Homology of Escherichia coli R773 arsA, arsB, and arsC Genes in Arsenic-Resistant Bacteria Isolated from Raw Sewage and Arsenic-Enriched Creek Waters

Abstract

The occurrence and diversity of the Escherichia coli R773 ars operon were investigated among arsenic-resistant enteric and nonenteric bacteria isolated from raw sewage and arsenic-enriched creek waters. Selected isolates from each creek location were screened for ars genes by colony hybridization and PCR. The occurrence of arsA, arsB, and arsC determined by low-stringency colony hybridization (31 to 53% estimated mismatch) was 81, 87, and 86%, respectively, for 84 bacteria isolated on arsenate- and arsenite-amended media from three locations. At moderate stringency (21 to 36% estimated mismatch), the occurrence decreased to 42, 56, and 63% for arsA, arsB, and arsC, respectively. PCR results showed that the ars operon is conserved in some enteric bacteria isolated from creek waters and raw sewage. The occurrence of the arsBC genotype was about 50% in raw sewage enteric bacteria, while arsA was detected in only 9.4% of the isolates (n = 32). The arsABC and arsBC genotypes occurred more frequently in enteric bacteria isolated from creek samples: 71.4 and 85.7% (n = 7), respectively. Average sequence divergence within arsB for six creek enteric bacteria was 20% compared to that of the E. coli R773 ars operon. Only 1 of 11 pseudomonads screened by PCR was positive for arsB. The results from this study suggest that significant divergence has occurred in the ars operon among As-resistant E. coli strains and in Pseudomonas spp.

The contamination of drinking water sources with arsenic (As) poses a potential threat to human health. Inorganic As, including the highly toxic trivalent form [arsenite; As(III)] and less toxic pentavalent As [arsenate; As(V)], is associated with increased cancer risk in a number of geographic areas (16, 40, 41). The toxicity of As is attributed to the substitution of As(V) for phosphate, affinity of As(III) for protein thiol groups, and protein-DNA and DNA-DNA cross-linking (23). Arsenic enrichment and pollution of environmental waters originate from either natural or anthropogenic sources. Sodium arsenite, monomethylarsonate (MMA), dimethylarsinic acid (DMA), and lead arsenate (PbHAsO₄) have been extensively used as herbicides and pesticides (13, 29, 30). Geological processes such as geothermal activity and weathering of As-containing rocks also contribute significantly to As enrichment of aquatic environments (15, 43).

Many bacteria have been isolated that exhibit resistance to lethal concentrations of arsenic (greater than 5 mM sodium arsenite) (8, 15), yet little is known about the genetics involved in As resistance (As^r) in environmental bacteria. Plasmids have been detected in some bacteria exhibiting high levels of resistance to arsenate, arsenite, and antimonate (5, 9, 19). In addition, As^r loci have also been found on the chromosomes of Pseudomonas aeruginosa and E. coli. The most-well-characterized genetic system for resistance to arsenicals is known as the ars operon. Escherichia coli and Staphylococcus ars operons have been thoroughly investigated at the genetic and biochemical levels (19, 33, 35). Additionally, similar homologs have been found during chromosomal sequencing of bacterial species, but the function of these is unknown (The Institute for Genomic Research <http://tigr.org/tdb/;> and National Center for Biotechnology Information < http://www.ncbi.nlm.nih.gov/>). The E. coli plasmid R773 ars operon contains five genes, arsRDABC, encoding an arsenate reductase (ArsC) that reduces arsenate to arsenite, a membrane-bound anion-translocating ATPase (ArsA), and ArsB, an inner membrane protein that forms the anion-conducting channel (11). The ars operon functions as a detoxification mechanism by lowering the intracellular arsenic concentration, thus conferring resistance to As(V) and As(III). While ArsR is a trans-acting inducer-responsive repressor, ArsD is an inducer-independent protein controlling basal and upper-level expression.

Gram-negative plasmid ars operons share highly homologous sequences, yet are highly divergent from their gram-positive counterparts. Sequences homologous to the E. coli chromosomal ars operon are also highly conserved among enterobacterial genera (12). In addition, the recently discovered P. aeruginosa ars operon containing arsRBC appears to be conserved in P. fluorescens, but not in other arsenic-resistant Pseudomonas spp. (4). Open reading frames sharing homology with the E. coli arsRDABC have also been found on the Acidiphilium multivorum plasmid pKW301 (37) and the archeal Halobacterium sp. strain NRC-1 plasmid pNRC100 (10, 14).

In other metal resistance genetic systems, such as the copper resistance operon pcop, the enteric model represented by E. coli has been shown to be structurally and functionally equivalent to other copper-resistant systems (5, 31). We therefore used a well-characterized genetic system, the E. coli R773 ars operon, as a model to investigate the diversity of the ars operon in As^r bacteria isolated from raw sewage and natural waters. The overall goal of our study was to determine the prevalence of this genetic model in relating the As^r phenotype among selected populations of enteric and nonenteric bacteria. DNA-DNA hybridization was done under different stringency conditions to address the issue of diversity of ars genes in As^r isolates originating from natural waters. In addition, PCR was evaluated for detection of ars genes in enteric and nonenteric bacteria. A phylogenetic analysis of novel ars-like sequences detected in environmental enteric bacteria is also presented.

MATERIALS AND METHODS

Bacterial strains, plasmids, and culture conditions.

