The facultative anaerobic bacterium Sinorhizobium sp. GW3 (Genome GenBank No. AUSY00000000.1) was mainly used in this study. This strain was isolated by our research group from arsenic contaminated groundwater sediments in Shanyin City, Shanxi Province, China. It could oxidize As(III) aerobically (Fan et al., 2008). The anaerobic As(III) oxidation of strain GW3 was performed using KMnO₄ method as described by Cai et al. (2009). For aioA gene knock-out and complementation study, strain A. tumefaciens GW4 was used (Fan et al., 2008; Wang et al., 2015). Genomic analyses of putative genes involved in denitrification, As(III)/Sb(III) oxidation/resistance, anaerobic selenite-, sulfate-, iron- and chromate- reduction and transportation were conducted through BlastN and BlastP in the genome of Sinorhizobium sp. GW3 on the NCBI website¹.

For aerobic Sb(III) oxidation experiments, bacterial strains were grown in chemically defined medium (CDM) (Weeger et al., 1999) containing 0 or 20 μM K₂Sb₂(C₄H₂O₆)₂ [Sb(III)] with shaking at 28°C. Overnight cultures of Sinorhizobium sp. GW3 strain were inoculated into 5 mL of CDM in the presence of 20 μM Sb(III) and incubated at 28°C with 120 rpm shaking. When the OD₆₀₀ reached 0.5–0.6, the cultures were each inoculated into 100 mL of CDM with 20 μM Sb(III) and different amounts (0, 1, and 5 mM) of KNO₃⁻. At designated times, the culture samples were centrifuged (13,400 × g, 5 min) and subsequently filtered (0.22 μm filter) to measure Sb(III) oxidation and denitrification. The Sb(III)/Sb(V) content was measured by HPLC-HG-AFS (Beijing Titan Instruments Co., Ltd., China) (Li et al., 2013), and the concentration of NO₃⁻/NO₂⁻ was measured by HPLC (HPLC 2690 series, Waters, MA, United States) according to the method described by Croitoru (2012).

For denitrification and anaerobic Sb(III) oxidation experiments, bacterial strains were cultured in CDM with 0 or 100 μM Sb(III) under a gas phase of N₂/CO₂ (80/20). The cultures were incubated at 28°C without shaking, and all operations were carried out in an anaerobic chamber (Vinyl Glove Box, Coy Laboratory Products) (Wang et al., 2018). First, strain GW3 was cultured anaerobically in CDM with 1 mM nitrate to investigate whether nitrate could be used as the electron acceptor for anaerobic growth. Samples were taken regularly to measure the NO₃⁻ and NO₂⁻ content. Then, bacterial cells were harvested by centrifugation (6,000 × g, 5 min) and washed twice with anoxic physiological saline after 10 days of cultivation. The resuspension was used as an inoculum for anaerobic Sb(III) oxidation. For nitrate addition, we used 1 mM KNO₃, which is similar to the amount of nitrate used in anaerobic As(III) oxidation (Zhang et al., 2017). To further investigate the effect of nitrate on anaerobic Sb(III) oxidation, different concentrations (0, 1, and 5 mM) of nitrate were added to the culture. Additionally, to observe the nitrate-induced oxidation of anaerobic Sb(III), 100 μM Sb(III) was used for the Sb(III) oxidation analysis instead of 20 μM Sb(III). Thus, the rate of anaerobic Sb(III) oxidation was determined in CDM amended with 100 μM Sb(III) and 0, 1, or 5 mM nitrate as the electron acceptor. The serum bottles (100 mL) containing 50 mL of CDM were flushed with a N₂ gas stream and sealed with butyl rubber stoppers and aluminum covers. After autoclaving, the stock cultures were each inoculated into the sealed serum bottles via injection to a final OD₆₀₀ of 0.2, and the cultures were incubated at 28°C without shaking. At designated times, 100 μL of culture was withdrawn anoxically with a syringe for measuring the Sb(III)/Sb(V) and NO₃⁻/NO₂⁻ content, as described above.

