Genes and Enzymes Involved in Bacterial Oxidation and Reduction of Inorganic Arsenic

The human use of toxic heavy metals is here to stay. In addition to intentional poisoning with arsenic (40) (arsenic levels in the hair of Napoleon Bonaparte approached 40 ppm, more than 1,000 times above allowable levels), medical, agricultural, and industrial uses of arsenic present major human problems (5, 40, 41). Arsenic catastrophes are occurring today, most notably as the high arsenic levels present in drinking water in East Bengal and West Bengal, where over 40 million people are being exposed to more than 50 ppb of arsenic (8); this level was the limit accepted by the World Health Organization before that organization (and later the U.S. government) set the allowable limit fivefold lower, at 10 ppb. Bangladesh maintains an allowable limit of 50 ppb, and the limit in Canada is 25 ppb. The arsenic found in Bengalese drinking water is present naturally in the sediment but is released only by humans, primarily from the digging of shallow wells. Similarly unacceptably high arsenic levels are found in shallow-well domestic drinking water in the American Midwest (7, 7a, 21). It seems likely that microbial metabolism is involved in mobilizing (18, 62) previously immobile “natural” subsurface arsenic in these areas.

Living cells (microbial or human) are generally exposed to arsenic as arsenate or arsenite. Arsenate, As(V), is frequently written as AsO₄³⁻, which is similar to phosphate, and has a pK_a of 7.0, with HAsO₄²⁻ and H₂AsO₄¹⁻ being equally abundant at pH 7.0. Although arsenate is thought to be highly soluble, in many environments with calcium or insoluble iron compounds, arsenate is precipitated as phosphate would be. Arsenite, As(III), is frequently erroneously written as AsO₂⁻, although with a pK_a of 9.3, it occurs at a neutral or acidic pH as As(OH)₃. Arsenite in water can be thought of as an inorganic equivalent of nonionized glycerol and is transported across cell membranes from bacterial cells to human cells by glyceroporin membrane channel proteins (36, 47).

Nealson et al. (38) introduced the phrase “eating and breathing” minerals in the context of seeking evidence of inorganic signatures for life on other planets (39), in contrast to earlier searches for life on Mars that emphasized organic carbon-containing signatures. For arsenic, “eating” means arsenite functioning as an electron donor at the start of a membrane respiratory chain (Fig. 1A), and “breathing” means arsenate functioning as a terminal electron acceptor for an anaerobic respiratory chain (Fig. 1B), as do other minerals, such as Fe and Mn with some microbes.

FIG. 1.

Cellular locations and functions of bacterial respiratory arsenite oxidase, respiratory arsenate reductase, and cytoplasmic arsenate reductase.

The enzyme responsible for the respiratory oxidation of arsenite, As(III), to arsenate, As(V) (Fig. 2A), has been found widely in various groups of Bacteria and Archaea and has been studied in detail (see below). Indeed, some microbes gain energy from oxidizing arsenite (43, 50, 60), although this activity may be an exception limited to chemolithotrophic bacteria; heterotrophic bacteria have not been shown to derive major energy from arsenate in growth experiments (12).

FIG. 2.

Models for heterodimer arsenite oxidase (data are from reference [13]) and arsenate reductase (data are from reference [22]). Funnel-shaped active sites are shown at the top. Also shown are embedded Mo-pterin and [Fe-S] cofactors, proposed two-electron (2 e⁻) transfer pathways, and amino acids (Cys or His) linking the [Fe-S] cofactors to the polypeptides. HIPIP, high-potential iron protein.

Anaerobic arsenate respiration was discovered (2) with a bacterial isolate that coupled anaerobic heterotrophic growth to arsenate as the terminal electron acceptor (replacing oxygen in an anaerobic respiratory process) (Fig. 1B). Since then, diverse bacterial types with anaerobic respiratory arsenate reductase have been discovered (42, 44, 49, 55). In fact, it might be argued that anaerobic terminal electron acceptors, such as arsenate, and other minerals, such as nitrate and ferric cations, occurred in early cellular life, preceding aerobic oxygen-utilizing respiratory electron transport chains. Oxygen respiration could not occur until cyanobacterial photosynthesis resulted in an atmosphere with molecular O₂, hundreds of millions of years after the first anaerobic bacteria arose from prebiotic forms. While recent reports (12, 17, 27, 42, 58) have described the broad diversity of microbes able either to reduce arsenate or to oxidize arsenite, the emphasis here is on genes that determine these transformations and a new understanding of enzyme structure.

