Abstract
Although arsenic is generally a toxic compound, there are a number of ferns in the genus Pteris that can tolerate large concentrations of this metalloid. In order to probe the mechanisms of arsenic hyperaccumulation, we expressed a Pteris vittata cDNA library in an Escherichia coli ΔarsC (arsenate reductase) mutant. We obtained three independent clones that conferred increased arsenate resistance on this host. DNA sequence analysis indicated that these clones specify proteins that have a high sequence similarity to the phi class of glutathione-S-transferases (GSTs) of higher plants. Detoxification of arsenate by the P. vittata GSTs in E. coli was abrogated by a gshA mutation, which blocks the synthesis of glutathione, and by a gor mutation, which inactivates glutathione reductase. Direct measurements of the speciation of arsenic in culture media of the E. coli strains expressing the P. vittata GSTs indicated that these proteins facilitate the reduction of arsenate. Our observations suggest that the detoxification of arsenate by the P. vittata GSTs involves reduction of As(V) to As(III) by glutathione or a related sulfhydro compound.
Funding The authors acknowledge support from the Indiana 21st Century Research and technology Fund (912010479) to DES and LNC, from the U.S. Department of Energy (grant no. DE-FG02-03ER63622) to DES, and from BBSRC-DFID (grant no. BBF0041841GJN) to AAM. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. There are no financial, personal, or professional interests that could be construed to have influenced the paper.
Introduction
Several members of the fern genus Pteris can accumulate high levels of arsenic [1,2]. The first arsenic hyperaccumulator discovered was Pteris vittata (Chinese brake fern), which can contain arsenic at levels in excess of 20 mg/g in its fronds [1,3]. Such high level accumulation, and hence, tolerance of arsenic is remarkable in view of the toxicity of this element [4]. The exceptionally high arsenic tolerance of the hyperaccumulating ferns not only makes them good candidates for use in biological remediation of arsenic-contaminated environments, but it also makes them interesting systems to unravel the basic mechanisms of resistance to this widespread environmental pollutant.
Paradoxically, detoxification of arsenate involves reduction to arsenite [5,6], which has potentially higher human and ecological toxicity than arsenate [7,8]. Arsenite produced from arsenate is known to be removed from the cytoplasm by various efflux mechanisms. In Escherichia coli, it is pumped out by an ATP-dependent export system, and in yeast it is eliminated by a combination of efflux into the medium by a plasma membrane-bound carrier and transfer into the vacuole as a glutathione adduct [5,6]. In Arabidopsis thaliana, an arsenic sensitive non-hyperaccumulating plant, arsenite is transferred to the vacuole, as a phytochelatin adduct via two paralogous ABCC-type transporters (AtABCC1 and AtABCC2) [9]. Three different families of arsenate reductases have been identified, represented by the ArsC protein of E. coli [10], the ArsC protein of Staphylococcus aureus [11], and the Acr2 protein of yeast [12]. The E. coli and yeast arsenate reductases use reduced glutathione (GSH) as the direct electron donor for the reduction of arsenate, whereas the S. aureus enzyme employs thioredoxin. Ellis et al. [13] cloned the P. vittata PvACR2 arsenate reductase gene from a cDNA library of arsenate-treated gametophytes. This enzyme shows 25% amino acid sequence identity / 47% similarity to the yeast ACR2 protein and likewise uses GSH as substrate. Like the yeast enzyme, the P. vittata ACR2 lacks phosphatase activity [13]. Arsenate reductase has been shown to be induced 7-fold by arsenate in the roots of P. vittata [14]. The reduction product of arsenate is accumulated as free arsenite in the fronds of this fern probably in the vacuoles [3,13,15]. Arsenate reductases similar to the P. vittata ACR2 protein have been cloned from rice [16] and the arsenate-tolerant grass Holcus lanatus [4], and orthologous sequences have been identified in the genomes of other plants and fungi that are not arsenic hyperaccumulators [16], though all these enzymes also have phosphatase activity.
