Deinococcus radiodurans R1 and D. indicus Wt/1aT (MTCC 4913) were grown either in: M53 medium, 1.0% (w/v) casein yeast peptone (Sigma-Aldrich), 0.5% (w/v) yeast extract (VWR Chemicals), 0.5% (w/v) glucose (Carl Roth) and 0.5% (w/v) NaCl (Merck); or Tryptone Glucose Yeast extract (TGY) medium, 1.0% (w/v) tryptone (VWR Chemicals), 0.5% (w/v) yeast extract (VWR Chemicals) and 0.1% (w/v) glucose (Carl Roth). For all assays, cells were grown at 30°C and 150 rpm agitation; for solid cultures, 1.5% (w/v) agar-agar (Carl Roth) was added. For each assay, two sequential subcultures were grown for 14–16 h prior to the final culture. Final cultures were diluted to an early-exponential phase either (Optical density at 600 nm (OD₆₀₀) of 0.2) for arsenate exposure growth curves and STEM-EDX or (OD₆₀₀ = 0.5) for oxidative stress, in-plate arsenate stress and UV assays.

Oxidative and arsenate stress assays were performed by adapting the disk diffusion method (Bauer et al., 1966; Hudzicki, 2012). Briefly, 400 μl of cells grown in either M53 or TGY (at an OD₆₀₀ = 2) were evenly spread on the surface of an agar plate (120.5 × 120.5 mm) using a cotton swab and allowed to dry for 15 min before the application of the disks (6 mm, PRAT DUMAS). Afterwards, oxidative stress inducer compounds or arsenic (20 μl) were applied on top of the disk in increasing concentrations. Plates were incubated for 3 days and halos were measured using Fiji/ImageJ (Schindelin et al., 2012). Oxidative stress compounds: (1) Methyl viologen dichloride, MV (C₁₂H₁₄Cl₂N₂· xH₂O, Sigma-Aldrich); (2) Hydrogen peroxide 30% (w/w), H₂O₂ (Sigma-Aldrich). Arsenic stress compound: sodium arsenate dibasic heptahydrate 98%, As(V), (HAsNa₂O₄ · 7H₂O, Sigma Aldrich).

Ultraviolet (UV) irradiations were performed at UV-C (254 nm) with a fluence of 3 mW/cm² throughout the assay. Fluence was assessed using a MS-100 optical radiometer with a MS 125 UVC sensor (Ultra-Violet Products, Upland, CA). D. indicus cells grown in either M53 and TGY (at an OD₆₀₀ = 2) were serially diluted and plated in duplicate for each dose. Plates were air dried for 15 min and placed under the UV lamp to be exposed from 300 to 1800 J/m² (in increments of 300 J/m² which corresponds to an exposure of 10 s). Afterwards, plates were incubated for 3 days and the colony forming units (CFU) were assessed. The average CFU mL⁻¹ of non-irradiated sample aliquots represented 100% survival. The surviving fraction at a given dose is the average CFU mL⁻¹ from each irradiated sample (N) divided by the average CFU mL⁻¹ of the non-irradiated sample (N₀). Survival curves were obtained by plotting the logarithm of N/N₀ versus the dose. To determine the survival curve parameters, the multi-hit model was applied (equation 1), S(D) = Survival fraction (where 1 is 100%); D = Dose (J/m²); K = inactivation constant and n = extrapolation number (obtained by the intercept of the extrapolated semi-log straight-line), (Severin et al., 1983; Kowalski et al., 2000; Kowalski, 2009). For statistical intercomparison of the different D₁₀ values extrapolated from the survival curves of M53 and TGY a Student’s t-test was used and Ρ ≤ 0.05 were considered statistically significant (De La Vega et al., 2005).

Arsenic resistance was assessed by adding As(V) to broth cultures (OD₆₀₀ = 0.3) for both M53 and TGY. D. indicus cells were incubated with 10, 25 and 50 mM while D. radiodurans was incubated with 0.05, 0.5, 1, 5, 10, and 20 mM of As(V). Growth curves were obtained by plotting the measured OD₆₀₀ versus time.

