Significant concentrations of dissolved inorganic arsenic can be found in the waters of a number of lakes located in the western USA and in other water bodies around the world. These lakes are often situated in arid, volcanic terrain. The highest concentrations of arsenic occur in hypersaline, closed basin soda lakes and their remnant brines. Although arsenic is a well-known toxicant to eukaryotes and prokaryotes alike, some prokaryotes have evolved biochemical mechanisms to exploit arsenic oxyanions (i.e., arsenate and arsenite); they can use them either as an electron acceptor for anaerobic respiration (arsenate), or as an electron donor (arsenite) to support chemoautotrophic fixation of CO2 into cell carbon. Unlike in freshwater or marine ecosystems, these processes may assume quantitative significance with respect to the carbon cycle in arsenic-rich soda lakes. For the past several years our research has focused on the occurrence and biogeochemical manifestations of these processes in Mono Lake, a particularly arsenic-rich environment. Herein we review some of our findings concerning the biogeochemical arsenic cycle in this lake, with the hope that it may broaden the understanding of the influence of microorganisms upon the speciation of arsenic in more common, less “extreme” environments, such as drinking water aquifers.
Certain arsenic compounds are recognized as potent poisons since antiquity . In recent years, however, concern has arisen about prolonged human exposure to sub-acute levels of arsenic by ingestion of contaminated drinking water supplies. In the Ganges Delta, for example, groundwater in which naturally occurring arsenic has been mobilized into solution from the aquifer sediments has put millions of people at risk, while thousands have developed cases of severe arsenicosis (also called “arseniasis”) [2,3]. It is not clear at this juncture what specific mechanisms are responsible for this mobilization, but they appear to include a combination of chemical, physical, and microbial factors . These factors contribute to speciation changes in the arsenic itself, altering it back and forth from the more toxic and hydrologically mobile arsenite to the less toxic and less mobile arsenate. It should be borne in mind, however, that the dissolved arsenic concentrations in such aquifers are often quite low (∼1 μM) from the point of view of experimental microbiology, although not from one of human health. These low concentrations make detection of in situ microbial arsenic metabolism or identification of the causative microbial reactions and agents somewhat problematic. A further experimental difficulty with aquifer systems is the presence of a solid phase in contact with an aqueous phase. Both arsenate and arsenite have different adsorptive affinities to various common mineral surfaces (e.g., ferrihydrite, alumina) which strongly affect their concentration in the aqueous phase  and hence will complicate the interpretation of microbial phenomena . It is instructive, therefore, to study the microbial reactions that influence the speciation of arsenic in environments that are not only rich in this element, but are free of a sediment/aquifer solid phase. In these regards, the water column of Mono Lake, California is ideal.
Mono Lake is a closed basin lake in central California located on the eastern slope of the Sierra Nevada mountain range, lying within the western edge of the arid Great Basin (Fig. 1). It is only one of several alkaline lakes in this region that have naturally occurring high concentrations of dissolved inorganic arsenic, and hence this phenomenon is regionally widespread. These concentrations range from a freshwater end-member as represented by Crowley Lake  to the extreme case of salt-saturated Searles Lake. Crowley Lake receives As-laden hydrothermal waters from Hot Creek, an upstream tributary of the Owens River that drains through the lake . This lake's dissolved arsenic concentration is seasonally variable, but nonetheless the concentration shown in Fig. 1 is about sixfold higher than that the recommended maximum concentration limit (MCL) of 10 ppb (∼0.13 μM) for drinking water . Because Crowley Lake is part of the water supply system for Los Angeles, this is a public health concern. For perspective, contrast Crowley's concentration with Pavin Lake (∼0.14 μM) in France  or with Upper Mystic Lake (∼0.02 μM) in Massachusetts, the latter system being contaminated with anthropogenic arsenic . The arsenic concentrations in near-saturation, alkaline brines found in the western USA can be astonishingly high. For example, millimolar arsenic occurs in the dense brine of Searles Lake, which is located in the Mohave Desert. Similar levels of arsenic are present in the brines underneath the playa of Owens Lake, also located in eastern California [12,13]. The concentration of arsenic in Mono Lake lies between the two extremes of Crowley Lake and Searles Lake, but is certainly at the high end of what is typical of other closed basin soda lakes in this region and comparable to other volcanic/hydrothermal terrains around the world . The ground waters of these regions are also rich in arsenic [15,16]. In Mono Lake, arsenic occurs primarily as either the oxyanion arsenate [HAsO42− or As(V)] or arsenite [H2AsO3− or As(III)]. Organic forms of arsenic (e.g., monomethylarsonic acid, dimethylarsenic acid) are not significant relative to these inorganic forms . For this reason we have focused our research on inorganic arsenic species.
Approximate locations and chemical features of five arsenic-rich lakes located in western Nevada and eastern California.