Phytochelatins are cysteine-rich peptides derived from GSH that can bind to arsenite and several reactive heavy metal cations [17]. A. thaliana mutants unable to synthesize phytochelatins show enhanced sensitivity to arsenic [18], and overproduction of these compounds has been shown to increase arsenate resistance [4,19,20]. Phytochelatins have been detected in P. vittata and in the related species Pteris cretica, but unlike in non-hyperaccumulator plants, only a small portion of arsenic is complexed with phytochelatins in these hyperaccumulators. Phytochelatin production in these arsenic hyperaccumulators is limited [17,21], suggesting that arsenic-phytochelatin complexes do not form a major storage form for arsenic, although they still may play an important role in hyperaccumulation, as indicated by the occurrence of highly localized thiolate-coordination of arsenic in close proximity to the vein and mid-veins in the fronds of the P. vittata sporophyte [17,21].
Two other proteins that could play a role in arsenic tolerance or hyerpaccumulation in P. vittata have been identified by selection for clones from cDNA libraries that increased the arsenate tolerance of an arsenate reductase mutant (ΔarsC) of E. coli. One gene obtained by this approach was found to specify a triosephosphate isomerase [22]. In addition to its expected enzymatic activity, it was suggested that this enzyme promotes the reduction of arsenate by an unknown mechanism. The second enzyme that was identified in this manner was a glutaredoxin [23], which was able to confer increased resistance to both arsenate and arsenite by a mechanism that needs to be determined.
In order to identify other proteins that could contribute to the arsenic resistance of P. vittata, we carried out a similar procedure to isolate clones from a P. vittata cDNA library that elevated the arsenate resistance of a ΔarsC mutant E. coli. Nucleotide sequence analysis revealed that the cDNA clones in these derivatives encode proteins that have high sequence similarity to glutathione-S-transferases (GSTs).
Materials and Methods
Bacterial Strains and Growth Conditions
The E. coli K12 strain W3110 (IN[rrnD-rrnE]) and its arsenate reductase mutant derivative WC3110 (ΔarsC IN[rrnD-rrnE]) [12] were obtained from B. Rosen (Wayne State University), the γ-glutamyl-cysteine synthetase mutant JTG10 (gshA20::Tn10kan thr-1 araC14 leuB6[Am] Δ(gpt-proA)62 lacY1 tsx-33 glnV44galK2 hisG4 rfbC1 mgl-5 kdgK51 xylA5 mtl-1 argE3 thi-1 λ rpsL31) [24] was obtained from the E. coli Genetic Stock Center, and the glutathione reductase mutant BW31342 (Δgor::cat) and the arsenate reductase mutant BW39212 (ΔarsC::kan) were obtained from B. Wanner (Purdue University). The arsenate reductase-glutathione synthetase double mutant strain KC1680 was constructed by P1 transduction of the gshA20::Tn10kan from strain JTG10 into WC3110. The arsenate reductase-glutathione reductase double mutant strain KC1852 was constructed by transduction of the ΔarsC::kan and Δgor::cat insertions elements BW31342 and BW39212 into WC3110. Strains KC1858 (Δgor::cat ΔarsC::kan IN[rrnD-rrnE] / pTriplEx2), KC1860 (ΔarsC::kan Δgor::cat IN[rrnD-rrnE] / pPvGST35), and KC1862 (ΔarsC::kan Δgor::cat IN[rrnD-rrnE] / pPvGST41) were made by electroporating the indicated plasmids into KC1852.
Bacterial cultures were grown with aeration at 37° in the rich medium LB [25] or K medium, which has a phosphate concentration of 1 mM [26]. When used, ampicillin (Ap), chloramphenciol (Cm), and kanamycin sulfate (Km) were at 100, 25, and 75 mg l-1, respectively. Solid media contained 20 g l-1 Difco agar. Na2AsO4·7H2O and NaAsO2 were purchased from J. T. Baker and Sigma, respectively.
Arsenate resistance was assessed qualitatively on solid media by the radial streaking method [27]: Whatman GF/A filter disks (21 mm diameter) were placed in the middle of K medium plates containing 1 mM isopropyl-β-D-thiogalactopyranoside (IPTG; added to induce the expression of the cDNA genes from the lac promoter on the cloning vector), and 30 μl of 2 M Na2HAsO4 solution was placed on the disks and air-dried. Strains were streaked with sterile toothpicks outward from the disk, and their AsR was evaluated by the size of the zone of growth-inhibition next to the arsenate-containing filter disk compared to that seen with WC3110 (ΔarsC) carrying the empty cloning vector pTriplEx2. For a more quantitative determination of arsenate and arsenite resistance in liquid culture, we used the procedure of Mukhopadhyay et al. [12] with slight modifications: Cultures of strains carrying the P. vittata GST clones were grown overnight in LB + Ap and were subcultured at 1:100 dilution into K medium + 1 mM IPTG, plus different concentrations of Na2HAsO4 or NaAsO2 and grown with aeration. After 24.0 h of incubation at 37°, the cell density of the cultures was measured as light scattering in a Shimadzu UV-1700 spectrophotometer at 600 nm (OD600), with appropriate dilutions to correct for multiple scattering at OD600> 1.0 (where OD600 = 1 corresponds to ~1.5 x 109 cells/ml). For each arsenate or arsenite concentration tested, the measurements were performed with three independent cultures.
