Cryo-EM structure of the zinc-activated channel (ZAC) in the Cys-loop receptor superfamily

Significance

Cys-loop receptors are a class of pentameric ligand-gated ion channels with a crucial role in the human body, especially in the central nervous system. They have been implicated in various diseases, particularly neurodegenerative diseases. Among them, the zinc-activated ion channel (ZAC) is the most recently discovered member of the human Cys-loop receptor family. Because of its broad expression pattern and its activation by extracellular Zn²⁺, ZAC has been proposed to function as a Zn²⁺ sensor in vivo. However, only the structure of ZAC remains missing in the Cys-loop receptor family. Our structural and functional analyses not only provide the structural information on the ZAC family but also provide unique mechanistic insights into the ion-conducting mechanism of ZAC.

Abstract

Cys-loop receptors are a large superfamily of pentameric ligand-gated ion channels with various physiological roles, especially in neurotransmission in the central nervous system. Among them, zinc-activated channel (ZAC) is a Zn²⁺-activated ion channel that is widely expressed in the human body and is conserved among eukaryotes. Due to its gating by extracellular Zn²⁺, ZAC has been considered a Zn²⁺ sensor, but it has undergone minimal structural and functional characterization since its molecular cloning. Among the families in the Cys-loop receptor superfamily, only the structure of ZAC has yet to be determined. Here, we determined the cryo-EM structure of ZAC in the apo state and performed structure-based mutation analyses. We identified a few residues in the extracellular domain whose mutations had a mild impact on Zn²⁺ sensitivity. The constriction site in the ion-conducting pore differs from the one in other Cys-loop receptor structures, and further mutational analysis identified a key residue that is important for ion selectivity. In summary, our work provides a structural framework for understanding the ion-conducting mechanism of ZAC.

Cys-loop receptors are a class of pentameric ligand-gated ion channels (LGICs) that play a crucial role in the body, particularly in the central nervous system, by mediating ion flux in response to the binding of extracellular ligands (1, 2). They are associated with various diseases, especially neurodegenerative diseases, including Alzheimer’s disease (3), Parkinson’s disease (4), and epilepsy (5). Human Cys-loop receptors include the excitatory nicotinic acetylcholine receptor (nAChR) (6) and 5-hydroxytryptamine 3 receptor (5-HT₃R) (7, 8) and the inhibitory γ-aminobutyric acid type A receptor (GABA_AR) (9) and glycine receptor (GlyR) (10). The zinc-activated ion channel (ZAC) family was more recently identified (11). Furthermore, the Cys-loop receptor family has bacterial and invertebrate members (12–14), including recently identified octopus and squid chemotactile receptors (15).

ZAC proteins are Zn²⁺-activated, nonselective cation channels that are widely distributed in vertebrates, including mammals (11, 16). In humans, ZAC is expressed in a variety of organs, including the brain, thyroid, pancreas, placenta, and prostate. Because of the wide range of expression pattern and its activation by extracellular Zn²⁺, ZAC has been proposed to function as a Zn²⁺ sensor in vivo (17). Zinc is the second-most abundant heavy metal after iron in the human body and is involved in various physiological functions; thus, it has also been implicated in neurological disorders and neurodegenerative diseases such as Alzheimer’s disease, epilepsy, and ischemic diseases (18). In particular, in the central nervous system, synaptic Zn²⁺ plays an important role in the modulation of neurotransmitter receptors, including the N-methyl-D-aspartic acid receptor (NMDAR), the α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid receptor (AMPAR), and the γ-aminobutyric acid receptor (19). However, the specific physiological function of ZAC has not been determined. With recent technological advances in structural biology, increasing structural information has become available for each family within the Cys-loop receptor superfamily over the last twenty years (20–27), with the exception of ZAC. Therefore, the overall architecture, Zn²⁺-dependent gating mechanism, and ion permeation mechanism of ZAC have remained unclear, hindering further understanding of the ZAC family.

In this work, we determined the cryo-EM structure of ZAC from Oryzias latipes, the Japanese rice fish, performed structure-based mutational analyses. We identified a few residues in the extracellular domain whose mutations had a mild impact on Zn²⁺ sensitivity. In addition, in our structure, the constriction site differs from those in other cation-conducting Cys-loop receptors. Further mutational analysis revealed a key residue involved in ion selectivity. Therefore, our work provides the structural framework to understand the ion-conducting mechanism of ZAC.

Results

Functional Characterization and Structure Determination.

Our previous expression screening of ZAC orthologs identified O. latipes OAC (OlZAC) as a suitable candidate for structural studies (28). To verify whether OlZAC also functions as a Zn²⁺-activated channel, we performed whole-cell patch-clamp recording of OlZAC. Patch-clamp recording of HEK293 cells expressing OlZAC showed that OlZAC can be activated by extracellular Zn²⁺ ion, with an EC₅₀ value of 0.35 ± 0.02 mM (SI Appendix, Fig. S1), which is comparable to the EC₅₀ value of human ZAC for Zn²⁺ ion (11). Notably, the Zn²⁺ concentration at the synaptic cleft is less likely to reach the millimolar level because of diffusion; instead, it can reach from the nanomolar to the micromolar level (29, 30). Furthermore, because human ZAC is activated by proton in addition to Zn²⁺ (17), we tested whether the extracellular proton (pH 4.0) could activate OlZAC, but unlike human ZAC, OlZAC was not activated by low pH (SI Appendix, Fig. S1). Consistent with this, in the previous functional characterization of mammalian ZAC proteins (17), some mammalian ZAC proteins are also not activated by low pH, whereas all functionally expressed mammalian ZAC proteins are activated by extracellular Zn²⁺.