For anaerobic Fe(II) oxidation assays, stock cultures of strain GW3 were each inoculated into 50 mL of CDM containing 1 or 5 mM nitrate and 1 mM Fe(II). The FeCl₂ stock liquid was prepared in an anaerobic chamber and subsequently injected into the culture. Strain GW3 was inoculated as described above. The experiment was performed in triplicate with no GW3 inoculum as a control, and all cultures were incubated at 28°C in the dark. The samples were taken periodically to determine the degree of Fe(II) oxidation and nitrate reduction. The dissolved Fe(II) content was quantified by the ferrozine colorimetric assay at 562 nm using an EnVision^® Multimode Plate Reader (PerkinElmer) (Stookey, 1970).

To investigate the expression of genes involved in Sb(III) oxidation and resistance under oxic conditions, overnight cultures of strain GW3 were inoculated into 100 mL of CDM with 0 or 100 μM Sb(III) at 28°C with 120 rpm shaking. After 24 h of incubation, bacterial cells were harvested for total RNA extraction. For anoxic gene expression analysis, bacterial cells were each inoculated into oxygen-free CDM with 0 or 100 μM Sb(III) and incubated at 28°C without shaking. After incubation for 4 days, bacterial cells were harvested and treated with Trizol reagent (Invitrogen).

Total RNA was extracted according to the manufacturer’s instructions (Invitrogen). Genomic DNA was removed by treatment with RNase-free DNase I (Takara) at 37°C, and the concentration of RNA was monitored using a spectrophotometer (NanoDrop 2000, Thermo). Reverse transcription was performed with 300 ng of total RNA for each sample using the RevertAid First Strand cDNA Synthesis Kit (Thermo) (Wang et al., 2012). Subsequently, the obtained cDNA was diluted 10-fold and used as a template for qRT-PCR, which was carried out in an ABI VIIA7 system in 0.1 mL Fast Optical 96-well Reaction Plates (ABI). The primers used for qRT-PCR are listed in Supplementary Table S1. Each reaction was replicated three times to estimate error. The 16S rRNA gene of strain GW3 was used as an internal control. First, standard dilution curves were generated for each pair of primers. Then, qRT-PCR was conducted to obtain Ct values for each sample. The slope (S) of each standard curve was used to calculate the primer amplification efficiency (E) with [log (1+E) = 1/S]. The Ct values of target cDNAs were each normalized with the cDNA Ct value of 16S rRNA to generate the ΔCt value of each sample in the control and Sb(III)-treatment groups. Subsequently, the ΔCt values of cDNAs in the control groups were each subtracted from those in the Sb(III) treatment groups. The relative gene expression ratios were analyzed using the 2^−ΔΔCT method (Livak and Schmittgen, 2001). Significance analysis was performed by one-way ANOVA.

To investigate the role of aioA in anaerobic Sb(III) oxidation, A. tumefaciens GW4 and the corresponding aioA-deletion and aioA-complementation strains were used to determine the anaerobic Sb(III) oxidation efficiency. A. tumefaciens strains were each inoculated into 50 mL of CDM containing 1 mM nitrate and 100 μM Sb(III) under anoxic conditions and cultured as described above. At designated times, samples were taken for measuring the Sb(III)/Sb(V) content.

A stock culture of strain GW3 was inoculated into 50 mL of CDM with 1 mM Fe(II), 100 μM Sb(III) and 1 mM nitrate under anoxic conditions. CDM containing 100 μM Sb(III) and 1 mM nitrate was used as a control. All cultures were incubated for up to 10 days at 28°C in the dark without shaking. Changes in Sb(III), Fe(II) and nitrate concentrations were determined as described above. Precipitates from anaerobic Fe(II)- and Sb(III)-oxidizing cultures were collected at the end of the experiment by centrifugation (13,400 × g, 5 min) and freeze-dried overnight. The elemental composition of the samples was investigated by energy-dispersive X-ray spectroscopy (EDS) using an Oxford Instruments model Inca X-sight EDS detector, which is coupled with the SEM (FEI Inspect F50). Powder X-ray diffraction (XRD) analysis was conducted using a Bruker D8 Advance X-ray diffractometer with Cu Kα as the X-ray source (40 mA, 20 kV). The elemental chemical status and compositions were analyzed with X-ray photoelectron spectroscopy (XPS) using Al Kα (1486.6 eV) excitation.