The biochemical reactions mediated by microbes with regard to reduction of arsenate or oxidation of arsenite can be divided into two basic categories. These consist of either detoxification reactions that confer As-resistance, or redox reactions that conserve the energy gained by the reaction to provide for cell growth. Arsenic detoxification (resistance) reactions have been known for some years, and were reviewed recently by Rosen  and Mukhopadhyay et al. . However, they were not investigated in Mono Lake. Rather, the energy-conserving reactions that include respiratory arsenate reduction and chemoautotrophic arsenite oxidation, and which represent more recent discoveries, have become the focus of our efforts in Mono Lake. These processes have also been recently reviewed [20–24].
2 Limnological characteristics
Some of the salient facts concerning this lake's physical limnology, hydrochemistry, and biological properties are given in Fig. 1. The Mono Basin was formed about one million years ago by a local subsidence . Volcanic activity in the region began with the eruption of the Long Valley Caldera, about 700,000 years ago . The volcanism progressively migrated ∼40 km northward, eventually forming the Mono Craters on the lake's south shore as well as several very recent volcanic features located in or around the lake itself  including discharge from arsenic-rich hot springs. The lake has expanded and contracted in size and volume over the long period of its existence in response to changes in climate. It is currently about fivefold smaller in area and contains 18-fold less water than it did during the Pleistocene, which has resulted in the evaporative concentration of a number of constituents, including sodium, chlorine, sulfur, boron, arsenic, and dissolved inorganic carbon (as HCO3− plus CO32−). Mono Lake has exceptionally high concentrations of phosphate in its surface and bottom water. Although phosphate is a molecular analog of arsenate, it is not clear if this twofold difference in concentration between these elements ameliorates the toxic effects of the latter on various components of the lake's ecosystem. In contrast to phosphate, the inorganic nitrogen concentration of surface water is quite low, and provides a nutrient limitation on the extent of phytoplankton productivity. The lake's abundant dissolved carbonates and organic ligands help to retain high concentrations of various fallout radionuclides in solution, whereas they would normally precipitate into the sediments in lakes with circumneutral pH [28–30]. Numerous natural gas seeps (mainly CH4) are common in and around the lake , and both nitrification and methane oxidation occur in the water column .
Lake-wide vertical mixing is usually temperature-driven, occurring once per year in early winter (monomixis). However, strong El-Nino climatic events can cause episodic pulses of above average runoff into the lake from the Sierra Nevada, thereby diluting the salinity of the surface waters enough to offset the increased density caused by their normal cooling during winter. These events have resulted in periods without annual mixing (meromixis) that can last for several sequential years. The lake has been meromictic twice in recent decades, once in the 1980s  and again in the 1990s . The effects of meromixis include decreased algal primary productivity [35,36], resulting from decreased upward flux of inorganic nitrogen  and the establishment of highly reducing conditions in the anoxic bottom waters (monimolimnion) . During meromixis, the vertical distribution of arsenic species consists of predominantly arsenate in the oxic surface waters, followed by a change to mostly arsenite in the anoxic monimolimnion .
3 Microbial communities
Because of the unusual characteristics of alkaline, hypersaline environments in general and Mono Lake in particular, their bacterial communities have been studied for many years, in part in the quest for commercially valuable biotechnology [40,41]. These studies have been summarized recently by Oren , who includes a chapter reviewing the history and findings of microbiological studies in Mono Lake up to the date of publication. A paper published subsequent to Oren's review  reported the results of a phylogenetic characterization of Mono Lake bacteria. This study compared the distribution of ribotypes in libraries derived from samples collected from the mixolimnion (2 m, aerobic), the base of the oxycline (17.5 m, microaerophilic), the upper chemocline (23 m, seasonally anoxic) and the monimolimnion (35 m, perennially anoxic and highly reducing). The samples were collected during August 2001, when the lake was highly stratified and had been meromictic for six years. Because the analysis was based on PCR and cloning, the relative abundance of sequences in the libraries may deviate from the distribution in situ (e.g., see ).
Most of the 212 sequences retrieved (∼60 clones were analyzed per sample) fell into five major lineages of the domain Bacteria: α- and γ-Proteobacteria (6% and 10%, respectively); Cytophaga-Flexibacter (CF, 19%); high G+C gram-positive (Actinobacteria, 25%); and low G+C gram-positive (Bacillus Clostridium, 19%). Twelve percent were identified as derived from chloroplasts. The remaining 9% represented β- and δ-Proteobacteria; Verrucomicrobiales, Planctomycetales and candidate divisions. The populations of sequences retrieved from mixolimnion and oxycline libraries were dominated by sequences related to Actinobacteria (49% and 63%, respectively) distributed into only three distinct ribotypes (defined as sequence similarities ≥97%, most were ≥99%) while the population of sequences retrieved from the monimolimnion library was dominated (52%) by low G+C gram-positive bacteria distributed in 12 distinct ribotypes. In most cases, the sequences that are most similar to those retrieved from Mono Lake were retrieved from other alkaline, hypersaline environments . Archaea do not appear to be abundant in Mono Lake as we have not been able to retrieve archaeal ribotypes using published domain Archaea primers [45–47].
