Conserved His-Gly motif of acid-sensing ion channels resides in a reentrant ‘loop’ implicated in gating and ion selectivity

Acid-sensing ion channels (ASICs) are proton-gated members of the epithelial sodium channel/degenerin (ENaC/DEG) superfamily of ion channels and are expressed throughout central and peripheral nervous systems. The homotrimeric splice variant ASIC1a has been implicated in nociception, fear memory, mood disorders and ischemia. Here we extract full-length chicken ASIC1a (cASIC1a) from cell membranes using styrene maleic acid (SMA) copolymer, yielding structures of ASIC1a channels in both high pH resting and low pH desensitized conformations by single-particle cryo-electron microscopy (cryo-EM). The structures of resting and desensitized channels reveal a reentrant loop at the amino terminus of ASIC1a that includes the highly conserved ‘His-Gly’ (HG) motif. The reentrant loop lines the lower ion permeation pathway and buttresses the ‘Gly-Ala-Ser’ (GAS) constriction, thus providing a structural explanation for the role of the His-Gly dipeptide in the structure and function of ASICs.

between residues 433-436 within the upper third of the TMD ( Figure 1D-E). At pH 7.0, SMA-109 cASIC1a particles prepared at pH 7.0 populate a desensitized state that mirrors the overall 110 architecture of the existing x-ray structure (29), including the presence of a proton-bound 111 'collapsed' acidic pocket ( Figure 1F). In contrast, SMA-cASIC1a particles maintained at pH 8.0 112 occupy a high pH resting conformation, characterized by an expanded acidic pocket ( Figure 1G) 113 that resembles the high pH, resting state structures solved by x-ray crystallography and single 114 particle cryo- EM (27). We propose that the limited resolution of the resting channel structure, 115 while presumably impacted by sample conditions including thicker ice, may also be due to 116 structural flexibility inherent to the resting channel conformation in the absence of divalent 117 cations, which serve to stabilize an expanded acidic pocket at high pH (40) but which are 118 incompatible with current SMA-based purification strategies. 119

Amino terminal residues form a reentrant loop 120
Numerous experiments have implicated residues within the pre-TM1 region of ASICs 121 and ENaCs in both gating and selectivity (32,34,41). Indeed, the highly conserved HG motif is 122 located within the pre-TM1 region of ASICs and ENaCs and its disruption lowers the open 123 probability in ENaCs and underlies PHA type 1 disorder (32,33). In contrast to existing 124 structures of ASICs solubilized in detergent micelles, we observed strong protein density 125 corresponding to amino terminal residues in cryo-EM maps of both desensitized and resting 126 SMA-cASIC1a channels maintained in a lipid environment (Figure 2A-B). 127 The quality of the 2.8 Å density map of the desensitized channel was sufficient to model 128 the amino terminus of the existing model (28, 29) (PDB 4NYK) from residue 17 (Supplementary 129 Data Figure 2), demonstrating that pre-TM1 residues form a reentrant loop comprised of two 130 short helical segments separated by a turn, positioned on the cytoplasmic side of the GAS belt 131 ( Figure 2C-E). Interestingly, the presence of the reentrant loop does not noticeably impact the 132 position of either transmembrane helix from those observed in prior x-ray or cryo-EM structures. 133 Rather, the reentrant loop residues are 'pinned' within the inverted 'v-shaped' cavity formed 134 between the lower transmembrane helices and maintained primarily by virtue of intra-subunit 135 contacts with TM2b and TM1 (Supplementary Data Figure 6). 136 While the quality of the resting channel density map that includes the pre-TM1 residues 137 was not sufficient for unambiguous model building (Supplementary Data Figure 4), no 138 significant differences in reentrant loop conformation were observed between the desensitized or 139 resting channels at the current resolutions ( Figure 2F), allowing us to rigid body fit the pre-TM1 140 structural element derived from the desensitized state structure into the resting state map. In 141 contrast with the discovery of new density for the pre-TM1 region, we did not observe any 142 additional interpretable density associated with carboxy terminus in either the desensitized or 143 resting state maps, thus suggesting that even in SMA-solubilized protein, these regions are 144 disordered. 145