The strains and plasmids used in this study are listed in Table 1. Isolates were obtained from water samples collected from Hot Creek, South Haiwee Drain 5, and Irvine Ranch Water District and Orange County Sanitation District raw sewage samples; all sampling sites are located in California. Samples were stored in sterile polyethylene bottles and transported on ice packs. Sewage samples were plated on mTEC (Difco) and mENDO (Difco). Creek waters were plated on 10 or 50% strength Difco plate count agar (PCA).

TABLE 1.

Bacterial strains analyzed for arsA, arsB, and arsC

Strain or isolate	Type/identificationa	Phenotypeb	ars genotypec	Total As of water sample (μg/liter)
E. coli strains
	pUM3 ars operon subclone of R773	As(III)r As(V)r Sbr	arsA+arsB+arsC+
	JM109 cloning host for pGEM-T vector		ars PCR negative
Raw sewage isolates				<50
ECO1	E. coli	As(III)r	arsB+arsC+
ECO2	E. coli	As(III)r	arsB+arsC+
ECO7	E. coli	As(III)r	arsA+arsB+arsC+
EN-1	Enteric	Ass	ars PCR negative
EN-2	Enteric	Ass	ars PCR negative
EN-3	Enteric	Ass	arsC+
EN-4	Enteric	As(III)r As(V)s	arsB+arsC+
EN-5	Enteric	Ass	ars PCR negative
ENIII-1	Enteric	As(III)r As(V)r	ars PCR negative
ENIII-2	Enteric	As(III)r As(V)r	arsC+
ENIII-3	Enteric	As(III)r As(V)r	ars PCR negative
ENIII-4	Enteric	As(III)r As(V)r	arsA+arsB+arsC+
ENV-1	Enteric	As(III)r As(V)r	arsB+arsC+
ENV-2	Enteric	As(III)s As(V)r	ars PCR negative
ENV-3	Enteric	As(III)s As(V)r	ars PCR negative
ENV-4	Enteric	As(III)r As(V)r	arsB+arsC+
ENV-5	Enteric	As(III)r As(V)r	ars PCR negative
G2	E. coli	As(III)r	arsB+
G3	E. coli	As(III)r	arsB+
G4	E. coli	As(III)r	arsB+
G5	E. coli	As(III)r	arsB+arsC+
G6	E. coli	As(III)r	arsB+arsC+
G7	E. coli	As(III)r	arsB+arsC+
G8	E. coli	As(III)r	arsB+arsC+
G9	E. coli	As(III)r	arsB+arsC+
G10	E. coli	As(III)r	ars PCR negative
G11	E. coli	As(III)r	arsB+arsC+
G13	E. coli	As(III)r	ars PCR negative
G14	E. coli	As(III)r	ars PCR negative
G15	E. coli	As(III)r	ars PCR negative
G16	E. coli	As(III)r	ars PCR negative
G18	E. coli	As(III)r	ars PCR negative
Hot Creek Isolates				180
HCB2	Enterobacter cloacae	As(III)r	arsA+arsB+arsC+
HCSH1	Klebsiella sp.	As(III)r	arsAhyb+arsBhyb+arsChyb+
HCSH9	Yersinia intermedia	As(III)r	arsAhyb+arsB+arsC+
KOV2	Acinetobacter calcoaceticus	As(III)r	ars PCR negative
KOV3	Xanthomonas oryzae	As(III)r	arsB+
KOV4	Janthinobacterium lividium	As(III)r	ars PCR negative
KOV5	Pseudomonas corrugata	As(III)r	ars PCR negative
South Haiwee Drain Isolates				740
SHA1	Serratia fonticola	As(III)r	arsA+arsB+arsC+
SHA2	Pseudomonas corrugata	As(III)r	arsAhyb+arsBhyb+arsChyb+
SHA3	Acinetobacter genospecies 15	As(III)r	arsAhyb+arsBhyb+arsChyb+
SHA4	Pseudomonas corrugata	As(III)r	arsAhyb+arsBhyb+arsChyb+
SHA6A	Pseudomonas vescicularis	As(III)r	arsBhyb+arsChyb+
SHA8	Pseudomonas corrugata	As(III)r	arsAhyb+arsBhyb+arsChyb+
SHA10	Pseudomonas corrugata	As(III)r	arsAhyb+arsBhyb+arsChyb+
SHA17	Serratia fonticola	As(III)r	arsA+arsB+arsC+
SHA27	Pseudomonas cichorii	As(III)r	arsAhyb+arsBhyb+arsChyb+
SHA29	Serratia fonticola	As(III)r	arsA+arsB+arsC+
SHA35	Serratia fonticola	As(III)r	arsA+arsB+arsC+

^aE. coli refers to isolates plated on mTEC supplemented with arsenic. Enteric refers to isolates plated on mENDO amended with arsenic.

^bAs(III)^r, resistance to 500 mg of sodium arsenite per liter; As(V)^r, resistance to 1,000 mg of sodium arsenate per liter; Sb(III)^r, resistance to antimony.

^chyb⁺, negative by PCR and positive by hybridization with the corresponding ars probe; +, positive by PCR.

Arsenic-resistant bacteria were obtained by spread plating diluted samples onto PCA medium supplemented with 500 mg of As(III) and 1,500 mg of As(V) per liter. Selected isolates were subsequently streak purified twice on PCA (creek isolates) and Luria-Bertani-Miller (LB) agar (Difco) (raw sewage isolates) containing 500 mg of As(III) or 1,000 mg of As(V) per liter.