Pteris vittata cDNA library construction and screening
RNA was purified from P. vittata gametophytes grown for 6 weeks in liquid culture [28] containing 1 mM KH2AsO4. A cDNA library was synthesized using BD Biosciences Clonetech SMART® cDNA library Construction kit (User Manual PT3000-1 [PR15738]) and cloned between the SfiIA and SfiIB sites of plasmid pTriplEx2 (Clonetech). The ligation mixture was electroporated into E. coli strain DH10B [29] and transformants selected on LB Ap plates. Plasmid DNA was isolated from a pool of ~7x106 colonies and electroporated into WC3110 (ΔarsC). The cells were allowed to recover in 1 ml LB medium for one hour at 37°C with aeration. Aliquots of 100 μl of the transformation mixtures were inoculated on K medium plates containing 6mM Na2AsO4·7H2O and 1 mM IPTG. After two days, the plates were examined for the appearance of AsR colonies. Spontaneous AsR mutants arose in the untransformed control cells at a frequency of ~2 per 107 cells; the frequency of such derivatives was generally 2-to 10-fold higher in the cultures transformed with the library. For plates that contained fewer than ~50 AsR transformed colonies, the strains were single colony isolated on K medium plates containing 6 mM Na2AsO4·7H20, and then tested for Ap resistance (ApR) on LB plates, confirming the presence of the cloning vector (as opposed to a spontaneous mutation to AsR by the host). Plates that contained more than 50 AsR colonies were replica-plated to LB Ap plates, and colonies that were stably ApR were single colony isolated on K medium plates containing 6 mM arsenate. Plasmids were isolated from the transformants that were confirmed to be both AsR and ApR and were retransformed into WC3110, selecting ApR. The derivatives obtained in the latter step were tested for AsR by radial streaking. Plasmids were isolated from the AsR strains and the sequences of the cDNA inserts were determined at Purdue Genomics Core Facility. The sequences were analyzed with the blastn and blastp algorithms (http://blast.ncbi.nlm.nih.gov/Blast.cgi). The sequences of the three P. vittata ST cDNA clones pPvGST35, pPvGST41, and pPvGST57 that we obtained were entered into GenBank under accessions numbers EU259318, EU259319, and EU259320, respectively.
Speciation of arsenic in cell cultures
Strains KC1749 (pTriplEx2), KC1761 (pPvGST35) and KC1763 (pPGSTv41) were grown LB + Ap, 0.03 ml were inoculated into 3 ml of K medium containing 10 mM glucose + 0.1 mM Na2HAsO4 + Ap + 1 mM IPTG and incubated with aeration for 24 h. At the end of the growth, the cell density was determined by light scattering at 600 nm (OD600). The cells from 1.5 ml samples were sedimented by a microfuge, 0.5 ml of the supernatant culture media were mixed with 0. 5 ml of a buffer containing 13.36 mM NH4NO3, 13.36 mM NH4H2PO4, and 10 mM EDTA (pH 6.2) and flash-frozen in dry ice ethanol. EDTA was used in the extraction buffer as it has been shown to preserve the inorganic arsenic species [30]. The As(III) and As(V) content of the samples from 5 independent cultures for each strain were determined as described by the method of Williams et al. [31].
