Mapping the interaction surface of scorpion β-toxins with an insect sodium 1 channel 2

. The interaction of insect-selective scorpion depressant β-toxins (LqhIT2 and Lqh- dprIT 3 from Leiurus quinquestriatus hebraeus ) with the Blattella germanica sodium channel, 28 BgNa v 1-1a, was investigated using site-directed mutagenesis, electrophysiological analyses, and 29 structural modeling. Focusing on the pharmacologically-defined binding site-4 of scorpion β-toxins 30 at the voltage-sensing domain II (VSD-II), we found that charge neutralization of D802 in VSD-II 31 greatly enhanced the channel sensitivity to Lqh-dprIT 3 . This was consistent with the high sensitivity 32 of the splice variant BgNa v 2-1, bearing G802, to Lqh-dprIT 3 , and low sensitivity of BgNa v 2-1 33 mutant, G802D, to the toxin. Further mutational and electrophysiological analyses revealed that the 34 sensitivity of the WT = D802E < D802G < D802A < D802K channel mutants to Lqh-dprIT 3 35 correlated with the depolarizing shifts of activation in toxin-free channels. However, the sensitivity 36 of single mutants involving IIS4 basic residues (K4E = WT << R1E < R2E < R3E) or double 37 mutants (D802K = K4E/D802K = R3E/D802K > R2E/D802K > R1E/D802K > WT) did not 38 correlate with the activation shifts. Using the cryo-EM structure of the Periplaneta americana 39 channel, Na v PaS, as template and the crystal structure of LqhIT2, we constructed structural models 40 of LqhIT2 and Lqh-dprIT 3 -c in complex with BgNa v 1-1a. These models along with the mutational 41 analysis suggest that depressant toxins approach the salt-bridge between R1 and D802 at VSD-II to 42 form contacts with linkers IIS1-S2, IIS3-S4, IIIP5-P1 new mutational and electrophysiological experiments from our study and also 749 another recent study (Zhu et al. 2020). Future studies involving mutational and electrophysiological analyses of multiple toxin-channel contacts, or cryo-EM structural analysis of toxin-channel complexes are 751 needed to further validate our models.


5
where I is the peak sodium current, Imax is the maximal current evoked, V is the potential of the 138 voltage prepulse, V1/2 is the half maximal voltage for inactivation, and k is the slope factor. 139 To detect the effect of scorpion β-toxins, a train of conditional prepulses is required [34]. 140 Therefore, we applied a 20 Hz train of fifty 5 ms depolarizing prepulses to 50 mV followed by a 141 20 ms depolarizing test pulse between -80 and -65 mV from a holding potential of -120 mV (Fig.   142 1B). The channel modification by Lqh-dprIT3-c was determined as the percentage of channels with 143 the voltage dependence of activation shifted to negative membrane potentials, which was derived 144 from double Boltzmann fits of the conductance-voltage relationships. Data are presented as means 145 ± S.D. Statistical significance was determined by one-way analysis of variance (p < 0.05).  All calculations were performed with the ZMM program [36], which minimizes energy in the 156 space of generalized (internal) coordinates [37]. The models were visualized using the PyMol 157 Molecular Graphics System, version 0.99rc6 (Schrödinger, New York, NY). 158 Nonbonded interactions were calculated with the AMBER force field [38]. Electrostatic 159 interactions were calculated with the distance-and environment-dependent dielectric function 160 [36]. No specific energy terms were used for cation-π interactions, which were accounted for with 161 partial negative charges at the aromatic carbons [39]. No distance cutoff was used to calculate 162 electrostatic interactions involving ionized groups in amino acids. For other interactions the 163 distance cutoff of 9 Å and a shifting function [40] were used. 164 To incorporate the experimental data in our calculations we used constraints. A distance 165 constraint is a flat-bottom energy function that allows respective atom-atom distance to deviate the 166 energy-free between lower and higher limits and imposes a parabolic energy penalty if the distance 167 is beyond the limits. To maintain the template folding in the channel model we used "pin" 168 constraints. A pin allows an alpha carbon of an amino acid residue to deviate up to 1 Å from the 169 respective template position without a penalty and imposes a parabolic energy function to penalize 170 larger deviations. Pins were not used to preserve the mobile toxin fold stabilized by three disulfide 171 bridges. To bias some toxin-channel contacts we used residue-residue constraints. Such a 172 constraint specifies two residues and the target distance between their side-chains (which was set 173 to 5 Å). In the beginning of each cycle of energy minimization, the ZMM program selects the 174 closest pair of atoms between the two side-chains and applies a distance constraint to these atoms 175 so that the atom-atom contacts may switch during the MCM trajectory [41]. For all constraints, 176 the energy penalty was calculated with the force constant of 10 kcal•mol -1 •Å -2 . To minimize 177 possible bias due to the imposed constraints, each model was optimized in two stages. After the 178 first MC-minimization with constraints, the model was refined by a second MC-minimization, in 179 which all constraints were removed and all degrees of freedom, including backbone torsions, were 180 allowed to vary.

