Nucleotide- and metalloid-driven conformational changes in the arsenite efflux ATPase ArsA

Significance

Arsenic toxicity arises from the ability of trivalent arsenite to tightly bind thiol groups, thus disrupting cellular processes. Many bacteria can detoxify arsenite via efflux through the membrane transporter ArsB. In some species, a cytoplasmic ATPase, ArsA, enhances the efficiency of efflux by utilizing nucleotide binding and hydrolysis to capture intracellular arsenite for transfer to ArsB. Despite its role in arsenite detoxification, the mechanism by which ArsA facilitates these molecular events remains unclear. Using cryogenic electron microscopy, we characterized key nucleotide-dependent conformational changes in ArsA that are crucial for arsenite capture and removal. These findings enhance our understanding of a widespread resistance mechanism in bacteria that is broadly impactful for arsenic bioremediation.

Abstract

Arsenite (As^III) is toxic to all organisms due to its ability to tightly bind exposed thiols within cells. An important As^III resistance mechanism in prokaryotes involves proteins encoded by the ars operon. A central component of the ars operon in many bacteria is the cytoplasmic ATPase, ArsA, which orchestrates a series of nucleotide-dependent handoffs, starting with the capture of As^III by the ArsD metallochaperone and culminating in its removal from the cell by the ArsB efflux pump. Although the mechanism of ArsA has been widely studied, the molecular details of how nucleotide hydrolysis modulates these events remain unclear. ArsA is an archetypal member of the intradimeric Walker A (IWA) family of ATPases, implicated in a diversity of complex biological functions. Conformational changes typical of IWA ATPases have been postulated to drive these molecular events but have not been demonstrated. We report cryogenic electron microscopy (cryo-EM) structures of ArsA in MgADP-bound and MgATP-bound open states, as well as a distinct closed MgATP-bound state liganded to As^III. X-ray absorption spectroscopy (XAS) confirmed three-coordinate binding of As^III to the conserved cysteines at the metalloid-binding site of the closed state. Coupled with biochemical characterization, our cryo-EM structures reveal key conformational changes in the ArsA catalytic cycle consistent with other IWA ATPases and provide the structural basis for allosteric activation of nucleotide hydrolysis by As^III. This work establishes how the nucleotide state of ArsA transiently creates a high-affinity binding site that can sequester metalloid within the cell, followed by a nucleotide-driven handoff to ArsB for efflux.

P-loop NTPases, representing one of the most prevalent nucleotide-binding folds in nature, catalyze nucleotide triphosphate (NTP) hydrolysis to facilitate a broad range of biological functions, ranging from transcription regulation to nitrogen fixation (1, 2). One distinct lineage in the superfamily is the “intradimeric Walker A” (IWA) ATPases, which are characterized by a unique consensus P-loop sequence [X-K-X-X-X-G-K-(T/S)], formerly referred to as the “deviant Walker A” ATPase family (3, 4). These homodimeric ATPases couple ATP binding and hydrolysis to the transfer of a cargo (or substrate) to a partner protein involved in a distinct biological function. The conserved Lys residue at the N terminus of the Walker A motif, referred to as the “IWA lysine,” is critical for triggering nucleotide hydrolysis in each nucleotide binding site across the dimer interface (5, 6), much in the same way as the “arginine finger” motif of the G-protein-specific GTPase-activating protein (7). IWA ATPases mediate diverse cellular functions and have been mechanistically studied to understand their respective biological processes, including, but not limited to, the tail-anchored (TA) membrane protein targeting factor Get3, the reductase component of nitrogenase NifH, and the spatial regulation of bacterial cell division protein MinD (4).

A distinct member of the IWA family is the cytoplasmic ATPase ArsA, which is involved in the bacterial arsenic detoxification machinery. Trivalent arsenite (As^III) induces toxicity in organisms across domains of life by tightly binding to thiols of cysteine residues, resulting in inactivation of thiol-containing molecules, including lipoic acid and enzymes, and inducing protein misfolding and aggregation (8, 9). As such, arsenic toxicity is a widespread global health concern, and understanding the molecular mechanisms of detoxification has garnered broad interest due to their potential for arsenic bioremediation of contaminated environments (10, 11). In bacteria and some archaea, resistance to As^III is encoded in the ars operon, which involves the membrane efflux of As^III by the integral membrane protein ArsB (12, 13). While diverse combinations of ars genes are found across species, ArsB is most commonly found in association with genes for a DNA repressor, ArsR, which regulates the operon, and an aerobic reductase, ArsC, that reduces pentavalent As^V to As^III. In many species, the operon additionally encodes the ArsA ATPase and the metallochaperone ArsD. Under arsenic stress, ArsD sequesters As^III from the cytoplasm and transfers the toxic metalloid to ArsA in a nucleotide-dependent manner (14, 15). Subsequently, ArsA transfers As^III to ArsB by coupling metalloid transport to ATP hydrolysis. Notably, in the presence of ArsA, ArsB is considerably more efficient at exporting As^III (16) and has been collectively referred to as the “rsAB pump” (17, 18). This mode of resistance enables the bacteria to survive a wide range of environmentally contaminating As^III concentrations (14, 16).

In contrast to other homodimeric IWA ATPases, ArsA is the only known member of the family that is pseudodimeric, comprising two homologous nucleotide-binding domains covalently linked by a flexible ~25-residue linker (18, 19) (SI Appendix, Fig. S1A). The metalloid-binding site, composed of three conserved cysteine residues (20), is located at the pseudodimer interface and can bind As^III or antimonite (Sb^III). Both domains of ArsA, referred to as the N- and C-domains, can support the binding and hydrolysis of ATP; however, the N-domain does so at a significantly higher rate than the C-domain (21). Despite considerable work on understanding the catalytic mechanism of ArsA, the molecular details of its nucleotide-dependent mechanism, especially in the context of the broader IWA family, remain to be understood.

