The Receptor Site and Mechanism of Action of Sodium Channel Blocker Insecticides*

Sodium channels are excellent targets of both natural and synthetic insecticides with high insect selectivity. Indoxacarb, its active metabolite DCJW, and metaflumizone (MFZ) belong to a relatively new class of sodium channel blocker insecticides (SCBIs) with a mode of action distinct from all other sodium channel-targeting insecticides, including pyrethroids. Electroneutral SCBIs preferably bind to and trap sodium channels in the inactivated state, a mechanism similar to that of cationic local anesthetics. Previous studies identified several SCBI-sensing residues that face the inner pore of sodium channels. However, the receptor site of SCBIs, their atomic mechanisms, and the cause of selective toxicity of MFZ remain elusive. Here, we have built a homology model of the open-state cockroach sodium channel BgNav1-1a. Our computations predicted that SCBIs bind in the inner pore, interact with a sodium ion at the focus of P1 helices, and extend their aromatic moiety into the III/IV domain interface (fenestration). Using model-driven mutagenesis and electrophysiology, we identified five new SCBI-sensing residues, including insect-specific residues. Our study proposes the first three-dimensional models of channel-bound SCBIs, sheds light on the molecular basis of MFZ selective toxicity, and suggests that a sodium ion located in the inner pore contributes to the receptor site for electroneutral SCBIs.

Sodium channels are excellent targets of both natural and synthetic insecticides with high insect selectivity. Indoxacarb, its active metabolite DCJW, and metaflumizone (MFZ) belong to a relatively new class of sodium channel blocker insecticides (SCBIs) with a mode of action distinct from all other sodium channel-targeting insecticides, including pyrethroids. Electroneutral SCBIs preferably bind to and trap sodium channels in the inactivated state, a mechanism similar to that of cationic local anesthetics. Previous studies identified several SCBI-sensing residues that face the inner pore of sodium channels. However, the receptor site of SCBIs, their atomic mechanisms, and the cause of selective toxicity of MFZ remain elusive. Here, we have built a homology model of the open-state cockroach sodium channel BgNa v 1-1a. Our computations predicted that SCBIs bind in the inner pore, interact with a sodium ion at the focus of P1 helices, and extend their aromatic moiety into the III/IV domain interface (fenestration). Using model-driven mutagenesis and electrophysiology, we identified five new SCBI-sensing residues, including insect-specific residues. Our study proposes the first three-dimensional models of channelbound SCBIs, sheds light on the molecular basis of MFZ selective toxicity, and suggests that a sodium ion located in the inner pore contributes to the receptor site for electroneutral SCBIs.
Voltage-gated sodium channels are transmembrane proteins whose activation triggers fast inflow of sodium ions into the cell, causing the rising phase of the action potential. Eukaryotic voltage-gated sodium channels comprise the pore domain and four voltage-sensing domains within a single polypeptide chain of four homologous repeats. Each repeat includes six trans-membrane helical segments (S1-S6) connected by extra-and intracellular loops. A voltage-sensing domain contains helices S1-S4. The pore domain is formed by quartets of the outer helices (S5s), the pore-lining inner helices (S6s), and extracellular membrane re-entering P-loops, which are contributed by the four repeats. The voltage-sensing domains are connected to the pore domain by linker helices S4-S5 (L45). Upon membrane depolarization, the positively charged helices (S4s) move outward, inducing opening of the activation gate, which is formed by cytoplasmic parts of the S6s. The selectivity filter is composed of Asp, Glu, Lys, and Ala residues from the ascending parts of the four P-loops and divides the ion-conducting pathway into the following two parts: the outer pore, which is exposed to the extracellular space, and the inner pore, which in the open channel is exposed to the cytoplasm.
Sodium channels are excellent targets of both natural and synthetic insecticides. Several classes of sodium channel-targeting insecticides are highly insecticidal, but less toxic to mammals and are widely used for controlling arthropod pests and human disease vectors (1). Pyrethroids are a large class of synthetic compounds structurally derived from pyrethrins and are broadly used for control of mites, ticks, and other insect pests. The sodium channel blocker insecticides (SCBIs), 5 indoxacarb and metaflumizone (MFZ) (Fig. 1), are among the relatively new classes of insecticides with favorable selectivity (1)(2)(3). The selective toxicity of indoxacarb is due to insect-specific metabolism that decarboxylates indoxacarb to produce a more toxic derivative, DCJW (3). However, the mechanism of selective toxicity of MFZ is still unknown.
The major problem for the effective use of insecticides is the development of insecticide resistance. Intensive use of pyrethroids over the last decades has led to selection of numerous mutations in sodium channels in various arthropod populations, which confer resistance to pyrethroids (4). Identification of these sodium channel mutations has facilitated the development of molecular markers for early detection of pyrethroid resistance in various populations around the world (5). Furthermore, these mutations help define the receptor sites for pyrethroids on sodium channels and reveal the molecular mechanisms of the differential sensitivities of insect and mammalian sodium channels to pyrethroids (6).