181
The BgNav1-1a-toxin complex was used as the starting structure for construction of the 182 models of toxin-channel mutants, which included VSM-II and PD-III (both elements interact 6 directly with the toxin). Other parts remote from the toxin in the channel mutants were not 184 included. It should be noted that MC-minimization of each toxin-channel complex yielded several 185 low-energy structures, which had comparable energy characteristics. Respective figures show low-186 energy structures in which the maximal number of specific contacts is established between the 187 toxin and the channel. 188 The in-silico deactivation of II-VSM was computed as follows. C α atoms of IIS4 basic 189 residues were forced to progressively move through 21 sets of planes (5 planes for five C α atoms 190 of basic residues in each set), which are normal the pore axis. Two adjacent planes are distant from 191 one another by 0.5 Å. The starting conformation for the II-VSM deactivation transit corresponded 192 to the BgNav1-1a model with activated VSMs, which is based on the NavPaS cryo-EM structure. 193 At each step of the transit, the C α atoms of basic residues were free to move within the   Table   209 S1). The two components likely represent unmodified and toxin-modified channels, respectively.

223
At 300 nM, Lqh-dprIT3-c had almost no effect on a V0.5 of the WT channel or mutant D802E 224 (Fig. 2C,D), but rendered hyperpolarizing shifts of a V0.5 in mutants D802G and D802A (Fig. 2E,F) 225 and especially for D802K, where a V0.5 was shifted by -57 mV (Fig. 2G). The toxin-induced shifts 226 of a V0.5 correlated with the enhanced sensitivity of the channels to Lqh-dprIT3-c (Fig. 2H) properties and sensitivity to Lqh-dprIT3-c. In toxin-free channels, R1E and R2E rendered a -5 mV 234 shift in a V0.5, whereas R3E and K5E rendered a shift of +8 and +4 mV in a V0.5, respectively (Fig. 235 3 and Table S2). We then examined the sensitivity of the mutants to 300 nM Lqh-dprIT3-c, the 236 concentration that had no effect on the WT channel BgNav1-1a. In channel mutants R1E, R2E and       Table 1). The side-chain of W38 is located at the apex of the loop between two beta-strands, 315 analogously to F44 in Css4 (Figs. S3, S4). Besides W38, the side-chain of another functionally 316 important residue, N18 (Table 1), is located at the same basal side of the toxin (Fig. 5A). The 317 surface opposite to the basal side is exposed to the solvent and lacks functionally important 318 residues (Fig. 5B), and so most likely it does not interact with the channel. Two regions between 9 the basal and solvent-exposed sides bear several functionally important residues: W36, I16, K23 320 and E24 (Table 1 and Fig. 5 C,D). 321 We placed LqhIT2 above II-VSM and next to IIIS5-P1 with the basal side down, and imposed 322 a residue-residue distance constraint to force W38 of the toxin into the channel gorge to approach 323 R1. Then we sampled multiple orientations of the toxin along axis C α _W38 ---C α _A49 and MC-324 minimized each sampled structure. In these computations, the C α atoms of the channel were pinned 325 (see section S1 in the Supplemental data), the toxin backbone was rigid, and side-chains of both 326 the channel and toxin were flexible. Among several predicted structures we selected the one where 327 functionally important toxin residues established specific contacts (salt-bridges, cation-π, 328 hydrophobic, or knob-into-hole) with channel residues, whose homologs are presumably involved 329 in binding of Css4 to rNav1.2 or/and are in direct contact in the cryo-EM structure of the NavPaS-330 Dc1a complex (Table 1). Moreover, in the selected structure, a few polar residues of the toxin 331 approach polar residues of the channel, but do not establish specific contacts, suggesting potential

349
The toxin-channel complex is stabilized by several H-bonds that involve functionally in agreement with the fact that substitution N58D strongly diminished the toxin binding (Table 1) 353 likely due to electrostatic repulsion between the two acidic residues, N58D and D1443. E24 forms  (Table 1).

360
Two cation-π contacts are seen in the toxin-channel structural model. W38 penetrates most 361 deeply into the II-VSM gorge, forms hydrophobic contacts with a number of channel residues and 362 a weak π-cation contact with R1, which is salt-bridged with D802 (Fig. 7E). Another functionally 363 important tryptophan, W36, forms a π-cation contact with K808 (Fig. 7E).