The Mono Lake Bacteria assemblage appears to have low diversity. A standard coverage estimate, C=1−(n1/N), where n1 is the number of ribotypes that occurred only once in the clone library and N is the total number of clones analyzed  indicated that 59–98% of the diversity in these libraries was sampled by sequencing ∼60 clones. Mixolimnion and oxycline libraries had the lowest sequence diversity with only 9 and 12 distinct ribotypes respectively; whereas chemocline and monimolimnion libraries were more diverse, containing 27 and 25 ribotypes, respectively. The composition of mixolimnion and oxycline libraries were not significantly different (LIBSHUFF analysis; ), but they were significantly different from chemocline and monimolimnion libraries (p<0.001) and chemocline and monimolimnion libraries were not significantly different from each other (p=0.006/0.124).
We used the distribution of ribotypes in our clone libraries to estimate the total number of ribotypes present in the samples using an analytical approach described by Curtis et al. . This analysis indicated that the total number of different ribotypes per sample ranged from 200 (17.5 m sample) to 4700 (23 m sample). The number of species in oceanic plankton samples or soil samples estimated by the same  or other methods [51–53] is of the order of 104.
Our data indicate that Mono Lake libraries are strongly dominated by a few ribotypes. All of the clones from the 2 m sample could be assigned to just nine ribotypes, with the three most abundant ribotypes accounting for 63% of all sequences. Forty three percent of the clones in the 17.5 m library fell into one ribotype while the three most abundant ribotypes accounted for 73% of the sequence population. The most abundant ribotype in the most diverse library (23 m) accounted for 10% of the population while the three most abundant ribotypes in this library accounted for 27% of the sequence population.
A broader survey using fingerprinting techniques similar to Hollibaugh et al.  indicated little seasonal variation in the richness and composition of the Mono Lake microbial community although there were seasonal shifts in apparent dominance, at least in the surface layer. Spatial variation in the relative abundance of ribotypes was related to mixing characteristics of the station sampled .
Sequences closely related (similarity >99%) to isolates known to be active in As cycling in Mono Lake have not been retrieved from clone libraries; however sequences of other members of clades containing these sequences (low G+C gram-positive bacteria, δ- and γ-Proteobacteria) are found regularly. An exception was the η-Proteobacteria. Although they were not retrieved from the Mono Lake clone libraries, they were a dominant band in a lakewater-based arsenate-enrichment incubation experiment . Fig. 2 shows a phylogenetic tree of arsenic-metabolizing bacterial species found to date in Mono Lake.
Neighbor-joining tree of 16S rRNA sequences of bacteria thought to be involved in arsenic geochemistry of Mono Lake. Strain MLHE-1 is an anaerobic arsenite oxidizer, B. arsenicoselenatis, B. selenitireducens and strain MLMS-1 are arsenate reducers, and the rest are reference sequences. All sequences are about 1400 bp except the two obtained by DGGE bands: ARS2band7 (133 bp) and ARS2band9 (159 bp).
4 Water column arsenic speciation and in situ microbial activity
Vertical profiles of some physical, chemical, and biological constituents taken when the lake was meromictic in 1999  are shown in Fig. 3. An oxycline was apparent from 10 m which descended to the disappearance of dissolved oxygen at 18 m, beneath which a chlorophyll a peak was evident (Fig. 3(a)). Unlike Big Soda Lake where photosynthetic bacteria are a prominent feature beneath the summer oxycline , this was not evident in Mono Lake. Rather, the chlorophyll maximum was mostly composed of Picocystis sp., an unusual eukaryotic alga [54,59] whose small diameter small (diameter, 1–3 μm) make it part of the “picoplankton.”
Vertical profiles of physical, chemical, and biological parameters in Mono Lake obtained in July, 1999 when the lake was meromictic. Reproduced from Oremland et al.  with permission.
The speciation of arsenic changed from arsenate (HAsO42−) in the oxic mixolimnion to arsenite (H2AsO3−) in the anoxic monimolimnion (Fig. 3(b)). The monimolimnion waters were highly reducing (Fig. 3(c)), denser than the mixolimnion (Fig. 3(d)), and contained abundant bacterial cells (Fig. 3(e)). The high concentrations of arsenate reported in the monimolimnion were an artifact of the initial analytical methodology employed, which determined it as the difference between total arsenic and arsenite. In subsequent analyses, arsenate was measured directly, which generated the profile shown in Fig. 4. Low, but significant quantities of arsenate (e.g., 5–10 μM) were consistently detected in the highly reducing monimolimnion.
Vertical profiles of As(V), arsenate reduction, and sulfate reduction in Mono Lake during October, 1999 when the lake was meromictic. The waters below 17 m depth were anoxic. An additional profile of sulfate reduction from the 1986 meromixis is included, when the lake was 4 m shallower. Reproduced from Oremland et al.  with permission.