Resistance to SCBIs recently emerged in populations of various agricultural pests. Two mutations, F1845Y and V1848I, in the sodium channel of the diamondback moth, Plutella xylostella (7), have been confirmed to reduce the SCBI sensitivity of cockroach sodium channels expressed in Xenopus oocytes (8). Therefore, the two mutations could be used as molecular markers for resistance monitoring in-field populations of the diamondback moth and possibly other pest species. Identification of more residues that are critical for the binding and action of SCBIs would provide further molecular markers for more accurate and earlier detection of resistance genotypes and might delay the development of resistance in natural populations.
SCBIs are believed to bind to and trap sodium channels in non-conducting slow inactivated states (9), the mechanism similar to that proposed for local anesthetics (LAs) (1,10). Molecular interactions of LAs with mammalian sodium channels have been intensively studied (11,12). Furthermore, structural models of the Na v 1.4 channel are elaborated in which several pore-facing residues in the inner helix IVS6 contribute to the binding site of LAs (13,14). In particular, two LA-sensing residues in IVS6, i.e. Phe-1764 and Tyr-1771 in rat Na v 1.2 and Phe-1579 and Tyr-1586 in Na v 1.4, are consistently critical for the binding and action of a wide range of LAs and related drugs on mammalian sodium channels. To facilitate recognition of ligand-sensing residues among different sodium channels, here we labeled these two LA-sensing residues in IVS6 as F 4i15 and Y 4i22 using a nomenclature universal for P-loop ion channels (6,15). Similar to the effect on the action of LAs, alanine substitution, F 4i15 A, resulted in a significant reduction in the ability of DCJW and an experimental SCBI, RH3421, to inhibit Na v 1.4 sodium channels expressed in Xenopus oocytes (16). However, the same substitution F 4i15 A in a cockroach sodium channel, BgNa v 1-1a, did not reduce the potency of both DCJW and metaflumizone (17). Remarkably, one of the two mutations from the indoxacarb-resistant diamondback moth populations was a tyrosine substitution of F 4i15 , which reduced the action of A sodium ion is placed at the possible position in the 5-membered chelating ring. In a low-potent compound D, the ammonium group is attracted to the imino nitrogen, making unlikely chelation of a sodium ion. The different positions of the ammonium nitrogen versus that of a chelated sodium in other compounds may explain the low potency of compound D. Structural formulae of compounds D and RH are taken from Refs. 2, 36. both DCJW and metaflumizone on BgNa v 1-1a channels expressed in Xenopus oocytes (7,8). Furthermore, mutation of the tyrosine residue, Y 4i22 , to alanine in Na v 1.4 channels resulted in a significant increase in the potency of indoxacarb, DCJW, and RH3421 (16), and the same substitution, Y 4i22 A, also enhanced the action of metaflumizone on BgNa v 1-1a channels (17). These data strongly support the notion that SCBI receptors are located, at least partially, in the inner pore and may overlap with the LA receptors (2). However, specific atomic details of interactions between SCBIs with sodium channels are still unclear.
In this study, we generated an open-state model of the cockroach sodium channel, BgNa v 1-1a. The model has the wide open activation gate like in the x-ray structure of the open potassium channel K v 1.2 (18) and wide inter-repeat fenestrations like in the x-ray structure of the closed sodium channel Na v Ab (19). SCBIs are bigger and have more electronegative atoms than LAs like lidocaine. At physiological pH, LAs are protonated to bear a positive charge that enables the channel block by electrostatic mechanism (13,14). In contrast, SCBIs are electrically neutral non-ionizable compounds (Fig. 1). Driven by the hypothesis that certain electroneutral compounds and permeant ions may form cationic blocking particles (20), we docked sodium-bound DCJW and MFZ into the channel. We arrived at a model in which the bulky part of SCBIs binds in the inner pore, a halogenated aromatic moiety protrudes in the III/IV fenestration, and the ligand-bound sodium ion occurs in the cation-attractive region of the inner pore, at the focus of P1 helices. Model-driven mutational and electrophysiological analyses revealed five new SCBI-sensing residues. These include insect-specific residues in the III/IV repeat interface and at the turn of repeat IV P-loop. We also revealed residues that interact with either MFZ or DCJW. Our study provides atomic details of the MFZ and DCJW receptors and supports the hypothesis on direct interactions of permeant cations with electroneutral ligands of ion channels (20). These results contribute to better understanding of mechanisms of sodium channel modulation by various ligands and will assist in monitoring emerging SCBI resistance of pests and rational design of new SCBIs.