Creek isolates were grown at 30°C in broth similar to 50% PCA amended with 500 mg of sodium arsenite [As(III)] per liter. Raw sewage isolates were grown in LB broth supplemented with 500 mg of As(III) or As(V) per liter. Cultures were stored in 50% glycerol at −80°C.

DNA preparation.

Total DNA was obtained by a modified guanidine thiocyanate extraction procedure (36). Overnight cultures (0.5 ml) were centrifuged (Eppendorf 5415C) at 14,000 × g for 5 min. Pellets were resuspended in 0.6 ml of guanidine thiocyanate lysis buffer (5.3 M guanidine thiocyanate, 10 mM dithiothreitol, 1% Tween 20, 0.3 M sodium acetate, 50 mM sodium citrate [pH 7.0]), 53 μl of cetyltrimethylammonium (CTAB [10%], 0.7 M NaCl), and 8 μl of 0.5 M NaCl. After incubation of the mixture at 65°C for 10 min, 50 μl of Glassmilk (Bio 101, Inc., La Jolla, Calif.) was added, and this combination was mixed for 15 min at room temperature. Following centrifugation for 1 min at 14,000 × g, pellets were rinsed three times with wash buffer (50% ethanol, 10 mM Tris-HCl [pH 7.5], 100 mM NaCl). DNA was eluted off the Glassmilk by resuspending the pellet in Tris-EDTA (TE [pH 8.0]) for 5 min at 50°C with periodic mixing. The supernatant was removed and treated with RNase at a final concentration of 50 μg/ml.

Species identification.

Isolates were streak purified on Trypticase-soy agar (TSA) with or without 5% sheep’s blood. Following incubation at 30°C for 24 h, GN Microplates (Biolog, Hayward, Calif.) were used to identify the species. A similarity index greater than 0.5 was considered a positive identification.

PCR amplification.

Primers were designed for three of the structural genes arsA, arsB, and arsC of the ars operon from the E. coli pUM3 plasmid (7). The criteria used in the design of the primers included conservation of homologous sequences determined in multisequence alignments (data not shown) or inclusion of active sites for arsenic binding. Table 2 lists the primer sequences, the targeted regions for the specific gene, and the expected PCR fragment sizes. Primers were tested for cross-reactivity to other bacterial sequences by BLAST.

TABLE 2.

PCR primers designed from the E. coli R773 plasmid encoded ars operon genes, arsA, arsB, and arsC and PCR primers for constructing internal probes for the arsA-1, arsB-1, and arsC-1 primer sets

Targeted regiona (bp)	Name	Sequence (5′ to 3′)	Tmb (°C)	Product size (bp)
203–222	arsA-1-F	TCCTGGATTGTCGGCTCTTG
			58	186
367–386	arsA-1-R	ATCTGTCAGTAATCCGGTAA
266–285	I-arsA-1-F	CGTTGACCCTATTAAAGGCG
			60	103
347–366	I-arsA-1-R	ATTCATCAAAAGCCGCAATC
139–158	arsB-1-F	CGGTGGTGTGGAATATTGTC	59	219
336–355	arsB-1-R	GTCAGAATAAGAGCCGCACC
166–184	I-arsB-1-F	CGACGGCAACATTTATCGC	62	181
327–344	I-arsB-1-R	AGCCGCACCATCGTTGGC
46–67	arsC -1-F	GTAATACGCTGGAGATGATCCG	59	370
393–413	arsC -1-R	TTTTCCTGCTTCATCAACGAC
99–118	I-arsC -1-F	TACCTTGAAAACCCGCCTTC	62	240
316–335	I-arsC -1-R	AACCACTTCAGAAGGACGGC

^aReferenced from the coding sequence (CDS) of the E. coli R773 ars operon.

^bT_m, annealing temperature used in PCR cycle.

PCR was carried out with a Perkin-Elmer 9600 Thermocycler. Twenty-five or 50-μl reaction mixtures were composed of 10 mM Tris-HCl (pH 8.3), 50 mM KCl, 2.0 mM MgCl₂, 200 nM each primer, 0.625 U of Taq polymerase (Perkin-Elmer or Promega) per 25 μl, and 50 ng of template DNA. The protocol for each primer set consisted of an initial denaturation step (94°C for 3 min) followed by 30 to 35 cycles of 94°C for 30 s, 58°C for 30 s, and 72°C for 30 s (arsA-1) or 30 to 35 cycles of 94°C for 30 s, 59°C for 30 s, and 72°C for 30 s (arsB-1 and arsC −1). A final extension was done for 7 min at 72°C. The E. coli pUM3 purified plasmid served as a positive control for arsA-1, arsB-1, and arsC-1 primer sets. Negative controls included a deionized water reagent control, Pseudomonas putida, and E. coli JM109. The amplified products (5- to 10-μl aliquots) were separated on a 1× Tris-borate-EDTA (TBE)-1.5 or 2% agarose gel (0.5 μg of ethidium bromide per ml) by electrophoresis. The bands were visualized on a UV transilluminator.

Preparation of internal probe.