The reentrant loop harbors the HG motif 146
Separated by more than 400 residues in the amino acid sequence, the GAS and HG motifs 147 are highly conserved amongst ENaC/DEG and ASIC channels ( Figure 3A) and have been 148 implicated in gating (41) and ion selectivity (42-45). Interestingly, the HG motif, which contains 149 a well-characterized disease mutation at the universally conserved glycine residue (32, 33), is 150 situated on the turn between the reentrant helices where it buttresses the TM2 domain swap and 151 GAS belt residues from 'below' ( Figure 3B). Residing along the ion permeation pathway and at 152 a subunit interface, the HG motif is capped by the carboxyl terminus of TM2a via an intra-153 subunit hydrogen bonding interaction with Ile 442 and participates in an inter-subunit hydrogen 154 bonding interaction with a neighboring GAS belt residue via Ser 445 ( Figure 3C). This intricate 155 network of intra-and inter-subunit interactions, formed between highly conserved motifs via the 156 TM2 domain swap and amino terminal reentrant loop, underscores the critical nature of the 157 lower pore architecture to ASIC, and by extension, to ENaC function. 158 Pre-TM1 residues form the lower ion permeation pathway 159 In structures of both desensitized and resting SMA-cASIC1a channels, the 'upper' ion 160 permeation pathway is comprised of TM2a residues and contains a closed gate between Gly 432 161 and Gly 436, in agreement with existing x-ray and cryo-EM models ( Figure 4A). However, 162 where structures of ASIC1a in resting (27), open (28) and desensitized (29) conformations 163 highlight a lower ion permeation pathway comprised entirely of TM2b residues that expands 164 outwards to form a wide intracellular vestibule, pre-TM1 residues of the SMA-isolated cASIC1a 165 channels line a more narrow ion permeation pathway extending below the GAS belt ( Figure 4A-166 B). 167 The lower ion conduction pathway of resting and desensitized SMA-cASIC1a channels is 168 formed by reentrant amino terminal residues Ser 24 through His 29 ( Figure 4C), the latter of 169 which is situated immediately below the GAS belt and is oriented towards the threefold axis 170 where it forms a constriction below the gate in the desensitized channel ( Figure 4D). Our data 171 demonstrate that pre-TM1 residues line the lower ion conduction pathway in structures of resting 172 and desensitized cASIC1a channels, providing a structural rationale for earlier reports which 173 indicated that pre-TM1 residues may form part of the pore (46) and contribute to ion permeation 174 and Na + selectivity of ASICs (34). 175 In the x-ray structure of an open channel conformation, hydrated Na + ions encounter a 176 constriction at the GAS belt TM2 domain swap (28), which has long been thought to underpin 177 ion selectivity in ENaC/DEG channels (42,44,47). Recently, however, residues along TM2a and 178 TM2b both 'above' and 'below' the GAS belt have been demonstrated to be important 179 determinants of selectivity in ASIC1a (48). Despite the presence of an ordered reentrant loop and 180 a narrower pore, we did not observe a change in the position of either TM2a/b or the GAS belt 181 residues between the resting and desensitized conformations. Additional studies of the open 182 state, perhaps under SMA isolation conditions, will be required to illuminate the structure of the 183 activated, ion conducting state of the channel. 184

Density features suggest TMD-membrane interactions 185
The reconstitution of membrane proteins into lipid nanodiscs is a well-established 186 technique in structural biochemistry that permits the study of sensitive membrane proteins 187 embedded in phospholipid bilayers (49, 50). While a reconstitution approach provides for a 188 controlled and defined lipid environment, the necessity of an initial detergent-based extraction 189 step may disrupt protein-lipid interactions integral to the structural integrity of transmembrane 190 segments. In contrast with nanodisc reconstitution, SMA copolymers extract membrane proteins 191 directly from the lipid bilayer, eschewing detergent entirely and permitting the study of 192 membrane proteins in the presence of endogenous lipids (51, 52) and, in principle, maintaining 193 the protein-lipid interactions that occur at the cell membrane (53). 194 In our 2.8 Å reconstruction of a desensitized ASIC1a, we observed multiple ordered 195 elongated densities situated in lipophilic channels along the TMD ( Figure 5A) that we suggest 196 may correspond to bound lipids. Separated into spatially distinct clusters ( Figure 5B), putative 197 lipid densities reside near the top of the membrane sandwiched between TM2a and TM1 helices 198 ( Figure 5C) and near the bottom of the membrane between TM1 and TM2b ( Figure 5D). 199 Our results suggest that the local lipid environment is important for maintaining the 200 architecture of the reentrant pre-TM1 residues in ASIC1a and thus the integrity of the lower pore 201 pathway. Therefore, the location of at least one cluster of putative lipid densities within a 202 lipophilic cleft immediately adjacent to reentrant loop residues and the GAS belt ( Figure 5D) is 203 intriguing, especially given that the cryo-EM structure of a full-length cASIC1a channel in an n-204 dodecyl-β-D-maltoside lacked ordered amino terminal residues (27). However, given the 205 resolution of our SMA-cASIC1a reconstructions, we are unable to assign this density to any 206 specific lipid. Future experiments are needed to determine the molecular composition of SMA-207 cASIC1a particles and to explore relevant interactions between ASICs and the plasma 208 membrane. 209