Deinococcus indicus cells were grown in M53 or TGY in 24 well plates for 3 days at 30°C. Grown cells were washed with Phosphate Buffer Saline (PBS) and applied over a lamella. Samples were fixed using a mixture of 2.5% glutaraldehyde and 1% formaldehyde in 0.1 M sodium cacodylate buffer for 1 h at room temperature. Afterwards, samples were washed 3 times with 0.1 M of sodium cacodylate buffer and dehydrated by sequential washing with increasing concentration of ethanol solutions (50, 70, 90, and 100%). Ethanol was removed and tert-butyl alcohol was added. Samples were incubated at 4°C for 30 min followed by 1 h in a vacuum desiccator and kept at room temperature until the scanning electron microscopy (SEM) analyses were performed. Samples were incubated for 10 min with 2% solution of osmium tetroxide and gold sputtered with 8 nm gold in an electron sputter (Cressington 108). Imaging was performed using a Hitachi TM3030Plus scanning electron microscope, at 1.5 kV.

Deinococcus indicus cells were grown in M53 broth media (OD₆₀₀ of 0.3) and washed twice with M53 at 4000 rpm, 4°C. Afterwards, cells were diluted to an appropriated concentration and 4 μL deposited onto a glow-discharged Quantifoil copper grid. The grids were blotted, and plunged into liquid ethane with an automated plunger (EM-GP, Leica Microsystems). STEM imaging was performed under cryogenic conditions with a field-emission Talos F200X (S)TEM microscope (Thermo Fisher Scientific) at 200 kV accelerating voltage (extraction voltage 3,850 V, gun lens 4, spot size 9), with a condenser aperture of 70 μm in microprobe mode with a semi convergence angle of 2.1 mrad. Energy-dispersive X-ray spectroscopy (EDX) was performed using an energy dispersive spectrometer encompassing a Super-X G2 detector that features 4 silicon drift detector (SDDs) for substantially enhanced sensitivity, which is critical for trace element detection, with 8 μs dwell time per pixel, and dispersion of 5 eV per channel. Acquired images and EDX data were processed using Velox™ (Thermo Fisher Scientific) software followed by analysis in ImageJ (Schindelin et al., 2012).

To identify D. indicus cells and the granules residing within them, grids were imaged using a Talos Arctica 200 kV FEG STEM (Thermo Fisher Scientific) applying the same settings described above with a few variations: spot size 4, camera length 160 mm and 6.66 nm pixel size with 1 μs dwell time per pixel. To quantify the size and distribution of granules within the bacteria, we used tiling and stitching images taken using TFS Tomography software to cover large field of view with multiple bacteria. Stitching was done using Fiji Grid/Collection stitching plugin (Preibisch et al., 2009). Segmentation of the individual bacteria and individual granules residing within them was performed and the number of granules and their average and total size within each bacterium was measured. To segment the bacteria and granules, we used a workflow combining Ilastik (Berg et al., 2019), pixel classifier followed by further processing using dedicated Fiji macro (Schindelin et al., 2012). We used multiple fields of view from different conditions for training two-stage machine-learning classifier in Ilastik “autocontext” approach to classify pixels into four categories: bacteria, granules, background and unclassified. The trained classifier was applied to the stitched images in Fiji (Schindelin et al., 2012). Bacteria were segmented based on connected component analysis of a filled mask of all pixels classified as bacteria, followed by size (1.6 < size [μm²] < 6) and shape filtering (circularity >0.2). Granules were then segmented based on connected component analysis of a mask of all pixels classified as granules within segmented bacteria and further filtered by size (0.012 < size [μm²] < 0.12) and shape (circularity >0.3). Manual correction was applied to correct the segmentation of some missed or falsely detected bacteria. The size and number of all valid bacteria and granules were extracted and analyzed.