Nucleotide-dependent conformational changes around the dimer interface are a conserved mechanistic feature among IWA ATPases (4). The conformational landscapes of Get3 and NifH have been extensively characterized. The catalytic cycles of both ATPases involve conformational switching of the dimer between a “closed” state that is catalytically competent for ATP hydrolysis and an “open” state that is ATPase inactive (5, 6, 22–25). The open conformation is reported for either the apo or posthydrolysis MgADP-bound states and is characterized by P-loops that are far apart (24, 26). In contrast, the closed state is reported for the pre- or midhydrolysis states where the two P-loops are adjacent (5, 6), mediating the position of the IWA lysines across the dimer interface. Two additional loop motifs, Switch I and II (SI Appendix, Fig. S1 A and B), are conserved across IWA ATPases and adopt discrete conformations depending on the nucleotide state. These nucleotide-dependent conformational transitions are critical for the function of IWA ATPases, as ATP binding and hydrolysis are functionally coupled to the targeting of a substrate specific to each enzyme (4).

An unresolved question concerns how nucleotide-dependent conformational transitions during the ArsA catalytic cycle facilitate As^III capture and efflux. Biochemical and spectroscopic characterizations of ArsA suggest that the ATPase undergoes a series of conformational changes in the presence of nucleotide and metalloid (27, 28). X-ray crystal structures of Escherichia coli ArsA (EcArsA) in multiple nucleotide-bound states reveal only subtle conformational changes between the various structures with a rmsd of 0.38 Å over 536 residues (SI Appendix, Fig. S1C) (14, 29), supporting that crystallization has likely restricted the range of motion in the protein. Kinetic assays have shown that the metalloid substrate activates ArsA by stimulating ATP hydrolysis, which is the proposed rate-limiting step of the enzyme (27). Based on biochemical studies with Sb^III, the metalloid substrate binds at a single high-affinity site in ArsA (29). Of the three conserved cysteines, Cys172, located on a mobile loop between helices 7 and 8 of the N-domain of EcArsA, has been proposed to control the metalloid affinity of this site (29, 30). As the crystal structures are not consistent with the results above, the mechanism of metalloid binding and ArsA activation requires further investigation.

In this article, we report single-particle cryogenic electron microscopy (cryo-EM) structures of ArsA from the thermotolerant and acidophilic bacterium Leptospirillum ferriphilum, in nucleotide-bound open and closed conformations. In the MgADP-bound open state, the ArsA pseudodimer adopts the conformation of the previously reported ArsA crystal structures, all representing an ATPase inactive open state with low affinity for As^III. MgATP alone is unable to stabilize a major conformational change from this open state. In the presence of MgATP and As^III, a distinct closed conformation with high As^III affinity is adopted that is analogous to the closed conformations observed for other IWA ATPases. These structures provide views of the large-scale conformational transitions in the ArsA catalytic cycle that were not revealed in the crystal structures. Combined with X-ray absorption spectroscopy (XAS) analysis and supporting biochemical studies, our cryo-EM structures reveal the As^III activation mechanism of ATP hydrolysis in ArsA. Altogether, this work elucidates the nucleotide- and metalloid-dependent mechanism that enables a series of handoffs of toxic metalloids between ArsA and its partner proteins ArsD and ArsB by modulating metalloid affinity to efficiently carry out efflux, thus placing the protein within the mechanistic landscape of the biologically important IWA ATPase family.

Results

ArsA ATPase from L. ferriphilum Strain ML-04.

Following a bioinformatic analysis of ars operons resembling the well-studied E. coli R773 plasmid ars operon (31), we identified an ArsA homolog from L. ferriphilum strain ML-04 (LfArsA), an acidophilic and moderately thermophilic bacterial species, where some strains are found in arsenic-rich acid mine drainage ecosystems (SI Appendix, Fig. S2A) (32). The ArsA gene has been shown to express under arsenic stress conditions in this strain, and the operon confers arsenic resistance in closely related L. ferriphilum strains (32–34). LfArsA has 69% sequence identity to EcArsA, retaining all the characteristic features of ArsA (SI Appendix, Fig. S1B). We cloned, overexpressed, and purified LfArsA with a C-terminal 6x-His tag. Purified LfArsA hydrolyzes ATP in vitro with a pseudo-first-order rate constant of 0.5 ± 0.1 min⁻¹, and saturating concentrations of As^III stimulate the ATPase activity ~15-fold above the basal activity (SI Appendix, Fig. S2B, black plot). To elucidate the conformational landscape of the ArsA catalytic cycle, we sought to determine nucleotide-bound LfArsA structures both in the presence and absence of its substrate As^III using cryo-EM.

Structure of the open Conformation of ArsA Bound to MgADP.

To determine the conformation of ArsA in an MgADP and As^III-bound state, ArsA was incubated with 2 mM MgCl₂, 2 mM ADP, and 5 mM sodium arsenite, and grids were prepared for single-particle analysis. From a dataset of 4,732 movies collected on a 300 kV Titan Krios Transmission Electron Microscope (TEM) and several rounds of 2D and 3D classification (SI Appendix, Fig. S3), we obtained a 103,258-particle reconstruction of the ArsA·MgADP state (Fig. 1A). The final Coulomb potential map was refined to a 3.4 Å gold-standard Fourier shell correlation (FSC) resolution and sharpened using a uniform B-factor of −165 Ų to better visualize high-resolution features. AlphaFold2 (35) predictions for the N- and C-domains of LfArsA were docked into the sharpened map, followed by model building and refinement to obtain the ArsA·MgADP structure (Fig. 1B). The average Q-score for the structure was 0.64 (SI Appendix, Fig. S4A), indicating good resolvability of the cryo-EM map (36). About 95% of the protein chain was modeled into the map, with missing residues including part of the linker region between the N- and C-domains (residues 295 to 307), residues 167 to 168, and residues 464 to 476 (Fig. 1B). The N (residues 1 to 298) and C (residues 315 to 587) domains are pseudosymmetric with an r.m.s.d. of 1.0 Å over 139 residues (Fig. 1C). At both nucleotide-binding sites, ADP is coordinated to a Mg²⁺ ion (Fig. 1D). The catalytic Switch I aspartates—Asp46 (N) and Asp364 (C)—are oriented away from the Mg²⁺ ion (>4.5 Å). The Switch II aspartates—Asp143 (N) and Asp447 (C)—interact with the Mg²⁺ ion within 4 Å and form hydrogen bonds with Thr23 (N) and Thr341 (C), respectively. Notably, the P-loops are sufficiently separated such that the catalytic IWA lysines (Lys17 and Lys335) are oriented away from the ADP bound to the opposing domain, positioning the terminal amino groups of the lysines over 8 Å away from the nucleotide phosphate groups (Fig. 1E).