PCR was used to construct internal probes to confirm the identity of the amplified DNA bands. Nested primers were designed to amplify internal regions of the arsA-1, arsB-1, and arsC-1 fragments, respectively. Plasmid DNA from E. coli pUM3 was first subjected to PCR with arsA-1, arsB-1, and arsC-1 primer sets, respectively, according to the PCR protocol described above. The fragments were purified by electrophoresis through a 2.5% agarose gel and buffered in 1× TBE, which was followed by excision of the corresponding band. DNA was eluted from the gel slice with a Genelute (Supelco) spin column according to the manufacturer’s instructions. PCR with the internal primer sets I-arsA-1, I-arsB-1, and I-arsC-1, respectively, was then done. The amplified fragments were purified by agarose gel electrophoresis followed by band excision and Supelco spin column purification. Twenty-five to 50 ng of purified DNA was labeled with [³²P]dCTP by random priming with Klenow enzyme and random hexamers (Boehringer-Mannheim) according to the manufacturer’s instructions.

Southern blot and hybridization.

DNA was transferred to nylon membrane (MagnaCharge; MSI, Westboro, Mass.) by capillary action according to standard methods (2). DNA was fixed to the membrane by UV cross-linking at 1,200 J/cm² (FB-UVXL-1000; Fisher Scientific) and stored under a vacuum at ambient room temperature.

Membranes were hybridized by a method similar to that of Keller (17). Briefly, membranes were treated with prehybridization solution [50% formamide, 5× SSC (1× SSC is 0.15 M NaCl plus 0.015 M sodium citrate), 31 mM KH₂PO₄, 0.25% sodium dodecyl sulfate (SDS), 1× Denhardt’s solution, and 100 μg of poly(A) per ml] and incubated at 42°C for 2 h, followed by hybridization at 42°C overnight with 1 to 5 ng of denatured ³²P-labeled probe DNA per ml. For moderate stringency, the filters were washed four times for 5 min each in 2× SSC-0.1% SDS. Low-stringency washes were done two times for 15 min each in 0.1× SSC-0.1% SDS at 50°C. The filters were exposed to X-Omat (Kodak) X-ray film at −80°C for 4 to 24 h.

Colony hybridization with ars operon probes.

Colony lifts were done according to the method of Sambrook et al. (34). Colonies were transferred onto MagnaGraph nylon transfer membranes (0.22-μm pore size; MSI). DNA was fixed to the membranes by UV cross-linking. Negative controls included degrading P. putida and salmon sperm DNA. Probes for arsA, arsB, and arsC were constructed from PCR fragments amplified from E. coli pUM3. Radiolabeling was done with gel-purified PCR products (described above). Membranes were hybridized as described above with the following modifications: prehybridization and hybridization were done at 37°C and with washes at two stringencies: low (2× SSC-0.1% SDS at 37°C) and moderate (1× SSC-0.1% SDS at 50°C). The estimated percent mismatches for low- and moderate-stringency washes were 31 to 53 and 21 to 36%, respectively. Percent mismatch was calculated based on the method of Anderson (1). Filters were exposed to X-ray film for 2 to 4 days at −80°C. Colony lysis was confirmed with 0.2% methylene green staining of the membranes (25).

Cloning and sequencing.

Isolates that were PCR positive for arsB and arsC were used to amplify a 1.5-kb fragment by using primers I-arsB-1-F and I-arsC-1-R. PCR products were separated on 1% agarose gel followed by excision of the corresponding band. The DNA was purified as described above. Purified arsBC fragments were cloned with the pGEM-T Easy Vector System (Promega, Madison, Wis.) according to the manufacturer’s instructions. Ligation reactions were transformed into competent E. coli JM109 cells (Promega, Madison, Wis.). Blue and white screening was done on LB agar supplemented with 100 μg of ampicillin per ml, 0.5 mM IPTG (isopropyl-β-d-thiogalactopyranoside), and 80 μg of X-Gal (5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside) per ml. Plasmid DNA was extracted (Qiaprep spin kit; Qiagen) from overnight cultures of four white colonies per isolate. To verify the presence of the arsBC insert, PCR was used with I-arsB-1-F and I-arsC-1-R primers.

Automated sequencing was done on an ABI 377 by a dideoxynucleotide method at the Davis Sequencing Facility (Davis, Calif.). The m13(−21) forward sequencing primer was used to sequence the cloned PCR fragment. A single primer walk was performed on cloned arsBC fragments with the following primers: PW1 (5′-GAGCCACTCGGTATTCCTGTGAG-3′), specific for SHA1, SHA17, SHA35, and HCSH9; PW1-S29 (5′-TTTTGTCCTCAACCACTCG-3′), specific for SHA29; and PW1-B2 (5′-CTGGTTGGGTTCTTTGTCC-3′), specific for HCB2.

Phylogenetic analysis of ars operon genes.

DNA sequences were aligned by using Clustal W (39), and phylogenetic trees were generated by PAUP* 4.0b4a (38) with an optimality criterion set to minimum evolution. The Kimura two-parameter model was used to estimate pairwise distances. Stepwise addition and TBR branch-swapping algorithms were used to construct phylogenetic trees. Following a heuristic search and bootstrap analysis with 100 samples, the final tree was assembled with TreeView (28). The following sequences (accession numbers in parentheses) were used as references: E. coli R773 (J02591), E. coli R46 (U38947), E. coli chromosomal ars operon (X80057), Serratia marcescens pR478 (AJ288983), Yersinia enterocolitica (U58366), and Klebsiella oxytoca pMH12 (AF168737), as well as the South Haiwee Drain isolates SHA1, SHA17, SHA29, SHA35 (Serratia fonticola), HCSH9 (Yersinia intermedia), and HCB2 (Enterobacter cloacae).