Multiple sequence alignments were performed using ClustalW (Thompson et al., 1994). Aligned sequences were used as input and the phylogenetic tree was generated using maximum likelihood (ML) in IQ-TREE2 (version 2.2.2.6) (Minh et al., 2020), software with the mod LG + G4 substitution model. The model was selected by running the first tree using the model finder option (Kalyaanamoorthy et al., 2017). The final tree was generated by performing bootstrap analysis of 1,000 data sets and was treated and displayed using the Interactive Tree Of Life (iTOL) (version 6) web service (Letunic and Bork, 2021).

DiArsC2 gene was cloned into a pET28a(+) plasmid that contains an His-tag followed by a TEV cleavage site (GenScript Inc.). The overexpression of DiArsC2 was obtained in Escherichia coli BL21 (Gold) cells transformed with the plasmid pET28a(+)-DiArsC2 and grown in Luria-Bertani (LB) medium at 37°C, 180 rpm. Cells at an OD₆₀₀ of 0.7 were induced with 500 μM of isopropyl β-D-thiogalactopyranoside (NZYTech) and grown overnight at 20°C. Afterward, the cells were harvested and resuspended in lysis buffer [20 mM HEPES pH 7, 10% glycerol, 300 mM NaCl, 10 mM MgCl₂, 1 μg/ml DNase I and 0.1 mg/ml Lysozyme (Sigma-Aldrich)]. Cells were disrupted by 5 cycles of freeze and thaw, and the soluble fraction containing overexpressed DiArsC2 was obtained by centrifugation at 25,931 x g, 20 min at 4°C. The supernatant was loaded onto a His Trap excel column (Cytiva) using as binding buffer 20 mM HEPES pH 7, 250 mM NaCl, 10% (v/v) glycerol, and 10 mM imidazole, and the elution buffer was 20 mM HEPES pH 7, 250 mM NaCl, 10% (v/v) glycerol and 1 M imidazole. DiArsC2 eluted at 300 mM of imidazole and was further loaded onto a desalting column (Hiprep^tm 26/10 Desalting, Cytiva) using 20 mM HEPES pH 7, 250 mM NaCl, 10% glycerol (v/v). Afterward, the DiArsC2 His-tag was cleaved by adding Tobacco Etch Virus (TEV) protease (Sigma) at 12°C with gentle shaking for 14 h. DiArsC2 was loaded onto His Trap excel (Cytiva) using the same buffers as described above. The final purification step was done by size exclusion chromatography (Superdex 75 10/300 GL, Cytiva), with the buffer 20 mM HEPES pH 7 and 250 mM NaCl. The protein was pure as judged by sodium dodecyl sulphate polyacrylamide gel electrophoresis analysis and concentrated to 21 mg/mL prior to being used for crystallization trials.

Crystallization screens for DiArsC2 were setup at the nL scale in a Crystallization Robot Mosquito LCP (SPT Labtech) with triple sitting drop 96-well plate (TTP Labtech) and using commercial screens Structure 1 and 2 (Molecular Dimensions). Needle- and plate-like crystals appeared in several conditions, three of which were selected for further optimization: #37 (0.2 M Sodium acetate trihydrate, 0.1 M Tris pH 8.5 and 30% (w/v) PEG 4000); #46 (0.05 M Potassium phosphate monobasic and 20% (w/v) PEG 8000); and #47 (15% (w/v) PEG1500). These were optimized by vapor diffusion using 2 μl hanging drops equilibrated against 500 μl reservoir solution in XRL 24-well crystallization plates (Molecular Dimensions), at 20 and 4°C with varying ratios of protein and crystallization solution in each drop. For the final optimized crystallization condition, a ratio of 1:1 (protein to crystallization solution) with 0.2 M Sodium acetate trihydrate, 0.1 M Tris pH 8.5 and 20% (w/v) PEG 4000 for DiArsC2 native (henceforth referred to as DiArsC2) at 4°C. In the case of DiArsC2 bound to arsenic (DiArsC2-As), a ratio of 1:0.8 (protein to crystallization solution) was used using 0.2 M Sodium acetate trihydrate, 0.1 M Tris pH 8.5 and 30% (w/v) PEG 4000, plus 0.2 μl of 0.1 M As(V) at 20°C. DiArsC2 crystals appeared after 7 days while DiArsC2-As crystals appeared after 3 days. Crystals were dipped in a cryoprotecting solution that consisted of the crystallization buffer supplemented with 10% (v/v) PEG 400 and flash-cooled in liquid nitrogen.