AEROBIC ARSENITE OXIDASE

Bacterial oxidation of arsenite to arsenate has long been recognized (reviewed in references 11, 12, and 54), especially with aerobic isolates from arsenic-impacted environments (14, 17, 50, 51, 56-58). Similar isolates have also been found in soils and sewage not known to be exposed to elevated levels of arsenic (45, 46). It is not currently clear whether arsenite oxidation is limited to a few isolates in each species. For example, although two Alcaligenes faecalis isolates have this activity (45, 46), the activity has not been examined in culture collection isolates of this species. It seems that most environmental isolates lack this potential, although a range of Bacteria with arsenite oxidase enzyme activity have been isolated and genes apparently encoding arsenite oxidase are found widely in various groups of Bacteria and Archaea (39a).

The 71-kb DNA region encoding arsenite oxidase and associated functions in A. faecalis strain NCIB8687 was recently sequenced (GenBank accession number AY297781; http://www.uic.edu/depts/mcmi/faculty/silver.html; L. T. Phung et al., unpublished data). This is the strain for which the enzyme crystal structure has been obtained (13), and this is the only currently available structure for a respiratory arsenite oxidase or arsenate reductase.

This review summarizes the Alcaligenes arsenic-related genes in what we are calling the first “arsenic gene island.” The asoA and asoB genes (Fig. 3) encode, respectively, the large molybdopterin-containing and the small Rieske (spectroscopy-identified [2Fe-2S] cluster) (52) subunits of arsenite oxidase of A. faecalis (Fig. 2). Upstream of asoB are 15 genes tentatively considered to be involved in arsenic resistance and metabolism (2 are shown in Fig. 3, and the others are available at GenBank accession number AY297781 and at http://www.uic.edu/depts/mcmi/faculty/silver.html); downstream of asoA are 6 genes also tentatively identified as being involved in arsenic resistance and metabolism (3 are shown in Fig. 3). These putative genes encode a total of three presumed periplasmic oxyanion binding proteins that are likely to be components of two ABC oxyanion ATPase membrane transport systems and an ArsAB arsenite chemiosmotic efflux system (Fig. 4). The details are incomplete, but the overall conclusion is that arsenite oxidase is encoded in a “gene island” of over 20 functionally related genes, a major change in the understanding of cellular arsenite resistance, from a smaller operon to a larger “island” with multiple related phenotypes (see below).

FIG. 3.

Genes for the arsenite oxidase region in A. faecalis (NCBI accession number AY297781; http://www.uic.edu/depts/mcmi/faculty/silver.html), a Sargasso Sea meta-genome environmental isolate (NCBI AACY01082423) (59), C. arsenoxidans (NCBI AF509588) (37), and chemolithoautotrophic strain NT-26 (NCBI AY345225) (51) and upstream and downstream genes. Presumed gene product lengths (in amino acids [aa]) and functions are indicated, as are percent amino acid identities between homologous products and those of A. faecalis.

FIG. 4.

Proposed inorganic arsenic metabolism in A. faecalis, predicted from linked genes in the sequence of GenBank accession number AY297781 (http://www.uic.edu/depts/mcmi/faculty/silver.html). Heterodimeric arsenite oxidase is coupled to the aerobic respiratory chain by a small protein (perhaps azurin or cytochrome c₅₅₁), whose gene is not in the gene cluster. Two predicted oxyanion ABC ATPases with periplasmic oxyanion binding proteins are indicated, along with their predicted roles, as are an intracellular glutathione-linked ArsC-class arsenate reductase and the ArsAB arsenite efflux complex.