Results
Open-state Model of BgNa v 1-1a-Previously identified SCBIsensing residues F 4i15 , V 4i18 , and Y 4i22 (Fig. 2) face the inner pore (8), and F 4i15 may also face the III/IV fenestration (21). None of the currently available high resolution structures of P-loop channels can be directly used to build a homology model of the open sodium channel, which would be suitable for the computational search for binding sites of large SCBI molecules. Indeed, the x-ray structures of the closed sodium channels (19,22,23) have wide fenestrations between two S6 and P1-helix, but their narrow S6 bundle is a poor template to model the cytoplasmic half of the open inner pore. The x-ray structures of open potassium channels, e.g. K v 1.2 (18) have the wide inner pore, but their narrow inter-subunit interfaces are poor templates to model the wide fenestrations. The open-state structure of a prokaryotic sodium channel, Na v Ms (24), lacks the linker-helices S4-S5, which contribute to the binding sites of pyrethroids and 1,1,1-trichloro-2,2-bis(p-chlorophenyl)ethane and, a priori, could also participate in the binding of SCBIs. Moreover, the cytoplasmic half of the inner pore in Na v Ms is narrower than that in K v 1.2 (the distance between diagonally opposed atoms C ␤ _i19 in Na v Ms and K v 1.2 are 12.9 and 14.0 Å, respectively).
To resolve the problem, we in silico opened a Na v Ab-based model of the closed mosquito sodium channel AaNa v 1-1a (6), see under "Materials and Methods" and supplemental Fig. S1. Residues of the AaNa v 1-1a and BgNa v 1-1a channels within the inner pore and repeat interfaces are practically identical. Therefore, we call the obtained model the Na v Ab/K v 1.2-based model of the open BgNa v 1-1a channel. This model has a wide III/IV fenestration and a wide open activation gate (supplemental Fig. S2, A and B). The known SCBI-sensing residues F 4i15 , V 4i18 , and Y 4i22 face the inner pore (supplemental Fig. S2, C and D).
Docking DCJW into the Open BgNa v 1-1a Channel-We first docked DCJW in the sodium ion-free channel as described under "Materials and Methods." In the apparent global minimum and several local minima, the heterocyclic core of DCJW bound against the IVS6 inner helix forming close contacts with F 4i15 , V 4i18 , and Y 4i22 , whereas aromatic moieties at the opposed termini of the ligand approached the III/IV and I/IV fenestrations (supplemental Fig. S3). The model is consistent with the results from mutational analysis of the action of DCJW on rNa v 1.4 and BgNa v 1-1a channels (Table 1). However, it highlights three issues. First, there is a large imbalance between the number of polar atoms in DCJW and in the side chains of porefacing residues. Indeed, the central part of DCJW contains seven closely spaced oxygen and nitrogen atoms ( Fig. 1), but their binding partners in the channel are unknown. In the model, only tyrosine Y 4i22 approached DCJW but did not H-bond to it. Other polar residues, which face the inner pore (S 1i15 , N 2i15 , S 3i15 , and Q 1p49 ), were too far from the DCJW heterocyclic core. Thus, the model does not explain the role of the N-N-CϭO fragment, which is the "fingerprint" of SCBIs ( Fig. 1). Second, both LAs and DCJW are believed to bind preferably to slow-inactivated channels (10,16,25). In the case of LAs, the preference is likely due to the lack of electrostatic repulsion between the ligand charged group and the cationdeficient outer pore (13). In the case of electroneutral DCJW, the cause of the preference is unclear. Third, the model does not suggest an atomic mechanism for the Na ϩ current block by the ligand. DCJW, which binds tightly to the pore-lining helix IVS6, does not occlude the inner pore, leaving space for permeant ions to pass between the ligand and polar residues S 1i15 , N 2i15 , and S 3i15 (supplemental Fig. S3). Such a binding mode is proposed for sodium channel agonists like batrachotoxin (26,27), but cationic antagonists such as LAs block the permeation by the electrostatic mechanism (13,14). The fact that cationic LAs and electroneutral DCJW have common pore-facing ligand-sensing residues F 4i15 , V 4i18 , and Y 4i22 and that these ligands apparently target the same region in the inner pore is a paradox.
A possible solution to these three issues could be based on the hypothesis that electroneutral ligands block the cation-attractive pore of ion channels not per se but in complexes with permeant ions (20). We first explored whether DCJW can inter-act with a sodium ion located at position Na III , between the backbone carbonyls of residues p48 as seen in the Na v Ms x-ray structure (24). We fixed a sodium ion in this position and submitted several MCM trajectories with various electronegative atoms of DCJW constrained to the sodium ion. These calculations did not yield a low energy complex with DCJW approaching the experimentally known DCJW-sensing residues F 4i15 and V 4i18 .
We then explored whether DCJW can interact with a sodium ion in the inner pore. In the x-ray structures of the open and closed potassium channels, a potassium ion is seen in the inner pore, at the focus of P-helices (18,28). Because the exact position of a sodium ion in the inner pore is unknown, we considered many various possibilities as described below.