DiArsC2 crystals were initially screened in-house using Cu Kα radiation with 1.5418 Å wavelength in a Bruker AXS Proteum Pt135 CCD detector system coupled to an Incoatec Microfocus X-ray Source with Montel mirrors. Bruker Proteum software package were used to process and scale the images. The structure was solved by Molecular Replacement using MORDA (Vagin and Lebedev, 2015) and using as phasing model the structure of arsenate reductase from Bacillus subtilis (PDB code: 1JL3) (Bennett et al., 2001). In addition DiArsC2-As crystal was also tested in-house and the structure was solved by PHASER (McCoy et al., 2007) in the PHENIX suite (Adams et al., 2010; Liebschner et al., 2019) using the DiArsC2 as phasing model.

Diffraction data to higher resolution were collected at the XALOC beamline BL13 of the ALBA Synchrotron (Barcelona, Spain) (Juanhuix et al., 2014). The images were processed with autoPROC (Vonrhein et al., 2011), which makes use of XDS (Kabsch, 2010) and the CCP4 suite (Potterton et al., 2003) for integration and conversion of integrated intensities to structure factors. Each dataset was integrated with XDS, followed by POINTLESS (Evans, 2011) for space-group determination, scaling and merging with AIMLESS and application of an anisotropic resolution cut-off at CC^1/2 < 0.3 with STARANISO (Tickle et al., 2018). The crystals belonged to the monoclinic space group P2₁ and contained two molecules per asymmetric unit, as estimated by the Mathews coefficient probability (Matthews, 1968; Kantardjieff and Rupp, 2003). The data processing statistics are given in Supplementary Table S1. Higher resolution crystallographic structures for both DiArsC2 and DiArsC2-As were determined by molecular replacement using PHASER-MR (McCoy et al., 2007) via the CCP4 Graphics User Interface (Potterton et al., 2003) or PHENIX Suite (Adams et al., 2010; Liebschner et al., 2019), and using the DiArsC2-As structure obtained from the in-house data as the phasing model. For DiArsC2, an initial refinement was done with REFMAC (Murshudov et al., 1997) and continued with PHENIX.REFINE (Terwilliger et al., 2007; Adams et al., 2010; Afonine et al., 2012) while for DiArsC2-As refinements were done using PHENIX.REFINE. Throughout the refinement, the model was periodically checked and corrected with COOT against σ_A-weighted 2|F_o|-|F_c| and |F_o|-|F_c| electron-density maps. Solvent molecules were added manually by inspection of electron-density maps in COOT (Emsley and Cowtan, 2004). TLS (translation-libration-screw) reciprocal space refinement was carried out for both structures. Hydrogen atoms were included in calculated positions with the PHENIX.READYSET tool and isotropic displacement parameters (ADPs) were refined for all non-hydrogen atoms. The final structures were validated using MOLPROBITY (Chen et al., 2010). Refinement statistics are given in Supplementary Table S2. Structure factors and associated structure coordinates of DiArsC2 and DiArsC2-As were deposited in the Protein Data bank (Berman et al., 2003) with accession code 8P6M and 8P5N, respectively (Supplementary Table S2).