Although the arsenite oxyanion was the main arsenic species in the lake's anoxic waters, a novel recent finding by the research team led by Hollibaugh was the detection of dissolved arsenic(III)/sulfide species with stoichiometries consistent with monothioarsenite (H2AsO2S−), dithioarsenite (H2AsOS2−) and trithioarsenite (H2AsS3−) species. Collectively these three thioarsenites accounted for 25% and 67% of total arsenic at 25 and 38 m depth, respectively, during August 2002. Although sulfides can rapidly reduce arsenate to arsenite at low pH values, this does not occur at pH 9.8 . Production of arsenic(III) trisulfide as a precipitate was noted during growth of Desulfosporosinus auripigmentum, an arsenate-respiring sulfate-reducer, when incubated at circumneutral pH . In freshwater lakes any such arsenic sulfides formed in the anoxic hypolimnion accumulate as precipitates in the sediments and thereby represent a net sink for removal of arsenic from the water column . In contrast, arsenic sulfides in the anoxic waters of Mono Lake are soluble and do not constitute a sediment sink and are oxidized back to arsenate when dissolved oxygen returns to the deeper waters during mixing events. The lack of such an arsenic sink contributes to the high dissolved arsenic content of Mono Lake's (and other alkaline lake's) waters. Rochette et al.  reported chemical formation of various arsenic3+ sulfides under conditions of low pH. These reactions also take place at pH 9.8, but the thioarsenites remain in solution .
A radiotracer technique was devised to measure dissimilatory arsenate reduction in the lake's water column . Water samples from the various depths were placed into sealed syringes, injected with 73As-arsenate, and incubated in the dark at in situ temperatures. Subsamples were withdrawn periodically from the syringes to generate time progress curves. The 73As-arsenate was separated from 73As-arsenite using ion exchange resins. The 73As-arsenate was not reduced in poisoned or filter-sterilized controls, thereby demonstrating its biological nature. The arsenate reduction assays were run concurrently with those for sulfate-reduction using 35S-sulfate. A profile of activity for bacterial arsenate and sulfate reduction is shown in Fig. 4. Highest rates of arsenate reduction (∼5.5 μmoll−1d−1) were evident at the bottom of the oxycline (18 m), then decreased to ∼0.5–1.0 μmoll−1d−1 at greater depths. The highest sulfate reduction rates (∼2.3 μmoll−1d−1) occurred at the deepest depths assayed, spatially separated by 10 m from the maximum for arsenate reduction.
The change in water column arsenic speciation over the course of a year as a response to vertical mixing was conducted in 2002 by the team led by Hollibaugh. This afforded an opportunity to determine the reliability of the arsenate reduction rates made with the radiotracer using an independent means of verification. Fig. 5 shows this change in speciation with time at a depth of 24 m as the upper water column underwent winter mixing, followed by the sinking of the late winter phytoplankton bloom, and the subsequent onset of anoxia. The rate of arsenate reduction between March and April corresponds to ∼6 μmoll−1d−1, after which it declines to 1.0 μmoll−1d−1 from April to May. These calculations are in close agreement with rates determined experimentally during meromixis (∼5.5 μmoll−1d−1; Fig. 4). Following summer stratification, temperature-associated mixing began in the fall and a trace of oxygen was detectable at 24 m depth by December. This created sub-oxic conditions as evidenced by a steep decline in sulfide and arsenite, and the reappearance of arsenate. The rate of arsenite oxidation from September to November extrapolates to ∼2.5 μmoll−1d−1, which is comparable to that observed for arsenate reduction at the beginning of the year.
Temporal variation of dissolved oxygen, arsenate, arsenite, and sulfide at 24 m depth in Mono Lake during 2002 obtained by Hollibaugh.
An estimate of the contribution made by microbial arsenate reduction to the mineralization of organic carbon formed annually by the photosynthetic activity of phytoplankton was made . This was achieved by dividing annual primary productivity (as electron equivalents) by the extrapolated integrated rates for water column arsenate reduction as well as sulfate reduction (Table 1). The results indicated that 8–14% of carbon fixed by phytoplankton primary productivity would be oxidized back to CO2 by the in situ activities of arsenate-respiring bacteria in the anoxic monimolimnion. Another 41% could be mineralized by sulfate reduction. Ignoring any aerobic degradation in the oxic mixolimnion, this means that about half of the sinking phytoplankton bloom would be oxidized to CO2 in the anoxic monimolimnion before it settled onto the bottom sediments.
Potential contribution of integrated monimolimnion arsenate reduction and sulfate reduction to the mineralization of phytoplankton primary productivity in Mono Lakea
aBased on the equation : CH2O+2H2AsO4−→HCO3−+2H2AsO3−+H+.
5 Manipulation experiments with Mono Lake water
Lab-based time course incubation experiments with anoxic Mono Lake water were performed to gain greater insight into what types of electron donors fueled in situ arsenate reduction, and the anions that could inhibit this process [56,57,65]. This included attempts at using “specific” inhibitors of sulfate-reduction (e.g., molybdate, tungstate) to help reveal the physiological types of microbes involved in arsenate reduction [56,57]. Preliminary experiments  were conducted with bottom water (24 m) to which 0.25 mM arsenate was added, and arsenate reduction was measured with 73As-arsenate. No obvious stimulation of either arsenate- or sulfate-reduction was achieved with addition of lactate or acetate as substrate amendments, implying that these processes were not electron donor limited. Arsenate itself caused a significant inhibition of sulfate reduction, which was only partially inhibited by tungstate. Curiously, the rates of arsenate reduction were five to sixfold higher than the in situ rates for this depth, but equaled the maximum displayed at 17 m (Fig. 4). It appears that dissimilatory arsenate reduction at 24 m was primarily limited by the availability of arsenate.