Bacterial cells were cultured as described above. Stock cultures of strain GW3 were inoculated into 50 mL of CDM containing 1 mM nitrate and 1 mM Fe(II) under anoxic conditions. After 10 days of static cultivation at 28°C in the dark, the orange precipitates produced by strain GW3 were harvested through centrifugation (13,400 × g, 5 min); these precipitates included both Fe(III) minerals and bacterial cells. Subsequently, the mixture was resuspended in PBS (pH 7.0) and injected into oxygen-free CDM containing 100 μM Sb(III). Meanwhile, the precipitates were taken immediately, washed three times with PBS, and used for plate counting to detect the bacterial numbers. The bacteria cultured without Fe(II) were used as a control. During cultivation, 100 μL of culture was withdrawn anoxically with a syringe for measuring the Sb(III) content at designated times. To test whether nitrite or Fe(III) could oxidize Sb(III) abiotically, 50 mL of CDM was added to the serum bottles under an atmosphere of N₂. After sterilization and sealing, 1 mM NaNO₂ or 1 mM FeCl₃ were amended with 100 μM Sb(III) and incubated for up to 10 days at 28°C under anoxic conditions. Samples were regularly taken to monitor the Sb(III)/Sb(V) and NO₂⁻ content.

Strain GW3 is a facultative anaerobic bacterium which can oxidize As(III) under both oxic and anoxic conditions (Fan et al., 2008; Supplementary Figure S1). The genome of strain GW3 contains putative genes responsible for aerobic As(III) and Sb(III) oxidation [the As(III) oxidase genes aioAB, Sb(III) oxidase gene anoA and catalase gene katA] (Supplementary Table S2). However, it has no arxA gene, which encodes anaerobic As(III) oxidase (Zargar et al., 2012). The genes involved in As(III) and Sb(III) resistance were also found, such as As(III)/Sb(III) efflux gene acr3 and As(V) reductase gene arsC (Rosen, 1995; Supplementary Table S2). These As(III)/Sb(III) oxidation and resistance genes are all single copy. It has been reported that the multicopper oxidase may play a role in bacterial anaerobic Fe(II) oxidation (He et al., 2016). Expectably, we found three multicopper oxidase genes in strain GW3 (Supplementary Table S2). In addition, there are also genes involved in denitrification and sulfate reduction. Interestingly, the strain GW3 genome also harbors genes which might contribute to anaerobic selenite reduction, iron reduction and chromate reduction (Supplementary Table S2), suggesting that strain GW3 may have variable anaerobic functions.

We first assessed the effect of nitrate addition on bacterial Sb(III) oxidation under oxic conditions. Strain GW3 was inoculated into CDM containing 20 μM Sb(III) and different concentrations (0, 1, and 5 mM) of nitrate, and the culture was aerated by shaking at 28°C. As shown in Figure 1A, strain GW3 could oxidize Sb(III) aerobically; however, there was no significant difference in Sb(III) oxidation rates upon the addition of different concentrations of nitrate (Figure 1A). Nitrate reduction experiments showed that nitrate was only slightly reduced to nitrite regardless of the concentration of nitrate added (Figure 1B). These results suggested that nitrate has no effect on bacterial aerobic Sb(III) oxidation and strain GW3 tends to use O₂ as the electron acceptor rather than nitrate under oxic conditions.

Under anoxic conditions in CDM, strain GW3 could reduce NO₃⁻ to NO₂⁻, and the contents of NO₂⁻ decreased (Supplementary Figure S2A); this suggests that the NO₂⁻ produced flowed into subsequent denitrification processes to generate N₂O or N₂, consistent with the presence of nir, nor, and nos genes in the strain GW3 genome (Supplementary Table S2). Upon addition of Sb(III), the denitrification efficiency increased compared to the efficiency in conditions without Sb(III); this was apparent because the amount of NO₂⁻ produced with 1 mM nitrate and 100 μM Sb(III) was less than that produced with 1 mM nitrate only (Supplementary Figure S2B). This suggests that the denitrification and Sb(III) oxidation processes enhance electron transfer. Thus, the Sb(III) oxidation rate was significantly increased by the greater amounts of nitrate (Figure 2A). We did not observe any abiotic Sb(III) oxidation by nitrite (Supplementary Figure S3A), and the content of nitrite was constant during 10 days of cultivation (data not shown). Additionally, we tested whether Sb(III) could be oxidized abiotically by FeCl₃. The results showed that the Sb(III) content was not decreased and Sb(V) was not observed in the presence of FeCl₃ (Supplementary Figure S3B). The results indicated that the intermediate products of denitrification in strain GW3 and FeCl₃ could not oxidize Sb(III) abiotically.