A simple ensemble refinement protocol in PHENIX was run for both final structures DiArsC2 and DiArsC2-As considering one domain per independent molecule in the asymmetric unit, of the crystal structure, and using 80% of the non-hydrogen atoms to include in the TLS fitting of the atomic displacement parameters (ADPs) from a previous refinement with isotropic ADPs (Burnley et al., 2012). The secondary structure was determined running PROCHECK within CCP4 Graphics User Interface (Laskowski et al., 1993; Potterton et al., 2003). Figures of the structures were prepared using PyMOL (Delano, 2022).

Deinococcus species are known for their extreme tolerance to radiation and oxidative stress (Lim et al., 2019). To evaluate D. indicus tolerance, we subjected the cells grown in M53 and TGY to hydrogen peroxide (H₂O₂) and methyl viologen (MV). MV is known to catalyze the formation of superoxide anion (O^2•-), (Halliwell and Gutteridge, 1999; Halliwell, 2006). D. indicus cells presented higher resistance to H₂O₂ than to MV in both media (Figure 1). Upon exposure to MV, cells grown in M53 showed higher resistance than the ones grown in TGY. In fact, at 20 mM of MV, cells in M53 were neither inhibited nor had their cell growth compromised. However, the same was not observed when directly applying H₂O₂, in which case the cells responded in the same way independently of the culture medium (Figure 1B). D. indicus cells were also exposed to UV-C irradiation. It is known that when UV radiation is applied to microbial populations a slight delay is often observed before the onset of an exponential decay (Cerf, 1977; Munakata et al., 1991; Pruitt and Kamau, 1993; Kowalski et al., 2000; Kowalski, 2009). This effect is usually named as a shoulder curve due to its shape and for this reason the multi-hit model was applied to determine curve parameters and extrapolate the D₁₀ (i.e., the dose required to yield 10% survival) for the cells grown in both culture media (Severin et al., 1983; Kowalski et al., 2000; Kowalski, 2009). Interestingly, in M53 cells exhibited significant higher resistance to UV-C in contrast to the ones grown in TGY (Figure 2). This was verified by calculating the D₁₀ values (Ρ ≤ 0.05): in M53 a dose of 809 ± 66 J/m² (30s exposure) is required to eliminate 90% of cells, while in TGY 619 ± 82 J/m² (20s exposure) was sufficient to achieve the same result.

Deinococcus indicus Wt/1aT was originally isolated from an aquifer in West Bengal, India, and it was reported to survive 10 mM of As(V) in nutrient broth medium (Suresh et al., 2004). To evaluate its ability to cope with higher concentrations of As(V), D. indicus and D. radiodurans (non-arsenic resistant specie) were exposed to As(V) in their early exponential growth phase (OD₆₀₀ = 0.3). D. indicus was able to grow in the presence of 25 mM As(V) in both media cultures. However, an extended lag phase was observed and at 50 mM As(V), cells were unable to grow (Figure 3A). Nevertheless, cells harvested and applied to fresh media were able to recover their growth in normal conditions (Supplementary Figure S1). Regarding D. radiodurans, different As(V) concentrations were tested, namely 0.05, 0.5, 1, 5, 10, and 20 mM, our results show that in the presence of 5, 10, or 20 mM, the cells are not able to recover the growth after 24H (Figure 3B). Cells were only able to display normal growth in 50 μM As(V), a 500-fold lower concentration than applied to D. indicus. When exposed to 1 mM, an evident decrease in growth was observed (Figure 3B). This observation was more evident in cells grown in TGY than for those grown in M53.

To further assess their resistance to As(V) in a solid matrix, D. indicus and D. radiodurans were exposed to increasing concentrations of arsenate ranging from 200 mM to 1 M (Figure 4). Interestingly, in the solid matrix D. indicus was not fully inhibited by 1 M of As(V) in either culture medium. Nevertheless, partial inhibition halos (reduction of cellular density) were observed in cells grown in M53 upon submission to 600 mM to 1 M of As(V). In contrast, D. radiodurans is inhibited by 200 mM of arsenate (Figure 4).