Docking Library of Sodium-bound DCJW Conformers-We generated a library of 110 sodium-bound DCJW conformers (supplemental Fig. S4). In most of the structures, the imino nitrogen and carbonyl oxygen chelated the sodium ion forming an energetically favorable five-membered chelating ring (29). In addition, many structures were found with the sodium ion che- Open circles indicate SCBI-sensing residues that were identified in earlier studies. Filled circles indicate SCBI-sensing residues, which are identified in this study. B, sequence alignment of K v 1.2 and sodium channels. Positions of residues according to the universal labeling scheme of P-loop are indicated above aligned sequences of individual segments. Residues, which face the inner pore or III/IV fenestration and which are different between the insect and mammalian channels, are underlined. Insertion/deletions in the P-loops are shown according to the sequence alignment proposed in Ref. 50. lated between an aromatic ring and a carbonyl oxygen. For each sodium-bound DCJW conformer, 256 random starting positions/orientations in the inner pore were generated, and the starting points were MC-minimized. The in silico opened homology model of the pseudo-heteromeric eukaryotic channel is not expected to be precise enough to use energetics as the only criterion to select the specific binding mode. Therefore, we imposed distance constraints between DCJW and DCJW-sensing residues F 4i15 and V 4i18 to bias the docking and find a compromise between the computed energetics and the experimental data.
In the lowest energy complex (Fig. 3), the sodium ion was bound to two nitrogen and two oxygen atoms of DCJW. The ion occurred in the central cavity, at the level of the pore-facing residues S 1i15 , N 2i15 , S 3i15 , and F 4i15 . These residues are highly conserved in eukaryotic sodium channels and likely contribute to intermediate binding site(s) for hydrated permeant ions. The DCJW-distant hemisphere of the sodium ion faced N 2i15 and S 3i15 and would attract water molecules wetting these residues. The summed electronegative potentials from the C-ends of P1-helices would further stabilize the DCJW-bound sodium ion and thus the entire channel-blocking cationic particle. The dihydroindene moiety formed a stacking contact with F 4i15 and closely approached V 4i18 . Tyrosine Y 4i15 directed toward the carbonyl oxygen in the bridge between two cyclic moieties but was too far to donate an H-bond to DCJW. The chlorine atom extended into the III/IV interface and interacted with F 3p44 and L 4i8 , which are insect-specific residues (Fig. 2). The phenyl trifluoromethyl group at the other end of DCJW approached T 4p48 . The latter forms an interdomain H-bond with W 1p52 (19) that contributes to the stability of the P-loop domain.
Docking Library of Sodium-bound MFZ-MFZ and DCJW have common ligand-sensing residues F 4i15 and V 4i18 (8) and a common fragment CϭN-NH-CϭO that may chelate a sodium ion. We generated a library of 140 sodium-bound MFZ conformers that contained highly divergent structures (supplemental Fig. S4, E and F). In the lowest energy complex the ion was chelated by the carbonyl oxygen and the imino nitrogen with the involvement of -electrons from an aromatic ring (supplemental Fig. S4F). For each sodium-bound MFZ conformer, 256 random starting positions/orientations in the inner pore were generated, and the starting points were MC-minimized with distance constraints between the ligand and known MFZ-sensing residues V 2i18 , F 4i15 , V 4i18 , and Y 4i22 .
The lowest energy ternary complex MFZ/Na ϩ /BgNa v 1-1a is shown in Fig. 4. The imino nitrogen and the carbonyl oxygen chelated a sodium ion in the central cavity, at the focus of P1-helices. As in the case of DCJW, the MFZ-distant sodium ion hemisphere faced N 2i15 and S 3i15 and would interact with water molecules wetting these residues. The trifluoromethoxy-phenyl moiety stacked with F 4i15 and extended into the III/IV interface to reach the insect-specific F 3p44 and I 4i8 , and this moiety also approached T 3i18 . The trifluoromethyl-phenyl moiety extended toward IIS6 to reach the MFZ-sensing V 2i18 . The 4-cyanophenyl group extended toward the I/II interface. Thus, our calculations predicted new SCBIs sensing residues that are common for MFZ and DCJW (F 3p44 and I 4i8 ), a DCJW-specific T 4p48 , and an MFZ-specific residue T 3i18 .
Na v Ms-based Model of BgNa v 1-1a with SCBIs-Our computations did not reveal SCBI-binding residues in the S4-S5 linkers or at the S5 and S6 sides that face the linkers. Therefore, the Na v Ms x-ray structure, which lacks the S4-S5 linkers, appears suitable for computational docking of SCBIs. We used MCM to move C ␣ atoms of Na v Ab/K v 1.2-based model to positions of respective atoms in the Na v Ms x-ray structure as described elsewhere (30). In the obtained model, positions of the S4-S5 helices are similar to those in the Na v Ab/K v 1.2-based model. DCJW fits snugly in the Na v Ms-based model and forms contacts with the same residues as in the Na v Ab/K v 1.2-based model (Fig. 5). We used the same approach to obtain Na v Msbased model of BgNa v 1-1a with MFZ and also found that MFZ forms contacts with the same residues as in the Na v Ab/K v 1.2based model (data not shown).