To investigate the potential mechanism of Sb(III) oxidation in strain GW3, the transcript levels of aioA, anoA, and katA were measured under both oxic and anoxic conditions using qRT-PCR. Additionally, we assessed the transcript levels of the As(III)/Sb(III)-resistance genes acr3 and arsC. The original qRT-PCR data and primer efficiency are shown in Supplementary Table S3 and Supplementary Figure S3, respectively. Since the primer amplification efficiencies for 16S rRNA and other genes were similar (∼5%) (Supplementary Figure S4), primer amplification efficiencies were not included in the calculation of the relative gene expression ratios.

The results showed that the transcript levels of anoA, katA, acr3, and arsC all significantly increased upon the addition of Sb(III), while the transcript level of aioA was not increased by Sb(III) under oxic conditions (Figure 3A). These findings hint that the gene products of anoA and the cellular oxidative stress response may substantially contribute to aerobic Sb(III) oxidation in strain GW3, which is consistent with our previous observations in A. tumefaciens GW4 (Li et al., 2015, 2017a,b). In contrast, under anoxic conditions, the addition of Sb(III) had no significant effect on the transcript levels of anoA and katA, while aioA, acr3, and arsC levels were promoted by Sb(III) (Figure 3B). The results suggested that AioA may play a role in bacterial anaerobic Sb(III) oxidation, and Acr3 and ArsC may be involved in Sb(III) resistance under both oxic and anoxic conditions.

To further analyze the role of aioA in anaerobic Sb(III) oxidation, strain A. tumefaciens GW4 and the corresponding aioA-deletion and aioA-complementation strains were used in the anaerobic Sb(III) oxidation experiments. Strain GW4 is closely related to strain GW3 based on the phylogenetic analysis of the 16S rRNA genes (Fan et al., 2008); furthermore, the amino acid sequence similarity of AioA in these two strains is 77%. We found that the oxidation efficiency of GW4-ΔaioA was 73% lower than that of wild-type strain GW4, and the complementary strain rescued the oxidation efficiency phenotype (Supplementary Figure S5). This suggests that AioA contributes to bacterial anaerobic Sb(III) oxidation in strain GW4 and may also contribute in the closely related strain GW3.

The ability of strain GW3 to oxidize Fe(II) under anaerobic nitrate-reducing conditions was also tested. The Fe(II) stock culture was prepared anaerobically in an anaerobic chamber and then injected into the culture. Compared with the sterile control, the bacterial Fe(II) oxidation rate increased as the concentration of added nitrate increased (Figure 4A). At the highest concentration of nitrate tested (5 mM nitrate), all of the Fe(II) was consumed within 10 days (Figure 4A), indicating that 5 mM nitrate could accept all electrons from 100 μM Fe(II) oxidation. Correspondingly, the NO₂⁻ content increased first and subsequently decreased (Figure 4B), consistent with the trend of nitrate reduction during anaerobic Sb(III) oxidation (Figure 2B). The addition of nitrate significantly stimulates bacterial Fe(II) oxidation, indicating that strain GW3 is also a nitrate-dependent, anaerobic ferrous-oxidizing bacterium. Additionally, orange precipitates were observed during cultivation (Figure 4B), indicating that Fe-containing minerals were formed during the Fe(II) oxidation process.