A cluster of four contiguous genes including those encoding arsenite oxidase was identified (Fig. 3) (GenBank accession number AF509588) (37) from Centibacterium arsenoxidans, a newly named member of the β-Proteobacteria isolated from industrial wastewater (60). The four identified genes were named aoxABCD (for arsenite oxidase); a fifth, partial gene upstream of aoxA was also found (Fig. 3). Unfortunately, the current names for the A. faecalis genes and the homologous genes from C. arsenoxidans are different, although the asoA gene product is 73% identical at the amino acid level to the aoxB gene product (Fig. 3); similarly, the asoB and aoxA gene products are 62% identical (Fig. 3). Inactivation of the aoxA or aoxB gene eliminated arsenite oxidase activity (37). Santini and vanden Hoven (51) reported the DNA (and deduced protein) sequences encoding yet another set of AsoA and AsoB polypeptides (with yet a third mnemonic, Aro, instead of Aso) from chemilithoautotrophic α-proteobacterium strain NT-26; the sequences were homologous to those of A. faecalis AsoA and AsoB (Fig. 3). The DNA sequence similarities are too weak to be found by standard DNA Southern blotting or degenerate primer PCR analysis. Strain NT-26 is thought to produce an α₂β₂ heterotetramer arsenite oxidase (51) with a molecular mass approximately twice that for the α₁β₁ heterodimer (3, 13) of A. faecalis; yet another β-proteobacterium strain, NT-14, appears to produce an α₃β₃ heterohexamer enzyme (58). Chemilithoautotrophic strain NT-26 is thought to derive useful energy for growth from arsenite oxidation (51), although it is questionable whether obligately heterotrophic (that is, dependent on fixed carbon for energy) arsenite-oxidizing strains derive useful energy in this way. Researchers are clearly at an early stage of understanding arsenite-oxidizing microbes.

The evolutionary trees of gene and protein sequence relationships (Fig. 5) would not have been possible a year or two ago. New results will undoubtedly make such trees too complex for presentation within a year or two. Two conclusions relevant to environmental microbiology can be drawn at this time: asoA and asoB genes are being found broadly in a wide range of prokaryotes (see text above and Fig. 5); however, the diversity of sequences precludes the use of gene-specific universal probes or primers for identifying these genes in new isolates.

FIG. 5.

Phylogenetic trees of protein sequences for the large Mo-pterin subunit (A) and the small Rieske subunit (B) of arsenite oxidase (presumed orthologs) and selected parologous homologs. NCBI accession numbers are as follows: (A) C. arsenoxidans AoxB, gi 22758844; A. faecalis AsoA, gi 33469597; Thiomonas sp. strain VB-2002 AoxB, gi 23821270; C. aurantiacus Chlo2048, gi 22972154; Sulfolobus tokodaii ST2391, gi 15922722; A. pernix APE2556, gi 14602144; Pseudomonas syringae pv. syringae B728a Psyr020491, gi 23468844; Methanococcus jannaschii FdhF, gi 15669895; Wolinella succinogenes DSMZ 1740 WS0764, gi 34482881; D. hafniense Desu0744, gi 23112121; B. selenitireducens MLS10 ArrA, gi 33466104; Shewanella sp. strain ANA-3 ArrA, gi 33286384; Rhodobacter capsulatus DorA, gi 2981245; Sargasso Sea environmental sequence, gi 44367188; and arsenite-oxidizing α-proteobacterium strain NT-26 AroA, gi 37962697; (B) C. arsenoxidans AoxA, gi 22758843; A. faecalis AsoB, gi 33469598; C. aurantiacus Chlo2049, gi 22972155; S. tokodaii ST2392, gi 15922723; A. pernix APE2563, gi 14602146; Aquifex aeolicus VF5 SoxF, gi 2982941; Thermoplasma acidophilum SoxL, gi 10640539; Sulfolobus solfataricus SoxF, gi 13816354; S. tokodaii ST0108, gi 15920287; Sargasso Sea environmental sequence, gi 44367187; and arsenite-oxidizing α-proteobacterium strain NT-26 AroB, gi 37962696.

Upstream from asoAB and divergently oriented is a gene that encodes a presumed periplasmic oxyanion binding protein and that is homologous to a partial gene (open reading frame 253) that occupies the first 759 bp of the sequence of GenBank accession number AF509588 (Fig. 3). Downstream from asoBA and aoxAB, the A. faecalis and C. arsenoxidans sequences are unrelated, both at the DNA level and the protein level (Fig. 3). The A. faecalis chromosome (GenBank accession number AY297781; http://www.uic.edu/depts/mcmi/faculty/silver.html; Phung et al., unpublished) contains an moaA gene homolog thought to be involved in molybdopterin biosynthesis and a gene, designated phnD, encoding a second periplasmic oxyanion binding protein; the C. arsenoxidans chromosome contains genes encoding a putative oxyanion reductase and cytochrome c (aoxC and aoxD, respectively, in Fig. 3) (37).