Results
Isolation of cDNA clones from P. vittata that confer arsenate-resistance in E. coli
Plasmids pPvGST35, pPvGST41, and pPvGST57 were isolated by introducing a P. vittata cDNA expression library into the ΔarsC (arsenate reductase) mutant E. coli WC3110 and selecting derivatives that could grow in K medium containing 6 mM Na2AsO4. The arsenate resistance (AsR) phenotype of these transformants is illustrated in Figure 1: The three plasmids carrying the P. vittata cDNA clones conferred increased AsR on this host strain compared to the control carrying the empty vector pTriplex2. The phenotype conferred by these three plasmids was evaluated in a more quantitative form in Figure 2, which shows that in liquid medium, the plasmids conferred very similar increases in AsR above that seen with the control strain carrying the empty cloning vector. WC3110 derivatives expressing the P. vittata GST clones had an exponential doubling time of 1.6 h in K medium +10 mM Na2AsO4, whereas the empty vector control had a doubling time of >12 h (data not shown).
The cDNA clones conferring arsenate resistance encode GST-family proteins
DNA sequence analysis revealed that the three plasmids conferring AsR contain a 217-codon open reading frame whose deduced protein product shows high similarity to GSTs from higher plants. Plasmids pPvGST35 and pPvGST57 are identical in their nucleotide sequences in the central protein-coding region (Figure 3); the gene contained in these two plasmids was designated PVGST1. These two plasmids exhibited slight sequence differences at the 5’ and 3’ ends of the inserts, which could have been due to the fact that they were from different mRNAs, or differences could have been introduced during cloning. In any case, the fact that the plasmids are different indicates that they are independent isolates. The insert in pPGSTv41 shows a 13% nucleotide sequence difference from the inserts in pPvGST35 and pPvGST57, resulting in 23-amino acid difference in the predicted open reading frames (Figure 3). Consequently, the open reading frame in pPvGST41, which is designated PVGST2, specifies a GST that is different from PVGST1. The expression vector pTriplEx2 contains translation start signals for 5’ portions of the ompA and lacZ genes in different reading frames, followed by a run of T residues that provide a site for transcriptional frameshifting. Consequently, cloned inserts may be translated in all three reading frames as N-terminal fusion proteins with OmpA or LacZ peptides. DNA sequencing revealed that the inserts in pPvGST35 and pPvGST57 contain a TAG termination codon 24 bases upstream of the ATG initiation codon for the GST, indicating that these two gene products are not made as fusion proteins. However, there is no analogous in-frame termination codon upstream of the GST sequence in pPvGST41, and so the latter may be translated as a fusion product with an N-terminal peptide from LacZ or OpmA.
GSTs of higher plants have been assigned into four classes based on their amino acid sequences: phi, tau, theta, and zeta; representatives of the former two classes are also found in bacteria and members of the latter two classes are also present in animals [32]. Figure 4A shows a multiple alignment of the two P. vittata genes against four of the most closely related enzymes from other plant species and one representative of each of the four classes of GSTs from A. thaliana. This analysis indicates that the P. vittata enzymes cluster together and with other members of the phi-class of plant GSTs (Figure 4B). One of the diagnostic features of the phi class of GSTs is that their GSH-binding site contains the amino acid triplet ESR [32], which is present in the P. vittata clones (Figure 4A).
The P. vittata GSTs do not show significant similarity (blastp Expect value > 0.019) to the three types of arsenate reductases represented by the E. coli, S. aureus, and the yeast/P. vittata enzymes. Genes encoding highly similar phi class GST can be found in the Myxococcales bacteria Sorangium cellulosum ‘So ce 56’ and Myxococcus xanthus DK 1622 (blastp similarity to PVGST1 Expect value = 5e-43 and 6e-28, respectively) and in over 100 other bacteria with blastp Expect values between 1e-18 and 1e-13. The function of these bacterial enzymes has not yet been determined.
Nucleotide sequence features of the P. vittata GST clones
An interesting feature of the three P. vittata GST clones is that they contain 10 repeats of the dinucleotide GA immediately upstream and overlapping the ATG start (Figure 3). This motif is not present in the most closely related plant GSTs genes, including the one from the moss Physcomitrella patens. However, a cDNA glutaredoxin clone from P. vittata [23] also contains 12 GA repeats seven nucleotides upstream of the translation start site (see sequence under GenBank accession number EF052272). Whether these repeats are associated with the expression of arsenic tolerance-related genes in P. vittata or whether they have some other function in gene expression in this fern remains to be explored. There are precedents for strings of GA repeats upstream of genes in other eukaryotic systems. In Drosophila melanogaster, the so-called GAGA elements have been recognized as the binding sites of a chromatin remodeling factor that increases transcription from the hsp70 promoter [33 and references therein]. A GAGA element has been found between the TATA box and the transcription start sites of the Gsa1 gene of Glycine max, which has been suggested to be involved in the transcriptional control of the gene. Repeats of GA, or its complement CT, have also been found downstream of viral promoters, and it has been suggested that they play a positive role in transcriptional activation [34,35].