We then performed single-particle cryo-EM analysis to determine the structure of OlZAC in the apo state (SI Appendix, Figs. S2–S5). The purified full-length OlZAC protein was reconstituted into either NAPol amphipol or nanodiscs for single-particle cryo-EM analysis. The cryo-EM data of the amphipol-reconstituted sample revealed significant orientation bias, predominantly in the top view (SI Appendix, Fig. S2A), whereas the cryo-EM data of the nanodisc-reconstituted sample included different orientations, including a side view (SI Appendix, Fig. S2B). Accordingly, we were unable to achieve satisfactory 3D reconstruction solely from the cryo-EM data of the amphipol-reconstituted sample (SI Appendix, Fig. S2A). In addition, the cryo-EM data of the nanodisc reconstituted alone also did not yield a high-resolution cryo-EM map (SI Appendix, Fig. S2B). Merging the data from the amphipol-reconstituted and nanodisc-reconstituted samples allowed us to overcome the inherent preferred orientation issues in the amphipol-reconstituted sample, resulting in a final EM map resolution of 3.4 Å (SI Appendix, Table S1 and Figs. S3 and S5). Furthermore, with the use of the 2D reference from the cryo-EM data of the nanodisc-reconstituted sample, which contains a side view, the cryo-EM data of the amphipol-reconstituted sample also yield a reasonable cryo-EM map at 3.6 Å (SI Appendix, Table S1 and Figs. S4 and S5). Notably, the structures from the merged cryo-EM data and the structure from the cryo-EM data in amphipols are essentially identical, with an RMSD value of 0.2 Å for Cα atoms (SI Appendix, Fig. S6), as in the other cases of cryo-EM data merging (31). Thus, we mainly utilize the structure from the merged cryo-EM data, which were determined at a higher resolution in this study.

Furthermore, to obtain a Zn²⁺-bound structure of OlZAC, we also attempted single-particle cryo-electron microscopy (cryo-EM) analysis of OlZAC in the presence of Zn²⁺ ions, but OlZAC tends to aggregate in the presence of Zn²⁺. Notably, our expression construct of OlZAC contains an octahistidine tag that is also capable of binding to Zn²⁺ ions, but the protein still tends to aggregate after the removal of the octahistidine tag.

The overall structure of the OlZAC channel adopts a bullet-like pentameric architecture (Fig. 1 A–C). Each subunit of OlZAC contains four transmembrane (TM) helices and an extracellular domain with an N-terminal helix and ten β-strands (Fig. 1D), exhibiting the common architecture of the Cys-loop receptor superfamily. In our structure, the membrane-associated (MA) helix preceding the M4 helix was resolved to form a helix bundle, whereas most of the post-M3 loop was disordered (Fig. 1D), as in other cation-conducting Cys-loop receptor structures. A search by Foldseek (32) also showed that the overall structure of the monomeric structure of OlZAC is similar to that of known Cys-loop receptors, particularly the mouse 5-HT₃ receptor structure (PDB ID:6Y1Z), which has the highest TM score of 0.6976.

Fig. 1.

Cryo-EM structure of OlZAC. (A) 3D reconstruction map of the full-length O. latipes ZAC at 3.4 Å resolution. The view is parallel to the membrane (side view). Individual subunits are depicted in different colors. The thick gray lines denote putative membrane limits. (B) Cartoon representation of the OlZAC structure based on the EM reconstruction map. The view is the same as the orientation shown in A (side view). (C) Top and bottom views of the OlZAC structural model from the extracellular (Upper panel) and intracellular (Lower panel) sides. (D) Single-subunit structure of OlZAC oriented as is the cyan subunit in A and rotated 90° about the pore axis; colored rainbow. The N terminus is blue, the C-terminus is red, and the middle colors vary. Secondary structures are labeled.

Zn²⁺-sensitive Region Candidates.

Close observation of the OlZAC structure indicated two possible candidates for the Zn²⁺-binding site (Fig. 2). The electrostatic surface potential of OlZAC showed that the canonical neurotransmitter binding site in the Cys-loop receptor, consisting of the loop A-G regions, is negatively charged (Fig. 2A), which is favorable for attracting Zn²⁺ ions. In addition, the subunit interface containing the Cys-loop region (Fig. 2B), which is closer to the membrane boundary, is also negatively charged (Fig. 2A). Notably, the residues at the Ca²⁺-binding site of human nACh alpha7 (Asp41, Asp43, Glu44, and Glu172) (33, 34) are not well conserved in OlZAC (Ser63, Asn65, Thr66, and Asp191, respectively) (SI Appendix, Fig. S8). Furthermore, Asp191 in OlZAC is not conserved among ZACs (SI Appendix, Fig. S7). Therefore, this site may not be an attractive site for the binding of cations, including Zn²⁺.

Fig. 2.

