The basal growth medium had the following composition: 20 mM sodium fumarate, 20 mM NaNO₃, 4.7 mM NH₄Cl, 1.3 mM KCl, 2 mM MgSO₄, 0.2 mM NaCl, 1.2 mM NaHCO₃, 5 mM NaH₂PO₄, 0.1 mM CaCl₂ with sterile vitamins and trace elements prepared as described by Widdel and Bak (1992). Initial cultures were grown aerobically. These were then diluted 20-fold into the experimental growth medium. For each experiment, where applicable, the indicated amounts of exogenous sulfur sources, uranyl acetate (U[VI]) and K₂Cr₂O₇ (Cr[VI]) were added to the basal growth medium. Cultures were grown in a 100 well Bioscreen plate with each well containing 400 μL of diluted preculture. The Bioscreen plate was incubated anaerobically or aerobically as indicated at 30°C with continuous shaking in a Bioscreen C (Thermo Labsystems, Milford, MA). In the case of anaerobic growths, the Bioscreen C was placed within an anaerobic chamber (Plas Labs, Lansing, MI) in an atmospheric composition of 95% Ar and 5% H₂. Growth was monitored at an absorbance of 600 nm. All experiments were performed in biological duplicate or triplicate, errors bars represent standard deviations.

The P. stutzeri RCH2 RB-TnSeq mutant library containing 166,448 single transposon mutations with mapped genome locations (Wetmore et al., 2015) was recovered from a 1 mL, 10% glycerol stock at −80°C by incubating aerobically with shaking (150 rpm) at 30°C for 5.5 h in 125 mL of Luria broth with 50 μg/mL kanamycin in a shake flask to an OD₆₀₀ of 1.0. A sample of the recovery culture was saved as the pregrowth condition. The fitness growths were carried out in triplicate in the basal growth medium described above except 20 mM sodium lactate was used as a carbon source instead of fumarate, and 0.5 g/L yeast extract were added. No additional metals beyond the trace metals solution were added to the control cultures, while 120 μM K₂Cr₂O₇ (240 μM Cr[VI]) was added to the Cr challenge cultures and 3 mM uranyl acetate (U[VI]) was added to the U challenge cultures. The Cr and U concentrations were chosen as the concentrations that decreased growth (OD₆₀₀) under the fitness growth conditions approximately 50%. Cultures (5 mL) in sealed anaerobic Hungate tubes with an argon atmosphere were inoculated to an OD₆₀₀ of 0.02 before incubating with shaking (150 rpm) at 30°C for 5 h. At the end of growth, the OD₆₀₀ of each culture was recorded and the cultures were saved as postgrowth samples.

The processing of the cultures for DNA isolation, DNA sequencing, and sequence analysis were carried out as previously described with the BarSeq98 method (Wetmore et al., 2015). The Illumina HiSeq system was used to sequence PCR products. Strain fitness defined as the binary logarithm of the ratio of postgrowth to pregrowth relative abundances were calculated for each individual transposon insertion strain. Gene fitness values (w) were calculated as previously described (Wetmore et al., 2015), by averaging the fitness values for strains with insertions in a given gene. Quality control and normalization of data were performed as previously reported (Wetmore et al., 2015; Vaccaro et al., 2016a). The quality of each experiment was evaluated using several criteria. Including, the number of counts for the median gene needed to be greater than or equal to 50, and the median absolute difference in fitness between the two havles of the genes (mad12) was less than or equal to 0.5. Quality metrics for the fitness data are reported for each growth condition in Table S1.

The marker-exchange deletion strain, Δ2088, used in this study was constructed by conjugation of an unstable, marker-exchange plasmid into RCH2 with selection for Kan^r, in a manner previously described (Vaccaro et al., 2016a). The plasmid was made with the primers found in Table S2.

