from these channels in the concentration-response data. While the rise and decay times of rASIC1a were not significantly affected by Ma-3, both rASIC1b and the SQRE mutant showed concentration-dependent decreases in rise and decay times (Fig. 2a–d). Where the peptide altered the current kinetics, this effect was reversible and returned to baseline levels with the same timescale as the recovery from inhibition of peak current amplitudes. To further examine differences in how Ma-3 modulates these channels, we determined the pH of activation and steady-state desensitisation (SSD) curves in the absence and presence of 100 nM Ma-3 (Fig. 2e–g; Table 1). At rASIC1a, Ma-3 inhibited the channel by shifting the activation curve by 0.48 pH units towards more acidic values and causing a slight alkaline shift of 0.08 pH units in the SSD curve. In contrast, at rASIC1b, there was no shift in the activation curve, but the SSD curve shifted by 0.06 pH units towards alkaline values. The rASIC1a SQRE mutant exhibited intrinsic pH-gating properties and Ma-3 modulation that were intermediate between the two wild-type channels. Specifically, the control pH50 of activation in the SQRE mutant was more similar to that of rASIC1a, while the control pH50 of SSD was closer to that of rASIC1b (Table 1). Additionally, 100 nM Ma-3 shifted the activation curve in the SQRE mutant by 0.11 pH units towards acidic values, while the pH50 of the SSD curve remained unchanged. While the SQRE mutations appear to bestow the rASIC1b Ma-3 concentration-response phenotype (see Fig. 1), their effects on gating kinetics and pH-properties present a more nuanced profile that combines features of both receptor subtypes. The data on the mechanism of action underscore the allosteric nature of mambalgins and demonstrate how the inherent gating properties of different ASIC variants significantly influence how they are modulated by the peptide.

Table 1 Effect of Ma-3 on the pH-dependence of rASIC1a, rASIC1b, and rASIC1a SQRE mutant

We further investigated the mechanism of ASIC modulation by Ma-3 using two rASIC1a: rASIC1b chimeras, in which increasing segments of rASIC1b were introduced to replace corresponding rASIC1a sequences (C92 and C166, where the number represents the amino acids replaced from the rASIC1a N-terminus; Fig. 3a and b) [20, 21]. Beyond residue 186, both rASIC1a and rASIC1b share the same sequence. The C92 chimera includes the S83T and Q84R substitutions, and 300 nM Ma-3 only partially inhibited this channel, similar to its activity at rASIC1b. Unexpectedly, Ma-3 inhibited a greater proportion of current from the C166 chimera, resembling the activity observed with rASIC1a. Although C166 retains only 19 amino acids unique to rASIC1a, this region includes the palm loop with residues Arg175 and Glu177 from rASIC1a. This data could be interpreted to suggest that the sequence between positions 92 and 166 contains residues that are important in determining mambalgin interactions with the channel. However, we hypothesise that altering this region may change how proton-induced conformational changes during gating are conveyed through the channel, and mambalgins alter these properties in an unpredictable manner. Similar unexpected findings were observed for Ma-2 modulation of ASIC1a: ASIC2a chimeras that exchanged parts of the thumb and palm domains [22]. The unexpected results from the chimera data, together with Ma-3 primarily affecting ASIC1a activation gating, prompted us to compare the thumb-palm domain interface in the apo and Ma-1 bound hASIC1a cryo-EM structures (Fig. 3c). In the apo hASIC1a structure, Asp355 on the thumb domain is within 6Å of the four core ASIC pharmacophore residues, and over 15Å away from Arg175 (distance between side chains). In contrast, in the Ma-1 bound complex, Asp355 shifts to within 4Å of Arg175, possibly forming electrostatic and hydrogen bond interactions that stabilise a thumb: palm domain interaction and prevent further conformational rearrangements associated with activation gating. Although Asp355 does not directly interact with Ma-1 in the bound structure, the Ma-3 concentration-response curve for rASIC1a D355A reveals reduced potency compared to rASIC1a and decreased efficacy at higher concentrations (Fig. 3d). This is consistent with findings from subtype substitutions and chimeras, where domain interface mutations disrupt mambalgin’s ability to inhibit conformational rearrangements during activation gating. Lastly, focusing on the four residues that were crucial for imparting differential Ma-3 potency between rASIC1a and rASIC1b, we generated reverse mutations in the rASIC1b background (Fig. 3e). The rASIC1b T128S/E129Q (reversal of rASIC1a S83T/Q84E) and rASIC1b C208R/G210E (reversal of rASIC1a R175C/E177G) mutants had IC50 values similar to rASIC1b but exhibited more current inhibition at saturating concentrations. We then combined the four substitutions to create the reversal of the rASIC1a SQRE mutant. The rASIC1b quadruple mutant showed an IC50 of 18.5 nM, intermediate between rASIC1a and rASIC1b, with Ma-3 completely inhibiting currents at saturating concentrations, similar to its effect at rASIC1a.