Mutational Studies Are Consistent with the Predicted Binding Sites of DCJW and MFZ-We used an SCBI-sensitive cockroach sodium channel, BgNa v 1-1a (17), to generate six mutants of the four predicted SCBI-sensing residues that directly interact with SCBI in our models. We also generated two mutants of W 1p52 , which did not interact with SCBIs in our models but forms interdomain H-bonds with the predicted DCJW-sensing residue T 4p48 . We expressed the eight mutants individually in oocytes and examined the effect of the mutations on the action of SCBIs. Most of the mutations did not alter the voltage dependence of activation, fast inactivation, or slow inactivation ( Table 2). As shown previously (8,17), DCJW and metaflumizone cause state-dependent inhibition of BgNa v 1-1a channels at depolarized holding potentials during 30 min of insecticide exposure (Fig. 6). The inhibition of sodium currents by DCJW at the end of 30 min of the insecticide exposure was significantly reduced by mutations F 3p44 A, F 3p44 L, L 4i18 A, W 1p52 A, W 1p52 S, and T 4p48 A but not by L 4i18 C or T 3i18 A (Figs. 6 and 7 and Table 3). We then examined the action of MFZ on the same mutants. Mutations of four residues (F 3p44 A, L 4i18 A, W 1p52 A, and T 3i18 A) did decrease the potency of MFZ. Mutation T 4p48 A had no effect on the MFZ action.  and MFZ, and used model-driven mutagenesis and electrophysiology to map binding sites for these SCBIs. A channel mutation may affect ligand action directly, by changing the ligand-channel contacts or indirectly by changing the channel conformation and thus the state-dependent binding-site geometry. For example, tryptophans in positions p52 are exceptionally conserved in sodium and calcium channels. They form intersubunit H-bonds with threonines p48 (19) and are involved in other intersegment contacts that stabilize folding of the P-loop domain (31). Despite that W 1p52 does not form direct contacts with SCBIs in our models (Fig. 5), mutation W 1p52 A decreased the potency of DCJW and MFZ (Figs. 6 and 7) likely due to destabilizing folding of the P-loop domain that contributes to the SCBI-binding sites. However, it is unlikely that mutations of residues, which form direct contacts with SCBIs in our models, indirectly affected action of the ligands. Although the current reduction by the mutants is rather small (Fig. 7), it should be noted that the effects are comparable with those of naturally occurring kdr mutations F 4i15 Y and I 4i18 I that are found in diamondback moth (P. xylostella) in China (7) and demonstrated to reduce action of SCBIs in BgNa v 1-1a (8).

Discussion
Previous studies revealed four SCBI-sensing residues in the BgNa v 1-1a and Na v 1.4 channels ( Table 1). These experimental data along with the hypothesis on direct interaction of electro-neutral blockers of ion channels with the permeant ions (20) were used to focus our search for the SCBI-binding sites. We arrived to the models where sodium-bound ligands bind tightly against IVS6, approach the P-loops in repeats I and III, and extend a hydrophobic group into the III/IV interface. Our model-driven mutagenesis unveiled five previously unknown DCJW and MFZ-sensing residues, including the insect-specific residues F 3p44 and L 4i8 , as well as residues that individually contribute to the receptors of DCJW (T 4p48 ) and MFZ (T 3i18 ). The fact that mutations of the insect-specific residues F 3p44 and L 4i8 A (Fig. 7B) affect the MFZ action may explain the selective toxicity of MFZ. The previous finding that the V 2i18 K mutation in rNa v 1.4 significantly reduces the action of MFZ, but has no effect on the action of DCJW (Table 1), is consistent with our models where tri-polar MFZ, but not dipolar DCJW, reached IIS6.
Unlike other SCBIs, MFZ can bind to the resting sodium channels (2). The long predicted hydrophobic access pathway for LAs in the closed sodium channels (25) was first visualized in homology models of sodium channels where LAs are proposed to reach the closed sodium channels through the "sidewalk" in the III/IV repeat interface (21,32). The x-ray crystallography demonstrated a wide inter-subunit fenestration in a prokaryotic sodium channel Na v Ab (19)   the predicted sidewalk in the eukaryotic channels. Here we revealed two MFZ-sensing residues, F 3p44 and L 4i8 , in the III/IV fenestration (Table 1). In view of these data, we suggest that a highly flexible MFZ molecule would pass through the III/IV fenestration in the resting (closed) channel and block it. DCJW has the bulky tricyclic core and is much more rigid than MFZ. In addition, DCJW has three more oxygen atoms than MFZ (Fig. 1). These two factors may explain why the hydrophobic access pathway involving the membrane and the III/IV fenestration is impassible for DCJW.