A partial gene sequence for the large molybdopterin subunit of arsenite oxidase from an additional β-proteobacterial isolate, a Thiomonas sp., has been deposited in GenBank (accession number AJ510263) (V. Bonnefoy, personal communication) and is shown in Fig. 5A. In addition to the four sequences that have been recognized as representing arsenite oxidase, the “meta-genome” of numerous chromosomal fragments from microbial DNA isolated from the Sargasso Sea (59) contains a 6.5-kb sequence (NCBI accession number AACY01082423) with strong candidate genes encoding products homologous to AsoA (NCBI accession number EAI76965) and AsoB (NCBI accession number EAI76963) (Fig. 3 and 5). The draft genome of Chloroflexus aurantiacus, a green filamentous anoxygenic photosynthetic bacterium (Joint Genome Institute contig NZ_AAAH01000321), includes genes Chlo2048 and Chlo2049, the products of which appear to be subunits of still another arsenite oxidase (Fig. 5). The genes upstream and downstream of the C. aurantiacus putative arsenite oxidase genes are not related to those in this region of the A. faecalis or C. arsenoxidans chromosome. Figure 5 contains current trees for AsoA (presumed orthologs are shown above the broken line, and paralogs are shown below) and AsoB Rieske subunit sequences.

In addition to bacterial sequences, reasonable candidate genes for arsenite oxidase have been identified in the genomes of two hyperthermophilic Archaea, Aeropyrum pernix (GenBank accession number NC_000854) and Sulfolobus tokodaii (GenBank accession number NC_003106) (Fig. 5). These gene pairs are contiguous and are in the same order as asoA and asoB in Fig. 3.

The evolutionary relationship to other protein sequences in the dimethyl sulfoxide (DMSO) reductase family of microbial molybdopterin enzymes (29) is shown in Fig. 5A. Sequences above the broken line are presumed to be for the molybdopterin subunit of arsenite oxidase, while those below the broken line are for other members of the DMSO reductase family, including the distantly related respiratory arsenate reductase ArrA. The tree for candidate small Rieske subunits in Fig. 5B also includes presumed orthologs for other arsenite oxidases and other paralogous Rieske subunit sequences.

A comparison of the AsoB small Rieske subunit amino acid sequences predicted from gene sequences (GenBank accession number AY297781; http://www.uic.edu/depts/mcmi/faculty/silver.html; Phung et al., unpublished) and X-ray crystallography electron density maps (Protein Data Base deposits GI 1208496 to GI 1208506) (4, 13) shows a surprising 9% difference (12 of 133 amino acid positions) (Fig. 6). All differences are expected to be mistakes in prediction in Protein Data Base deposits, as was the case for the 9% difference (78 of 825 positions) between the sequences predicted for the molybdopterin subunit from the asoA gene and the electron density map (data not shown). Amino acid sequences predicted from electron density maps are rarely deposited today, since DNA sequences are generally available prior to the solution of the protein structure and primary amino acid sequences predicted from DNA sequences are generally consistent with those predicted from electron density maps and accepted.

FIG. 6.

Comparison of the amino acid sequences of the AsoB Rieske ([2Fe-2S]) subunit predicted from the DNA sequence translation and from the crystal electron density map. Asterisks indicate identical amino acids; triangles and diamonds indicate, respectively, the cysteine and histidine residues that anchor the [2Fe-2S] center. The TAT leader sequence, with its conserved twin arginines, is indicated, along with the predicted cleavage site.

The differences are not random, and of the 12 discordant calls in Fig. 6, 3 are lysines at the protein surface that are listed as alanine or serine in the crystal sequence, presumably because the long lysine side chain is mobile in the crystal, and 6 are secondary amines (asparagine or glutamine) that are listed as the corresponding dicarboxylic acid or alanine. This pattern of missing lysine side chains and secondary amines also dominated the discordant calls between the DNA translation and the electron density maps for the large molybdopterin subunit (data not shown).