RESULTS

Abundance and phenotypic As^r determination.

The diversity of the E. coli R773 ars model was investigated in As^r bacteria originating from a variety of sources. We first established the abundance of the As^r phenotype in bacteria isolated from industrial or domestic raw sewage, two sites in Hot Creek, and one site in South Haiwee Drain 5. Resistance to arsenic was defined as observable growth in liquid broth at 500 mg of As(III) or 1,000 mg of As(V) per liter (19). Table 3 summarizes the plating results for arsenate- or arsenite-resistant bacteria. The average abundance of E. coli was 4 × 10⁴ CFU/ml on mTEC without arsenic supplementation. As(III) addition to the medium had no effect on decreasing the abundance of CFU. Interestingly, when raw sewage was plated on mTEC supplemented with As(V), a log increase in E. coli abundance was observed compared to that with no arsenic supplementation. The average number of CFU per milliliter for creek samples plated on 50% PCA without arsenic ranged from 3 × 10³ to 9 × 10³. The abundance of CFU growing on nonselective medium (50% PCA) compared to As(V) medium showed little change for the South Haiwee Drain sample. Moreover, As(III) addition to the medium inhibited the growth of bacteria by 1 log for all creek sites. Hot Creek sites 1 and 2 had measurable colony growth when plated on As(III) and As(V); however, there was less than 300 CFU/ml. We conclude that raw sewage provided a good source for the isolation of phenotypically As^r enteric bacteria. In addition, the South Haiwee Drain microbial community was more resistant to As than sites within Hot Creek, as would be expected based on the high total As in the waters of the South Haiwee Drain site.

TABLE 3.

Bacterial counts on nonamended and As-amended media at various sampling locations

Location	Bacterial count (CFU/ml) ona:
	No arsenic	As(III) (500 mg/liter)	As(V) (1,000 mg/liter)
Raw sewageb	4.0 × 104	5.7 × 104	5.4 × 105
South Haiwee Drain 5	3.3 × 103	4.7 × 102	4.2 × 103
Hot Creek Swimming Hole (site 1)	8.7 × 103	<300 (230)	<300 (285)
Hot Creek Bridge (site 2)	6.3 × 103	<300 (147)	<300 (183)

^aNumbers in parentheses are the actual CFU per milliliter and are noted when counts were less than 300 CFU/ml.

^bSamples plated on mTEC—selective for E. coli. Samples from other locations were plated on PCA amended with the corresponding arsenic concentrations.

Isolates were randomly chosen from mTEC, mENDO, and PCA plates and further investigated for E. coli R773 ars-like sequences. A mixed phenotype of As^r isolates was found for the enteric bacteria isolated on mENDO. Table 1 reports the phenotypes of isolates screened by PCR. One enteric bacterium (EN-4), although isolated originally on no arsenic, was able to grow on As(III), but not As(V). Some bacteria isolated on arsenate media showed As(III)^r, while others were As(III)^s. E. coli, South Haiwee Drain, and Hot Creek isolates were only tested for arsenite resistance. These isolates were routinely grown in broth cultures containing 500 mg of sodium arsenite per liter.

Homology of ars genes in environmental isolates to the E. coli R773 genetic model.

The diversity of ars genes was investigated by DNA-DNA hybridization in a total of 84 gram-negative environmental isolates originating from two sites in Hot Creek and one site in South Haiwee Drain. We assessed how well the E. coli R773 ars operon described the genotype of phenotypically As^r isolates. The arsA probe included a region of the gene that encodes one of the three cysteine residues that form the As(III) binding site within the protein. Although there is no known active site for ArsB, we chose the probe based on conserved regions among the known enteric arsB genes found in the GenBank database. Primer construction for arsC was more restrictive due to methodological constraints (high G-C content), and therefore the first 45 bases were excluded in the design; unfortunately, this region contained Cys-14, which is required for protein activity.

In previous hybridization studies in our laboratory with the 370-bp arsC gene probe, high-stringency conditions resulted in less than 5% hybridizations among 50 creek isolates (unpublished data). Modifications to the wash conditions allowed increased detection of divergent arsC-like sequences. The results from this experiment formed the basis for investigation of ars genes under low- and moderate-stringency conditions. The results for the colony hybridization and effects of stringency are summarized in Table 4. For all ars probes, as stringency was relaxed, a 20 to 30% increase in hybridization was observed. There also appeared to be site-to-site variation in the distribution of divergent ars-like sequences. Interestingly, arsA was more prevalent and more conserved in South Haiwee Drain (total As, 740 μg/liter) isolates at moderate stringency. Hot Creek site 1 (HCSH) isolates exhibited a higher prevalence of ars genes at moderate stringency compared to site 2 (HCB) isolates. Site 1 is located in an area in which most of the geothermal activity occurs and approximately 50% of the dissolved As is As(III). At site 2, located approximately 1 km downstream, almost 100% of the As had been oxidized to As(V), the less toxic form. Interestingly, arsA was significantly less diverse in the South Haiwee isolates than those originating from Hot Creek (Chi square; P < 0.05). It is concluded that at a higher As concentration, genetic diversity is lower for genes that affect high-level resistance to As, such as in arsA, the arsenite-translocating ATPase gene.