SCBI Action in Insect and Mammalian Channels-Mutation F 4i15 A significantly impairs the action of DCJW and MFZ in the rNa v 1.4 channel, but it does not affect the action of DCJW or enhance the action of MFZ in BgNa v 1-1a (Table 1). These facts indicate that SCBI receptors in the Na v 1.4 and BgNa v 1-1a channels are not identical (2). In our models, SCBIs would be attracted to big hydrophobic residues F 3p44 and L 4i8 in BgNa v 1-1a more strongly than to their smaller cognates L 3p44 and C 4i8 in rNa v 1.4 (Table 4). In two other fenestration-lining positions, 4i11 and 4i12, the Na v 1.4 channel has isoleucines, whereas BgNa v 1-1a has smaller valines ( Fig. 2 and Table 4). These facts suggest that the trifluoromethoxy-phenyl moiety of the flexible MFZ would penetrate more deeply into the III/IV fenestration of BgNa v 1-1a than into the cognate fenestration of Na v 1.4. The deeper MFZ penetration would be opposed by the pore-facing F 4i15 . Mutation F 4i15 A would permit the flexible MFZ deeper into the III/IV fenestration (which is also lined by the MFZ-sensing T 3i18 ). In contrast, the more rigid DCJW would be unable to move more deeply into the III/IV fenestration, and the F 4i15 A mutation would not affect its action.
The fact that mutations V 4i18 I and F 4i15 Y decrease the potency of both MFZ and DCJW in the BgNa v 1-1a channel ( Table 1) is consistent with our models that suggest tight binding of the ligands against IVS6. The bigger V 4i18 I would sterically repel the ligand, whereas the hydroxyl group in F 4i15 Y would repel the hydrophobic moieties of the ligand. Mutation Y 4i22 A potentiates DCJW and MFZ block in both rNa v 1.4 and BgNa v 1-1a (Table 1). In our models, the side chain of Y 4i22 is oriented upward and its hydroxyl group approaches hydrophobic moieties of the ligands. This energetically unfavorable interaction would be relieved in the Y 4i22 A mutant.
The above rationale suggests that the binding modes of SCBIs in the rNa v 1.4 and BgNa v 1-1a channels are similar, whereas different effects of the F 4i15 A mutation are due to channel-specific residues in the III/IV fenestrations. However, our homology models are not precise enough to quantitatively evaluate the energetics of SCBI interactions in different channels and their mutants.    SEPTEMBER 16, 2016 • VOLUME 291 • NUMBER 38

JOURNAL OF BIOLOGICAL CHEMISTRY 20119
SCBI Interaction with Sodium-DCJW, MFZ, and other SCBIs have a conserved fragment N-N-CϭO that would chelate a sodium ion to form the energetically preferable 5-membered ring. Such chelating patterns are common. For example, in the gas phase, amino acids form stable 5-membered rings with a sodium ion chelated between the backbone amino and carbonyl groups (33). A sodium ion is readily chelated between the imino nitrogen and ketone oxygen to form a 5-membered ring in picolinic acid (34) and a four-membered ring in cytosine (35).
SCBIs are obviously not strong chelators, and we are not aware of experimental data on association of SCBIs with Na ϩ , K ϩ , Ca 2ϩ , or other cations in water. However, electroneutral ligands with multiple electronegative atoms should interact with cations within the channel much stronger than in the bulk solvent. Because of specific interactions with the channel proteins, local concentrations of ions and ligands within the channels are much higher than their overall concentrations in the bulk solvent. Therefore, the probability of ligand-metal association within the channel must be much higher than in the bulk Note that DCJW binds tightly against DCJW-sensing residues F 4i15 , V 4i18 , and Y 4i22 and forms close contacts with insect-specific residues F 3p44 and L 4i8 . C, view through the IV/I fenestration. D, view from the extracellular side with parts of P-loops removed for clarity.

TABLE 2
Voltage dependence of activation, fast, and slow inactivation of BgNa v 1-1a and mutant channels at the holding potential of ؊120 mV The voltage dependences of conductance and inactivation were fitted with a two-state Boltzmann equation to determine V 1/2 , the voltage for half-maximal conductance or inactivation, and k, the slope factor for conductance or inactivation. The number of oocytes was 6 -14, and values in the table represent the mean Ϯ S.E. The asterisks indicate significant differences from the BgNa v 1-1a channel as determined by one-way analysis of variance (pϽ0.05) with Scheffe's post hoc analysis. Asterisks indicate that the specific residue lines the inner pore and/or specific fenestration.