Putative Zn²⁺-binding pocket. (A) The electrostatic potential surface of OlZAC, contoured from −5 kT (red) to +5 kT (blue). The two black boxes indicate potential binding pockets for Zn²⁺ ions. (B) The putative zinc-binding pockets are the Loop C (Upper panel) and Cys-loop (Lower panel) regions. Glu148, Glu160, and Glu204 residues are shown as sticks, while the side chain of Glu148 is disordered. (C) Superimposition of the extracellular domain monomers of mouse 5HT3A (PDB ID: 6Y5A), and octopus CRT1 (PDB ID: 8EIS) onto that of OlZAC. (D) Sequence alignment of ZAC and other Cys-loop receptor subunits at the loop C region, including OlZAC, human ZAC (HsZAC, UniProt ID: Q401N2), mouse 5HT3A (Mm5HT3A, P23979), human GlyRa1 (HsGly_alpha1, P23415), human GABAA receptor alpha1 subunit (HsGABAA_alpha1, P14867), human GABAA receptor beta2 subunit (HsGABAA_beta2, P47870), human GABAA receptor rho1 subunit (HsGABAA_rho1, P24046), human nACh alpha7 (HsnACh_alpha7, P36544), human nACh alpha4 (HsnACh_alpha4, P43681), human nACh beta2 (HsnACh_beta2, P17787), and Octopus CRT1 (CRT1, A0A0L8FVQ9). These aberrations for the sequences were also used for the subsequent alignments. (E) Sequence alignment of ZAC and other Cys-loop receptor subunits in the Cys-loop region.

To investigate the importance of these regions in the Zn²⁺-dependent gating of OlZAC, we generated a FLAG-tagged OlZAC construct and alanine-substituted mutants (Glu148 and Glu160 in the Cys-loop and Glu204 in loop C) for whole-cell patch-clamp recording (Figs. 2 B–D and 3). We chose these residues because they are negatively charged residues in the regions of interest. The FLAG tag was inserted into the post-M3 loop because an insertion in this region is less likely to affect channel activity, as is the case for other Cys-loop receptors (35, 36). As expected, the FLAG-tagged OlZAC showed similar gating to that of the non-FLAG-tagged construct, with a comparable EC₅₀ value of 0.41 ± 0.02 mM for Zn²⁺ (Fig. 3). We employed this construct as a background construct for mutational analysis. In the loop C region, mutation of Glu204 to alanine increased the EC₅₀ for Zn²⁺ by threefold compared to that of the wild type (Fig. 3). In the Cys-loop region, the E148A mutant exhibited an approximately fivefold lower affinity for Zn²⁺, whereas the alanine mutation of Glu160 slightly decreased the EC₅₀ for Zn²⁺ (Fig. 3).

Fig. 3.

Mutational analysis of putative Zn²⁺-binding sites. (A) Representative Zn²⁺-evoked currents of FLAG-tagged OlZAC and its mutants at −60 mV. The values shown above the traces indicate the applied Zn²⁺ concentrations (mM). (B) Concentration–response curves of Zn²⁺ on OlZAC and its mutants (n = 3 for E148A-flag and E204A-flag, n = 4 for WT-flag and E160A-flag and n = 6 for D103A-flag). Peak current amplitudes were normalized to the maximum current amplitude in each construct. The data are presented as the mean ± SEM.

These results suggest that in addition to the loop C region in the canonical neurotransmitter binding site, the Cys-loop region is involved in the Zn²⁺ sensitivity of OlZAC. In other words, there might be multiple binding sites for Zn²⁺ ions at the subunit interface, which is distinct from the agonist-binding properties of other Cys-loop receptors. This notion of multiple Zn²⁺ binding sites for ZAC ions was originally proposed by the previous mutational analyses of human ZAC, in which several regions in loops A–F and the Cys loop were targeted (37).

However, the mutations tested in this study as well as the mutations of human ZAC tested in the previous study had only a mild effect on Zn²⁺ sensitivity and did not abolish it. Owing to the lack of a Zn²⁺-bound structure, it would be difficult to determine whether Zn²⁺ binds directly to these residues or whether these mutations allosterically affect the ability of Zn²⁺ to activate the channel. In particular, the Cys-loop is known to be important for communication between the extracellular domain and the TM domain in other Cys-loop receptor family members (1); thus, mutation of the Cys-loop may also alter channel function without direct binding to Zn²⁺. Overall, a Zn²⁺-bound structure would be required to unambiguously identify where Zn²⁺ ions bind to this receptor.

Notably, the loop C region of OlZAC is much shorter than those of other Cys-loop receptors and even shorter than that of the recently identified CRT1 chemotactic receptor (Fig. 2 C and D); moreover, its sequence is not highly conserved (Fig. 2D and SI Appendix, Figs. S7 and S8). These findings also highlight the unusual features of the Zn²⁺-dependent activation of ZAC proteins.

Pore Architecture.

In the apo-state OlZAC structure, both the extracellular and intracellular vestibule regions are wide open at the center of the extracellular domain, whereas the TM region provides the constriction site for ion conduction (Fig. 4 A–C). The TM pore consists of the M2 helix, and Gly250 (2’), Thr254 (6’), Leu257 (9’), Ser258 (10’), Met261 (13’), Asn264 (16’), Leu265 (17’), and Ser268 (20’) of the M2 helix contribute to pore formation (Fig. 4B). Met261 (13’) forms the narrowest part of the pore, with a pore radius of less than 2.76 Å, the hydration radius of the sodium ion (38) (Fig. 4 B and C).

Fig. 4.