Another series of experiments were conducted without radiotracer in which the consumption of added 1.0–2.0 mM arsenate was followed . The rates of arsenate reduction in these experiments were typically 30–50-fold faster (150–250 μmoll−1d−1) than those observed previously at ambient arsenate or with ∼0.25 mM arsenate additions, reinforcing the idea of kinetic limitation by arsenate. Rates of arsenate reduction were slightly slower in samples amended with tungstate or molybdate, implying the involvement of sulfate-reducers in a small portion of the overall activity. Since some sulfate-reducers have been shown to respire arsenate [66,67], this was a reasonable interpretation, especially in light of analyses of the ensuing populations of microbes which showed that sequences for sulfate-reducers appeared in some of the arsenate-amended lakewater samples (see Section 6). No strong stimulation of arsenate reduction was noted by amendment with acetate or lactate, or for that matter with glucose, succinate, or malate again reinforcing the concept that arsenate-reduction in the monimolimnion is not limited by the availability of organic substrates.
The waters of Mono Lake contain an unusually high concentration of dissolved organic carbon, ∼6.7 mM. Although DOC is usually composed of refractory organics that are resistant to bacterial attack, it is possible that some portion of this DOC is “labile” in that it fuels arsenate reduction, especially when associated with the senescence of phytoplankton blooms. Alternatively, the monimolimnion also contains high levels of reduced inorganic compounds (sulfide, ammonia) as well as methane (Fig. 3) and it is possible that one or more of these substances serve as electron donors for dissimilatory arsenate reduction. Preliminary experiments have shown that this occurs with sulfide, and a pure culture of a bacterium (strain MLMS-1) was isolated by S.E. Hoeft that demonstrated growth via the sulfide-arsenate couple.
Arsenate reduction was strongly inhibited by the addition of nitrate ions to incubated bottom water . However, rather than acting as a preferred electron acceptor to arsenate by virtue of its higher electrochemical potential (Table 2), nitrate was found to couple to the biological oxidation of arsenite back to arsenate. Hence, the rate of arsenite oxidation essentially equaled that of arsenate reduction, thereby giving the appearance of no loss of added arsenate. In addition to nitrate, other potential biological oxidants included nitrite and to a lesser extent, chelated Fe(III) . Rapid oxidation of arsenite with MnO2 addition was also observed, but this proved to be of a chemical rather than biological nature. Investigation of in situ arsenite oxidation in lakewater, either with oxygen or nitrate as the electron acceptor, remains a topic for future investigation.
Electrochemical potential of some environmentally important redox pairsa
aCompiled from Bard et al. , Thauer et al.  and Macy .
6 Arsenic-metabolizing bacteria in Mono Lake
Two As(V)-respiring anaerobic bacilli, B. arsenicoselenatis (strain E1H) and B. selenitireducens (strain MLS10), were isolated from Mono Lake sediments using lactate as the electron donor and either selenate (E1H) or selenite (MLS10) as the electron acceptor. Both strains were also capable of anaerobic growth using As(V) as the electron acceptor for the oxidation of lactate 
The bacilli proved to be haloalkaliphiles as they exhibited growth maxima at ∼60 gl−1 salinity and pH at ∼9.5, and were therefore well adapted to withstand the chemical conditions of the lake (Fig. 1). Other electron acceptors and donors used for growth are shown in Table 3.
aCompiled from Switzer Blum et al. ; Oremland et al. , and Herbel et al. .
Arsenate-respiring bacteria, as well as sulfate reducers, were also found in the water column of Mono Lake as determined by most probable number culturing methods (Table 4). Not surprisingly, arsenate respirers were most abundant in the anoxic bottom waters where they were about 20-fold more abundant than sulfate reducers. Both types of bacteria were present in the anoxic water column during meromictic as well as monomictic periods. The phylogenetic affiliation of the arsenate reducers cultivated by MPN methods was not pursued. To gain some insight into this question, we performed cold (6 °C) incubations of anoxic bottom water amended with 1 mM arsenate followed by denaturing gradient gel electrophorsis (DGGE) of the extracted DNA . Unique bands that appeared in the arsenate-enriched water yielded sequence data that affiliated them with either η- or δ-Proteobacteria in the Thiomicrospira and Desulfovibrio groups, respectively, clades in which arsenate respirers have not as yet been described. However, some 16 species of arsenate-respirers have been reported in the literature since 1994 , and the list is growing even as this manuscript was being written (e.g., ). Taxonomically, these organisms can be found within several diverse divisions of prokaryotes (e.g., β-, γ-, δ-, and η-Proteobacteria, low G+C gram-positive bacteria, and Crenarchaea), and therefore it is not surprising that novel clades within these broad divisions responded to arsenate-enrichment, implying that there are new species within these divisions that are yet to be cultivated in pure culture.