Cultures (300 mL) of P. stutzeri RCH2 wild-type and Δ2088 were grown in triplicate anaerobically on basal medium in sealed bottles at 30°C with or without 50 μM K₂Cr₂O₇ as indicated to late log phase. The cultures were harvested by centrifugation for 10 min at 4°C and 7,500 RPM. Cell pellets were moved into an anaerobic chamber (Coy Laboratory Products, Grass Lake, MI), with an atmospheric composition of 95% Ar and 5% H2, where they were washed once with 50 mM Tris pH 8.0, and suspended in the same buffer at a volume of 3 mL to 1 g of pellet (wet weight). Lysozyme (0.1 mg/mL) and deoxy ribonuclease (0.1 mg/mL) were added to the cell suspensions, which were then lysed by sonication and centrifuged at 10,000 RPM for 10 min. The supernatant was saved as cell free extract, and the Bradford assay was used to quantitate protein concentration (Bradford, 1976).

Total free thiol concentrations were measured from the cell free extracts using the Ellman assay (Sedlak and Lindsay, 1968). Briefly, 20 μL of cell free extract was combined with 75 μL of 30 mM Tris, 3 mM EDTA pH 8.2; 25 μL of 150 μM DTNB dissolved in methanol; and 400 μL of methanol. Samples were spun at 7.5 rpm for 5 min and 270 μL were transferred to a microplate for absorption measurement at 412 nm. A standard curve of reduced glutathione dissolved in 20 mM triethanolamine-HCl was used to convert absorbance values to moles of reduced thiol groups/g protein. Error is reported as the standard deviation between biological triplicates.

RCH2 cultures (5 mL) were grown anaerobically in crimp sealed Hungate tubes under a 100% argon atmosphere with shaking (150 rpm) at 30°C for 5 h. Cultures in triplicate contained either no additional metal, 3 mM uranyl acetate, or 120 μM K₂Cr₂O₇. Preparation of whole cells and nitrite reductase assays were performed as previously described (Thorgersen and Adams, 2016). Nitrite reductase values are reported as Units/mg protein, where a unit corresponds to 1 nmol of nitrite reduced/min.

Fitness experiments were conducted using the previously described RCH2 RB-TnSeq mutant library containing 166,448 single transposon mutations with mapped genome locations (Wetmore et al., 2015). The library was grown anaerobically using fumarate (20 mM) as the carbon source under denitrifying conditions with 20 mM nitrate in the presence of either no additional metal (control), 120 μM K₂Cr₂O₇ (Cr[VI]) or 3 mM uranyl acetate (U[VI]). Metal concentrations were selected that would inhibit growth of RCH2 by approximately 50%. Gene fitness values (w) are a measure of the population of mutants with disruptions in an individual gene relative to the overall mutant library population and these were calculated for each gene under all growth conditions as previously described (Wetmore et al., 2015). An increase in the relative abundance of mutants in a given gene in the test condition compaired to the pregrowth condition results in a positive fitness value and a decrease results in a negative fitness value. Both the Cr[VI] and U[VI] challenge growth gene fitness values were compared to the control grown in the absence of these metals to determine genes whose fitness was increased or decreased as a result of the metal challenge. Gene fitness values for the Cr[VI] (w_Cr) and U[VI] (w_U) challenges that have been corrected by the control gene fitness values (w_Cont) will be referred to by the symbols Δw_Cr and Δw_U. Genes with Δw_Cr and Δw_U ≤ −1.0 are listed in Tables 1, 2 respectively. Genes with Δw_Cr and Δw_U values ≥1.0 are listed in Tables S3, S4 respectively. Analysis of the potential influence of polar effects in the RCH2 mutant library was previously conducted (Wetmore et al., 2015), and it was conluded that they do not have a major effect. The false discovery rate for genes with phenotypes was also estimated for the library, and the rate was under 2% (Wetmore et al., 2015).

In response to exposure to Cr[VI], negative Δw_Cr values were observed for a large number of genes involved in DNA repair. These included recF (−1.9), recO (−1.9), recR (−1.8), and recA (−1.3), which are all part of the RecFOR pathway involved in homologous recombination to repair single strand breaks (Morimatsu and Kowalczykowski, 2003). The gene uvrC involved in nucleotide excision repair also had a large negative Δw_Cr (−1.4), as did the genes of the other proteins involved in excision repair, uvrA and uvrB, that had Δw_Cr values of −0.8 each. The gene of a helicase involved in double stranded repair processes, RecC, had a Δw_Cr value of −1.6. Interestingly, the gene encoding the SOS response repressor of LexA also had a large negative Δw_Cr value (−1.7) indicating that uncontrolled SOS repair under conditions of Cr[VI] stress is detrimental. The SOS-response cell division inhibitor gene sulA, also had a large negative Δw_Cr (−1.3).