Fig. 1.

Cryo-EM structures of LfArsA in the nucleotide-bound open conformation. (A) Two views of the sharpened cryo-EM map colored based on local resolution (3.4 Å overall) for LfArsA·MgADP open state (ADP-open) in the presence of As^III. (B) Cartoon representation of the ADP-open state colored in viridis. Unmodeled regions are indicated as dotted lines. (C) Left, surface representation of the ArsA pseudodimer highlighting the N- (blue) and C- (green) domains. Right, cartoon representation of the overlay of the two domains based on the pseudo-2-fold symmetry. (D) Nucleotide-binding sites of the ADP-open state showing the Coulomb potential map around the MgADP shown as sticks. Switch I & II Asp residues of both N- and C-domains are shown as sticks. (E) The pseudodimer interface of the ADP-open state highlighting the distance between the IWA lysines. (F) Structure of LfArsA·MgADP open state in the absence of As^III. Left, 3.8 Å cryo-EM map; middle, ribbons model (pale green) overlayed with the ADP-open state model from B; Right, the metalloid-binding sites from both structures. (G) Cryo-EM map (6 to 7 Å overall resolution) of the LfArsA nonhydrolyzing variant (D46N/D364N) solved in the presence of MgATP with the ADP-open state model from B docked in the map. (H) Nucleotide-binding sites of the LfArsA·MgATP state showing Coulomb potential map likely corresponding to ATP. MgATP was manually fit using Coot.

We propose that our ArsA·MgADP structure represents an open form of ArsA, designated as the “ADP-open” state for the rest of the discussion. Based on the general mechanism of the IWA ATPase family, this conformation represents an inactive form of ArsA unable to hydrolyze ATP. Comparison of the cryo-EM structure to the X-ray crystal structures of EcArsA reveals the structures adopt similar conformations (SI Appendix, Fig. S1C); for example, the cryo-EM structure and the EcArsA·MgADP crystal structure (PDB: 1F48) have an r.m.s.d. of 0.95 Å over 475 residues. Despite the presence of millimolar concentrations As^III in the samples, there was no evidence of As^III at the metalloid-binding site in either our EM or the X-ray structures (SI Appendix, Fig. S5) (18, 37).

Structures of Nucleotide-Bound ArsA in the Absence of As^III.

To clarify the absence of As^III at the metalloid-binding site in the ArsA·MgADP structure, we solved a cryo-EM structure using identical experimental conditions but in the absence of As^III. The resulting 3.8 Å resolution map of the ArsA·MgADP state is in the open conformation (Fig. 1F and SI Appendix, Figs. S4B and S6), similar to the structure solved in the presence of As^III, with an r.m.s.d. of 0.57 over 545 residues. Importantly, the configuration of the metalloid-binding site remains unchanged (Fig. 1 F; Inset on Right), confirming that both states lack As^III. This suggests that the ADP-open state does not effectively bind the metalloid.

The presence of As^III stimulates steady-state ATP hydrolysis by ArsA above the basal activity (SI Appendix, Fig. S2B) (28, 29). To characterize the ATP-bound ArsA conformation in the absence of metalloid, we generated a nonhydrolyzing variant of LfArsA by mutating the catalytic Switch I aspartates—Asp46 and Asp364 of the N- and C-domains, respectively—to asparagines (SI Appendix, Fig. S2B, red plot). We used this variant to capture the MgATP-bound state, similar to the approach used to obtain the MgATP-bound Get3 structure (PDB: 7SPY) (6). For the rest of the discussion, structures solved using the D46N/D364N variant will be simply referred to as ArsA, and the Asn residue will be simply referred to as Asp residue. Under experimental conditions similar to those used to solve the ADP-open structures, we could only obtain a low-resolution (6 to 7 Å) EM map that revealed an open state of ArsA (SI Appendix, Fig. S7). We could confidently fit the model of the ArsA·MgADP structure into the map (Fig. 1G). While the density observed at the nucleotide-binding sites could not be unambiguously distinguished at this resolution, it is consistent with MgATP binding (Fig. 1H). This is reminiscent of NifH that is refractory to crystallization in the presence of ATP but readily crystallizes in the presence of both ATP and its partner protein, NifDK (MoFe protein) (26). It is plausible that binding of ATP makes ArsA conformationally flexible, preventing high-resolution cryo-EM reconstruction, and that a partner protein or As^III is required for stabilization. Importantly, this structure is consistent with the low basal ATPase activity of ArsA. Together, these As^III-free structures support that nucleotide binding alone is unable to stabilize a conformation competent for ATP hydrolysis.

Structure of the closed Conformation of ArsA Bound to MgATP and As^III.

