Bacteria were routinely grown in LB medium (Sambrook and Russell, 2001) with shaking at 37°C. Growth was started by inoculating fresh LB medium with 1:100 dilutions of overnight cultures. For growth in solid medium, LB-agar (2%) plates were incubated overnight at the same temperature. When required, cells were grown in the presence of chloramphenicol (25 μg ml^-1) or ampicillin (100 μg ml^-1; Table 1). Expression of cloned genes (ASKA collection) was induced with 1 mM IPTG for 6–10 h.

Staphylococcus haemolyticus BNF01 (Arenas et al., 2014b) ahpF and trxB genes were amplified using specific primers (Supplementary Table S1). Since E. coli lacks the MerA gene and given that bioinformatic data suggested that MerA could display TR activity, the merA gene was amplified from the environmental plasmid pTP6¹. PCR products were individually inserted into the vector Champion^TM pET101 Directional TOPO Expression (Invitrogen^®) to generate plasmids pET/ahpF, pET/trxB, and pET/merA. Correct insertion of genes was checked by PCR using specific primers (Supplementary Table S1). Their identity was confirmed by DNA sequencing.

Escherichia coli flavoproteins were purified using cells from the ASKA collection (Kitagawa et al., 2005). S. haemolyticus BNF01 genes encoding the selected flavoproteins and merA from plasmid pTP6 were cloned into the pET101/D-TOPO vector (Invitrogen^®) and transformed into E. coli BL21 (DE3). Cells were grown at 37°C to OD₆₀₀ ∼ 0.6 and induced with 1 mM IPTG for 6–10 h with vigorous shaking. Cells were suspended in binding buffer (20 mM sodium phosphate, pH 7.4, 0.5 M NaCl, 20 mM imidazole), supplemented with 0.1 mM PMSF and disrupted by sonication. The cell debris was discarded by centrifugation at 14,000 × g for 15 min at 4°C and His-tagged proteins present in the crude extracts were purified by affinity chromatography columns (HisTrap HP, GE Healthcare^®). After extensive washing with binding buffer, bound proteins were eluted with elution buffer (same as binding buffer but containing 0.5 M imidazole). Proteins were dialyzed against 50 mM Tris-HCl buffer pH 7.4 for 2 h. Protein concentration was determined as described earlier (Bradford, 1976), and SDS-PAGE was carried out to assess enzyme purity.

Tellurite reductase activity was determined in a final volume of 200 μl of 50 mM Tris-HCl buffer pH 7.4, 0.15 mM K₂TeO₃, 1 mM NAD(P)H, 1 mM β-mercaptoethanol (TR buffer), and the enzyme (50 μg protein). Production of elemental tellurium was monitored at 500 nm using a Tecan Infinite^® M200 PRO plate reader. One enzyme unit was defined as the amount of enzyme required to increase the OD₅₀₀ by 0.001 in 1 min under the assay conditions as described earlier (Chiong et al., 1988; Castro et al., 2008; Arenas et al., 2014a). The effect of pH on tellurite reduction was assessed by determining TR activity at 37°C using the following buffers at 50 mM: Na₂HPO₄-citric acid (pH 3.0–6.0), Tris-HCl (pH 7.0–9.0), glycine/NaOH (pH 10.0), carbonate/NaOH (pH 11.0) and KCl/NaOH (pH 12.0). The effect of temperature on TR activity was determined at the optimal pH for each enzyme in a temperature range that included 25, 30, 37, 42, and 50°C. The apparent kinetic parameters were determined in triplícate using the same reaction mixture as for tellurite reduction at pH and temperature optima for each enzyme; tellurite concentrations varied from 0 to 2 mM. Maximal velocity, apparent K_m and K_i were determined by fitting non-linear regression using the GraphPad Prism Version 7.01 program (Table 1).

In vivo synthesis of TeNS was performed using E. coli (ASKA collection, Supplementary Table S1) grown to exponential phase (OD₆₀₀ ∼ 0.3), induced with 1 mM IPTG, treated with 0.5 μg/ml TeO₃^2- for 4 h and centrifuged at 6,000 × g for 10 min. The bacterial pellet was sent to the Advanced Microscopy Unit (AMU) at Pontificia Universidad Católica de Chile for thin sectioning and transmission electron microscopy (TEM) analysis. In vitro TeNS production was carried out for 60 min in TR buffer at pH and temperature optima using 250 μg/ml of purified enzyme. Samples were analyzed by TEM using a Philips Tecnai 12 TEM.