TABLE 4.

Summary of colony lift and hybridization with E. coli probes arsA-1, arsB-1, and arsC-1

Location (n)	% of hybridization at stringencya
	arsA		arsB		arsC
	Low	Moderate	Low	Moderate	Low	Moderate
South Haiwee (34)	79.4	70.6	94.1	38.2	94.1	64.7
Hot Creek Swim Hole (site 1) (26)	76.9	26.9	75.0	76.9	84.6	69.2
Hot Creek Bridge (site 2) (24)	87.5	16.7	87.5	58.3	75.0	54.2
Total isolates screened (84)	81.0	41.7	86.9	56.0	85.7	63.1

^aLow, proportion of hybridizations at low stringency (31 to 53% estimated mismatch); moderate, proportion of hybridizations at moderate stringency (21 to 36% estimated mismatch).

Use of PCR to detect ars genes associated with enteric bacteria and pseudomonads.

PCR was evaluated for its usefulness to detect ars-like genes in the isolates listed in Table 1. Because low-stringency hybridization was required to detect divergent ars-like sequences, we hypothesized that PCR would only detect closely related sequences with less than 10% sequence divergence at the primer binding sites. Average levels of nucleotide divergence for known enteric arsA, arsB, and arsC genes are 0.22 (standard deviation [SD] = 0.03), 0.21 (SD = 0.06), and 0.19 (SD = 0.054) nucleotide substitutions per site, respectively. At this level of divergence within enteric ars genes, the use of nondegenerate PCR primers would most likely not capture divergent ars genes with greater than 10% sequence differences.

In general, arsA, arsB, and arsC genes were detected in only enteric bacteria and not in pseudomonads (Table 5). Eight isolates were found to contain arsA homologs. Hybridization with the I-arsA-1 internal probe showed that highly homologous DNA was detected in these isolates. Interestingly, arsA in E. cloacae (HCB2) exhibited an amplicon of similar size, yet failed to hybridize to the R773 arsA internal probe at high stringency. Decreasing the stringency of the final wash step resulted in the confirmation of the amplicon (data not shown). However, use of low-stringency colony hybridization also allowed detection of divergent ars-like sequences in many of the pseudomonads.

TABLE 5.

Summary of PCR results grouped as enteric or nonenteric bacteria phenotypically resistant to arsenite or arsenate

Gene	% Positive
	Nonenteric genea (n = 11)	Enteric
		Sewage (n = 32)	Hot Creek (n = 7)
arsA	0	9.4	71.4
arsB	9.1	69	85.7
arsC	0	46.9	85.7

^aIsolates originated from Hot Creek samples.

In contrast to our results for arsA, arsB (Fig. 1) was conserved in 69% of the raw sewage bacteria and 85.7% of the creek enteric bacteria (Table 5). However, the E. coli R773 ars operon was not a good model for describing the As^r genotypes of Hot Creek and South Haiwee Drain pseudomonads. Surprisingly, KOV3 (Xanthomonas oryzae AV) exhibited a faint 219-bp band following Southern blotting and hybridization with the arsB-specific internal probe. Successive PCR attempts failed to generate the 219-bp band.

FIG. 1.

(A) Ethidium bromide-stained agarose gel (top panel) and Southern blot analysis with the internal arsA probe (bottom panel). Electrophoretic analysis of PCR-amplified products was done with arsA-1 primers and genomic DNAs of South Haiwee Drain isolates. Lanes: 1, 123-bp ladder (GIBCO BRL); 2, Serratia fonticola (SHA1); 3, Pseudomonas corrugata (SHA2); 4, Acinetobacter genospecies strain 15 (SHA3); 5, Pseudomonas corrugata (SHA4); 6, Pseudomonas vescicularis (SHA6A); 7, Pseudomonas corrugata (SHA8); 8, Pseudomonas corrugata (SHA10); 9, Serratia fonticola (SHA17); 10, Pseudomonas cichorii (SHA27); 11, Serratia fonticola (SHA29); 12, Serratia fonticola (SHA35); 13, deionized water negative control; 14, E. coli pUM3-positive control. (B) Ethidium bromide-stained agarose gel (top panel) and Southern blot analysis with the internal arsB probe (bottom panel). Electrophoretic analysis of PCR-amplified products with arsB-1 primers and genomic DNAs of Hot Creek and E. coli isolates. Lanes: 1, 100-bp ladder (Boehringer/Mannheim Marker XIV); 2, E. cloacea (HCB2); 3, Klebsiella (HCSH1); 4, Y. intermedia (HCSH9); 5, A. calcoaceticus (KOV2); 6, X. oryzae (KOV3); 7, J. lividium (KOV4); 8, P. corrugata (KOV5); 9, E. coli (1ECO); 10, E. coli (2ECO); 11, E. coli (7ECO); 12, P. putida; 13, deionized water negative control; 14, E. coli pUM3 positive control.