To understand the structural basis of As^III activation, we sought to characterize the conformation of ArsA in the presence of both MgATP and As^III. We incubated the nonhydrolyzing ArsA variant with 2 mM MgCl₂, 2 mM ATP, and 2 mM sodium arsenite and then performed single-particle cryo-EM (SI Appendix, Fig. S8). The resulting Coulomb potential map was resolved to an overall gold-standard FSC resolution of 3.0 Å and sharpened using a uniform B-factor of −124 Ų (Fig. 2A). AlphaFold2 models of N- and C-domains of LfArsA were docked into the sharpened density map. The model was subsequently rebuilt and refined to obtain the structure of ArsA·MgATP·As^III with clear density for MgATP at both nucleotide-binding sites (Fig. 2 B and C). Significantly, at the metalloid-binding site, we observed density enclosed between the three conserved cysteines—Cys114, Cys173, and Cys422—into which we modeled an arsenic atom (Fig. 2D). The average Q-score for the structure was 0.71 (SI Appendix, Fig. S4C), indicating high resolvability of the cryo-EM map. About 97% of the protein was modeled into the map, while the regions with poor density, such as most of the linker region (residues 299 to 305) and residues 474 to 479, were omitted from the final model.

Fig. 2.

Cryo-EM structure of LfArsA closed conformation bound to MgATP and As^III. (A) Two views of the sharpened cryo-EM map colored based on local resolution (3.0 Å overall) for LfArsA·MgATP·As^III closed state (ATP-closed) of the nonhydrolyzing variant (D46N/D364N). (B) Cartoon representation of the ATP-closed state colored in viridis. Unmodeled regions are indicated as dotted lines. (C) Nucleotide-binding sites of the ATP-closed state showing the Coulomb potential map around MgATP and Switch I catalytic residues (Asn46 and Asn364) shown as sticks. Switch II Asp residues of both N- and C-sites are shown as sticks. The nucleophilic water molecule (W_cat) is highlighted. (D) Metalloid-binding site of ATP-closed state showing Coulomb potential map for As^III (purple atom) coordinated by Cys114, Cys173, and Cys422. The view is rotated 110º clockwise relative to B. (E) The pseudodimer interface of the ATP-closed state highlighting the distance between the IWA lysines (Lys17 and Lys335). (F) Top–down view, relative to A, comparing the ADP-open and the ATP-closed states of LfArsA colored as in Fig. 1C, showing both states in cartoon and then overlayed in surface representation. The overlay is aligned to the N-domain P-loop (residues 16 to 23) showing the relative change of the C-domain. (G) Structure of LfArsA·MgATP·As^III closed state solved using wild-type LfArsA. Left, 3.0 Å map cryo-EM map; Right, ribbons model (yellow) overlayed with the model from B. (H) Comparison of Switch I conformations between the two ATP-closed structures. Differences in interactions are highlighted.

Within both nucleotide-binding sites, ATP is stabilized by electrostatic interactions of the γ-phosphate group with Mg²⁺ and Lys (22 and 340, respectively), and by hydrogen bonds involving the ⍺- and β- phosphates with the P-loop backbone (Fig. 2C and SI Appendix, Fig. S9A). The ribose sugar and the adenine base are stabilized by hydrogen bonds contributed by side chains from both domains, and by backbone hydrogen bonds from the loop region corresponding to the adenine-binding loop (A-loop), as described in Get3 structures (SI Appendix, Fig. S9A) (24). Additionally, ATP bound to the N-domain is stabilized by potential cation–π interactions formed between the adenine base and both Arg207 (N) and Arg544 (C) (SI Appendix, Fig. S9 A, Left). However, in the C-domain, the corresponding interaction is only observed between the adenine and Arg256 (N). In this structure, the Switch I aspartates (mutated to asparagines here)—Asp46 (N) and Asp364 (C)—are positioned adjacent to each Mg²⁺ ion and anchor the catalytic water molecule (W_cat) near the γ-phosphate of ATP via a 3.0 Å hydrogen bond for nucleophilic attack, priming the enzyme for hydrolysis (Fig. 2C). Switch II aspartates—Asp143 (N) and Asp447 (C)—are positioned at 3.2 and 3.4 Å from the Mg²⁺ ion, respectively, and form short hydrogen bonds with the Thr residue coordinated to the respective Mg²⁺ ions, similar to the MgADP-bound structure.

This structure adopts a closed pseudodimer conformation that we define as the “ATP-closed” state. Notably, the P-loops of the two domains are in close proximity (Fig. 2E). The terminal amino group of the conserved IWA lysine of each domain—Lys17 (N) and Lys335 (C)—is positioned ~3 Å from the bridging oxygen atom between the β- and γ-phosphate groups of the ATP bound to the opposing domain. This enables stabilization of negative charge build-up on the nucleotides in the transition state. Together with the Switch I aspartates, which are positioned to activate W_cat for nucleophilic attack on ATP, this closed pseudodimer conformation represents a catalytically competent state of ArsA at both nucleotide-binding sites. The ATP-closed state is analogous to the closed dimer states previously reported for Get3 (PDB: 7SPY) and NifH (PDB: 1M34) (SI Appendix, Fig. S10) (6, 38). Notably, in our structure, similar to the Get3 closed structure, both IWA lysines appear to form a hydrogen bond with a water molecule found at the pseudodimer interface; these waters may assist with anchoring the terminal amino groups close to the nucleotides in preparation for ATP hydrolysis (Fig. 2E).

Relative to the ADP-open state, the two domains rotate toward the pseudodimer interface, forming a tighter interface in the ATP-closed state (Fig. 2F). This generates new interdomain contacts not observed in the ADP-open state (SI Appendix, Fig. S8B). Glu215 and Glu507, from the N- and C-domains, respectively, form hydrogen bonds with the catalytic water molecule and Switch I residues of the opposing nucleotide-binding site. Electrostatic contacts are formed between Arg214 of the N-domain and Asp371 of the C-domain. This results in a buried surface area of 2,380 Ų between the domains, ~500 Ų more than that of the open conformation.