Differences in the Ma-3 pharmacophore between rat ASIC1a and rASIC1b

Having confirmed that Ma-3 has a different mechanism of action on rASIC1a and rASIC1b and identifying the channel residues underlying this effect, we next investigated if there was any difference in the Ma-3 pharmacophore for rat ASIC1a and ASIC1b by testing a panel of 20 Ma-3 variants at both subtypes (Fig. 4a). Mambalgins are members of the three-finger toxin family, consisting of three finger loops protruding from a central core (Fig. 4a an b). To test our Ma-3 variants, peptides were applied at pH 7.45 and channels were stimulated at pH 6 (Fig. 4, Supplementary Figs. 1 and 4 and Supplementary Tables 4 and 5). Consistent with previously published reports on activity at rASIC1a [13, 14], we observed an increase in IC50 values of greater than 10-fold for Ma-3 H6A in finger I, as well as F27A, R28A, L32A, I33A, and L34A in finger II. Residues in finger III (Ser39, Ser40, Lys48 and Arg54) that are spatially distant from the core pharmacophore of Ma-3 are not important for activity at rASIC1a. Next, we evaluated the activity of our Ma-3 mutants at rASIC1b, where the pharmacophore has not been previously analysed. All alanine mutants that lost activity at rASIC1a also lost activity at rASIC1b, indicating a largely conserved pharmacophore between the two subtypes. However, we observed a significant increase in pIC50 at rASIC1b (estimated at > 30-fold loss in potency) for K8A and M16A, in contrast to the modest effect these mutations had on the potency of Ma-3 at rASIC1a and hASIC1a (~ 4-fold for K8A, and no change for M16A) (Fig. 4d–f). Notably, these two residues are located outside the core Ma-3 pharmacophore that is shared between rASIC1a and rASIC1b and are spatially very distant from each other (Fig. 4d). This also positions Lys8 and Met16 relatively distant from channel residues that have been identified as the core binding interactions for potency at ASIC1a (Fig. 4g) [15, 16, 22].

At 300 nM, WT, K8A and M16A Ma-3 strongly inhibit rASIC1a currents. All three peptides showed similar levels of current inhibition (although significantly different for K8A; I/Icontrol: WT = 0.48, K8A = 0.65, M16A = 0.56), as wild-type Ma-3 at the rASIC1a SQRE mutant, in contrast to largely abolishing Ma-3’s inhibitory activity as they do at rASIC1b (I/Icontrol: WT = 0.40, K8A = 0.91, M16A = 0.80) (Fig. 4h). We then tested these peptides for activity at three rASIC1b mutants (Fig. 4i). At rASIC1b T128S/E129Q, and rASIC1b C208R/G210E, the K8A and M16A mutants were significantly less potent than wild-type Ma-3, akin to the results seen when testing the peptide mutants against wild-type rASIC1b. However, when assaying the peptides at the rASIC1b combined quadruple mutant (TECG), there was minimal difference in inhibition levels between Ma-3 WT, K8A, and M16A. Therefore, the TECG mutant behaved more like rASIC1a, where alanine mutations of Ma-3 at positions Lys8 and Met16 have little effect on peptide potency. Interestingly, the rASIC1a SQRE mutation could effectively swap the lower potency and efficacy of WT Ma-3 at rASIC1b into rASIC1a, with full recovery of efficacy but only partial recovery of potency with the reciprocal mutations in rASIC1b TECG. In contrast, these quadruple channel mutations could only swap the striking subtype dependent effects of Ma-3 K8A and M16A peptide mutations when introduced into the rASIC1b background (TECG mutant), but not reciprocally when the rASIC1b residues were introduced into rASIC1a (SQRE mutant). Figure 2 shows that Ma-3 has very different effects on rASIC1a gating compared to rASIC1b. The lack of correlation between findings in the reciprocal quadruple mutants with Ma-3 WT, K8A, and M16A further supports the idea that mambalgins bind to these two subtype variants in a subtly different orientation and stabilise distinct non-conducting states to achieve inhibition.