Previously, it was demonstrated that D. indicus presents growth media-induced morphotypes displaying a reversible behavior from single cells to multi-cell chain morphology (Chauhan et al., 2019b). To explore this capacity and understand if differences in morphotypes account for the behavior disparities observed against cell insults, D. indicus grown in M53 and TGY media were observed through Scanning Electron Microscopy (SEM). SEM data revealed small differences on the size of the cells, where D. indicus grown in TGY presented slightly lower cell lengths. However, no differences in cell chain clustering were observed (Figure 5). Nevertheless, the addition of 500 μM MnCl₂ seems to affect cells grown in TGY, where cell septation is compromised and cell length is vastly increased (Supplementary Figure S2).

Polyphosphate-like granules (pPLGs) are known to be involved in the sequestration of cation ions as well as in heavy-metal detoxification (Peverly et al., 1978; Urech et al., 1978; Scott and Palmer, 1990; Keasling, 1997; Kulakovskaya, 2018). D. radiodurans, Deinococcus murrayi and Deinococcus proteolyticus present pPLGs in their native compositions, but no information is available for D. indicus (Thornley et al., 1965; Sleytr et al., 1976; Ferreira et al., 1997; Eltsov and Dubochet, 2005; Copeland et al., 2012). Wolf et al. (2015) was carried out to image D. indicus in native conditions at the early-exponential phase in M53 medium. Through the ImageJ macro previously trained using ilastik “autocontext” approach, we were able to rapidly quantify and segment D. indicus cells and their granules using Cryo-STEM stitched images (Figure 6A). We observed a wide heterogeneity of electron dense granules distribution ranging from 0 to 8 granules per cell (Figure 6B). Overall, from the 435 cells analyzed 31.8% of the cells had no granules, followed by 17.2% of the cells containing 1 or 2 granules. Only 1.4 and 0.5% of the cells contained 7 and 8 granules, respectively (Figure 6B). The average granule area (n = 803) was 0.040 ± 0.019 μm² with the smallest granules measuring 0.012 μm² while the biggest granule measuring 0.107 μm² (Figure 6C). EDX spectral analysis detected the enrichment in phosphorus of the intracellular granules, confirming that these structures are indeed polyphosphate-like granules (pPLGs) (Figure 7). The main counter ion found was magnesium (Supplementary Table S3).

In order to explore the difference between the two species, D. radiodurans and D. indicus, to the arsenate resistance we analyzed the arsenate reductases that belong to the ars operon, which are able to convert As(V) into As(III) (Mukhopadhyay and Rosen, 2002). It is predicted that D. indicus possesses three genes that encode cytosolic arsenate reductases, DiArsC1, DiArsC2, and DiArsC3 while D. radiodurans possesses two arsenate reductases, DrArsC1 and DrArsC2 (Figure 8). There are three families of cytosolic arsenate reductases: one belongs to the E. coli R773 plasmid and comprises glutaredoxin (Grx)-coupled enzymes, while another class belongs to the Staphylococcus aureus Plasmid pI258 comprising the thioredoxin (Trx)-coupled enzymes (Ji et al., 1994; Oden et al., 1994; Shi et al., 1999; Messens et al., 2002a; Messens and Silver, 2006) and the final class belongs to eukaryotic organisms, encompassing the ACR2p from Saccharomyces cerevisiae (Mukhopadhyay and Rosen, 1998; Mukhopadhyay et al., 2000). A sequence analysis of the different protein sequence, shows that the three arsenate reductases from D. indicus, DiArsC1, DiArsC2 and DiArsC3 belong to three branches (Figure 8). Interestingly, DiArsC1 and DrArsC1 branch closely to the E. coli ArsC from the Grx-linked prokaryotic family (Figure 8; Gladysheva et al., 1994; Rosen, 1999). DiArsC3 and DrArsC2 branch closely to Thermus thermophilus ArsC that belongs to the Trx-linked prokaryotic ArsC reductases family. DiArsC2 branch closely to Desulfovibrio alaskensis G20 ArsC3 (Nunes et al., 2014), which are in the main branch with S. aureus pI258 ArsC (Ji et al., 1994; Rosen, 1999; Figure 8). Since DiArsC2 did not cluster together with any of D. radiodurans ArsCs, we decided to further unveil the structural mechanism for arsenate detoxification.