We next sought to establish the conformation of ArsA under turnover conditions. Wild-type ArsA (wtArsA) was incubated with 2 mM MgCl₂, 2 mM ATP, and 2 mM sodium arsenite at room temperature for ~1.5 min before cryo-EM grid preparation. This resulted in a Coulomb potential map at 3.0 Å overall gold-standard FSC resolution (Fig. 2 G, Left and SI Appendix, Figs. S4D and S11). The final refined structure revealed a closed conformation similar to the ArsA nonhydrolyzing variant (r.m.s.d = 0.53 Å over 565 residues) with MgATP present at both nucleotide-binding sites and arsenic modeled at the metalloid-binding site (Fig. 2 G, Right). The IWA lysines in this structure are oriented toward the nucleotides (SI Appendix, Fig. S12A), supporting a catalytically competent prehydrolysis state consistent with the variant structure. While the Switch I aspartate in each nucleotide-binding site is positioned next to the Mg²⁺ ion anchoring the W_cat for nucleophilic attack, subtle differences are observed in the conformations of the Switch I loop between the wtArsA and variant closed structures. In the N-domain, the Switch I loop moves toward the nucleotide in wtArsA such that Asn50 now interacts with the Mg²⁺ ion via an intervening water molecule (Fig. 2 H, Left). Likewise, in the C-domain, Switch I loop shifts toward the nucleotide, causing His368 to interact with Mg²⁺ via an intervening water molecule (Fig. 2 H, Right). Additionally, we see evidence for a hydrogen-bonding network composed of ordered water molecules including W_cat, that connects the Switch I aspartates at the N- (Asp46) and C- (Asp364) sites, potentially coupling the two active sites (SI Appendix, Fig. S12B). Together, these subtle structural changes observed in the prehydrolysis state prime the enzyme to initiate hydrolysis in both N- and C-domains.

The As^III-Binding Site of ArsA.

We observed density for arsenic at the metalloid-binding site in the ATP-closed state, coordinated by the thiolate groups of Cys114, Cys173, and Cys422 in a pyramidal geometry (Fig. 2D and SI Appendix, Fig. S12C). Cys114 and Cys422 are homologous residues within the pseudodimer, located at the N-terminal end of helix 6, denoted as H6_N and H6_C, respectively, while Cys173 is located on the intervening loop (residues 155 to 183) between helix 7 and helix 8 in the N-domain.

To confirm the As^III coordination environment, we performed solution XAS. The ArsA nonhydrolyzing variant (1.95 mM) supplemented with 15 mM MgCl₂, 15 mM ATP, and 1.5 mM sodium arsenite was subjected to XAS analysis. X-ray absorption near-edge spectra (XANES) confirmed the presence of trivalent arsenic (Fig. 3A, black spectra), as the K-edge at 11865.9 eV was consistent with the first inflection edge energies of As^III model compounds (11,867 eV). Best-fit simulations of the extended X-ray absorption fine structure (EXAFS) spectra supported the presence of As^III coordinated by three sulfur ligands (AsS₃) at an average As-S bond length of 2.27 Å (Fig. 3 B, Top panel and SI Appendix, Table S2), consistent with cysteine coordination in the cryo-EM structure. Long-range carbon scattering was also observed in the sample, likely from secondary and tertiary sphere carbon atoms associated with the β- and β-carbon atoms of the cysteines, respectively. The oxidation state and coordination environment of the ArsA sample are consistent with that of a control sample consisting of As^III and L-cysteine (Fig. 3A, purple spectra; 3B, middle panel; and SI Appendix, Table S2). Similar As^III coordination has been reported for the ArsR repressor and ArsD metallochaperone, based on XAS analysis (15, 39, 40). We quantified the amount of As^III bound to ArsA under conditions similar to those used in cryo-EM sample preparation using inductively coupled plasma mass spectrometry (ICP-MS). In the presence of MgATP, ArsA binds As^III with 1:1 stoichiometry (Fig. 3C, black plot). Taken together, cryo-EM, XAS, and ICP-MS analysis support the binding of a single three-coordinate As^III at the ArsA pseudodimer interface in the presence of MgATP.

Fig. 3.

Characterization of the As^III binding site of ArsA. (A) X-ray absorption near edge spectra (XANES) for LfArsA·MgATP·As^III (“As^III-ArsA”; black), As^III-L-cysteine control (“As^III-Cys”; purple), and LfArsA(C173A)·MgATP·As^III (“As^III-ArsA C173A”; teal). (B) Left column, extended X-ray absorption fine structure (EXAFS) spectra (black) with best-fit simulations (green) for each complex. Right column, corresponding Fourier transforms. A replicate of each XAS sample was individually prepared and analyzed (SI Appendix, Fig. S13). (C) ICP-MS analysis for As^III quantification in nucleotide-bound LfArsA samples (ArsA·MgATP, black; ArsA C173A·MgATP, teal; ArsA·MgADP, gray; ArsA C114A/C173A/C422A·MgADP, yellow). Bars represent mean As/ArsA molar ratio for n = 3 individually prepared samples with error bars representing SD. (D) Steady-state ATPase activity of LfArsA (nmol min⁻¹ mg⁻¹) plotted against varying As^III concentrations (mM) in the presence of 5 mM MgCl₂ and 5 mM ATP at 37ºC (wild-type, black; C173A, teal; C114A/C173A/C422A, yellow). Data points represent mean of n = 3, and error bars represent SD. The data were fit to the Michaelis–Menten equation. (E) Comparison of metalloid-binding sites of ADP-open (Left) and ATP-closed (Right) highlighting the relative positions of Cys114, Cys173, and Cys422 that enable As^III (purple) binding.