Ion permeation pathway of OlZAC. (A and B) The ion permeation pathway of OlZAC calculated by HOLE (A) and a close-up view (B). For clarity, the cartoon representation is shown for only two subunits. The side chains of the residues lining the pore are shown in stick representation. (C) The pore radius of OlZAC, closed 5HT3A (PDB ID: 4PIR) and open 5HT3A (PDB ID: 6DG8) is plotted as a function of distance along the pore axis. The dashed line indicates the approximate radius of a hydrated Na⁺ ion. (D–G) MD simulations of the OlZAC structure. The plots of the Rms deviations (RMSDs) for Cα atoms (D) and the distances between the Cα of Met261 and the center of the pore (E) and between the CE atom (the tip carbon atom of the methionine side chain) of Met261 from the central axis of the pore (F). The pore center is defined as the center of five Cα atoms of Met261 from each subunit. The distance was defined as the average distance between the Cα of Met261 from each subunit and the center of the pore. (G) Snapshot structure from the MD simulations at 800 ns. The cross-sectional surface of OlZAC is shown, and water molecules within the transmembrane region are shown as spheres.

To be noted, the TM domain conformation of a Cys-loop receptor is known to depend on the membrane-mimicking environment in structure determination (39). Thus, to validate our structure, especially its pore conformation, we performed a molecular dynamics (MD) simulation of OlZAC embedded in a lipid environment (Fig. 4 D and E). The overall structure of OlZAC was stable throughout the MD simulation (Fig. 4E). The distance of the Cα atoms of Met261 (13’), which constricts the pore, from the central axis of the pore was also stable (Fig. 4E), supporting the stability of the pore conformation in our cryo-EM structure. Furthermore, the distance between the CE atoms (the tip carbon atoms of the methionine side chain) of Met261 (13’) from the central axis of the pore was also mostly stable during the simulation (Fig. 4F), but the pore was filled with water molecules (Fig. 4G), reflecting the potential flexibility of the Met261 side chain. Considering such observations in the MD simulation and the quality of the EM density map for the side chain of Met261 at the current resolution in our structure (SI Appendix, Fig. S5B), we cannot conclude whether the pore is closed or open in our cryo-EM structure.

We next compared the pore architecture of our OlZAC structure with those of other Cys-loop receptor structures in the apo state, particularly those with high structural similarity to OlZAC, by using Foldseek (32) (5-HT₃R and α7 nAChR) (Fig. 5). In the 5-HT₃R and α7 nAChR structures, the pore constriction region is located at the conserved Leu residue at the 9’ position (Leu260 in 5-HT₃R and Leu247 in α7 nAChR) (Fig. 5 B and C). In general, in the Cys-loop receptor superfamily, the Leu residue at the 9’ position of each protomer forms the “leucine” ring in the pentameric channel, which acts as a constriction site in the resting state (40, 41), and mutations at this site are known to severely affect channel gating in Cys-loop receptors (42–45). In contrast, the corresponding residue in OlZAC is Leu257 (9’) (Fig. 5F), while the constriction site in OlZAC is different (Met261, 13’) (Fig. 5A), again highlighting a noncanonical feature of our OlZAC structure.

Fig. 5.

Pore constriction of OlZAC. (A–C) Extracellular view of the pore constriction regions of OlZAC (A), 5HT3A (PDB ID: 4PIR) (B), and α7 nAChR (PDB ID: 7KOO) (C). (D and E) Superimposition of the transmembrane helices of 5HT3A (D) and α7 nAChR (E) onto that of OlZAC. For clarity, only the M2 helix is shown in cartoon representation. The side chains of the residues at the constriction site are shown in stick representation. (F) Sequence alignment of ZAC and other Cys-loop receptor subunits at the M2 helix.

Mutational Analysis in the Pore Region.

To investigate the functional roles of the pore-lining residues of OlZAC, we generated a series of mutants for whole-cell patch–clamp recording (G247E, T254A, L257A, S258A, M261A, L265A, S268A, and S268Q) (Fig. 6). We designed G247E and S268Q mutants of OlZAC because Gly247 (-1’) at the cytoplasmic entrance of the pore and Ser268 (20’) at the extracellular end of the pore in OlZAC are replaced by Glu and Gln residues, respectively, in human ZAC (Fig. 5F). In excitatory Cys-loop receptors, residues corresponding to Gly247 are highly conserved as Glu residues, whereas residues corresponding to Ser268 are conserved mainly as Glu residues and occasionally as Gln residues (46).

Fig. 6.

Mutational analysis of pore-lining residues. (A) Representative Zn²⁺-evoked currents of OlZAC and its mutants at −60 mV (1 mM Zn²⁺ was tested). (B) Current densities of Zn²⁺-evoked currents of OlZAC and its mutants (n = 17, 7, 6, 6, 7, 5, 5, 8, and 5 for WT, G247E, T254A, L257A, S258A, M261A, L265A, S268A, and S268Q, respectively). (C) The desensitization ratio (I_rem/I_max) of OlZAC and its mutants (n = 17, 7, 6, 7, 8, and 5 for the WT, G247E, L257A, S258A, S268A, and S268Q, respectively). (D) Representative I–V curves for OlZAC and G247E in asymmetric solutions (pipette: 150 mM NaCl; bath: 15 mM NaCl with glucose and mannitol). (E) P_Cl/P_Na of OlZAC and its mutants estimated from I–V curves (n = 5, 5, 5, 4, and 8 for the WT, G247E, S258A, S268A, and S268Q, respectively). (F) Representative I–V curves for OlZAC and G247E in asymmetric solutions (pipette: 150 mM NaCl; bath: 150 mM NMDG with glucose and mannitol). (G) P_NMDG/P_Na of OlZAC and its mutants estimated from I–V curves (n = 6, 7, 5, 5, and 5 for the WT, G247E, S258A, S268A, and S268Q, respectively). The data are presented as the mean ± SEM (ANOVA followed by post hoc test, *P < 0.05, **P < 0.01 vs. WT).