Some of the DNA repair fitness data seen anaerobically with RCH2 in the Cr[VI] challenge mirror what has been seen in other microorganisms under aerobic conditions. For example, in E. coli, several SOS genes, including those encoding RecA and a cell division inhibitor, had increased transcription upon Cr[VI] challenge (Llagostera et al., 1986) and in C. crescentus, whole-genome transcriptional analysis in response to chromate toxicity revealed upregulation of some of the components of the excision repair system (Hu et al., 2005). In S. oneidensis MR-1, Cr[VI] exposure resulted in the upregulation of numerous DNA repair related genes including recO, that had a large negative Δw_Cr value (−1.9) here but homologs of several upregulated genes in S. oneidensis had no significant fitness value change in RCH2 including uvrD (0.1) (Chourey et al., 2006).

When grown aerobically in the presence of Cr[VI], many organisms induce production of enzymes known to detoxify ROS, including SOD, catalase (Ackerley et al., 2006), thioredoxin and glutaredoxin (Hu et al., 2005; Chourey et al., 2006). None of the homologs of these genes in RCH2 had significantly negative Δw_Cr values when exposed to Cr[VI] anaerobically, including those encoding SOD (0.2), five catalases (0.0–0.3), seven glutaredoxins (0.1–0.2), and five thioredoxins (0.0–0.2). This indicates that the anaerobic approach described herein is an efficient means to deconstruct the direct effects that chromium has on DNA damage from the indirect effects that are mediated by ROS when Cr[VI] exposure occurs in the presence of O₂.

In the present study, the RB-TNSeq library of RCH2 was challenged with Cr[VI] and U[VI] while growing under denitrifying conditions. As a consequence, genes encoding nitrate reductase and other accessory proteins required for nitrate reductase synthesis and function were expected to be important for fitness in RCH2 (Vaccaro et al., 2016a,b). This was confirmed as in the absence of either Cr or U (control) the nitrate reductase structural genes narGHI (psest_3483, psest_3485, and psest_3486) had w_Cont ranging from −1.8 to −2.4 (Tables 1, 2). In addition, genes encoding proteins involved in synthesis of the Mo-cofactor (Mo-co), a required component of the catalytic site in nitrate reductase, were also expected to have negative w_Cont, and this proved to be the case. The two Mo-co genes (psest_3479 and psest_3480) that are part of the nitrate reductase operon and encode MoeA and MoaB had w_Cont of −2.1 and −1.6 respectively. Similarly, two unlinked Mo-co genes that encode MoeB and MobA (psest_1115 and psest_1961) had w_Cont values of −2.6 and −2.4 respectively.

In the Cr challenge experiments, many genes involved in nitrate reduction had even larger negative fitness values than those that were measured in the absence of Cr (Table 1). These included the nitrate reductase structural genes narG (−1.1) and narH (−1.0), Mo-co biosynthesis genes moaA (−1.5), moaB (−1.4), moeA (−1.3), and a component of the molybdate ABC transporter (−1.2) (Table 1). These results show that Cr[VI] interferes with RCH2 growth under denitrifying conditions when nitrate reductase activity is decreased or eliminated. We propose that RCH2 strains lacking nitrate reductase activity survive in the library control growth, albeit with lower fitness, by using the rest of the denitrification pathway to respire (Figure 1). In this case, the presence of Cr[VI] could interfere with a component of the denitrification pathway whose action takes place after nitrate is reduced to nitrite. The likely target(s) of Cr[VI] toxicity in the remainder of the denitrification pathway are one or more of the several cytochrome containing enzymes involved, such as NirS (cytochrome cd₁), a nitrite reductase, or NorBC, the cytochrome b and c subunits of Nor, required for nitric oxide respiration (Figure 1). From a previous study, the reduction of Cr[VI] in RCH2 under anaerobic conditions was shown to require the presence of nitrate, indicating that a component of the denitrification pathway is involved (Han et al., 2010). It has also been observed that other organisms reduce Cr[VI] under anaerobic conditions using cytochrome components of electron transport chains (Mangaiyarkarasi et al., 2011; Joutey et al., 2015). If Cr[VI] is reduced by RCH2 denitrification cytochromes at the expense of their normal activities, this could represent a Cr[VI] toxicity target in RCH2 (Figure 1).