The hydrodynamic diameter of TeNS (in vitro synthesis) was determined at room temperature (25°C) using a Zetasizer Nano ZS Marlvern instrument. Values were calculated from three independent measurements of 20 repetitions each. Tellurium in TeNS generated in vitro was quantified by optical emission spectrometry-inductively coupled plasma (ICP-OES) using a Perkin Elmer 2000 DV optimum with a wavelength of 214.281 nm corresponding to tellurium, as described previously (Pugin et al., 2014). A calibration curve (1–200 μg/ml) in 2% ultrapure HNO₃ was constructed using pure, commercially available tellurium (Sigma–Aldrich). In vitro-generated samples were sedimented (13,000 × g for 90 min), washed two times with milliQ water and suspended in 10% HNO₃. Once completely solubilized, they were filtered through 0.2 μm pore membranes and analyzed by ICP-OES.

As mentioned, several enzymes able to reduce tellurite have been identified. In this line, it is intriguing that enzymes catalyzing very different biological reactions are capable of tellurite reduction. These proteins show very low amino acid sequence identity (<30%) and no obvious conserved motifs.

To look for domains and/or functional sites that may be common to these proteins, they were independently characterized using bioinformatic resources that included Prosite (Sigrist et al., 2013), InterPro (Mitchell et al., 2014), SCOP (Andreeva et al., 2008), CATH (Sillitoe et al., 2013), and Pfam (Finn et al., 2014). InterPro and Pfam showed the presence of two groups of proteins, one exhibiting the pyridine nucleotide-disulphide oxidoreductase [FAD/NAD(P)-binding] domain (characteristic of flavoproteins) and the second exhibiting the molybdopterin oxidoreductase 4Fe–4S domain, which is found in a number of reductase/dehydrogenase families (Supplementary Table S2).

The Prosite database allowed identification of the pyridine nucleotide-disulphide oxidoreductase class-I active site (PS00076), encompassing the amino acid pattern G-G-x-C-[LIVA]-x(2)-G-C-[LIVM]-P, and also the class-II active site C-x(2)-C-D-[GAS]-x(2,4)-[FYA]-x(4)-[LIVMAT]-x(0,1)-[LIVM](2)-[GI]-[GDS]-[GRD]-[DN] (PS00573). Another interesting organization found in some of these proteins was the “ferredoxin-type iron-sulfur binding” domain, which displays the pattern C-x-{P}-C-x(2)-C-{CP}-x(2)-C-[PEG]. All the above mentioned domains contain nearby cysteine residues at the active site, which play a critical role in catalysis (see below).

Since it was previously shown that flavoproteins such as GorA (Pugin et al., 2014) and E3 (Castro et al., 2008) display TR activity, we decided to carry out a more in-depth analysis of the FAD/NAD(P)-binding and pyridine nucleotide-disulphide oxidoreductase domains as being -at least- partly responsible for tellurite reduction. Using these domains as a signature pattern to identify putative TR enzymes, a cross-search for them in E. coli proteins was carried out. Utilizing different data bases (Uniprot, SCOP, PFAM, Prosite) a number of enzymes bearing the referred motifs were identified. Eight of them, namely TrxB, AhpF, glutamate synthetase (GltD), putative oxidoreductase (YkgC), GorA, nitrite reductase (NirB), flavorubredoxine reductase (NorW), and mercury reductase (MerA) were selected for further analysis. Unfortunately, we were unable to purify NirB and GltD thus hampering their characterization.

Given that numerous enzymes display these motifs, starting point was defined strains from the E. coli ASKA collection that overproduced enzymes exhibiting vicinal Cys residues as well as the FAD/NAD(P) motif; these were then used to purify enzymes to be tested for TR activity (see below). Enzymes lacking the FAD and/or NAD(P)⁺ binding domains or the catalytic cysteines (SthA, PreT, PreA, HcaD, and Glf) were also tested for TR activity. Unfortunately, we were unable to purify NirB and GltD thus hampering their characterization.