Among these mutants, T254A, M261A and L265A almost completely abolished Zn²⁺-dependent currents, and the S258A mutant exhibited a slightly lower current density, whereas there was no significant difference between the current density of the wild type and the other four mutants (G247E, L257A, S268A, and S268Q) (Fig. 6 A and B). Notably, Western blotting revealed that the mutants that almost completely abolished Zn²⁺-dependent currents (T254A, M261A and L265A) or resulted in a slightly smaller current (S258A) were also expressed at the cell surface, similar to the wild type (SI Appendix, Fig. S9).

We examined the ion selectivity of G247E, S258A, S268A, and S268Q, which still exhibit robust Zn²⁺-dependent currents. Among the mutants with robust Zn²⁺-dependent currents, we were not able to examine the ion selectivity of the L257A mutant because of its rapid desensitization as compared to WT (Fig. 6C). One possible explanation is that Leu257 may be important for stabilizing the open state of the channel. The alanine-substituted mutation at Leu257 may destabilize the open state, making it easier for the channel to close.

Among these mutants, the G247E mutant presented altered ion selectivity compared with that of the wild type, whereas other mutants presented little change in ion selectivity (Fig. 6 D–G). Under NaCl conditions, the reversal potential of the G247E mutant was negatively shifted by ~60 mV (Fig. 6D), and the estimated P_Cl/P_Na value was significantly decreased (Fig. 6E). In addition, in reversal potential experiments under bi-ionic NMDG_out/Na_in conditions (Fig. 6F), P_NMGD/P_Na decreased (Fig. 6G). These results indicate that Gly247 (-1’) is involved in the ion selectivity of OlZAC.

Discussion

In this work, we showed that the ZAC protein from Japanese rice fish (OlZAC) functions as a Zn²⁺-activated channel (Fig. 1), like human ZAC, and we determined the structure of OlZAC in the apo state by single-particle cryo-EM analysis (Fig. 2). Structure-based mutational analysis via whole-cell patch–clamp recording identified a few residues in the extracellular domain whose mutations had a mild impact on Zn²⁺ sensitivity (Fig. 3). However, the exact locations of the Zn²⁺-binding sites are still unknown owing to the lack of a Zn²⁺-bound structure. The narrowest portion in our structure is located at Met261 (13’), which is distinct from the pore constriction site in other Cys-loop receptors (Figs. 4 and 5). Further mutational analysis of the pore region revealed that Gly247 (-1’) is important for cation selectivity (Fig. 6).

Consistent with previous mutational analyses of human ZACs (37), our structure and associated mutational analysis of OlZAC via patch–clamp recording also suggested that multiple sites may be involved in Zn²⁺ sensitivity (Fig. 3), in contrast to other canonical Cys-loop receptors that have a single agonist binding site per protomer. In previous mutational analyses of human ZAC (37), multiple mutations that weakly affect Zn²⁺ sensitivity were identified; here, we discuss those previous results on the basis of our structure (SI Appendix, Fig. S10). Among such mutants, the EC₅₀ value of the H79A mutant for Zn²⁺ was ~2-fold greater than that of WT ZAC (37). His79 in human ZAC corresponds to Tyr93 in OlZAC and is not well conserved among ZAC proteins (SI Appendix, Fig. S7), as in the case of other mutated residues in human ZAC. In addition, Tyr93 is surrounded by hydrophobic residues in the OlZAC structure (SI Appendix, Fig. S10). Similarly, the E130A mutant of human ZAC also weakly affects Zn²⁺ sensitivity (37). While Glu130 in humans corresponds to Glu144 in OlZAC, it is not well conserved among ZACs (SI Appendix, Fig. S7). Consistently, the environment surrounding Glu144 also lacks other residues to attract Zn²⁺ and thus is seemingly not suitable for Zn²⁺ (SI Appendix, Fig. S10). Considering these situations, these residues, Tyr93 and Glu144, might be less likely involved in Zn²⁺ binding in OlZAC. On the other hand, Glu89 in human ZAC, whose alanine substitution also weakly altered Zn²⁺ sensitivity, is highly conserved among ZAC proteins (SI Appendix, Fig. S7). The corresponding residue Asp103 in OlZAC is located in favorable regions, clustering the hydrophilic residues nearby in the OlZAC structure, unlike Tyr93 and Glu144 (SI Appendix, Fig. S10). Thus, Asp103 in OlZAC might contribute to a potential consensus Zn²⁺ binding site between human ZAC and OlZAC. Our mutational analysis consistently revealed that the mutation of Asp103 to alanine increased the EC₅₀ for Zn²⁺ by threefold compared with that of the wild type in OlZAC (Fig. 3).