DISCUSSION 210
Here we present structures of chicken ASIC1a solubilized by SMA in high pH 'resting' 211 and low pH 'desensitized' conformations. While the conformation of both resting and 212 desensitized channels throughout the ECD faithfully mirrors those solved previously via 213 detergent-based methods, our structures demonstrate that amino terminal residues prior to TM1 214 form a reentrant loop that comprises the lower portion of the ion permeation pathway. In both 215 resting and desensitized structures, the conserved HG motif is situated within the reentrant loop, 216 immediately below the GAS belt TM2 domain swap, where it forms a constriction along the ion 217 permeation pathway and is stabilized by a complex network of inter and intra subunit 218 interactions. Finally, we detected elongated ordered densities within lipophilic channels of the 219 TMD, some of which are adjacent to the reentrant amino terminus, that may correspond to bound 220

lipids. 221
These results provide a structural basis for contributions of the amino terminus to both 222 ion permeation and proton-dependent gating of ASICs, reveal the location of the conserved HG 223 motif along the ion conduction pathway and expose a role for the plasma membrane in 224 maintaining the TMD architecture of ASICs. Given the structural similarities between ENaCs 225 and ASICs, as well as the highly conserved and functionally-important nature of the HG and 226 GAS belt residues, these results provide detailed structural information pertaining to a pair of 227 motifs central to gating and ion permeation and of possible therapeutic relevance to the entire 228 superfamily of ENaC/DEG ion channels. 229

METHODS 231
Expression and purification of cASIC1a channels: Recombinant full-length acid-sensing ion 232 channels (Gallus gallus) were expressed in HEK293S GnTI-cells and membrane fractions were 233 isolated as previously described (27) The Ni-NTA bead suspension was then transferred to a XK-16 column and subject to two 240 washes, first with three column volumes of TBS containing 10 mM imidazole and last with three 241 column volumes of TBS containing 30 mM imidazole. The SMA-cASIC1a protein was eluted 242 with TBS containing 250 mM imidazole, and peak fractions were pooled and concentrated to ~ 5 243 mg/ml. The His 8 EGFP tag was removed via thrombin digestion using a ratio of cASIC1a to 244 thrombin of 25:1, overnight at room temperature (RT). The following day, the SMA-cASIC1a 245 protein was purified via size-exclusion chromatography (Superose 6 10/300) using a mobile 246 buffer composed of TBS supplemented with 1 mM DTT. A single peak fraction was collected 247 and concentrated to ~ 1 mg/ml for cryo-EM sample preparation. 248 Cryo-EM of SMA-cASIC1a: Quantifoil holey carbon grids (R1.2/1.3 200 mesh Au) were glow 249 discharged for 1 minute at 15 mA, carbon side facing up. For structure determination of SMA-250 cASIC1a particles at high pH, purified protein at ~ 1 mg/ml was used immediately for grid 251 preparation. For structure determination at low pH, the pH of the sample was adjusted to 7.0 by 252 addition of MES, pH 6.0, following concentration of purified protein to ~ 1.0 mg/ml. A 4 μl 253 droplet of sample, applied to the carbon side of the grid, was blotted manually with pre-cooled 254 filter paper (Whatman, grade 1) and the grids were vitrified in ethane/propane mix using a 255 custom-built manual-plunge apparatus housed in a 4°C cold room with 60-70% relative 256

humidity. 257
Cryo-EM data acquisition for SMA-cASIC1a: For the resting channel structure at high 258 pH, data were collected on a Titan Krios cryo-electron microscope (ThermoFisher) operated at 259 300 keV. Images were recorded on a Gatan K3 camera positioned after an energy filter (20 eV 260 slit width) operating in super-resolution mode with a binned pixel size of 0.648 Å. Data were 261 collected with SerialEM (54) and dose-fractionated to 50 frames for a total exposure time of 2-3 262 s and a total dose of 40-50 e -Å -2 . 263 For the desensitized state structure at pH 7.0, data were recorded on a Titan Krios cryo-264 electron microscope operated at 300 kV and equipped with a spherical aberration corrector. 265 Images were recorded on a Gatan K2 Summit camera in super-resolution mode with a binned 266 pixel size of 1.096 Å. Data were acquired using Leginon (55) and dose-fractionated to 48 frames 267 at 0.15 s per frame for a total exposure time of 7.25 s and a total dose of 50 e -Å -2 . 268 Cryo-EM data processing for SMA-cASIC1a: Images were motion corrected using UCSF 269 MotionCor2 (56) and CTF estimation was performed using Gctf (57). Particles picked using 270 DoG Picker (58) were subjected to reference-free 2D classification in cryoSPARC V2 (59). 271 Following initial classification, an ab-initio model was generated in cryoSPARC V2 and used for 272 iterative rounds of 3D classification and refinement in cryoSPARC V2. For the pH 7.0 dataset, 273 per-particle CTF estimation was performed using Gctf. Final reconstructions for both datasets 274 were obtained via non-uniform refinement (C3 symmetry) in cryoSPARC V2.

AUTHOR INFORMATION 300
The authors declare no competing financial interests. Correspondence and requests for material 301 should be addressed to E.G. (gouauxe@ohsu.edu). 302

DATA AVAILABILITY 304
The data that support these findings are available from the corresponding author upon request. 305 The coordinates and associated cryo-EM map for the desensitized SMA-cASIC1a channel at pH