Cell densities of arsenate- and sulfate-reducing bacteria in the water column of Mono Lake as determined by MPN culture serial dilutionsa
aOremland et al. ; values in parentheses represent 95% confidence intervals.
These results, of course, did not prove that populations found in these enrichments were representative of the natural arsenate-respiring flora in Mono Lake water. Indeed, as reviewed in Section 3 above, a survey of microbial diversity made by extraction of ambient DNA in the lake water (i.e., no arsenate enrichment or incubation) demonstrated that anoxic bottom waters were dominated by low G+C gram-positive bacteria of the Bacillus/Clostridium families . This would suggest that organisms like B. arsenicoselenatis and B. selenitireducens are more typical of water column arsenate-reducers. Nonetheless, to prove this point, future studies will need to be performed using molecular techniques that are more specific than phylogenetic probes like fluorescence in situ hybridization (FISH). One possibility would be to target the RNA for expression of dissimilatory arsenate reductase. However, before this can be attempted it will first require the purification of their arsenate-reductase and identification of this enzyme's encoding genes (see Section 7).
An anaerobic arsenite-oxidizing bacterium, strain MLHE-1, capable of chemoautotrophic growth with arsenite, sulfide, or hydrogen as its electron donor and nitrate as its electron acceptor, was isolated from Mono Lake bottom water . Growth in mineral salts medium on arsenite plus nitrate conforms to the following stoichiometry:
Strain MLHE-1 was capable of dark 14CO2 fixation with arsenite, and contained a Form 1 RuBisCo gene most similar to that of an unidentified Thiobacillus sp. (86% similarity). Strain MLHE-1 was also able to grow as a heterotroph with acetate as its electron donor either as an aerobe or with nitrate under anaerobic conditions. However, no growth on arsenite or its oxidation occurred under aerobic conditions. Strain MLHE-1 is a member of the Ectothiorhodospira clade of the γ-Proteobacteria, but is incapable of photosynthetic growth and lacks pigments. Previously, all other bacteria capable of autotrophic growth on arsenite were found in the α-Proteobacteria, and they are obligate aerobes .
7 Arsenic enzymes and genes
The isolation of the two Bacillus species from Mono Lake (B. selenitireducens, B. arsenoselenatis) provided the opportunity to investigate the pathway of arsenate reduction in these strains. Initial investigations indicated that they both possessed a membrane-associated enzyme that could couple the oxidation of methyl viologen to arsenate reduction. The respiratory arsenate reductase from B. selenitireducens was subsequently purified . Its characterization has revealed some features in common with the enzyme that has been previously purified from the arsenate-respiring bacterium Crysiogenes arsenatis, but has some distinctions. It is tightly associated with the cytoplasmic membrane and requires solubilization with detergent (i.e., 2% Triton X-100). The active enzyme was purified in a heterodimeric state, with subunits of 110 and 34 kDa. This compares with the heterodimer of 87 and 29 kDa for C. arsenatis. The large subunit, ArrA, contains the catalytic site, while the smaller subunit, ArrB, is an iron-sulfur-cluster protein assumed to be involved in electron shuttling. The enzyme has a significantly higher affinity for arsenate (Km of 34 μM as compared to 300 μM for C. arsenatis) and exhibits an optimal activity at pH 9.5 and salinity of 90 gl−1. The enzyme is also able to oxidize methyl viologen with selenate, selenite, and arsenite but not nitrate, nitrite, or fumarate. Thus it shows broader substrate specificity than C. arsenatis which was reported to be arsenate specific . Comparison of the N-terminal amino acid sequence of B. selenitireducens indicated a high degree of identity with the known amino acid sequence of C. arsenatis. Thus, the B. selenitireducens ArrA (20 amino acids) had a 50% identity and 85% similarity to that of C. arsenatis, while the ArrB (31 amino acids) had a 48% identity and 87% similarity.
Using the N-terminal sequence data, degenerate primers were designed to PCR amplify the gene encoding arrA. A 2.5 kb amplicon was obtained which contained all but the 5′ end of arrA, an intergenic region of 29 nucleotides, and 30 nucleotides of the 5′ end of arrB. The inferred protein had a molecular mass of 90,980 Da and a pI of 5.12. It shares a 56.7% identity with the arrA from Shewanella sp. strain ANA-3 . The highest degree of identity was with a gene that encodes a hypothetical protein from the genome of Desulfitobacterium hafniense. D. hafniense, a low mole G+C gram-positive bacterium, was shown to be capable of using arsenate as a terminal electron acceptor, although its arsenate reductase was not identified . Analysis of the contig revealed that the putative arr operon contains at least seven genes. The first three genes are putative regulatory elements and include a sensory kinase and a response regulator. These are followed by a gene that encodes a protein of unknown homology but is predicted to be membrane integral, then arrA and arrB. The last gene before the contig ends encodes a chaperon-like protein similar to TorD and SerD. This operon is unlike the arr operon from Shewanella sp. strain ANA-3 that consists only of arrA and arrB. Whether the arr operon from B. selenitireducens contains additional genes remains to be determined.