The crystal structure for the Rieske subunit of arsenite oxidase starts with R43 (Fig. 6) and lacks eight N-terminal amino acids expected to be present in the processed protein. G. L. Anderson (personal communication) obtained the KAPADA hexapeptide sequence in Fig. 6 by direct N-terminal sequencing. Q35, which is predicted to follow the twin arginine translocation (TAT) protease cleavage site (6), was not found. The N-terminal methionine is present in the 826-amino-acid-long AsoA molybdopterin subunit sequence from the translated DNA, but not in the crystal structure (electron density) map. Direct experiments are needed to test hypotheses from DNA sequence-based protein similarity searches.

The alignments of polypeptide gene product sequences and the phylogenetic trees showing relationships (23) indicate that functionally related arsenite oxidases in both Bacteria and Archaea start with the determinant for a TAT leader sequence (6, 15, 23, 29) at the N terminus of each Rieske subunit gene. The first 34 amino acids of the AsoB Rieske subunit sequence (Fig. 6) form the canonical TAT leader sequence, including the RRGFLK hexapeptide sequence, which is highly conserved and followed by about 20 hydrophobic amino acids before the predicted protease cleavage site. The main characteristic for a TAT-transported periplasmic enzyme such as arsenite oxidase is that the nascent polypeptide is folded into its tertiary structure in the cytoplasm, rather than transported in an unfolded form, as in the sec-dependent pathway (as happens for perhaps 90% of exported proteins in a bacterium such as Escherichia coli) (6, 15). Cofactors such as the molybdopterin complex and the [Fe-S] clusters are incorporated within the relatively anaerobic environment of the cytoplasm. Then the preassembled protein is translocated to the periplasm by utilizing the TAT leader sequence, which can be on either subunit of the heterodimer, apparently on the AsoB small subunit (Fig. 2A) of arsenite oxidase or on the ArrA large molybdopterin subunit of respiratory arsenate reductase (49) (Fig. 2B). After transport, the TAT leader sequence is cleaved by a signal protease (6).

Determination of the protein crystal structure of arsenite oxidase (13) followed extensive analytical and functional studies (3, 4). The molybdopterin- and [Fe-S] cluster-containing enzyme is thought to reside in the periplasmic space associated with the outer surface of the cytoplasmic membrane and to be associated with the aerobic respiratory chain via an azurin or a c-type cytochrome (Fig. 4) (3, 54). Both [3Fe-4S] and [2Fe-2S] Rieske-type iron-sulfur centers have been found, on different subunits (Fig. 2) (3, 13). Four domains are recognized in the structure of the large arsenite oxidase subunit (13), with the first domain containing the [3Fe-4S] cluster, coordinated to the protein by C21, C24, and C28 (Fig. 2). S99 occupies the position that might anchor the fourth Fe if this were a [4Fe-4S] center. The Rieske subunit consists of a single domain with the [2Fe-2S] cluster coordinated by C68, H70, C86, and H89 (indicated in Fig. 6 with numbering from the unprocessed sequence).

Two guanosine dinucleotide pterin cofactors coordinate each molybdenum center, with one oriented upward and the other oriented downward in the structure (13), as expected for bacterial molybdopterin proteins. Unlike most other members of the DMSO family of molybdopterin iron-sulfur oxidoreductases, for which the Mo is anchored with pterin sulfurs and a polypeptide serine hydroxyl or cysteine sulfhydryl, the fifth position of arsenite oxidase is occupied by double-bonded oxygen, and the corresponding amino acid is A199. Extended X-ray absorption fine-structure spectra and resonance Raman spectroscopy of the active-site molybdenum of arsenite oxidase (9) show the four Mo-S interactions and a single Mo double bonded to O in reduced arsenite oxidase; an additional Mo—O single bond has been found in the oxidized enzyme. Additional electrochemical studies (16) have established several key properties of arsenite oxidase. Electron spin resonance measurements have demonstrated the movement of electrons from the [3Fe-4S] center of the large subunit to the Rieske [2Fe-2S] center of the small subunit (3, 13).