In-silico deactivation of II-VSM bound to Lqh-dprlT 3 -c.
There is no structure available

411
In the 3D-aligned crystal structures of NavAb with VSMs in the resting (PDB ID: 6P6W) and 412 activated (PDB IDs: 6p6x, 6p6y) states, the z-coordinates of C α _R1 differ by ~10 and ~11 Å, 413 respectively. In the 3D-aligned cryo-EM structures of the Nav1.7-NavPaS channel with IV-VSM 414 in the resting (PDB ID: 6nt4) and activated (PDB ID: 6nt3) states, the z-coordinates of C α _IVR1 415 differ by ~10.5 Å. However, employing the resting-VSM structures of NavAb or Nav1.7-NavPaS 416 as templates to model the resting II-VSM in BgNav1-1a would be inappropriate due to two major 417 reasons. Firstly, the mutual disposition of PD and VSMs in different P-loop channels varies. 418 Secondly, such models would not suggest intermediary structures between the resting and 419 activated II-VSM. 420 To overcome these problems, we in-silico deactivated II-VSM in the Lqh-dprlT3-c-BgNav1- and S870 that in the activated II-VSM make strong contacts with the toxin (Fig. S5C). In particular, 436 I16, the functionally most important residue in LqhIT2 (Table 1) 456 In the model of Lqh-dprIT3-c bound to WT BgNav1-1a, D802 and R1 form a salt-bridge (Fig. 9A).

457
R2 forms a salt bridge with E860 in IIS3 and approaches R1. Due to the electrostatic repulsion 458 between the positive charges of R1 and R2, the D802:R1 salt bridge is fortified. This prevents R1 459 from changing its conformation to form a salt-bridge with the nearby E864 in IIS3. The close 460 proximity between R1 and R2 seems electrostatically unfavorable, yet analogous orientations of  In channel model D802A, R1 interacts with E864 and E860 and repels R2 that forms a salt-478 bridge with E864 (Fig. 9B). As R1 and R2 no longer limit W38 from entering deeper into the II-479 VSM gorge, this tryptophan now establishes a strong π-cation contact with K815 at IIS1, donates 480 an H-bond to D811 at IIS1, and retains van der Waals contacts with M806 at IIS1-S2 (Fig. 9B).  Figure 2H).

488
In channel mutants D802A and D802K (Fig. 9 C,D), most of the toxin-channel interactions are 489 similar, but an additional cation-π interaction is formed between K802 of channel mutant D802K 490 and W38 of Lqh-dprIT3-c. This interaction provides a unique cation-π-cation 'sandwich' involving 491 W38 and two lysines of the channel (Fig. 9C).

492
In channel mutant R1E the glutamate substitution did not establish any salt bridge, yet it formed 493 H-bonds with S870 at IIS3-S4 and Q1419 at IIIS5 (Fig. 9D). The R1E substitution released D802 494 from its salt-bridge with R1 so that D802 formed an H-bond with W38 of Lqh-dprIT3-c, which in 495 turn forms a π-cation contact with K815. Still, W38 makes fewer contacts with II-VSM than in the 496 model of Lqh-dprIT3-c-D802K complex. This may explain the lower sensitivity of the R1E 497 channel mutant to the toxin compared to channel mutant D802K (Fig. 4F).

498
In channel mutant R2E, the glutamate substitution does not restrict the conformational 499 mobility of R1 as in the WT channel (Fig. 10A). The side-chain of R1 turns towards IIS3 to form 500 salt bridges with E860 and E864 and π-stack with W38 of Lqh-dprIT3-c. Such arginine-tryptophan 13 π-stacking pairs are seen in crystal structures of some proteins, where they contribute to the 502 structure stabilization [53, 54]. Furthermore, W38 donates an H-bond to D802. The above 503 interactions of W38 with R1 and D811 can explain why channel mutant R2E is much more 504 sensitive to the toxin than the WT channel (Fig. 4F). 505 Among the most paradoxical observations in the present study is the high sensitivity of channel 506 mutant R3E to the toxin (Fig. 4F). The glutamate substitution is very far from the toxin, from D802 507 and from R1 (Fig. 10B). In terms of toxin-channel contacts, channel mutants R2E and R3E are 508 rather similar (cf. Figs. 10A and 10B). It seems in the R2E channel model that the conformation 509 of R1 is firmly restrained by the salt-bridges with E864 and E860 as well as by an electrostatic 510 attraction to R2E, whereas in the R3E channel model arginine R2 electrostatically repels R1, likely 511 increasing its conformational flexibility. This may facilitate the π-stacking interactions of R1 with 512 W38 of Lqh-dprIT3-c.