The phylogenetic diversity of prokaryotes capable of arsenate respiration minimizes the usefulness of molecular probes based on 16S rDNA sequence. However, the high degree of identity between the respiratory arsenate reductases from phylogenetically distant organisms suggests it will possible to develop biochemical (i.e. immunohistochemical) and molecular probes. Indeed, alignment of arrA from B. selenitireducens, Shewanella sp. strain ANA-3, and D. hafniense reveal several regions with high identity that differentiate this protein from other related members of the dimethyl sulfoxide (DMSO) reductase class of enzymes. We are currently investigating whether these regions will provide useful epitopes for both Western blot analysis and fluorescent in situ labeling of cells. The development of molecular probes may present a greater challenge as coding bias is evident in the few gene sequences currently available. While comparison of arrA from two species, such as B. selenitireducens and D. hafniense may show a high degree of sequence identity (e.g., 60%), this identity drops to 43% when the sequence from Shewanella sp. strain ANA-3 is added to the mix. Nevertheless, degenerate oligonucleotide primers have been used successfully by D.K. Newman to clone the putative arrA from a number of diverse species.
While there are more than 30 species of bacteria known to oxidize arsenate, only seven are chemolithoautotrophs that can grow by coupling the energy gained from this oxidation to the incorporation of CO2 into cellular material . The arsenite oxidizing bacterium strain MLHE-1, isolated from Mono Lake, was the first to couple the oxidation of arsenic to the reduction of nitrate . To date only the arsenite oxidase from the heterotrophic bacterium Alcaligenes faecalis has been purified and fully characterized (including the crystal structure) . A mononuclear molybdenum enzyme belonging to the DMSO class, it is comprised of two subunits. The catalytic subunit (∼85 kDa) contains molybdenum bound to two pterin cofactors and a [3Fe-4S] cluster. The associated subunit (∼14 kDa) presumably functions as an electron shuttle and has a Rieske-type [2Fe-2S] cluster. More recently, the genes encoding the catalytic and Rieske subunits have been cloned and sequenced from the beta Proteobacterium Cenibacterium arsenoxidans strain ULPAs1  and identified in the genomes of several other organisms including Chloroflexus auranticus, Sulfolobus tokodaii, and Aeropyrum pernix. A comparison of the amino acid sequence of the catalytic subunit (AoxA) from these five organisms shows they share only 15.7% identity and 42% similarity. While the Rieske subunit (AoxB) shows slightly better homology (23% identity and 48% similarity), this is due to the highly conserved nature of the iron sulfur cluster binding site. Thus, whether biochemical and molecular probes that specifically recognize arsenite oxidases can be successfully developed remains to be determined. An interesting finding was that the evolutionary lineage of the AoxA subunit seemed to occur prior to the taxonomic split between the Bacteria and Archaea . This suggests that arsenite oxidation is an ancient trait that developed quickly in the Earth's early biosphere.
8 Organic arsenicals
In this review we have focused upon the energy-yielding transformations of inorganic arsenic in Mono Lake that are mediated by prokaryotes. This has constrained the discussion to the reduction of arsenate to arsenite and the corresponding reverse oxidation reaction. We were justified in this approach because Anderson and Bruland  have shown that organo-arsenicals are not abundant in the lake relative to its large inorganic pool. However, the analytical technique employed by these workers was essentially overwhelmed by the great size of the inorganic arsenic pool, thereby limiting its ability to directly detect any organo-arsenicals in Mono Lake surface water. Thus, their reported values were “less than 100 nM” for monomethylarsonic acid (MMAA) and dimethylarsenic acid (DMAA). For comparison, they also reported quite respectable concentrations of MMAA (8.6 nM) and DMAA (33 nM) in the saline Salton Sea, where the inorganic arsenic concentration is only 0.13 μM, or over three orders of magnitude less than that of Mono Lake. It seems possible that significant concentrations of organo-arsenicals occur in Mono Lake, and that a future lake-wide survey aimed at describing their molecular identification, distribution, and seasonal variability would be justified.
Although little is known about organoarsenicals in alkaline soda lakes, the marine environment has been well-studied in this regard. Therefore, we can use what we know about marine systems as a departure point from which to formulate hypotheses concerning the biogeochemical cycling of organoarsenicals in Mono Lake. For example, it is well known that methylated arsenicals are synthesized by marine phytoplankton  and that these substances are present in ocean water  and anoxic marine sediments [84,85]. Of the various marine organoarsenicals, one of the most widespread forms is arsenobetaine. Arsenobetaine, an analog of the organic osmolyte glycine betaine, is commonly found in a diversity of marine animals including fish and invertebrates . It would make considerable “sense” if the microalgae and zooplankton in Mono Lake employed arsenobetaine as an osmolyte, because its biosynthesis would additionally confer a protective mechanism to counter the toxicity of inorganic arsenic ions entering the cell. Preliminary measurements of samples of the brine shrimp Artemia monica taken from Mono Lake failed to detect much arsenobetaine; however, high levels of MMAA and DMAA were detected by W.C. Cullen in the extracted tissues. Glycine betaine is an important osmolyte in the eukaryotic picoplankter Picocystis strain ML from Mono Lake , but it is not known if cells contain arsenobetaine. Strain ML-D, an aerobic bacterium isolated from Mono Lake accumulated arsenobetaine when it was provided externally, accounting for as much as 28% of the internal osmolyte pool which also consisted of ectoine, DMSP (dimethylsulfoniopropionate), and glycine betaine . It is not known as yet if the organisms living in Mono Lake actually synthesize and re-cycle arsenobetaine in situ.