The molybdopterin center occurs at the bottom of a shallow funnel-shaped cavity formed on the subunit surface by domains I, II, and III, providing solvent access for entry of the arsenite substrate and exit of the arsenate product (4, 13). The active-site surface in the crystal structure contains highly polar amino acid side groups, with H195, E203, R419, and H423 of the large subunit being considered to form the arsenite binding site (4). Chemical modification of one histidine residue, perhaps H195 or H423, by diethylpyrocarbonate inactivates the enzyme activity (30); hydroxylamine restores the activity. The predicted binding of arsenite places it adjacent to the molybdenum center, at a distance suitable for nucleophilic attack of Mo(VI)=O by an arsenite electron pair, a two-electron reaction (Fig. 2A) (4, 13).

After the reduction of molybdenum from Mo(VI) to Mo(IV) (Fig. 2A), the electrons are transferred to the [3Fe-4S] center of the same subunit, then to the [2Fe-2S] center of the Rieske subunit, and then to the first coupling protein of the aerobic respiratory chain, possibly azurin or cytochrome c (3, 4, 13). Within the crystal structure, the [3Fe-4S] cluster is 12 Å distant from the molybdenum and requires an electron pathway involving intermediate amino acid residues and/or positions in the pterin (13). At present, intraprotein electron pathways are only proposals and require experimental testing. Electron transport from the [3Fe-4S] cluster to the [2Fe-2S] cluster occurs over a comparable distance and must involve H bonds with amino acids and/or water molecules. There is precedent for such intra- and intermolecular electron transfer pathways with better-studied enzymes, such as nitrogenase (4, 29).

RESPIRATORY ARSENATE REDUCTASE

Starting with the discovery of an anaerobic bacterial isolate able to use arsenate as a terminal electron acceptor for a heterotrophic respiratory chain (2), many diverse bacteria with this potential have been isolated (26, 27, 27a, 42, 44, 55). However, no protein structures are available, and the DNA sequences of the genes (arr) involved have only recently appeared (1, 49).

The most detailed report of the purification and characterization of the respiratory arsenate reductase enzyme (22) (Fig. 2B) showed that anaerobic respiratory arsenate reductase, like arsenite oxidase, is a heterodimer periplasmic or membrane-associated protein consisting of a larger molybdopterin subunit (ArrA) which contains an iron-sulfur center, perhaps a high-potential [4Fe-4S] cluster, and a smaller [Fe-S] center protein (ArrB) that is not homologous to the Rieske polypeptide of arsenite oxidase. The smaller ArrB subunit is approximately twice the size of the AsoB subunit of arsenite oxidase (Fig. 2) (1, 22, 49) and may contain up to four [4Fe-4S] clusters (27). The respiratory arsenate reductase of gram-positive Bacillus differs from that of gram-negative bacteria in that it is anchored to the membrane of the gram-positive cell (1), which lacks a periplasmic compartment. There may also be anaerobic respiratory arsenate reductases that are unrelated in structure and based on chemistry other than molybdopterin and [Fe-S] centers (27). This scenario remains to be shown.

Recently, the genetic determinants of two respiratory arsenate reductase systems became available; one is from γ-proteobacterium gram-negative Shewanella strain ANA-3 (NCBI accession number AY271310) (48, 49), and the other is from gram-positive Bacillus selenitireducens (NCBI accession number AAQ19491) (1). About 11 kb of DNA from the respiratory arsenate reductase region of Shewanella strain ANA-3 was sequenced. The DNA sequence consists of two divergently transcribed operons (Fig. 7) (48, 49) followed by a predicted transcriptional termination signal and two unrelated genes. Upstream of arrA and arrB, encoding the two subunits of respiratory arsenate reductase, is a standard arsDABC operon for arsenate and arsenite resistance (Fig. 7) (48). The Shewanella and Bacillus ArrA sequences for the large molybdopterin subunit are 47% identical (Fig. 5A), consistent with orthologous proteins from gram-positive and gram-negative bacteria (see also reference 27a). The transcriptional regulatory gene, arsR, is absent in the Shewanella sequence and is predicted to be present elsewhere on the chromosome; this situation is unusual, as arsR generally occurs at the beginning of the operon, upstream of arsD, the determinant of a minor secondary regulatory protein (53). The arsC gene product of Shewanella appears (from its sequence) to be a member of the cytoplasmic glutathione-glutaredoxin-dependent arsenate reductase clade (36). The arsB and arsA genes of Shewanella probably determine the membrane carrier and ATPase components of an arsenite efflux pump (48), like that shown in Fig. 4 for the Alcaligenes strain, that removes arsenite from the cytoplasm. The Shewanella ars operon confers arsenite resistance when transferred to a different Shewanella strain or to E. coli. A mutation disrupting arsB in Shewanella led to arsenite sensitivity, although the arsB mutant strain could still respire arsenate to arsenite anaerobically (48).