The As^III-coordinating cysteines are conserved among ArsA homologs (SI Appendix, Fig. S14). Alanine mutations of these residues ablate As^III activation of steady-state ATP hydrolysis (Fig. 3D, yellow plot). By contrast, mutation of Cys173 located on the loop region of N-domain alone appears to only decrease the binding of As^III to ArsA, as the C173A variant requires higher As^III concentrations than the wild-type enzyme to approach half-maximal activation of ATPase activity (Fig. 3D, teal plot). XANES and EXAFS spectra for the C173A variant incubated with MgATP and As^III suggest AsO₂S coordination, indicating that As^III is no longer coordinated to three cysteines (Fig. 3A, teal spectra; 3B, Bottom panel; and SI Appendix, Table S2). Moreover, ICP-MS analysis of this variant shows over 50% decrease in As^III-ArsA molar ratio (Fig. 3C, teal plot). As previously reported (29), Cys173 regulates the binding affinity of As^III and is not critical for the stimulation of nucleotide hydrolysis.

Based on our ICP-MS analysis, ArsA·MgADP binds As^III far less tightly than ArsA·MgATP (Fig. 3C, black and gray plots), corroborating the absence of As^III density in the ADP-open structure. Mutating the As^III-coordinating cysteines completely disrupts As^III binding to ArsA·MgADP (Fig. 3C, yellow plot). How do the conformational changes between open and closed states modulate the affinity for As^III? Our cryo-EM structures reveal that switching from open to closed state causes reorientation of helix 6 of both domains such that Cys114 on H6_N and Cys422 on H6_C shift ~9 Å away from the P-loops (Fig. 3 E, Top panels). This makes the metalloid-binding site more accessible to the solvent for As^III binding. Additionally, a change in the relative positions of the side chains of Cys114 and Cys442 accompanies this conformational rearrangement. As shown in Fig. 3 E, Bottom panels, the distance between C_β atoms of Cys114 and Cys422 reduces from 6 Å in the open state to 4 Å in the closed state, facilitating As^III binding. Therefore, by controlling the relative positions of the three cysteines at the metalloid-binding site, nucleotide-dependent conformational changes in ArsA modulate the binding affinity for As^III.

Intradomain Conformational Rearrangements Between open and closed States.

In addition to the nucleotide-dependent conformational changes at the pseudodimer interface, large-scale structural rearrangements are observed within each domain between the ADP-open and the ATP-closed states (Fig. 4 A and B). These changes are primarily linked to the Switch I and II motifs. Compared to the ADP-open state, the Switch I loop of both the N- and C-domains rearranges to shift toward the nucleotide in the ATP-closed state (Fig. 4 C and D). This enables the positioning of Asp46 and Asp364 to activate W_cat and initiate ATP hydrolysis. Differences in the Switch I conformation between the two domains allow Ser49 to additionally stabilize the catalytic water in the N-domain (Fig. 4 C and G). While the carboxylate group of Asp364 shifts about 5 Å toward the active site from open to closed state, the carboxylate of Asp46 shifts only 3 Å (Fig. 4 C and D).

Fig. 4.

Intradomain conformational changes between ADP-open and ATP-closed states. (A and B) Cartoon representations of overall structural rearrangements within N- and C-domains, respectively. Arrows indicate direction of motion of the corresponding motif from the ADP-open (Left) to the ATP-closed state (Right). All models in these and subsequent panels are colored in viridis. Highlighted residues are shown as sticks. All overlayed structures are aligned on the respective P-loops (residues 16 to 23 for N-domain and 334 to 341 for C-domain). (C and D) Conformational changes in Switch I motif between open (50% opacity) and closed (100% opacity) states at both N- and C-nucleotide binding sites, respectively. Shift in the carboxylate group of catalytic Asp from open to closed states is indicated in Å. (E) Conformational changes in Switch II and H7_N of the N-domain between open (50% opacity) and closed (100% opacity) states, shown in two views rotated 40° apart. The “metalloid affinity loop” connecting H7_N and H8_N is highlighted. (F) Conformational changes in Switch II and H7_C of the C-domain between open (50% opacity) and closed (100% opacity) states. In both panels E and F, helix 6 has been omitted for visual clarity. (G) Comparison of Switch I and II conformations at the N- and C-nucleotide binding sites of the ATP-closed state.

Switch II, acting as a link between the nucleotide-binding site and the metalloid-binding site, modulates a series of dramatic conformational changes within each domain between the open and closed states. Upon As^III binding to the metalloid-binding site of the ATP-closed state, both H6_N and H6_C undergo rigid body rearrangement such that the N-terminal end of each helix, bearing the As^III ligands—Cys114 and Cys422, respectively—rotates away from the nucleotide-binding sites (Fig. 4 A and B). Reorientation of helix 6 results in the restructuring of the downstream Switch II motif in the ATP-closed state. Although the Switch II aspartates (Asp143 and Asp447) of the respective domains remain anchored near the Mg²⁺ ion in both states, downstream residues 144 to 147 of the N-domain (Switch II_N), residues 448 to 451 of the C-domain (Switch II_C), and both helix 7 s shift toward the nucleotide in the ATP-closed state (Fig. 4 E and F). Switch II_N shifts toward the nucleotide to a greater extent than Switch II_C (Fig. 4G). Hydrogen-bonding interactions between Thr147 (Switch II_N) and Glu117 (H6_N) and between Thr451 (Switch II_C) and Glu425 (H6_C), in the ADP-open state, are destabilized in the ATP-closed state to facilitate the conformational change coordinated by helix 6 and Switch II (SI Appendix, Fig. S15). In the wtArsA ATP-closed state, Switch II_C further shifts toward the nucleotide, adopting the same conformation as Switch II_N (SI Appendix, Fig. S12D). This symmetric conformation of Switch II results in an additional contact between the two domains where His149 (H7_N) and His453 (H7_C) form a π–π stacking interaction (SI Appendix, Fig. S12E).