513
Many toxin contacts in channel mutant D802K/R1E (Fig. 10C) are similar to those in channel 514 mutant D802K. However, the cation-π-cation interactions involving W38 of Lqh-dprIT3-c in 515 channel mutant D802K/R1E are weaker than in channel mutant D802K because K802 is attracted 516 by R1E, whereas K815 is attracted by E864 and Y812. This may explain why double-mutant 517 channel D802K/R1E is less sensitive to the toxin than the single-mutant channel D802K (Fig. 4F).

518
In channel mutant D802K/R2E, the side-chain of R1 turns in the cytoplasmic direction and 519 towards IIS3 to form salt bridges with R2E and E864 (Fig. 10D) and electrostatic interaction with

528
The interaction of scorpion β-toxins with their target voltage-gated Na-channels has been 529 studied for more than two decades [10, 12, 15, 16, 28, 34, 42, 43, 46, 55-59] and yet, despite the 530 elucidation of their pharmacological mode of action, bioactive surfaces and putative binding site 531 at the channels, molecular details of the toxin-channel interactions have not been fully described.

532
To this end we combined the data accumulated on toxin binding, electrophysiological effects, and 533 structure with mutational analysis and structural modeling of an insect sodium channel with the 534 objective to better understand at the atomic level the mechanism by which β-toxins obstruct the  state, shifting a V0.5 by +50 mV [62]. In the bacterial sodium channel NaChBac, neutralization of 550 E43 at the extracellular end of S1 also resulted in a large (+47 mV) shift of a V0.5 [51]. Thus, our 551 conclusion that the activated state of II-VSM of the BgNav1-1a channel is stabilized by the salt 552 bridge between D802 and R1 (Fig. 6 C,D) is consistent with the experimental data available for a 553 variety of voltage-gated ion channels.

554
When the voltage-sensing helix IIS4 is in the activated, outward position, D802 is the only 555 acidic residue that forms a close contact with R1 (Fig. 9A). Therefore, contribution of the salt 556 bridge to stability of the II-VSM activated conformation is particularly important. Such a salt 557 bridge is likely formed also between R1 and D802E as seen in the cryo-EM structure of the brain  (Table S2) suggests that the inactivation process starts when II-VSM is 569 activated and IIS4 still has not returned to its resting position. of Lqh-dprIT3-c into the II-VSM gorge and formation of an H-bond with D811 as well as a π-575 cation contact with K815 (Fig. 9B). These two contacts, which are absent in the WT channel, may 576 explain why mutant D802A is more sensitive to Lqh-dprIT3-c than the WT channel. Similar 577 contacts are likely formed between the toxin and channel mutant D802G. Moreover, channel 578 mutant D802G is more sensitive to Lqh-dprIT3-c than channel mutant D802A (Fig. 2H) probably 579 because the flexible loop IIS1-S2 adjusts more easily to the toxin surface.