Our investigation into the diversity of arsC resulted in similar findings to those with arsB. The total occurrence of arsC for all of the isolates was 42% (total n = 50). Southern blotting and hybridization with the internal probe I-arsC-1 verified that highly homologous DNA was amplified in the creek isolates (data not shown). All isolates that exhibited PCR positivity were enteric bacteria originating from raw sewage and creek waters. When examining the occurrence of arsC within As(III)^r E. coli, only 50% of the isolates were positive. This finding was much lower than we expected in light of the known sequence data for arsC. For the As^r enteric isolates originating from creek samples, arsC was observed in 86% of the isolates, which was much higher than that observed in the E. coli strains isolated from sewage. However, only seven isolates were investigated for the creek samples (Table 5). In contrast, arsC was not detected in any of the 11 creek pseudomonads, suggesting that divergence is too great to use the E. coli R773 ars operon as a model for arsC. We concluded that the PCR-negative As^r enteric and nonenteric isolates represent divergent groups of ars operons.

A significant difference (chi-square test; P < 0.05) in the distribution of the genotypes was found between raw sewage and creek waters. The arsA gene was generally conserved in enteric creek isolates and not in the raw sewage enteric isolates. We concluded that the low occurrence of arsA in raw sewage isolates resulted from the lack of a selective pressure for maintaining arsA.

Determination of diversity of ars genes among enteric isolates.

The final investigation into the diversity of ars genes was done by phylogenetic inference of arsB sequences obtained from South Haiwee Drain and Hot Creek isolates. They were compared to known arsB genes obtained from the GenBank database. The arsB fragments of the creek Serratia spp. and Yersinia sp. were almost identical: the average sequence similarity for this group was 98%. However, they appeared to be of a divergent lineage compared to the recently published Serratia marcescens (accession no. AJ288983) and Yersinia enterocolitica (accession no. U58366) sequences. Phylogenetic analyses placed the creek Serratia and Yersinia spp. into a distinct cluster (Fig. 2). The arsB fragments of this group appeared to be more closely related to the E. coli R773 arsB; the average nucleotide sequence similarity to this group was 84%. The creek isolate HCB2 (Enterobacter cloacae) shared greater sequence similarity to E. coli R46 and Klebsiella oxytoca arsB genes—89 and 94%, respectively. In addition, HCB2 sequence similarity to the South Haiwee Drain cluster was about 81%. The overall topology of the tree indicated that clustering of arsB was not species dependent, as would be expected for a phylogeny inferred from 16S rRNA genes. However, it should be noted that the bacteria were identified by phenotype.

FIG. 2.

Phylogenetic relationships among the arsB genes in enteric bacteria and Hot Creek and South Haiwee Drain isolates. The phylogenetic tree was constructed according to distance criterion (Kimura two-parameter substitution model with equal distribution of sites). The percentages of 100 bootstrap replicates that supported the branching order are shown above or near the relevant nodes. Clones from this study are indicated in boldface type.

DISCUSSION

In recent years, the study of metal resistance in bacteria has led to the discovery of many metal-specific genetic models, including czc (Cd, Zn, and Co) (22) cop (Cu) (42), mer (Hg) (26), cadA (Cd) (24, 46; G. Nucifora, L. Chu, S. Silver, and T. K. Misra, Abstr. 88th Annu. Meet. Am. Soc. Microbiol. 1988, abstr. H-209, p. 179, 1988), and chr (Cr) (6, 21;C. Cervantes, H. Ohtake, and S. Silver, Abstr. 88th Annu. Meet. Am. Soc. Microbiol. 1988, abstr. H-213, p. 180, 1988). In addition, molecular approaches are more frequently being used to study microbial communities in metal-contaminated environments. PCR and gene probes are often used to characterize a community for the prevalence of a particular genetic model. Advantages to using molecular approaches included enhanced assessment of specific microbial populations in contaminated environments (32) and increased sensitivity in monitoring specific microbial populations during remediation efforts (18).

When investigating the diversity of ars, we used the E. coli R773 ars operon as a model for describing the genotype of phenotypically As^r bacteria. We first established the abundance of As^r bacteria in various water sources, including raw sewage and creek waters containing elevated levels of total arsenic. Raw sewage was used as a source of enteric bacteria, in which we hypothesized that ars genes would be relatively conserved in comparison to isolates from natural waters. Raw sewage exhibited an abundance of As^r enteric bacteria, despite low total As concentrations (<50 ppb; Y.-L. Tsai, personal communication.). We found that approximately 1% of the culturable E. coli species and enteric bacteria can resist the toxicity of As(III). However, all bacteria appeared to be resistant to As(V), because no reduction in the number of CFU compared to background CFU was obtained. Similar results were found for Hot Creek and the South Haiwee Drain. These results are similar to Hysman’s and Frankenberger’s (15) investigation of the abundance of As^r CFU in agricultural drainage waters and evaporation pond sediments. No CFU were observed on medium supplemented with ≥500 mg of As(III) per liter. In addition, Zelibor et al. (47) isolated As(V)^r bacteria in well water samples. These isolates tolerated up to 2,000 μg of As(V) per ml. However, they did not test for As(III) resistance. Interestingly, bacteria indigenous to these monitoring wells did not precipitate or volatilize dissolved As. This suggests that the predominant resistance mechanism either involves a nonspecific physiologically based oxidation or reduction of As or is determined by an As-specific genetic system.