Phenotypic characterization of the P. vittata GST clones in E. coli
We tested whether the P. vittata GST clones can confer increased AsR in an arsenate reductase proficient (arsC+) host (Figure 5). Our results reproduce the finding of Mukhopadhyay et al. [12] that loss of the E. coli arsenate reductase greatly increases sensitivity to arsenate (compare arsC+/empty vector with ΔarsC/empty vector). Although the P. vittata clones substantially elevate AsR in the ΔarsC mutant, they resulted only in a marginal, statistically insignificant increase in the already high AsR of the arsC+ strain. These results indicate that although the P. vittata GSTs can functionally substitute for the E. coli arsenate reductase in the ΔarsC mutant, they do not provide improved resistance against this toxic chemical above the level conferred by the endogenous enzyme.
The sequence similarity of our clones to known GSTs suggests that the P. vittata enzymes probably use glutathione as a substrate. To test this, we determined whether blocking glutathione synthesis with a gshA (γ-glutamyl-cysteine synthetase) mutation would diminish the arsenate resistance conferred by these plasmids. The gshA ΔarsC double mutant carrying the empty cloning vector was considerably more sensitive to arsenate in the absence of GSH than the ΔarsC gshA+ counterpart (Figure 6A). More importantly, the gshA mutation cancelled the ability of the P. vittata clones to confer arsenate resistance in the absence of glutathione. Supplementation with GSH, which can be taken up by E. coli [36], restored the ability of the clones to confer AsR in the gshA mutant background (Figure 6B). These results suggest that the P. vittata GSTs use GSH as substrate for the detoxification of arsenate.
Finally, we examined whether the P. vittata GSTs could confer resistance to arsenite and found that that the clones were unable to do so either in the ΔarsC mutant (Figure 7) or in the arsC+ background (data not shown), indicating that the P. vittata GSTs can detoxify arsenate but not arsenite.
We carried out GST activity assays of the E. coli strains carrying the P. vittata clones using 1-chloro-2,4-dinitrobenzene (CDNB) as substrate, which has been used to assay phi class GSTs from plants [32,37]. However, we were not able to detect statistically significant GST activity in the E. coli strains carrying the P. vittata clones (data not shown), indicating that the P. vittata GSTs are unusual among the phi class enzymes in that they have no detectable activity with CDNB as substrate.
The P. vittata GSTs promote the reduction of arsenate in E. coli
In order to probe the mechanism by which the P. vittata GST clones confer increased arsenate resistance, we carried out two experiments to determine whether this phenotype involves the reduction of this toxic compound. In one of these experiments, we tested the effect of a glutathione reductase (gor) mutation on the ability of the P. vittata GSTs to impart increased arsenate resistance. The result was that the ability of the GST plasmids pPvGST35 and pPvGST41 to confer increased arsenate resistance was compromised by the Δgor::cat mutation (Figure 8). This result suggests that arsenate detoxification involves the oxidation of GSH, and glutathione reductase is required to regenerate this reduced sulfhydro compound.
We also tested whether the P. vittata GSTs increases the formation of As(III) from arsenate. For this experiment, we grew derivatives of strain WC3110 (ΔarsC) carrying the empty cloning vector or plasmids pPvGST35 and pPvGST41 in medium containing an initial 1 mM Na2HAsO4, and after 24 h of growth, we quantified the proportions of arsenic as As(III) and As(V) in the culture medium. The final biomass density of the cultures was OD600 = 0.96 ± 0.05 for the strain carrying the empty cloning vector, and OD600 = 1.71 ± 0.10 and 1.82 ± 0.15 for the strains carrying pPvGST35 and pPvGST41, respectively, confirming the ability of the GST plasmids to confer increased arsenate resistance. The cells were sedimented by centrifugation, and the amounts of arsenic in the medium as As(III) and As(V) were determined as described in Materials and Methods. The results are shown in Table 1. For strain KC1749, which carried the empty cloning vector, approximately 30% of the input arsenic was converted to As(III), with the balance remaining as As(V). It is not clear whether the reduction of arsenate in this culture was due to an enzymatic process or to some spontaneous reaction with GSH or some other sulfhydro compound. However, in the cultures of the strains carrying pPvGST35 and pPvGST41, the proportion of arsenic that was recovered as As(III) increased to approximately 60% and 69%. This result indicates that the P. vittata GSTs promoted the formation of an As(III) species, suggesting that the increased arsenate resistance conferred by these enzymes involves the reduction of arsenate.