Nitrite reductase activity was measured for RCH2 whole cells grown under denitrifying conditions in the absence and presence of 120 μM K₂Cr₂O₇ and 3 mM uranyl acetate (Figure 2). This was to test the model that Cr[VI] interferes with cytochrome-containing components of the denitrification pathway downstream of nitrate reductase activity, leading to the large negative Δw_Cr values seen for nitrate reductase related genes. In support of the model, cells grown in the presence of Cr[VI] had greatly decreased nitrite reductase activity (Figure 2). This indicates that Cr[VI] indeed interferes with nitrite reductase activity, likely through interaction with one or more of the involved cytochrome containing enzymes (NirS, NirB, NirC, NirM, and/or NirT; Figure 1). Many contaminated sites like Hanford, WA, where RCH2 was isolated, and Oak Ridge Reservation, TN are contaminated with both heavy metals like Cr[VI] or U[VI] and nitrate (Riley and Zachara, 1992; Fruchter, 2002). Remediation efforts at sites like these may be complicated by this interaction between metals and the denitrification pathway.

Chromate is known to interacts with the sulfur assimilation pathway at multiple levels. One of the major ways in which Cr[VI] enters the cell is through the sulfate transporter system (Cervantes et al., 2001) thereby competing with sulfate transport. In addition, sulfide (H₂S) can directly react with and reduce Cr[VI] to the less mobile and less toxic Cr[III] (Kim et al., 2001). Intracellularly, reduced thiols such as glutathione and ascorbic acid can also reduce Cr[VI] (Arslan et al., 1987; Costa, 2003; Xu et al., 2004). These effects of Cr[VI] on sulfur assimilation were demonstrated in anaerobically-grown RCH2 where the growth defect caused by 80 μM K₂Cr₂O₇ was mitigated by the exogenous addition of various sulfur sources (Figure S1). High concentrations (1 mM) but not low concentrations (0.1 mM) of sulfate partially restored growth, presumably by competing with Cr[VI] uptake through the sulfate transport system. Addition of 100 μM H₂S also restored growth, presumably by reducing Cr[VI] in the growth medium to the less cell permeable and thus less toxic Cr[III] (Figure S1).

One of the largest negative Δw_Cr values observed during the anaerobic Cr[VI] fitness challenge was for the gene encoding the hypothetical protein Psest_2088 (−2.3), which is located directly downstream of the gene encoding the β subunit of sulfite reductase (psest_2089) (no gene fitness data is available for psest_2089), and upstream of another gene of unknown function encoded in the opposite direction. Psest_2088 is predicted to be an intracellular protein with a mass of 19.2 kDa and contains a conserved domain of unknown function found in several bacterial proteins (DUF934 superfamily). A deletion mutant lacking psest_2088 was constructed (Δ2088) and this strain was more sensitive than the wild-type to Cr[VI] (Figure 3A). When grown in the presence of 0.5 g/L yeast extract, the Δ2088 mutant had a growth defect compared to wild-type. However, growth of Δ2088 was severely impaired compaired to wild-type if 25 μM K₂Cr₂O₇ was added to the growth medium (Figure 3A).

Interestingly, the Δ2088 mutant strain is unable to grow in the absence of 0.5 g/L yeast extract unless an additional sulfur source is added to the growth medium. This is similar to how wild-type behaves when Cr[VI] is present (Figure 3B and Figure S1). Amounts of sulfate and sulfite (1 mM) are unable to correct the growth defect, but 0.3 mM cysteine and sulfide partially restore growth (Figure 3B). Thiosulfate (0.1 mM), a sulfur compound that is enzymatically broken down into sulfide and sulfite (Haschke and Campbell, 1971), also restores growth. This growth restoration profile of sulfur sources is consistent with Psest_2088 having an integral role in sulfite reductase activity since sulfur sources after sulfite reductase in the sulfur assimilation pathway restore growth (sulfide and cysteine) where those before (sulfate and sulfite) do not (Figure 4). One possibility is that Psest_2088 is involved in siroheme biosynthesis, a cofactor required by sulfite reductase. These growth phenotypes of Δ2088 were also seen when the cells were grown aerobically (Figure S2), indicating that the function of Psest_2088 is required both under aerobic and anaerobic growth conditions. Cysteine did not restore growth as well as sulfide likely due to a deleterious effect that cysteine has on RCH2 growth in general.