580
In channel mutant D802K, additional cation-π contacts are formed between K802 and W38 of 581 the toxin resulting in a cation-π-cation 'sandwich' (Fig. 9C). Analogous interactions of tryptophan 582 with two basic residues (Arg-Trp-Arg) appear, for example, in a ubiquitin variant (PDB ID: 583 5TOG). The stronger attraction of the toxin W38 to II-VSM may explain why channel mutant 584 D802K is even more sensitive to Lqh-dprIT3-c than channel mutant D802A (Fig. 2H). Overall, the 585 sensitivity of the channels to the toxin (WT = D802E < D802A < D802G < D802K) qualitatively 586 correlates with the number of favorable toxin contacts with the activated II-VSM (Fig. 2H), and it 587 also qualitatively correlates with the negative shifts of a V0.5 in toxin-bound vs. toxin-free channels 588 (Fig. 2C-G). The structural models of the WT and D802A/K channel mutants (Fig. 9A-C) provide  Substitutions R1E, R2E and R3E dramatically increased the channel sensitivity to Lqh-dprIT3-627 c (Fig. 4F). The structural models of channel mutants R1E, R2E and R3E with bound Lqh-dprIT3-628 c ( Fig. 9D and 10A,B) predict stronger toxin-channel interactions due to elimination of the 629 D802:R1 salt bridge, leading to increased number of specific toxin-channel contacts. The 630 sensitivity of channel mutant K4E to the toxin is comparable with that of the WT channel (Fig.   631 4F) likely because K4E is far from salt bridge D802:R1 and from R2, which stabilizes the salt 632 bridge (Fig. 9A). Channel mutants R2E, R3E and K4E are as sensitive to the toxin as the 633 background channel mutant D802K (Fig. 4F). The reason is that in the D802K channel mutant, 634 W38 of the toxin is engaged in several specific contacts with II-VSM (Fig. 9C), and so charge 635 reversal of the basic residues in IIS4 has a small impact on these contacts as seen in the structural 636 model of the D802K/R2E channel mutant (Fig. 10C). dprIT3-c (Fig. 1A,C) are located in the loop between two beta-strands that penetrate deeply into 648 the gorge of II-VSM (Fig. 4E). F48 between two beta-strands in Dc1a also penetrates deeply into 649 the II-VSM gorge of NavPaS; (ii) Most of the ionizable residues in Css4, LqhIT2 and Dc1a are  interacts with the toxin cavity formed by T3, K11 and W53. In our LqhIT2-BgNav1-1a model, 687 Q867 (homolog of Q604 in NavPaS) accepts an H-bond from K23, whereas K11 and W53 form 688 specific contacts, respectively, with D1443 and N1445 in loop IIS5-P1 (Fig. 7D). Furthermore, in 689 the LqhIT2-NavPaS model, N58 interacts with G602 of the channel [26], whereas in our model 690 N58 donates an H-bond to D1443 (Fig. 7D). VSM. Instead, L872 forms a strong hydrophobic contact with L856 in the resting (Fig. S7B), but 696 not in the activated state of II-VSM (Fig. S7A). Analogs of the two leucine residues in BgNav1-1a 697 are phenylalanine residues, F609 and F593, in NavPaS (Fig. 1D), which likely form a stacking

722
The strong activity of the scorpion β-toxin Lqh-dprIT3-c on the cockroach splice variant  Therefore, this study focused on the structural entity surrounding residue 802, its putative 729 interactions with the positively-charged IIS4 residues and the channel interactions with Lqh-dprIT3-c. Structural models of the channel in complex with the depressant scorpion β-toxin LqhIT2 731 or its super-active homolog Lqh-dprIT3-c were generated using the crystal structure of LqhIT2 and 732 the cryo-EM structure of NavPaS as template. The reliability of these models is supported by the 733 results from previous binding studies using toxin mutants [16] as well as electrophysiological 734 assays using BgNav1-1a mutants (present study). In our model, all functionally important LqhIT2 735 residues [16] make specific contacts with channel residues in the II-VSM extracellular loops IIS1-736 S2 and IIS3-S4, as well as with the III-PD extracellular loops IIIS5-P1 and IIIP2-S6. The models 737 suggested that the low sensitivity of BgNav1-1a to the toxin is due to the salt-bridge between D802 738 and R1. All substitutions that eliminated the salt bridge enabled formation of additional toxin-739 channel contacts that increased the channel sensitivity to Lqh-dprIT3-c. BgNav1-1a residues, which 740 interact with the toxins, are homologous to residues in the mammalian brain channel rNav1.2, that -775 a Shown are specific contacts (H-bonds, salt bridges, hydrophobic, cation-π) of the toxin with the 776 channel side-chains within 4 Å from the toxin residue. b Lqh-dprIT3-c residues identical to 777 sequentially matching residues of LqhIT2 (Fig. 1A)

809
Residues whose substitution reduced toxin binding to the Periplaneta americana channel over 810 100-fold (Table 1) are colored. Names of the channel segments proximal to the toxin are 811 shown. A, The basal side of the toxin is exposed to the II-VSM gorge. B, Solvent-exposed 812 side of the toxin opposite to the basal side. C, The lower part of the toxin side is exposed to 813 linker IIS1-S2. D, The lower part of the toxin side, which is exposed to linker IIS3-S4, bears 814 several functionally important residues (Table 1).   (Table 1) is engaged in π-cation contacts 823 with K808 in IIS2. D802 is salt-bridged with R1, which forms a weak cation-π contact with 824 W38. F and G, Side and extracellular views of ionizable residues whose substitutions had 825 only a minute impact on binding of LqhIT2 to the Periplaneta Americana membrane 826 preparations [16]. The residues exposed to the extracellular space, do not interact with the 827 toxin, but enable its hydration. Note that these residues are most different between toxins 828 LqhIT2 and Lqh-dprIT3-c (Table 1 and Fig. S3).         IIS3  IIS4  IIS1  IIS2   D802A   D802K  R1E   IIS3  IIS1  IIS4  IIS2  IIS3  IIS1 IIS4 IIS2