It should also be noted that the effect of these mutations is relatively weak (Fig. 3), and the amino acid sequence of these regions, including loop C, is not highly conserved among ZAC proteins (Fig. 2D and SI Appendix, Fig. S7). Consistent with these findings, in previous mutational analyses of human ZAC that investigated the Zn²⁺ binding sites, none of the mutations severely affected or abolished the Zn²⁺ sensitivity of human ZAC, even though several regions in loops A–F and the Cys loop were targeted for mutational analysis (37). Thus, these suggest the possibility that Zn²⁺ may not be a physiological ligand for ZAC.

To further support this notion, the affinity of ZAC for Zn²⁺ is quite low. Our patch-clamp recording of OlZAC showed an EC₅₀ value of 0.35 ± 0.02 mM for Zn²⁺ ion (SI Appendix, Fig. S1), which is comparable to that of human ZAC (11). However, these EC₅₀ values may be too high for Zn²⁺ ions to bind under physiological conditions. Although the Zn²⁺ concentration at synaptic vesicles can reach or exceed 1 mM (47), the Zn²⁺ concentration at the synaptic cleft is less likely to reach the millimolar level due to diffusion; instead, it can reach from the nanomolar to the micromolar level (29, 30). Consistently, other ligand-gated ion channels, such as GlyR and NMDAR can be modulated by Zn²⁺ at nanomolar to micromolar concentrations (10, 48, 49). In contrast, recent electrophysiological recordings of ZAC proteins from different mammalian species also revealed a weak affinity for Zn²⁺ ion (17), as we observed for fish ZAC in this study. Overall, these might raise the possibility that additional cofactors enhance the Zn²⁺ affinity of ZAC proteins or that Zn²⁺ might not be a physiological ligand for ZAC. Our OlZAC structure could be useful for future in silico investigations of such cofactors or other ligands.

Our OlZAC structure and associated mutational analysis also revealed ion selectivity features in the transmembrane pore (Figs. 4 and 6). Human ZAC is highly cation selective and does not permeate NMDG (11), whereas OlZAC is not cation selective and can permeate NMDG. Gly247 is replaced by glutamate in human ZACs, and our mutational analysis revealed that the glutamate substitution at Gly (-1’) decreased P_Cl/P_Na and P_NMGD/P_Na (Fig. 6). Thus, our results indicate that this residue is involved in the different ion selectivity properties between OlZAC and human ZAC. Consistently, the glutamate residue in 5HT3A and α7 nAChR corresponding to Gly247 (-1’) is implicated in their cation selectivity (50, 51). These mutations convert these cationic channels to anionic channels. Thus, these results suggest that the position of Gly247 (-1’) is important for the cation selectivity of ZAC, as in the case of other cationic Cys-loop receptors, such as 5HT3A and α7 nAChR (52).

In summary, our work provides the structural framework of the ZAC family, which will facilitate future mechanistic and functional investigations of the ZAC family.

Methods

Expression and Purification of OlZAC.

The O. latipes OAC (OlZAC, GI: 765127633) gene containing ALFA and octahistidine tags at the C-terminus was synthesized by Genewiz (Suzhou, China) and was then subcloned and inserted into the pEG Bacmam vector for expression in HEK293S GnTI^- cells using a baculovirus-mediated gene transduction system (28, 53, 54). HEK293S GnTI^- cells were grown to a density of 2.5 × 10⁶ ml⁻¹ and infected with 1% (v:v) P2 BacMam virus. After 16 h of culture at 37 °C, 10 mM sodium butyrate was added, and the cell culture temperature was maintained at 37 °C for another 72 h. Then, the cells were harvested and washed with TBS buffer [50 mM Tris-HCl (pH 8.0) and 150 mM NaCl]. All purification steps were performed at 4 °C. Cells were lysed by sonication with protease inhibitors (1 mM PMSF, 5.2 μg/ml aprotinin, 2 μg/ml leupeptin, and 1.4 μg/ml pepstatin A). The membrane fractions were collected by ultracentrifugation (200,000×g, 60 min). The membrane was solubilized in buffer A [20 mM HEPES (pH 7.5), 150 mM NaCl] containing 2% (w/v) 2% n-dodecyl-beta-d-maltopyranoside (DDM) (Anatrace), 0.4% cholesteryl hemisuccinate (CHS) (Sigma Aldrich), and protease inhibitors (1 mM PMSF, 5.2 μg/ml aprotinin, 2 μg/ml leupeptin, and 1.4 μg/ml pepstatin A) for 2 h. The cell debris was removed by ultracentrifugation (200,000×g, 60 min). The supernatant was loaded onto TALON resin (Takara, Japan) equilibrated with buffer A containing 0.01% (w/v) lauryl maltose neopentyl glycol (LMNG) (Anatrace) and 10 mM imidazole and then washed with the same buffer. The protein was eluted with buffer A containing 0.01% (w/v) LMNG and 300 mM imidazole. The eluted protein was loaded on a Superdex 200 Increase 10/300 GL column (Cytiva) equilibrated with buffer A containing 0.01% (w:v) LMNG for size exclusion chromatography (SEC). The main peak fractions were pooled and concentrated to ~1 mg/ml using an Amicon Ultra 100 K filter (Merck Millipore). The SEC and FSEC profiles and the SD–PAGE gel after SEC were reported previously (28).

Amphipol Reconstitution.