Discussion
In this work, we isolated three independent clones from a P. vittata cDNA library that imparted increased arsenate resistance to an arsC mutant E. coli. Sequence analysis indicated that these clones encode proteins that have high sequence similarity to the phi family of higher plant GSTs. GSTs have diverse functions related to the detoxification of harmful substances [32]. Some have glutathione dependent peroxidase or isomerase activities [38,39]. Others are involved in detoxification of xenobiotics by glutathione conjugation. It has been suggested that some GSTs function as ligands or carrier proteins for a variety of molecules [40]. They have also been implicated in signal transduction pathways [41] and in responses to a wide range of stresses [32].
Our phenotypic analysis in a ΔarsC mutant E. coli indicated that the P. vittata GSTs require GSH for arsenate resistance. The ArsC, Acr2p, and PvACR2 arsenate reductases of E. coli, yeast, and P. vittata use GSH as electron donor and produce oxidized glutathione (GSSG) [12,13,42]. Arsenate reduction in E. coli and yeast requires the participation of a glutaredoxin, which regenerates GSH from GSSG. However, the human monomethylarsonic acid reductase, which is a member of the omega family of mammalian GSTs [43], can carry out the GSH-dependent reduction of methylarsonate to methylarsenite without the direct involvement of a glutaredoxin [44]. It is possible that the P. vittata GSTs may be able to catalyze the analogous reduction of arsenate to arsenite.
Another type of mammalian GSTs, which belong to the omega subfamily, can detoxify arsenite by forming an arsenite-(GSH)3 complex that is excreted from the cells [45,46]. Our observation that the P. vittata GSTs do not provide increased resistance to arsenite suggests that they do not operate by this mechanism. It is possible that the P. vittata GSTs might facilitate the formation of an arsenate-GSH adduct that could be expelled from E. coli. However, the latter mechanism would necessitate the stoichiometric consumption of at least one molecule of GSH for one molecule of arsenate that is detoxified. In view of the fact that our GST cDNA clones could provide resistance to at least 10 mM arsenate in a medium that contains only 10 mM glucose as carbon source, we feel that this mechanism is unlikely. The requirement for glutathione reductase also supports the conclusion that GSH has a catalytic rather than a stoichiometric role.
Arsenate resistance in the hyperaccumulator P. vittata is thought to involve reduction of As(V) to As(III) and transport into the vacuoles in fronds and gametophytes. An arsenate reductase has been cloned from P. vittata [13], and it has been recently reported that a triosephosphate isomerase [22] and glutaredoxin [23] from this organism can increase the arsenic resistance in E. coli. Our observations that the P. vittata GSTs promote the GSH-dependent conversion of As(V) to As(III) in the arsC mutant E. coli suggest that these proteins participate in an arsenate reduction pathway, which also involves the bacterial glutathione reductase. However, we do not have any direct evidence that the P. vittata GSTs by themselves have arsenate reductase activity and cannot rule out other possible mechanisms, such as that these enzymes might detoxify arsenate by promoting the formation of a GSH complex. It will be necessary to purify these enzymes from E coli and establish whether they have one of the above enzymatic activities in vitro. Finally, although the fact that the GSTs conferred increased arsenate tolerance in E. coli makes it attractive to hypothesize that these proteins are likely to contribute to the arsenic tolerance or hyperaccumulation in P. vittata, it will be necessary to confirm the roles of these enzymes in arsenic metabolism in the ferns.
Author Contributions
Conceived and guided the project: DES LNC. Isolated P. Vittata GST cDNA clones in E. coli and carried out initial phenotypic characterization: AAK. Constructed the P. vittata cDNA expression library: DRE. Carried out the arsenic speciation and related data analysis: GJN and AAM. Performed the GST assays: XH. Wrote the manuscript: LNC. All authors read and approved the final manuscript.
Acknowledgments
We thank A. Murphy for helpful discussions, and N. Csonka for help with the photography and image processing.