FIG. 7.

Genes for respiratory (arrAB) and cytoplasmic (arsC) arsenate reductases of Shewanella strain ANA-3 (from GenBank accession number AY271310) (48, 49). Upstream and divergently oriented from the arrA and arrB genes for respiratory arsenate reductase is a four-gene canonical ars operon with the arsC gene for cytoplasmic arsenate reductase. aa, amino acid.

Upstream and in an orientation opposite that of the ars operon of Shewanella is a two-gene operon determining the subunits of the periplasmic heterodimer respiratory arsenate reductase (Fig. 2 and 7) (49). The predicted ArrA sequence for the large molybdopterin-containing subunit starts with a 42-amino-acid TAT leader sequence. Deletions in either arrA or arrB result in a loss of the ability to grow on arsenate and to reduce arsenate anaerobically (ArsC reduces arsenate aerobically). Anaerobic growth on arsenate is restored by complementation by functional genes. The ArrA sequence includes a CX₂CX₃CX₂₇C motif predicted to anchor a [4Fe-4S] cluster. The shorter, 234-amino-acid ArrB sequence contains four four-cysteine motifs as candidates for binding [Fe-S] centers and no TAT leader sequence. The phylogenetic tree of ArrA and related sequences (Fig. 5A) indicates that ArrA is distantly related to AsoA in the DMSO oxyreductase family of molybdopterin- and [Fe-S] cluster-containing enzymes (27a). The available genome sequence of Desulfitobacteria hafniense contains genes that encode the closest currently available homologs of ArrA and ArrB (AY271310) (49). ArrB appears to be an iron-sulfur protein related to DmsB of DMSO reductase and NrfC of nitrite reductase.

CYTOPLASMIC ARSENATE REDUCTASE FOR INTERCELLULAR DEFENSE

ArsC cytoplasmic arsenate reductase (Fig. 1C) is found widely in microbes, and the arsC gene occurs in ars operons in most bacteria with total genomes measuring 2 Mb or larger as well as in some archaeal genomes. It can be argued that ars operons for arsenic resistance are found more widely in microbes than, for example, trp operons for tryptophan biosynthesis. The literature on ArsC enzymes has been repeatedly reviewed (e.g., references 36, 53, and 54); therefore, the available and newer understanding will be briefly summarized. arsC is almost always found next to the arsB gene for the arsenite membrane pump (as shown in Fig. 7). Surprisingly, three unrelated clades (trees) of ArsC sequences are currently recognized, and these share a common biochemical function but have no evolutionary relationship (36). These three are (i) a glutaredoxin-glutathione-coupled enzyme, like that found associated with both the arsenite oxidase of Alcaligenes (NCBI accession number AY297781) and the respiratory arsenate reductase of Shewanella (Fig. 7), as well as many plasmids and chromosomes of gram-negative bacteria; (ii) a less-well-defined glutaredoxin-dependent arsenate reductase found in yeasts; and (iii) a group of thioredoxin-coupled arsenate reductases found initially in gram-positive bacteria but more recently also in many gram-negative proteobacteria (28, 31-35, 53, 54). Thioredoxin-coupled arsenate reductase cannot use glutaredoxin, and glutaredoxin-coupled arsenate reductase cannot use thioredoxin.

ArsC arsenate reductase is a small monomeric protein of about 135 amino acid residues and containing three essential cysteine residues that are involved in a cascade sequence of enzyme activity. There are no inorganic or other bound cofactors in the ArsC enzyme. In the glutaredoxin-glutathione-coupled ArsC reductases, the first cysteine residue is located at about position 11 from the N terminus of ArsC, but the other two catalytic cysteines are provided by glutathione and glutaredoxin rather than the ArsC polypeptide. After arsenate associates with C11 (E. coli plasmid R773 numbering), forming an As-S covalent bond, glutathione provides a second cysteine that forms an oxidized Cys-S-S-Cys pair with the ArsC cysteine; and that disulfide is reduced by a cysteine on glutaredoxin to form a cascade of reduced cysteines to oxidized cysteines (28, 36).