The difference in abundance between As(III)^r and As(V)^r phenotypes may be due to the expression of different mechanisms for the detoxification of arsenicals. Because As(V) is structurally similar to phosphate, in E. coli, As(V) can be taken up through the phosphate transport system (44, 45). In the ars system, intracellular As(V) is reduced to As(III) and pumped out of the cell by the ArsAB membrane-bound pump. Alternatively, physiological As(V)^r arises as a result of down regulatio of the inorganic phosphate transport system (pit), yielding a decrease in As(V) uptake and an insensitivity to As(V) toxicity (45). The apparent ubiquity of As(V)^r enteric bacteria suggests that physiological resistance may play a significant role in this observed phenotype. However, the proportion of physiological As(V)^r to ars-specific resistance cannot be determined from our results.

In a study detailed in reference 3, Barkay et al. investigated the diversity of mercury resistance genes of the mer operon. Low-stringency hybridization with probes for merA permitted the detection of isolates that previously failed to hybridize with a Tn21 mer probe at high stringency (27). Similarly, when we investigated the diversity of ars within selected South Haiwee Drain and Hot Creek isolates, we found that colony hybridization at low and moderate stringency provided a useful tool for assessing diversity. Moreover, the relative frequencies of occurrence for arsA, arsB, and arsC were determined. The occurrence of arsA was much lower than that of arsB and arsC. On the other hand, arsB and arsC occurred in approximately equal frequencies. This suggests that either (i) arsA is more divergent than arsB and arsC, and the detection is therefore limited based on the current lack of arsA sequence data, or (ii) the arsBC genotype predominates in environments with low arsenic concentrations. The latter is supported in enteric bacteria for several reasons. Many known ars operons are composed of arsRBC (12). Pseudomonas aeruginosa PAO1 and E. coli K-12 have chromosomal ars operons composed of arsRBC. Staphylococcus spp. similarly have arsRBC. To date, there are only a few plasmid-associated arsA sequences: E. coli R773 and R46 and Acidiphilium multivorum pNCR-1. Because the arsenite-ATPase ArsA provides high-level resistance to arsenite, as was needed at the South Haiwee and Hot Creek sites, we propose that at low As concentrations, as in sewage, arsA is maintained in a few bacteria within a population, possibly on a plasmid or transposable element. The genetic determinant is then disseminated throughout a population of bacteria when a necessity arises for accelerated detoxification of As (i.e., during As enrichment events). The presence of arsA-like sequences in enteric isolates isolated from As-enriched waters supports this hypothesis. Our investigation of the diversity of the ars operon by PCR suggests that this technique is useful for detecting ars-like sequences in enteric bacteria. The high prevalence of the ars operon genes among the E. coli, mENDO, and creek enteric isolates is comparable with those in other studies that used filter hybridization with ars operon probes. Approximately 50% of As^r E. coli and Klebsiella spp. have been shown to contain homologs of the R773 ars operon, determined by hybridization with a 4.3-kb ars operon probe (20). Similarly, ars sequences have also been detected by Southern blotting and hybridization in the following bacteria: E. coli 40, Shigella sonnei, Citrobacter freundii, Enterobacter cloacae, Erwinia carotovora, Salmonella enterica serovar Arizonae, Klebsiella pneumoniae, and Peudomonas aeruginosa PAO1 (12).

The E. coli model for arsenic resistance appears to be an appropriate genetic system to use when screening for closely related isolates, when sequence identity is ≥80%. The enteric bacterium-related Hot Creek and raw sewage isolates showed strong PCR amplification products for arsB and arsC. Phylogenetic analysis of arsB fragments of South Haiwee Drain and Hot Creek isolates revealed significant divergence from the R773 ars model (Fig. 2). Possible mechanisms for this include horizontal gene transfer via plasmids or transposons. In addition, genetic recombination of ars genes and geographic isolation may have ultimately resulted in the formation of a distinct lineage of ars sequences.

The As^r pseudomonads listed in Table 1 showed little homology to the arsA, arsB, and arsC genes of the E. coli R773 model. In comparison to the P. areuginosa ars operon, the R773 ars operon exhibits significant differences at the nucleotide level. Pseudomonas arsB has 67% genetic similarity to E. coli R773 arsB. In addition, the arsC arsenate reductase gene shares only 29% sequence similarity to the E. coli R773 arsC gene. This high degree of sequence divergence is a limiting factor in the detection by PCR of distantly related sequences. Unrelated isolates did not show visible PCR amplification products. In some cases, smears were seen that indicated poorly matched primers to the target sequence. In addition, X. oryzae (KOV3) showed a weak hybridization signal with an internal probe for arsB, although no distinct DNA band was visualized on the agarose gel. The failure to detect ars genes in creek Pseudomonas spp. can be attributed to several factors, including limitations in the primer design and the high divergence of the target DNA. The use of degenerate primers may have allowed us to detect divergent ars-like genes. The ars primer sets were originally designed to amplify regions containing active sites or to encompass the most homologous sections of DNA indicated in multiple gene alignments with E. coli R773, R46, and chromosomal ars operons.

To detect divergent genes in nonenteric environmental bacteria, future research should focus on the development of hybridization methods based on model organisms that most appropriately reflect the environment to be studied. The abundance of pseudomonads in the aquatic environment makes them the most suitable for this type of methodology. P. aeruginosa and P. fluorescens are commonly found in aquatic environments. These species may provide better models for the development of PCR-based methods for the detection of As^r genes specific for the ars operon. The E. coli model will be most appropriately applied to environments in which fecal contamination is expected.