Six of the above proteins (TrxB, AhpF, YkgC, GorA, E3, and NorW) were purified after being overproduced in E. coli (Supplementary Table S1, Supplementary Figure S1A). Since E. coli lacks merA, it was amplified from the environmental plasmid pTP6 (Smalla et al., 2006), cloned and over- expressed in this bacterium (Supplementary Table S1, Supplementary Figure S1B). To assess if there were differences between TR enzymes from tellurite-sensitive (E. coli) and tellurite-resistant (S. haemolyticus BNF01; Arenas et al., 2014b) organisms, the ahpF and trxB genes from this strain were cloned, over-expressed and the respective proteins purified (Supplementary Table S1, Supplementary Figure S1C). Excepting for NorW (Supplementary Figure S1A, lane 7), all other proteins were obtained with a purity of >90%, as judged from denaturing polyacrylamide gel electrophoresis.

Purified proteins were assessed for TR activity in the presence of NADH or NADPH as electro donor. While GorA, E3, and TrxB used preferentially NADPH, AhpF, YkgC, and NorW used NADH as cofactor (Table 1). Next, the effect of pH and temperature on TR activity was determined for each flavoprotein using the preferred pyridine cofactor.

Most enzymes showed maximal tellurite reduction at pH 8.0–10.0 (Figure 1); the exception was E3, which showed maximal TR activity at a rather acidic pH (Figure 1B). While GorA exhibited the highest TR activity (∼30,000 U/mg protein, Figure 2C), NorW and YkgC displayed the lowest (∼660 and 870 U/mg protein, respectively; Table 1). Tellurite-reducing activity was extremely low and almost undetectable at pH 3.0–4.0 (Figure 1). As expected, PreT, PreA, HcaD, and Glf did not show TR activity under these conditions (not shown).

The effect of temperature was assessed for GorA, AhpF, YkgC, TrxB, NorW, and E3. At their optimal pH values, all of them exhibited a similar behavior in the range of 25–50°C, with peak activity at ∼37°C (Figure 2). GorA showed ∼11.5-, 22.6-, 34.5-, 36.8-, and 48.5-fold more tellurite-reducing activity than E3, TrxB, AhpF, YkgC, and NorW, respectively (Table 1). Kinetic parameters such as K_m, K_i, and V_max are shown in Table 1. While AhpF, YkgC, and NorW showed Michaelis–Menten kinetics, GorA, E3, and TrxB exhibited a behavior compatible with substrate (tellurite) inhibition (not shown). As expected, enzymes exhibiting higher TR activity displayed lower K_m and higher V_max values.

Staphylococcus haemolyticus BNF01 genes encoding TrxB and AhpF were cloned and the proteins purified to determine if they displayed TR activity. TrxB and AhpF displayed maximal NADH-dependent TR activity at pH 9.0 and 40°C (Figures 3A,B). As with GorA, AhpF, YkgC, TrxB, NorW, and E3, tellurite reduction by TrxB and AhpF from BNF01was inhibited by divalent metals such as Zn²⁺, Ni²⁺, and Co²⁺ (not shown). In turn, pTP6-encoded MerA showed TR activity at pH 7.0–9.0 in the presence of NADPH (Figure 3C).

The primary amino acid sequence of these flavoproteins was examined using the PROSITE database. Three common pattern groups were identified: PS00573, PS51354, and PS00076. TrxB and AhpF displayed a similar active site, with the class-II pyridine nucleotide-disulphide oxidoreductase, and GorA, E3, and YkgC with the class-I pyridine nucleotide-disulphide oxidoreductases (Supplementary Table S2).

In spite that the amino acid sequence of the two classes of enzymes showed low identity (Supplementary Figure S2), they actually did exhibit high structural similarity exhibiting a global RMSD of 3.61 Å (Figure 4A). Although the SSAP score within proteins of the same class was close to 90 points, that of proteins of different classes is still high (72 points; Supplementary Table S3).