Taken together, these data support a model in which Cr[VI] toxicity is caused at least in part by oxidation of reduced sulfur compounds in the cell or in the oxidation of cellular components that require a reduced sulfur species for function (Figure 4). Strains lacking Psest_2088 have a decreased capacity to reduce sulfite to sulfide and therefore would be more sensitive to Cr[VI] oxidation of intracellular reduced sulfur species. Extracts of RCH2 wild-type cells grown anaerobically in the presence and absence of 50 μM K₂Cr₂O₇ and of Δ2088 were prepared anaerobically and were measured for reduced thiol concentrations. As shown in Figure S3, all of the extracts had comparable concentrations of reduced thiol groups. Although, on the surface the model would predict that the wild-type extract (no Cr[VI] added) should have higher concentrations of reduced thiols than the other extracts, this result is still consistent with the proposed model (Figure 4). RCH2 may strictly maintain a minimum intracellular concentration of reduced thiols during growth and oxidation of thiols by Cr[VI] may result in a slower growth rate rather than in an oxidized intracellular environment.

In addition to psest_2088, a negative Δw_Cr value was exhibited by psest_4063 (−1.0), the gene encoding the ATP-binding subunit of the sulfate ABC transporter. The other two genes encoding components of the sulfate transporter psest_4061 (−0.6) and psest_4062 (−0.8) also had negative Δw_Cr values. If Cr[VI] enters the cell by multiple transporters and depletes the reduced thiol pool, deletion of a sulfate transporter could have the deleterious effect of decreasing sulfate uptake, even though it may also decrease Cr[VI] uptake. In addition, a gene (psest_0494) annotated as a rhodanese-related sulfurtransferase, putatively involved in the transfer of sulfur containing groups, had a large negative Δw_Cr (−2.1). All three genes of the methionine import system, which could provide a source of reduced sulfur, display large negative Δw_Cr, psest_4314-4316 (−2.0, −1.5, and −3.0). From the combined data, we propose that the reduced sulfur pool of RCH2 is a target of chromate toxicity (Figure 4).

In our study, large negative Δw_Cr values were seen for the genes of both cytochrome cd₁ nitrite reductase NirS (−1.6), and for NirF (−1.5), a protein needed for NirS maturation (Nicke et al., 2013). If Cr[VI] interferes with nitrate reduction, nitrite reduction, requiring NirS, would become important for respiration and survival. Anaerobic Cr[VI] reduction has been associated with electron transport systems in many organisms with electrons being transferred to Cr[VI] by various cytochromes (Mangaiyarkarasi et al., 2011; Joutey et al., 2015). For example, Desulfovibrio vulgaris soluble cytochrome c₃ was shown to be involved in Cr[VI] reduction (Lovley and Phillips, 1994). In a previous study, the reduction rates of Cr[VI] were measured for cell suspensions of RCH2 under both aerobic and anaerobic conditions (Han et al., 2010). In both cases, similar Cr[VI] reduction rates (4–5 × 10⁻¹² μmol. h⁻¹. cell⁻¹) were observed, however, the presence of O₂ (in the case of aerobic conditions) or nitrate (in the case of anaerobic conditions) was required for activity in the assay. Cr[VI] reductase activity was not inducible by adding Cr[VI] to the growth medium, and it was concluded that RCH2 could not reduce Cr[VI] unless reducing equivalents for Cr[VI] are first generated in the presence of the physiological electron acceptor (Han et al., 2010).