The crystal structures of DiArsC2 and DiArsC2-As were determined and refined at resolutions of 1.65 and 1.50 Å, respectively. The space group of both crystal structures is P2₁, presenting two molecules in the asymmetric unit. DiArsC2 and DiArsC2-As structures were refined to final R_factor/R_free values of 0.224/0.254 and 0.152/0.172, respectively (Supplementary Table S2). The superposition of both chains based on the secondary structure matching algorithm yields a root-mean-square deviation (r.m.s.d.) of 0.6 Å between the superimposed C^α carbon atoms and 127 aligned amino acid residues. The largest deviations are in a loop region between residues 88–100 that presents high flexibility and for DiArsC2 it lacks electron density for this region (Figures 9A,B). In fact, in DiArsC2 it was not possible to build the protein chain between the residues 89–96 for Chain A and 89–93 for Chain B (Figure 9B). Thus, the structural analysis for both structures were performed with Chain B. Ensemble refinements (Burnley et al., 2012) were performed for DiArsC2 and DiArsC2-As (leading to R_factor/R_free values of 0.221/0.255 and 0.160/0.177 respectively) where the most flexible regions were residues 31–36 and the loop 86–97 that contains the two conserved cysteines (Cys87 and Cys94) involved in As(V) reduction (Figures 9C, 10). The loop that contains the conserved catalytic Cys15 does not show a high degree of structural flexibility between the two structures here presented.

DiArsC2 structures are composed of an α/β domain and four parallel β-strands which are flanked by three helices and two small helices (Figures 10, 11). These crystal structures are similar to the arsenate reductases from B. subtilis (Bennett et al., 2001), (PDB 1JL3) and S. aureus, (PDB 1LJL) (Messens et al., 2002b). The crystal structures of DiArsC2 and DiArsC2-As presented an r.m.s.d. value of 1.3 Å with 127 aligned amino acid residues with the arsenate reductases from B. subtilis (Bennett et al., 2001) (PDB 1JL3). The arsenate reductase from S. aureus, (PDB 1LJL) presents an r.m.s.d. value of 1.0 Å and 119 aligned amino acid residues with DiArsC2 and an r.m.s.d. value of 1.2 Å and 121 aligned amino acid residues with DiArsC2-As.

The reduction mechanism of arsenate involves three cysteines, Cys15, Cys87, and Cys94. In the case of the DiArsC2-As structure, a disulfide bridge between Cys87 and Cys94 is formed. Although the protein was co-crystallized in the presence of arsenate (AsO₄³⁻), the density observed close to the catalytic Cys15 suggests the presence of AsHO₃²⁻. This molecule is located in a positively charged pocket and establishes a network of hydrogen bonds with the main-chain nitrogen atoms of residues G₁₇NTARS₂₂, Thr16 and Ser22 hydroxyl and Arg21^Nε groups, as well as water molecules (Figures 11A,B). Cysteine 15 is part of a well-known phosphate-binding loop (motif, CX₅R) of LMW PTPases which is conserved in the Trx-linked ArsC family (Figure 10; Bennett et al., 2001). In the native structure, there is no species present in this pocket, and all the residues in the phosphate binding loop G₁₇NTARS₂₂, are in a similar position as in the Arsenate-bound structure. Although no electron density was observed for the region 89–93 (Chain B), weak density for Cys94 was obtained (Figure 11C). Cys15 and Cys87 are involved in the reduction of As(V) to As(III), and in DiArsC2 structure both residues were refined with a double side-chain conformation, suggesting the presence of a mixed state (Figure 11C).