In all cases, the FAD molecule was bound to the active site by the vicinal cysteine residues and adopted a similar conformation (Figures 4B,C). Although both motifs I and II were characterized using FAD to convey reducing power, the spatial localization of the disulfide bridge was different between them (Figures 4B,C, Supplementary Table S4).

It is intriguing that to date no biological function has been reported for E. coli YkgC, one of the flavoenzymes exhibiting TR activity. Since there are no data regarding its spatial structure, the 3D structure of the protein was built by means of comparative modeling, which resulted in a configuration very similar to that of dihydrolipoyl dehydrogenase (4JDR), with a SSAP score of 89.6 (Supplementary Table S3). Although not exactly in the same position, FAD was predicted to bind to the same zone as in the other flavoproteins, with the flavin ring located near to the two cysteines (Figure 5), as predicted by docking (Autodock software).

Next, the relevant distances at the active site of the flavoproteins were determined from the 3D structure. Supplementary Table S4 summarizes the distances that are predicted to be important for the displayed TR activity. In general, shorter distances between cysteine residues were associated with higher TR activity (distance 4, Supplementary Table S4), as it occurs in GorA and E3. This could be even more relevant for TR activity that interaction between cysteines and the FAD moiety (distances 2, 3, 5, Supplementary Table S4).

The synthesis of tellurium-based nanostructures is a relatively new field that has gained interest because of the multiple potential applications of the TeNS. Along this line, after characterizing TR activity, it was very interesting to analyze the reduction products. In vitro synthesis of nanostructures was carried out using GorA, YkgC, E3, and AhpF, as described in Section “Materials and Methods.” Dynamic light scattering was used to determine the hydrodynamic diameter of the in vitro generated TeNS.

Structures generated by E3 showed a wide size distribution, from a few nanometers to micrometers. Although with large deviations, the most abundant structure observed had a hydrodynamic diameter of 254.2 nm, with an average PDI (polydispersity index) of 0.159 ± 0.041 (Figure 6B).

Structures synthesized using GorA exhibited a Gaussian size distribution (Figure 6C). The most abundant nanostructure showed a maximum hydrodynamic diameter of 70.37 nm (PDI 0.217 ± 0.004). Similarly, AhpF (Figure 6A) generated structures of 78.06 nm (PDI 0.345 ± 0.013). However, their size distribution showed a smoother attenuation at larger sizes, resulting in a higher number of structures exceeding 200 nm. On the other hand, tellurium structures generated by YkgC showed a size distribution that was similar to that generated by E3, with a peak at 162.2 nm and a PDI of 0.125 ± 0.011 (Figure 6D).

To determine if there is a correlation between TR activity and the amount of produced NS, the amount of tellurium present in the NS was quantified by ICP-OES. In general, Te content of NS correlated well with TR activity (Table 1, Figure 6E). For instance, TeNS produced with GorA contained 3.25 times as much Te as those produced with YkgC (Figure 6E).

Te-containing nanostructures synthesized using GorA, YkgC, and AhpF were analyzed by TEM (Figure 7). The results showed that they varied in size and morphology depending on the synthesizing enzyme. In this line, AhpF generated rather rounded structures of 50–100 nm with irregular edges that doubled the size of the center core (Figure 7A, right). On the other hand, GorA-synthesized TeNS exhibited a compact morphology with a roundish shape and amorphous edges. Size and shape differences did occur, maintaining the general characteristics described above (Figure 7B). Structures produced by YkgC showed larger average sizes, often exceeding 100 nm. The morphology was rather elongated and irregular rods with numerous tips outlining its contour were seen (Figure 7C).

To determine if such particles were also synthesized in vivo, the synthesis of TeNS was assessed by electron microscopy in ultrathin sections of E. coli overproducing tellurite reductases (TRs; Supplementary Figure S3). Membrane damage and electron-dense elements (probably Te⁰-containing structures) were seen only in tellurite-exposed cells that over synthesized GorA, YkgC, or TrxB (Supplementary Figures S3C–E). As seen in vitro, strains overproducing YkgC also exhibited larger electron-dense structures (Supplementary Figure S3D).

Finally, electron-dense elements were not observed in cells overproducing AhpF and E3, indicating that TR activity does not necessarily correlate with the generation of electron-dense deposits in vivo (Supplementary Figures S3A,B).