A striking difference is seen between the N- and C-domains in the conformation of helix 7 following the Switch II loop. In the ADP-open state, H7_N is composed of residues Gly148-Leu154. In the ATP-closed state, this helix is extended by incorporation of formerly loop residues Gln155-Ala166 into the C-terminal end of H7_N transition, resulting in a twisted helix composed of residues Thr147-Ala166 (Fig. 4E). By contrast, H7_C transitions from residues Gly452-Ala460 in the ADP-open state to a discontinuous helical segment between Thr451 and Gln471, connected by a loop formed by residues Asp459-Gly462 in the ATP-closed state (Fig. 4F). Complete helical rearrangement of H7_N arises by virtue of Cys173 on the loop region between H7_N and H8_N, that serves as the third ligand for As^III binding in the ATP-closed state. This loop, designated as the “metalloid affinity loop,” undergoes a considerable rearrangement between the open and closed states to stabilize three-coordinate As^III binding (Fig. 4E). An analogous residue is not found on the C-domain; in fact, the corresponding loop region in the C-domain is poorly resolved in both states. Notably, such helical transition of helix 7 is also observed upon substrate (TA protein) binding at the homologous site in Get3 (PDB: 7SQ0) (6). Taken together, the conformational changes associated with As^III binding in the ATP-closed state demonstrate how three-coordinate As^III binding at the metalloid-binding site of ArsA is allosterically coupled to the nucleotide-binding sites via Switch II.

Interaction of ArsA with Partner Proteins, ArsD and ArsB.

We next inquired how the nucleotide- and As^III-driven conformational changes of ArsA described here modulate its interaction with its partner proteins, ArsD and ArsB. Genetic and biochemical studies suggest that ArsA and ArsD interact independent of metalloid and nucleotide and that ATP hydrolysis facilitates metalloid transfer from ArsD to ArsA (15). In fact, ArsD stimulates ArsA ATPase activity by increasing its affinity for As^III from millimolar to micromolar range (14). On the other hand, there is no biochemical evidence to suggest a direct interaction between ArsA and ArsB, primarily due to challenges in overexpressing ArsB, except for one study that reported saturable binding of ArsA to membrane vesicles isolated from E. coli cells that expressed ArsB (17). In either case, a complex structure has been elusive. To understand whether the open or closed state of ArsA preferentially interacts with its partner proteins, we performed AlphaFold 3 modeling of the LfArsA·ArsD and LfArsA·ArsB complexes in apo, MgADP-bound, and MgATP·As^III-bound states (41). The latter two predictions were intended to mimic the ADP-open and ATP-closed states, respectively. For predictions with ArsD, all three cases revealed ArsA·ArsD complexes with good “predicted template modeling” (pTM) scores ≥ 0.6 and reasonable “interface pTM” (ipTM) scores > 0.6 (SI Appendix, Fig. S16 A and B). In each case, ArsD is docked proximal to the metalloid binding site of ArsA (SI Appendix, Fig. S16A). The As^III-binding residues of ArsD—Cys12, Cys13, and Cys18—interface with the As^III-binding residues of ArsA—Cys114, Cys173, and Cys422 (SI Appendix, Fig. S16B). Notably, ArsA approaches a closed conformation in each case, supporting the notion that ArsD transfers As^III to ArsA by preferentially interacting with the closed state of ArsA. This is consistent with our experimental results, indicating that the closed state has a higher affinity for the metalloid than the open state. The predictions for the ArsA·ArsB complex were relatively less informative due to poor ipTM scores (<0.6) for all three predictions (SI Appendix, Fig. S16 D and E). Although ArsA consistently docks near the predicted cytoplasmic face of ArsB, and some interactions between the ArsA N-domain residues and ArsB residues can be observed in the MgATP·As^III prediction (SI Appendix, Fig. S16F), a nucleotide- or metalloid-dependent complex is not implied.

Discussion

Among the IWA ATPases implicated in various complex biological functions, ArsA was one of the first proteins in the family to be studied biochemically and structurally. ArsA plays a central role in ensuring the survival of many bacterial species under toxic concentrations of As^III by coupling ATP binding and hydrolysis, not only to capture As^III from the cytoplasm via ArsD but also to facilitate its transport across the membrane via ArsB (15, 16). The underlying nucleotide hydrolysis mechanism of ArsA, which regulates these events, is not fully understood. Based on pre-steady-state kinetic analysis, the pseudodimeric ATPase undergoes nucleotide-dependent conformational changes throughout its catalytic cycle (27, 28, 42). MgATP-bound ArsA has been reported to alternate between two conformations that differ in their affinities for the metalloid substrate (27). Metalloid binding stabilizes one of these conformations, followed by hydrolysis and phosphate release. Previous X-ray crystal structures were broadly inconsistent with this mechanism. Our high-resolution cryo-EM efforts provide the structural basis for the conformational states proposed in this mechanism. We have characterized open and closed conformations of ArsA. The open state can exist bound to either ADP or ATP (Fig. 1 B, F and G), has low metalloid affinity, and is ATPase inactive. The closed state is stabilized as a ternary complex when As^III binds ArsA·MgATP (Fig. 2B) and is competent for ATP hydrolysis. The structures enable us to propose a catalytic mechanism for ArsA (Fig. 5). ATP binds both N- and C-sites of ArsA where the enzyme exists in equilibrium between open and closed states (states 1 and 2). ATP alone is unable to drive the equilibrium to favor the closed state; as a result, ArsA hydrolyzes nucleotide at a basal rate. When As^III is coordinated by three cysteines at the metalloid-binding site (Fig. 2D), the open conformation transitions into the catalytically competent closed conformation (state 3). Coordination of As^III to Cys114 and Cys422 allosterically triggers the transition of the pseudodimer from open to closed state via helix 6 and Switch II (Fig. 4 A and B). Cys173 on the “metalloid affinity loop” acts as a switch controlling the binding affinity of As^III for ArsA and stabilizes the ternary complex in the closed state. As^III binding by this motif is coupled to the nucleotide-binding site of the N-domain via H7_N and Switch II_N (Fig. 4E). This mechanism reveals how three-coordinate As^III binding allosterically stimulates ATP hydrolysis, which is the proposed rate-limiting step of ArsA (27). Following ATP hydrolysis, the enzyme loses affinity for As^III, releasing the metalloid and phosphate, and returns to the open state (state 4). Nucleotide exchange can then reset the enzyme for another catalytic cycle. The conformational cycling of ArsA between high-affinity closed and low-affinity open states facilitates coordination of a series of handoffs to capture the metalloid from ArsD, followed by transfer to ArsB at the membrane for export.