Na ؉ channel Activation Fast inactivation Slow inactivation Ϫ58.1 Ϯ 1.7 10.9 Ϯ 1.9 solvent. By the same cause, SCBIs are unlikely to bring a metal ion into the pore from the cytoplasm or extracellular environment, but they would readily associate with a metal ion within the pore. Structure-Activity Relations of SCBIs-At physiological pH, compound D (Fig. 1) would be protonated. The fact that compound D blocks the sodium channel, although with a weaker potency than classical SCBIs (2), supports the concept that electroneutral SCBIs chelate a sodium ion to form cationic blocking particles. The relatively low potency of compound D is understandable: the ionized ammonium group would reach the imine nitrogen, but not the ketone oxygen so that position of the positive charge would differ from that in DCJW and MFZ. In the series of six RH dihydropyrazoles (Fig. 1), four compounds readily inhibited sodium uptake into mouse brain vesicles, whereas compounds RH5654 and RH2643 were inactive (36). A possible cause is that the hydrophobic butyl group in RH5654 or a partial positive charge in the hydroxy proton of RH2643, which is oriented toward the imino nitrogen, repels a sodium ion from the ligand.
Possible Mechanism of State-dependent Channel Block by LAs and SCBIs-Mutational studies suggest that LAs bind preferably to the fast-inactivation channels (37,38), but the slowinactivation states are also considered as targets for the action of LAs (10,39) and SCBIs (1). A possible cause of this state dependence is that in the inactivated channels the permeant ions would not compete with the pore-bound cationic blockers (13). To some extent, the proposed DCJW binding mode is reminiscent to that of tetracaine, which is predicted to approach the outer pore by the cationic group and extend the other end in the III/IV fenestration (21). The above predictions are consistent with the x-ray structure of the LA-bound Na v Ms  To measure the inhibition of peak current by SCBIs, test pulses (20 ms) to Ϫ10 mV from a depolarizing holding potential (as indicated, e.g. Ϫ55 mV for BgNa v 1-1a) were given once every minute to record the remaining sodium current. The remaining sodium current was then normalized to the current measured prior to application of insecticide. Reduction in Normalized I Na reflects the progress of channel inhibition by SCBIs. The values of percentage of inhibition were determined by values of "normalized I Na " at the end of the 30 min of recording (see Fig. 6). The values represent the mean Ϯ S.E., and the number of oocytes was 6 -14. The asterisks indicate significant differences from the BgNa v 1-1a channel as determined by one-way ANOVA (p Ͻ 0.05) with Scheffe's post hoc analysis. Asterisks indicate that the specific residue lines the inner pore and/or specific fenestration. channel in which the outer pore is sodium-deficient, and a heavy bromine atom of the ligand is seen in a fenestration (40).
Our study proposes common features of the binding modes for cationic LAs and sodium-bound SCBIs in the channel as follows: a positive charge at the focus of P1-helices and a hydrophobic moiety protruding into the III/IV interface. Another possible explanation for the preferable binding of LAs and SCBIs to the inactivated channels would be a state-dependent three-dimensional complementarity between the ligands and the channel. The pore geometry in the inactivated sodium channels is unknown, and our open-state model is just an approximation to the inactivated state structures. However, the hypothesis on the state-dependent three-dimensional complementarity would not explain why many structurally different ligands of different channels bind preferably to inactivated states. The electrostatic mechanism of the channel block, which assumes competition between ligands and the permeant ions, seems universal to both cationic ligands and metal-bound electroneutral ligands.
Conclusions-In this study we docked DCJW and MFZ in the open-state model of an insect sodium channel. The models predict that the central heterocyclic moiety of SCBIs interacts with a sodium ion in the central cavity, a halogen-substituted aromatic ring extends into the fenestration lined by helices IIIP1, IIS6 and IVS6, and another end of the ligand interacts with the pore helix IP1. Model-driven mutagenesis allowed us to discover five new SCBI-sensing residues, four of which directly interact with the ligands. These include residues in IIIP1 and IVS6, which are specific for insect sodium channels and thus may explain the selective toxicity of MFZ, as well as residues in IVP1 and IIIS6, which interact, respectively, only with DCJW and MFZ. The models imply the channel-blocking mechanism, which is consistent with the hypothesis on direct interactions of electroneutral ligands with permeant ions in the pore of inactivated channels. Because many electroneutral ligands have binding sites that overlap with cationic ligands, the latter conclusion may be of general importance for the pharmacology of ion channels and for developing new selective ligands.

Materials and Methods
Site-directed mutagenesis was performed by PCR using mutant primers and Pfu Turbo DNA polymerase (Stratagene, La Jolla, CA). All mutants were verified by DNA sequencing.

Expression of BgNa v 1-1a channels in Xenopus laevis
Oocytes-The procedures for oocyte preparation and cRNA injection are identical to those described previously (41). For robust expression of the BgNa v 1-1a sodium channel, cRNA was co-injected into oocytes with Drosophila melanogaster tipE cRNA (1:1 ratio), which enhances the expression of insect sodium channels in oocytes (42,43).