All steps were performed at 4 °C. Purified OlZAC was mixed with NAPol amphipol (Anatrace) at a mass ratio of 1:20 and incubated for 16 h. The detergent was removed with Bio-Beads SM-2 (Bio-Rad) for 4 h, and the beads were subsequently removed by using a disposable Poly-Prep column. The eluent was applied to a Superdex 200 10/300 GL column equilibrated with buffer A, and the main fractions containing the amphipol-reconstituted OlZAC were pooled and concentrated to ~3 mg/ml using an Amicon Ultra 100 K filter for electron microscopic analysis.

Nanodisc Reconstitution.

OlZAC was mixed with MSP2N2 and soybean polar lipid extract (Avanti) at a molar ratio of 1:3:180 for 1 h at 4 °C. The sample was mixed with Bio-Beads SM-2 and rotated gently for 4 h, after which the Bio-Beads were removed by using a disposable Poly-Prep column. The eluents were mixed with Ni-NTA resin (QIAGEN) and rotated gently for 1 h. The Ni-NTA resin was washed with buffer A containing 30 mM imidazole to remove excess MSP2N2. The nanodisc-reconstituted OlZAC protein was eluted with buffer A containing 300 mM imidazole and loaded onto a Superose 6 10/300 size-exclusion chromatography (SEC) column (Cytiva) equilibrated with buffer A. The SEC fractions corresponding to the nanodisc-reconstituted OlZAC were collected and concentrated to ~2.0 mg/ml with an Amicon Ultra 100 K filter.

EM Data Acquisition and Analysis.

A 2.5 μl sample of OlZAC was applied to a glow-discharged holey carbon-film grid (Quantifoil, Au 1.2/1.3, 300 mesh) blotted with a Vitrobot (Thermo Fisher Scientific) system using a 1 to 4 s blotting time with 100% humidity at 9 °C, after which the sample was plunge-frozen in liquid ethane. Cryo-EM data collection was performed using a 300 kV Titan Krios microscope (Thermo Fisher Scientific) equipped with a K3 direct electron detector (Gatan, Inc.). The specimen stage temperature was maintained at 80 K. Images were recorded by beam-image shift data collection methods (55) under superresolution mode with a pixel size of 0.41 Å (a physical pixel size of 0.82 Å), a magnification of 29,000, and a defocus ranging from −1.5 µm to −2.3 µm. The dose rate was 20 e⁻s^–1, and each movie was 1.758 s long and was dose-fractioned into 40 frames with an exposure of 1.3 e⁻Å^–2 for each frame. The cryo-EM data are summarized in SI Appendix, Table S1.

Image processing.

A total of 14,105 movies of the amphipol-reconstituted OlZAC and 6,680 movies of the nanodisc-reconstituted OlZAC were collected and merged for data processing. Motion correction and further image processing were performed using RELION-3 (56). The movies were motion-corrected and binned with 5 × 5 patches, producing summed and dose-weighted micrographs with a pixel size of 0.82 Å. Contrast transfer function (CTF) parameters were estimated by CTFFIND 4.1 (57). A total of 7,709,521 particles were automatically picked and extracted with a box size of 256 × 256 pixels. After several rounds of 2D classification, 2,874,167 particles were selected for 3D classification, which was performed with C1 symmetry and then with C5 symmetry. Finally, 49,759 particles were recentered and extracted for 3D refinement with C5 symmetry imposed. After CTF refinement and Bayesian polishing, the final map reached a resolution of 3.4 Å. The final 3D reconstruction was calculated from 2 × 2 binned images (0.82 Å). The final resolution was estimated using the Fourier shell correlation (FSC) = 0.143 criterion on the corrected FSC curves, in which the influence of the mask was removed. The local resolution was estimated using RELION. The workflow for image processing and for 3D reconstruction and angular distribution plotting are shown in SI Appendix, Fig. S3.

Processing of the cryo-EM data of the amphipol-reconstituted sample with a 2D reference generated from manual picking was aborted at 2D classification (SI Appendix, Fig S2A). Processing of the nanodisc-reconstituted sample was aborted at 3D classification (SI Appendix, Fig. S2B). When the cryo-EM data of the amphipol-reconstituted sample were processed with the 2D reference from the nanodisc-reconstituted sample, a total of 5,832,776 particles were autopicked. After several rounds of 2D classification, 2,211,319 particles were selected for 3D classification. Finally, 54,435 particles were used for the final refinement, which reached a resolution of 3.6 Å. Its workflow for image processing and for 3D reconstruction and angular distribution plotting are shown in SI Appendix, Fig. S4.

Model Building and Refinement.

The initial OlZAC model was manually built starting from the homology model generated by SWISS-MODEL (58). Manual model building was performed using Coot (59). Real-space refinement was performed using PHENIX software (60). The final atomic model included residues 25 to 315 and 350 to 406 for each protomer. The figures showing the structures were generated using PyMol (https://pymol.org/) and UCSF Chimera X (61). The ion permeation pathway shown in Fig. 4 was analyzed by HOLE2 (62). The sequence alignment was created by Clustal Omega (63) and ESPript 3.0 (64).

Electrophysiology.