Fig. 5.

Mechanistic model for the ArsA catalytic cycle. ATP-bound ArsA alternates between open (state 1) and closed (state 2). As^III binding stabilizes the catalytically competent closed conformation (state 3). Following ATP hydrolysis and release of As^III and phosphate (P_i), ArsA switches to the ADP-bound open conformation (state 4). Exchange of the nucleotide resets the enzyme for another catalytic cycle. States 1, 3, and 4 were characterized in this study. ArsA is colored by N (blue) and C (green) domains. Nucleotides are shown as space-filling models. As^III (As(OH)₃ at physiological pH) and P_i are shown as ball and sticks.

The conformational changes of ArsA reported here are consistent with the general structural and mechanistic framework of IWA ATPases (SI Appendix, Fig. S17). Despite each IWA ATPase binding a unique substrate at the dimer interface, the structural changes corresponding to the substrate, e.g., As^III binding to ArsA, remain conserved. In the case of Get3, TA protein binding induces a conformational change in helix 6 (substrate helix) and Switch II, thus regulating the dimer dynamics (6, 25). Likewise in NifH, association with NifDK affects the positioning of the [4Fe:4S] cluster for electron transport, which then induces conformational changes in the substrate helix and Switch II, stabilizing the closed dimer conformation (5). Notably, the pseudodimeric architecture of ArsA is distinct from other ATPases in the family. While both N- and C-domains have been shown to be competent for ATP hydrolysis (21), structural differences between the domains in the ATP-closed structures highlight the functional asymmetry within ArsA. Both Switch I and II loops adopt distinct conformations relative to the nucleotide in the N- and C-domains in the closed state (Fig. 4G and SI Appendix, Fig. S18A), reflecting the difference in ATPase activities, where the N-site is more competent for hydrolysis than the C-site (SI Appendix, Fig. S18B). These structural differences become less obvious in the wtArsA closed state under turnover conditions where a potential hydrogen-bonding network linking the two active sites may coordinate hydrolysis at the two sites (SI Appendix, Fig. S12 B and D). Further evaluation of the role of pseudodimer would be insightful for understanding functional asymmetry as a generalized mechanistic feature across homodimeric IWA ATPases.

How ArsA enhances the efficiency of As^III removal by ArsB is an outstanding question. Properties of the “ArsAB pump” are distinct from other ATP-dependent transporters, including ABC transporters and P-type ATPases, both in terms of the protein fold of the ATPase domain and that the ATPase activity is only required to improve the efficiency of transport (1, 43, 44). ArsB can transport the metalloid by itself as a secondary carrier (16). An N-domain P-loop mutant of EcArsA (G20S) is hydrolysis-deficient and inhibits the ability of ArsB to remove As^III (45), implying that the inactive ATPase may trap the transporter in a nonfunctional state. ArsA may enhance efflux either by triggering a transient conformational change in ArsB to facilitate metalloid transport, or by increasing the local concentration of As^III at the membrane, thereby reducing the entropic penalty of delivering the metalloid to ArsB by diffusion. In either case, the nucleotide-dependent conformational landscape of ArsA enables the formation of a transient high-affinity metalloid binding site to direct the toxic metalloid for removal, and therefore, ensures survival of the bacteria under arsenic stress.

Materials and Methods

LfArsA (Uniprot Accession Number: J9ZFA3) was cloned into a pETDuet-1 backbone. The resulting LfArsA with a C-terminal 6x-His tag was used for overexpression in E. coli BL21 Gold (DE3) competent cells (Invitrogen). Wild-type LfArsA and variants were expressed and purified as described in SI Appendix, Materials and Methods, and subsequently used for cryo-EM and biochemical characterization. Cryo-EM data were collected using a 300 kV Titan Krios TEM and processed using cryoSPARC versions 4.4.1 to 4.5.1(46). Model building and refinement were performed using Phenix (47) and Coot (48). Detailed procedures for cryo-EM sample preparation, data collection and processing, model building and refinement, XAS, ATPase assays, inductively coupled plasma-mass spectroscopy, and AlphaFold 3 modeling are available in the SI Appendix, Materials and Methods.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Cryo-EM maps and the corresponding atomic models generated in this study have been deposited into the Protein Data Bank (PDB) and the Electron Microscopy Data Bank (EMDB) under the following accession codes: 9NBL (49) and EMD-49231 (50) (LfArsA·MgADP in the presence of As^III); 9NBM (51) and EMD-49232 (52) (LfArsA·MgADP in the absence of As^III); EMD-49233 (53) (LfArsA·MgATP); 9NBO (54) and EMD-49234 (55) (LfArsA (D46N/D364N)·MgATP·As^III); and 9NBW (56) and EMD-49237 (57) (wtLfArsA·MgATP·As^III). AlphaFold 3 model cif files for ArsA·ArsD and ArsA·ArsB complexes and corresponding JSON input files can be accessed at https://github.com/smahajan82/ArsAmanuscript.git (58). Scripts for the ATPase assay and ICP-MS plots can be accessed in the same GitHub repository as well.