Binding interactions in the lower thumb domain determine the rat to human ASIC1a potency difference for Ma-3

Ma-3 inhibits both rat and human ASIC1a via shifting the pH-dependence of activation curve to more acidic values (Fig. 5a and b; Supplementary Table 7). Furthermore, these species variants are 98% identical in sequence [8, 12, 15], yet rat and human ASIC1a exhibit a difference in IC50 values of approximately 6-fold in our experiments (rASIC1a IC50 3.9 nM and hASIC1a IC50 25.0 nM, statistics in Supplementary Table 8). This is not due to differences in the time-course of Ma-3 induced inhibition, or recovery of current from inhibition, which are both comparable between hASIC1a and rASIC1a (Supplementary Fig. 5). Comparison of the amino acid sequences around the Ma-3 binding site reveals two areas that may contribute to the difference in potency (Fig. 5c). To investigate this, we introduced mutations in rASIC1a corresponding to the residues in human ASIC1a at these positions: N291K, D298, and L299 (Fig. 5d). Human ASIC1a contains an Asp298 and Leu299 insertion relative to rASIC1a. Inserting this DL sequence into the rASIC1a background (mutant named 298-DL) had no effect on Ma-3 potency compared to wild-type rASIC1a. However, we found that with the introduction of a positive charge in rASIC1a N291K, the potency of Ma-3 shifted to overlap with that of hASIC1a. In contrast, the alanine mutant at this position, rASIC1a N291A, had no effect on potency (Fig. 5e). To further validate the importance of this channel region, we mutated the spatially neighbouring Glu362 to alanine or arginine to investigate the role of charge at this position (Fig. 5c and e). Ma-3 was equipotent on wild-type and rASIC1a E362A but introducing a positive charge in rASIC1a E362R resulted in a ~ 5-fold decrease in Ma-3 activity. Lastly, we produced the reverse position 291 mutation in the hASIC1a background (hASIC1a K291N), which was inhibited by Ma-3 with a potency comparable to that at wild-type rASIC1a (Fig. 5f). These results suggest that the presence of either a positive charge or additional bulk in the region surrounding Asn291 and Glu362 disrupts Ma-3 activity at ASIC1a through a repulsive and/or steric clash effect. This observation shows that the lower thumb domain contributes to the interaction between ASICs and mambalgins, and the substitution between species variants at position 291 determines rat to human ASIC1a potency of Ma-3. With publication of the Ma-1:hASIC1a cryo-EM structure, it was suggested Lys8 from mambalgin can form a salt bridge interaction with Asp300 of hASIC1a [15], however we find this explanation to be unlikely from our results detailed above.

Combined binding and allosteric substitutions determine the rat to human ASIC1b difference in Ma-3 activity