We assessed the effects of Fe(II) on bacterial anaerobic Sb(III) oxidation in CDM containing 100 μM Sb(III) and 1 mM nitrate with or without 1 mM Fe(II) under anoxic conditions. Sb(III) oxidation experiments showed that up to approximately 15.9 μM Sb(III) was reduced by the end of 10 days of cultivation in the presence of Sb(III) and nitrate, and 82.4 μM Sb(III) was reduced when Fe(II) was added (Figure 5A); Fe(II) significantly increased bacterial Sb(III) oxidation efficiency by 66.5%. Interestingly, the Sb(V) content increased first and subsequently decreased, and orange precipitates were observed in the culture with Fe(II). The increase in Sb(V) concentration did not compensate for the decrease in Sb(III) concentration, suggesting that Sb may adsorb or coprecipitate on the Fe-containing minerals. The Fe(II) oxidation experiment showed that up to approximately 0.75 mM Fe(II) was consumed by the end of 10 days of cultivation, and this value was significantly lower than the sterile control (Figure 5B). As the electron acceptor, nitrate was reduced to nitrite and other denitrification intermediates. Fe(II) addition accelerated the consumption of nitrate, since Fe(II) oxidation also needs nitrate as an electron acceptor (Figure 5C).

To investigate the chemical components of the precipitate, XRD analysis was conducted. However, we did not identify any characteristic peaks, indicating that the secondary mineral formed by strain GW3 was poorly crystalline (Supplementary Figure S6A). Thus, XPS analysis was further employed to determine the valence state of elements. The results showed that the valence state of iron mineral produced by strain GW3 in the presence of Fe(II) and nitrate was positive trivalent (Fe2p peak at 711.25 eV; Supplementary Figure S6B), indicating that strain GW3 could produce Fe(III)-containing minerals during anaerobic Fe(II) oxidation. In the presence of Fe(II), Sb(III) and nitrate together, the XPS spectra showed Fe2p and Sb3p peaks at 711.05 (Supplementary Figure S6C) and 531.71 eV (Supplementary Figure S6D), respectively, which were attributed to Fe(III) and Sb(V), respectively (Wu et al., 2018). These observations indicated that the removed Sb(III) was its oxidized product Sb(V) on the Fe(III)-containing minerals.

To identify the role of microbially produced Fe(III)-containing minerals in anoxic Sb(III) oxidation, the orange precipitates were harvested after 10 days of cultivation and then added to CDM in the presence of 100 μM Sb(III) under anoxic conditions. Since the precipitates need to be maintained in their original state with their original chemical properties, no extra steps were taken to remove cells. To determine whether there were living bacteria in the Fe(III)-containing mineral precipitates, a plate counting assay was employed to compare the number of bacteria in the precipitates and the control cells after 10 days of cultivation. The results showed that there was no difference in the cell counts between the precipitates and the cultures with only cells, and no Sb(III) oxidation was observed in the supernatant of the culture (data not shown).

To compare the contribution of biotic and abiotic Sb(III) oxidation, bacterial cells cultured without Fe(II) were also collected as a control. Compared with the control cells without precipitates, the amount of Sb(III) significantly decreased in the culture with microbially produced Fe(III)-containing precipitates (Figure 6A). However, the increased Sb(III) oxidation efficiency is due to the function of both Fe(III)-containing minerals (abiotic) and bacterial cells (biotic). As shown in Figure 6A, after 10 days of cultivation, the Sb(III) oxidation efficiency of bacterial cells was 18% (the biotic effort), while the Sb(III) oxidation efficiency of the precipitates which included Fe(III)-containing minerals and bacterial cells was 89%. Thus, the contribution of abiotic Sb(III) oxidation in this system was approximately 71%.

The precipitates were harvested for SEM and EDS analyses. Consistent with the previous results, the precipitates contained Fe and Sb (Figure 6B), indicating that the microbially produced Fe(III)-containing minerals greatly contributed to the increased anaerobic Sb(III) oxidation and the immobilization of Sb(V). As shown in Figure 6C, the surface of the precipitate was porous, which provided a sorption phase for Sb(V) immobilization. Additionally, Sb(III) could not be oxidized abiotically by FeCl₃, indicating that the artificial Fe(III) solution without secondary mineral formation from microbes has no functional roles in Sb(III) oxidation.