The first recognized arsenate reductase was found on a gram-positive Staphylococcus plasmid (19, 20). From protein crystallography, enzymology, and mutational studies (31-35, 61), it is known that the thioredoxin-coupled ArsC arsenate reductases utilize three cysteines, all in the ArsC polypeptide primary sequence, again for a cascade of oxidizing and reducing cysteine residues, with thioredoxin reducing the final Cys82-S-S-Cys89 oxidized bond. Arsenate covalently bound to the N-terminal Cys10 residue of Staphylococcus is reduced and released as the first internal Cys10-S-S-Cys82 cysteine is formed. The thioredoxin-coupled clade of arsenate reductases is found widely among plasmids and genomes of gram-positive bacteria as well as in some gram-negative bacteria (36). The Pseudomonas aeruginosa genome, for example, has separate genes for glutaredoxin- and thioredoxin-coupled ArsC reductases, while that for cyanobacteria appears to be an unusual hybrid with strong sequence similarity to thioredoxin-dependent reductase but functioning with glutaredoxin and glutathione instead (24). The cyanobacterial arsenate reductase also occurs as a homodimer (24), different from other known bacterial enzymes but similar to the yeast enzyme (36). The glutaredoxin-glutathione-coupled and thioredoxin-coupled ArsC arsenate reductases represent convergent evolution, in which a similar chemical solution has been “invented” more than once, analogous to the wings of birds and insects, which have no evolutionary relationship but which both allow for animal flight. The third clade of cytoplasmic arsenate reductases has been found so far only in fungi (36).

FUTURE AND BIOTECHNOLOGY POTENTIALS FOR PRACTICAL USE

Television and newspaper reports related to microbial inorganic arsenic transformation occur frequently; for example, a recent advertisement promised “arsenic-free lumber” for replacement of home decks, but the U.S. Environmental Protection Agency shortly thereafter decided against a requirement to use such lumber. The problems of drinking water containing high arsenic levels in Bangladesh are reported regularly in newspapers and magazines (8) and have their own mailing list (arsenic-crisis@yahoogroups.com) and sites (see, e.g., http:bicn.com/acid/ and http:groups.yahoo.com/group/arsenic-crisis-news/).

Among the many websites related to arsenic in drinking water and public health and environmental problems, readers may wish to access http://www.who.int/inf-fs/en/fact210.html and http://www.nlm.nih.gov/medlineplus/arsenic.html.

Microbial metabolism undoubtedly exacerbates environmental arsenic problems (14, 17, 18, 41), perhaps by releasing arsenic into drinking water in shallow wells (as in East Bengal, West Bengal, the American Midwest, and the Canadian Maritime Provinces). Understanding the mechanisms may help minimize the impact. It is proposed that microbial anaerobic respiratory arsenate reductase releases previously immobilized underground As(V) into water in newly drilled wells (18). It is possible that microbial metabolism (arsenite oxidase coupled with precipitation in mineral deposits) (18) can be harnessed for practical bioremediation of drinking water arsenic, although this prospect is just beginning to be recognized and no sustained efforts in this direction have been made. Microbial batch reactors to remove arsenic by oxidation of As(III) to As(V) (25) and the use of bacterial arsenate reductase genes in transgenic plants for potential phytoremediation by intracellular sequestration after reduction from As(V) to As(III) (10) were reported recently.

A bacterial arsC gene for cytoplasmic arsenate reductase and a gene whose enzyme product leads to the overproduction of glutathione in the plant model Arabidopsis were tested for potential phytoremediation (10). Synthesized alone in Arabidopsis, the bacterial arsenate reductase led to hypersensitivity to arsenate. However, when arsenate reductase was present together with an increased level of glutathione, greater arsenate resistance, along with the hyperaccumulation of an arsenite adduct of glutathione, was found (10). It was suggested that the arsenite adduct of glutathione is transported into the vacuole compartment of the plant cell, as is known to happen in yeast cells. This compartmentalization effectively removes the arsenic from the plant cytoplasm and places it in a harmless subcellular location. Further work is needed to apply these genetically modified plants to the removal of arsenic from polluted soils.