Electrophysiological Recording and Analysis-Sodium currents were recorded using the two-electrode voltage clamp technique. Electrodes were pulled from borosilicate glass and filled with 3 M KCl and 0.5% agarose. Resistances ranged between 0.5 and 1.5megohms. Currents were measured with an oocyte clamp amplifier OC725C (Warner Instrument Corp., Hamden, CT), Digidata 1440A (Axon Instruments, Foster City, CA), and pClamp 10.2 software (Axon Instruments). Capacitive transient leak currents were subtracted using the P/N (n ϭ 4) subtraction method.
Examination of BgNa v 1-1a Channel Sensitivity to SCBIs-The methods for measuring the effects of SCBIs on BgNa v 1-1a channels are similar to those described previously (17). Briefly, we measured the onset of block by SCBIs at or near the potential of 50% steady-state inactivation. After establishing a stable voltage clamp near the half-inactivation potential specific to a channel variant, insecticide-containing solution was perfused into the bath at a rate of 3 ml/min over the first 7-8 min, and the time course of onset of block was recorded for 30 min. All experiments were performed at room temperature.
Indoxacarb and DCJW were provided by K. D. Wing and D. Cordova (DuPont Agrochemicals), and metaflumizone was provided by Vince Salgado (BASF Agricultural Products). Insecticides were perfused onto oocytes in a manner similar to that previously described (44).
Data are presented as mean Ϯ S.E. Statistical analysis was determined using a one-way analysis of variance test and Scheffe's post hoc analysis. Significance values were set at p Ͻ 0.05 or as indicated in the table and figure legends.
Molecular Modeling-Sequence alignment of K v 1.2 and BgNa v 1-1a channels is shown in Fig. 2. Homology modeling and ligand docking were performed using the ZMM program (45) and Monte Carlo-minimization protocol (46) as described elsewhere (6). Molecular images were created using the PyMOL Molecular Graphics System, Version 0.99rc6 (Schrödinger, LLC, New York, NY). We docked S-DCJW because the R isomer of indoxacarb is completely inactive (47).
In Silico Opening of the Closed BgNav1-1a Model-We superimposed the closed BgNa v 1.1 model with the K v 1.2 x-ray structure by minimizing root mean square deviation between the C ␣ atoms in positions o21, which usually contain small residues Ser, Gly, or Ala (Fig. 2). In both K v 1.2 and BgNa v 1-1a, o21 residues make close contacts with the H ␣ backbone atoms in the p40 positions that contain big residues (Phe, Leu, Tyr, or Met). In the superimposed structures (supplemental Fig. S1), the extracellular halves of the P1, S5, and S6 helices overlap rather well. As expected, the cytoplasmic parts did not overlap; the C ␣ _i30 atoms of the model deviated from the cognate atoms of K v 1.2 by 7.6 -8.7 Å (supplemental Fig. S1C). To open the closed BgNa v 1-1a model, we applied centrifugal forces to the C ␣ _i30 atoms, allowing these atoms (along with the cytoplas- Organism-specific sodium channel residues, which line the inner pore and fenestrations around repeat IV Asterisks indicate that the specific residue lines the inner pore and/or specific fenestration.

III/IV fenestration IV/I fenestration 3p44
Leu Phe * 4p47 Thr Ser * 4i8 Cys Leu * 4i11 Ile Leu * * 4i12 Ile Val * 3i18 Val Thr * 1i19 Ile Val * 4i19 Val Ile * mic halves of the S6 helices) to move with the step of 0.5 Å and MC-minimized energy at each step as described previously (48). To preserve general folding of the model, we constrained the ␣-helical H-bonds in helices L45 and the inter-repeat H-bonds between side chains of residues W p52 and T/C p48 (19). Unbiased Docking DCJW into the Open BgNa v 1-1a Channel-We generated a library of 36 DCJW conformers (supplemental Fig. S4, A and B) using a force field that disregards the intramolecular van der Waals attractions, thus favoring extended conformations that would interact stronger with the protein (45). Each conformer was docked from 1280 random starting positions/orientations. A total of 36 ϫ 1280 MCM trajectories, each including 10 energy minimizations, were calculated with rigid backbones without imposing any ligandchannel distance constraints.
Biased Docking of Ligands to Experimentally Known Ligandsensing Residues-This was performed using ligand-side chain distance constraints (49). To preserve the channel folding during docking of flexible ligands to the flexible protein, pin constraints were imposed between matching ␣-carbons in the template and the model. A pin constraint is a flat-bottom parabolic energy function that allows a C ␣ atom to deviate, penalty-free, up to 1 Å from the template and imposes a penalty of 10 kcal/mol/Å for larger deviations. The pin constraints are necessary because the initial relaxation of the channel model with a bulky ligand would cause large deviations of the protein backbones from the template due to sterical ligand-protein clashes.