The Cr[VI] efflux protein ChrA has been shown to confer Cr[VI] resistance to P. aeruginosa and other microorganisms (Aguilera et al., 2004; Ramírez-Díaz et al., 2008). RCH2 contains three homologs of ChrA, Psest_1915 and two proteins that are identical in sequence, Psest_0945 and Psest_4025. There was not a negative Δw_Cr value for psest_1915 (0.1) and the other two ChrA homologs do not have gene fitness values as no transposon insertions were present in these genes in the RB-TNSeq library. Hence, without further information, the role of efflux in RCH2 Cr[VI] resistance cannot be determined.

There were ten genes of unknown function with large negative Δw_Cr values ranging from −1.0 to −1.6 (Table 1). The function of these proteins in Cr[VI] resistance could be the basis of a future endeavor. Tools such as RB-TNSeq can be used to discover multiple fitness defects for genes of unknown function under diverse conditions, elucidating the roles these proteins have in cellular metabolism.

In the U challenge experiments, as with the Cr[VI] challenge, genes related to nitrate reduction had large negative Δw_U values (Table 2). These large negative Δw_U values were seen for the denitrification required nitrate reductase structural genes narGHI (−1.4 to −1.8), Mo-co biosynthesis genes moaB (−3.5), moeA (−2.2), mobA (−1.7), moaC (−1.3), and two genes of the molybdate ABC transporter (−1.1 and −1.2). In general, the Δw_U for nitrate reduction related genes were more pronounced and widespread than the Δw_Cr. This may simply indicate that the U[VI] challenge was more effective than the Cr[VI] challenge, or uranium toxicity may have a greater effect on this area of metabolism. We propose a similar model for the effect of U[VI] on nitrate reduction gene fitness values as we did for Cr[VI], in that U[VI] likely interacts with cytochromes required for the steps in the denitrification pathway after nitrate is reduced to nitrite (Figure 1). U[VI] reduction by bacteria is not catalyzed by specialized reductases but by redox enzymes that normally function in other cellular processes (Wall and Krumholz, 2006). These include cytochromes such as cytochrome c3 from D. vulgaris (Payne et al., 2002). If U[VI] is reduced by or interacts with one or more of the denitrification cytochromes of RCH2, this could disrupt the later steps of denitrification explaining the large negative Δw_U seen for genes involved in nitrate reduction. Althought U[VI] does not decrease RCH2 whole cell nitrite reductase activity (Figure 2), U[VI] may still interfere with the cytochrome containing nitric oxide reductase resulting in the large negative Δw_U observed for nitrate reductase related genes.

A large negative Δw_U value was observed for psest_3488 (−1.9), which encodes the universal stress protein UspA. Expression of the uspA gene has been shown to be up-regulated under several stress conditions in E. coli including starvation, heat, oxidants, metals, uncouplers, ethanol and antibiotics (Kvint et al., 2003). This stress protein also appears to play a role in uranium resistance in RCH2. In E. coli oxidative stress genes including SOD and catalase were also connected with resistance to uranium toxicity under aerobic conditions (Khemiri et al., 2014). In our RCH2 experiment, the gene for SOD and the five annotated catalase genes did not have large negative Δw_U (0.0–0.2), presumably due to the anaerobic growth conditions used in the experiment.

Two genes involved in the formation of polysaccharides had large negative Δw_U values, psest_2232 (−4.3) and psest_0993 (−2.3). They encode UTP-glucose-1-phosphate uridylyltransferase and glucose-6-phosphate isomerase, respectively. These enzymes are involved in the synthesis of UDP-glucose, which is a building block needed to form the polysaccharide glycogen (Alonso et al., 1995). We propose that RCH2 produces an exopolysaccharide using both Psest_2232 and Psest_0993 that is important for uranium resistance. Previously Pseudomonas sp. EPS-5028 and Acidithiobacillus ferrooxidans were shown to accumulate uranium on exopolysaccharides under aerobic conditions (Marqués et al., 1990; Merroun et al., 2003), but it was not demonstrated that this accumulation prevented U toxicity. The data presented here supports the idea that under anaerobic conditions, accumulation of uranium on exopolysaccaride constitutes a defense mechanism.

There were five hypothetical proteins that had large negative Δw_U values ranging from −1.0 to −1.5 (Table 2). The function of these proteins in uranium resistance is not known and these will be the basis of future research.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.