The activity and mechanism of action of mambalgins at hASIC1b is different to rat and human ASIC1a and rASIC1b. First, we confirmed that Ma-3 inhibits pH 5 and 6 evoked rASIC1b currents equally and we observed no significant shift in the activation curve with 300 nM peptide application (Fig. 6a–c). In contrast, at hASIC1b, Ma-3 inhibits pH 5 evoked currents and potentiates pH 6 evoked currents, an effect underpinned by an alkaline shift in the pH-dependence of activation (Fig. 6c). To explore the molecular basis for this distinct pharmacology, we examined the effects of mutating the only two non-conserved residues between rat and human ASIC1b in and around the mambalgin binding site: C208Q (equivalent to Arg175 in rASIC1a) and N324K (equivalent to Asn291 in rASIC1a) (Fig. 6a). For rASIC1b C208Q, Ma-3 inhibited pH 5 evoked currents similarly to wild-type rASIC1b, but potentiated pH 6 evoked currents comparable to its effect on hASIC1b, with an alkaline shift in the pH-dependence of activation (Fig. 6d). These effects mostly resemble the effect of Ma-3 on hASIC1b, but we see a greater alkaline shift in the activation curve with the C208Q mutant. Strikingly, this single mutation causes an acidic shift in the activation curve without any peptide present (pH50 act: rASIC1b = 6.15, rASIC1b C208Q = 5.84, and hASIC1b = 5.91), further highlighting the importance of this region in activation gating properties. The effect of Ma-3 on rASIC1b N324K was very similar to that on rASIC1b but significantly less potent (Fig. 6e). This is similar to the loss of potency observed in the equivalent mutation when looking at the differences in Ma-3 activity at rat and human ASIC1a (N291K substitution), suggesting that this is also a binding site for Ma-3 in the ASIC1b background. Notably, this loss in potency resulted in Ma-3 inhibiting pH 5 currents with a similar IC50 as at hASIC1b. Together this is reflected in the pH-activation data showing no shift in pH50 but less inhibition at all activating pH values compared to wild-type rASIC1b, indicating that Asn324 is likely not important in the pH-sensing properties of the channel. Finally, we combined both mutations to make rASIC1b C208Q/N324K (Fig. 6f), resulting in concentration-response curves of Ma-3 with pH 5 and 6 stimuli, as well as a peptide induced shift in the pH-dependence of activation, comparable to wild-type hASIC1b. In summary, these two mutations significantly alter the response of the channel to Ma-3, with the C208Q mutation primarily responsible for causing an alkaline shift in the activation curve, and the N324K mutation reducing the potency of Ma-3. Thus, the unique pharmacology of Ma-3 at hASIC1b is determined by the combined effect of these two substitutions.

The Ma-3 inhibitory pharmacophore at hASIC1b resembles rASIC1a more than rASIC1b but the potentiating pharmacophore appears to contain less residues

We tested our panel of Ma-3 mutants on both pH 5 and pH 6 evoked currents of hASIC1b using a single concentration of 1 µM, which results in maximal inhibition of pH 5-induced currents and maximal potentiation of pH 6-induced currents (Fig. 7a). Like rASIC1a and rASIC1b, Ma-3 mutants H6A, F27A, R28A, L32A, and L34A lost activity at both pH 5 and pH 6 stimulated currents (Fig. 7b). Interestingly, L30A, K31A, and I33A mutants at 1 µM lost inhibitory activity for pH 5 currents, but still potentiated pH 6 currents like wild-type Ma-3. However, it is important to note that there is a ~ 10-fold difference in potency for wild-type Ma-3 between these pH stimulus conditions, and the activity/potency is highly dependent on the stimulating pH used. Notably, M16A and K8A had no effect on activity for both pH 5 and 6 activity, making this part of the pharmacophore more similar in profile to rat and human ASIC1a than rASIC1b (where these mutations substantially decreased activity; see Fig. 4e). In conclusion, our results suggest that the pharmacophore of Ma-3 at hASIC1b appears to be more like rASIC1a than rASIC1b, despite the higher similarity between ASIC1b species variants than the ASIC1a and ASIC1b subtypes. Interestingly, the more potent effect of potentiating pH 6-induced currents seems to result from a smaller number of interacting residues.

Fig. 7

The Ma-3 pharmacophore at hASIC1b. (a) Concentration-response curve of Ma-3 at hASIC1b activated by pH 6 and pH 5 (n = 5–6). (b) Ma-3 and mutants (1 µM) tested for activity at hASIC1b activated by pH 6 (top) and pH 5 (bottom). Violet indicates statistically significant differences from Ma-3 (Welch’s one-way ANOVA with Dunnett’s multiple comparisons test). Data are n ≥ 5 with all points shown to give exact replicates. All data use a conditioning pH of 7.45, and are mean ± SEM. See Supplementary Tables 12 and 13 for details of inhibition and statistical comparisons of panel b data