Whole-cell patch-clamp recording was performed at 24 to 48 h posttransfection at room temperature (25 ± 2 °C) (65–67). The patch pipettes were pulled from glass capillaries by a two-stage puller (PC-100, Narishige, Japan) and filled with pipette solution containing 120 mM KCl, 30 mM NaCl, 0.5 mM CaCl₂, 1 mM MgCl₂, 10 mM HEPES, and 5 mM EGTA. The pipette resistance (65) ranged from 3 to 5 megaohms. The bath solution was composed of 150 mM NaCl, 5 mM KCl, 10 mM glucose, 2 mM CaCl₂, 10 mM HEPES, and 1 mM MgCl₂, and the pH was adjusted to 7.35 to 7.40. The data were acquired using an Axon 200B amplifier (Molecular Devices) in combination with a Digidata 1440A digitizer (Molecular Devices) at a sampling rate of 10 kHz and filtered at 2 kHz. In the gap-free protocol, the membrane potential was clamped at −60 mV. Zn²⁺ was administered through a Y-tube delivery system. In the ionic selectivity experiments, the patch pipette was filled with a solution containing 150 mM NaCl and 10 mM HEPES. The bath solution used to determine the P_Cl/P_Na ratio consisted of 15 mM NaCl, 10 mM glucose, 10 mM HEPES, and 255 mM mannitol, and the bath solution used to determine the P_NMDG/P_Na ratio consisted of 150 mM NMDG, 10 mM glucose, 10 mM HEPES, and 10 mM mannitol. The reversal potential was determined by applying a voltage ramp and identifying the voltage at which the current was zero on the basis of the current–voltage relationship. The permeability ratios were then calculated based on the change in reversal potential using the simplified Goldman–Hodgkin–Katz (GHK) equation for monovalent permeant ions (68).

Molecular Dynamics (MD) Simulation.

The MD simulation was performed using the Desmond software package (version 2.3) (67, 69). The OlZAC protein was embedded in a lipid bilayer composed of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine (POPC) according to the OPM server (https://opm.phar.umich.edu/). The system was neutralized by the addition of 20 Na⁺ ions to maintain overall charge neutrality. The OPLS3e force field was employed to describe the parameters of the protein and lipids, while the SPC model was used for water molecules (70). Prior to the simulation, the system underwent a series of equilibration steps to relax the initial configuration: 1) a 100 ps simulation in the NVT ensemble with Brownian kinetics at a temperature of 10 K, with solute heavy atoms constrained; 2) a 12 ps simulation in the NVT ensemble using a Berendsen thermostat at a temperature of 10 K, employing small time steps and with solute heavy atoms constrained; 3) a 12 ps simulation in the NPT ensemble using a Berendsen thermostat and barostat at 10 K and 1 atm, with solute heavy atoms constrained; 4) a 12 ps simulation in the NPT ensemble using a Berendsen thermostat and barostat at 300 K and 1 atm, with solute heavy atoms constrained; and 5) a 24 ps simulation in the NPT ensemble using a Berendsen thermostat and barostat at 300 K and 1 atm without any constraints. Following equilibration, each MD simulation was performed for a duration of 800 ns. Long-range electrostatic interactions were computed using the smooth particle mesh Ewald method. The trajectory recording interval was set to 200 ps, and other default parameters of DESMOND were used during the MD simulation runs. The simulation was carried out on a DELL T7920 graphics workstation equipped with an NVIDIA Tesla K20C-GPU. The preparation, analysis, and visualization tasks were performed on a 12-CPU CORE DELL T3610 graphics workstation.

Western Blotting.

HEK293 cells were transfected with 8 μg of OlZAC or a mutant plasmid and cultured for 24 h (66). Surface biotinylation was performed by incubating the cells with sulfo-NHS-LC-biotin dissolved in PBS (pH 8.0). The reaction was stopped by adding glycine to the PBS. Cell lysis was carried out using RIPA lysis buffer supplemented with cocktail protease inhibitors. The cell lysate was collected and centrifuged, and the supernatant was collected. Twenty percent of the volume of the supernatant was diluted with SDS loading buffer to obtain the total protein fraction. The remaining lysate was then adsorbed to NeutrAvidin agarose resin by incubation at 4 °C. The resin was washed three times with chilled PBS and diluted with SDS loading buffer to obtain the surface protein fraction. The protein samples were denatured by heat treatment. Equal volumes of protein were loaded onto an SD–PAGE gel, which included a molecular weight marker for size estimation. Protein transfer from the gel to a PVDF membrane was performed using a wet transfer system. The PVDF membrane was then blocked with 5% nonfat milk and incubated with anti-FLAG, anti-GAPDH, or anti-Na⁺/K⁺ ATPase antibody overnight at 4 °C. Subsequently, a secondary antibody conjugated to HPR was used for a 1-h incubation at room temperature. The protein bands on the PVDF membrane were visualized using chemiluminescent/fluorescent detection systems, and images were captured with a highly sensitive camera. Analysis of band intensity and quantification of protein expression were performed with ImageJ software (71). The analysis of protein expression was repeated for at least three independent experiments. All the western blotting images are deposited in Mendeley Data (https://doi.org/10.17632/zpxrts59g2.1).

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

PDB and cryo-EM map files, current traces, numeric data, WB images data have been deposited in Protein Data Bank (PDB), Electron Microscopy Data Bank (EMDB), Mendeley Data (PDB: 8WGE, 8ZTS EMDB: EMD-37511, EMD-60471 Mendeley Data: doi: https://doi.org/10.17632/zpxrts59g2.1) (72–76).