The pelagic and littoral sediments of Mono Lake have active methanogenic activity [88–90]. The high concentration of sulfate in the lake competitively eliminates acetate and H2 from methanogenic pathways, which shifts methane precursors to non-competitive methylotrophic substrates like trimethylamine or dimethylsulfide. Anaerobic degradation of arsenobetaine could follow pathways analogous to that for glycine betaine  resulting in the formation of trimethylarsine rather than trimethylamine. Methanogenic degradation of 14C-trimethylamine was detected in Mono Lake pelagic sediments  but was inhibited when littoral sediments were incubated with 10 mM trimethylamine, probably due to the buildup of ammonia as an end product, which is toxic at this pH . It is not known if methylotrophic methanogens can degrade trimethylarsine since the endproduct, arsine gas (H3As), would be even more toxic than the corresponding concentration of ammonia.
The subject of microbial methylation of arsenic and related metalloids was recently reviewed by Bentley and Chasteen . Methylotrophic methanogens can demethylate metalloids like dimethylselenide, which is an analog of dimethylsulfide, one of their recognized growth substrates . However, in the case of arsenic, McBride and Wolfe  demonstrated the opposite, namely that whole cells and extracts of hydrogen-oxidizing Methanobacterium bryantii could form dimethylarsine from externally supplied arsenate ions. It appears that the ability of various species of methanogenic archaea to either methylate or demethylate toxic elements like arsenic is an area ripe for closer scrutiny. Certainly novel isolates from arsenic rich environments like Mono Lake would be worth testing in this capacity.
9 Future directions
Mono Lake has provided us with a wealth of new insights with regard to the diversity of microbes carrying out arsenic metabolism, their novel biochemical reactions, and the importance of these reactions to the biogeochemical cycles occurring in the lake, including the carbon cycle. The first two haloalkaliphiles isolated from the lake, B. selenitireducens and B. arsenicoselenatis grow as conventional heterotrophs while respiring arsenate. However, two novel strains, MLHE-1 and MLMS-1, provide an autotrophic-based linkage of arsenic to the nitrogen and sulfur cycles of the lake. In the case of MLHE-1, nitrate-linked oxidation of arsenite has already been demonstrated for more conventional freshwater environments, such as Upper Mystic Lake , and a subsurface aquifer in Bangladesh . Clearly, work in such an extreme environment like Mono Lake has technical transferability to our understanding of hydrologic systems of more concern to human health.
Future research work in the lake should focus on using molecular tools, such as functional gene probes, for enumerating arsenate-respiring and arsenite-oxidizing populations. Such efforts will directly rely on progress made in the laboratory on the biochemistry and genetics of arsenic metabolism of Mono Lake arsenotrophs, as well as with isolates from freshwater systems. It also will be important to develop radiotracer assays to quantify the rates of arsenate reduction and arsenite oxidation occurring in the lake's shallow sediments. Although a potential impediment may be the adsorptive binding of the 73As-arsenate radiotracer with any Fe(III) present in the sediments , for Mono Lake in particular this may be more than offset by the large amount of available phosphate that would compete with arsenate for binding sites. Such research may ultimately lead to practical techniques that can be adapted to freshwater systems and drinking water aquifers. Therefore, Mono Lake and similar soda lakes are a good departure points to initiate such an endeavor.
Funding in support of this research was provided by the USGS National Research Program (R.S.O.), the NASA Exobiology Research Program (R.S.O. and J.F. Stolz), and by the NSF Microbial Observatories Research Program (J.T.H.). We are grateful to J. Hering and L.G. Miller for their constructive criticisms of an earlier draft of the manuscript.
(1998) Bacillus arsenicoselenatis sp. nov., and Bacillus selenitireducens sp. nov.: two haloalkaliphiles from Mono Lake, California that respire oxyanions of selenium and arsenic. Arch. Microbiol. 171, 19–30.
(1994) Biochemistry of selenium metabolism by Thauera selenatis gen. nov., sp. nov., and use of the organism for bioremediation of selenium oxyanions in San Joaquin Valley drainage water. In: Selenium in the Environment (FrankenbergerW.T.Jr.. Ed.), pp. 421–444. Marcel Dekker, New York.
(2003) Reduction of elemental selenium to selenide: experiments with anoxic sediments and bacteria that respire Se-oxyanions. Geomicrobiol. J. 20, 587–602.
(1993) Aspects of the biogeochemistry of methane in Mono Lake and the Mono Basin of California. In: Biogeochemistry of Global Change. Radiatively Active Trace Gases (OremlandR.S. Ed.), pp. 704–741. Chapman & Hall, New York.