An Escherichia coli chromosomal ars operon homolog is functional in arsenic detoxification and is conserved in gram-negative bacteria

Arsenic is a known toxic metalloid, whose trivalent and pentavalent ions can inhibit many biochemical processes. Operons which encode arsenic resistance have been found in multicopy plasmids from both gram-positive and gram-negative bacteria. The resistance mechanism is encoded from a single operon which typically consists of an arsenite ion-inducible repressor that regulates expression of an arsenate reductase and inner membrane-associated arsenite export system. Using a lacZ transcriptional gene fusion library, we have identified an Escherichia coli operon whose expression is induced by cellular exposure to sodium arsenite at concentrations as low as 5 micrograms/liter. This chromosomal operon was cloned, sequenced, and found to consist of three cistrons which we named arsR, arsB, and arsC because of their strong homology to plasmid-borne ars operons. Mutants in the chromosomal ars operon were found to be approximately 10- to 100-fold more sensitive to sodium arsenate and arsenite exposure than wild-type E. coli, while wild-type E. coli that contained the operon cloned on a ColE1-based plasmid was found to be at least 2- to 10-fold more resistant to sodium arsenate and arsenite. Moreover, Southern blotting and high-stringency hybridization of this operon with chromosomal DNAs from a number of bacterial species showed homologous sequences among members of the family Enterobacteriaceae, and hybridization was detectable even in Pseudomonas aeruginosa. These results suggest that the chromosomal ars operon may be the evolutionary precursor of the plasmid-borne operon, as a multicopy plasmid location would allow the operon to be amplified and its products to confer increased resistance to this toxic metalloid.

Arsenic is a metalloid found in the environment, and it exists commonly in trivalent and pentavalent ionic forms (24). Its toxic properties are well-known and have been exploited in the production of antimicrobial agents, such as the first specific antibiotic (Salvorsan 606) and the African sleeping sickness drug Melarsen, in addition to the commonly used wood preservative chromated copper arsenate (5,45). Because of increasing environmental concentrations as a result of industrialization, perhaps it is not surprising that plasmid-located genes which confer resistance to arsenic have been isolated from bacteria (21,35). These arsenic resistance determinants (ars), isolated from both gram-positive and gram-negative bacterial species, have been found to be very homologous and generally consist of either three or five genes that have been organized into a single transcriptional unit (38). In the wellstudied ars-containing plasmid R773, isolated from Escherichia coli (7,34), the operon consists of five genes that are controlled from a single promoter located upstream of the first cistron (arsR). These cistrons, arsRDABC (in that order), encode an arsenic-inducible repressor (arsR) (46), a negative regulatory protein that controls the upper level of transcription (arsD) (48), an ATPase plus membrane-located arsenite efflux pump (arsA and arsB, respectively) (20,29,35,42), and an arsenate reductase (arsC) (18). In the well-studied ars-containing plasmids isolated from Staphylococcus species (plasmids pI258 and pSX267), the arsR, arsB, and arsC cistrons are conserved, while the arsD and arsA cistrons are absent (17,30). In this case, the ArsB protein is believed to use the cell's membrane potential to drive the efflux of intracellular arsenite ions (18). The origin of these homologous plasmid-borne arsenic resistance determinants has not yet been defined.
Given the ubiquitous presence of arsenic, we sought to determine if bacteria contain chromosomally located genes whose expression is induced at elevated arsenic ion concentrations and which aid cells in detoxification of episodic increases in extracellular arsenic (39). We have previously reported the presence of aluminum (12)-and nickel (14)-inducible genes in E. coli by screening a library of 3,000 single-copy Vibrio harveyi luciferase gene fusion chromosomal insertion clones (13) for changes in light emission upon addition of these metals. Using a collection of lacZ chromosomal gene fusions prepared with MudI (6), we report here the identification of an arsenicinducible operon in the chromosome of E. coli located at 77.5 min. The cloning and sequencing of this operon revealed that it can encode proteins that are highly homologous to plasmidencoded ars determinants and that its expression is inducible at arsenic ion concentrations just above the environmental background (9,39). We also show that this operon is present in the chromosomes of a wide variety of gram-negative bacterial species and that it is a functionally important determinant in detoxification of arsenic ions in E. coli.

MATERIALS AND METHODS
Bacterial strains and phages. The following bacterial strains used were all derivatives of E. coli K-12: E. coli 40 (⌬pro-lac rpsL trp), E. coli BU5029 (a recA mutant derivative of strain 40), and those described (including sources) by Autexier and DuBow (1). Phages MudI and P1vir were kind gifts of M. Casadaban (University of Chicago) and R. Stewart (McGill University), respectively. DNA manipulations. All restriction endonuclease hydrolyses and DNA ligations were performed as described by Tolias and DuBow (43). DNA sequencing of both strands (see Fig. 2A) was performed by the dideoxy DNA sequencing method with single-stranded DNAs from cloned fragments in plasmids pUC118 and pUC119 (44) by using the Sequenase version 2.0 kit from United States Biochemicals. Southern blotting and hybridization, as well as isolation of total cellular DNA, were performed according to the method of Autexier and DuBow (1), while P1 transduction was done according to the method of Miller (25). DNA was isolated from stationary-phase cells grown in Luria-Bertani (LB) (25) broth (E. coli 40 and Pseudomonas aeruginosa PA01) or nutrient broth (Difco Laboratories) (Shigella sonnei, Citrobacter freundii, Enterobacter cloacae, Salmonella arizonae, Erwinia carotovora, and Klebsiella pneumoniae) at 37ЊC, except for E. coli 40, P. aeruginosa PA01, and S. arizonae, which were grown at 32ЊC. For Southern blotting, 10 g of total cellular DNA was digested with the appropriate restriction enzyme and blotted to nylon (Hybond-N; Amersham) membranes following electrophoresis through 0.75% agarose gels (31). Membranes were probed with 2 ϫ 10 8 to 4 ϫ 10 8 cpm of an ␣-32 P-labelled 1,188-bp EcoRV (bp 1664 to 2852 [see Fig. 2A]) or a 587-bp NdeI-EcoRV (bp 1077 to 1664 [see Fig.  2A]) DNA fragment (prepared by the random priming method [31]) per ml under high-stringency conditions (1). After being washed, membranes were exposed to Agfa Curix RPI X-ray film.
Construction of strains LF20001 and LF20018. lacZ fusions to chromosomal genes were constructed by infecting E. coli 40 with MudI (amp lac) bacteriophages as described by Casadaban and Cohen (6). The resultant clones were picked to a master plate and replicated on LB agar plates that contained ampicillin and 5-bromo-4-chloro-3-indolyl-␤-D-galactopyranoside (X-Gal; Research Organics, Inc.) plus increasing concentrations of sodium arsenate (0.1 to 10 g/ml). One clone, which became blue when it was grown in the presence of sodium arsenate and remained white in its absence, was named LF20001 and isolated for further study. E. coli LF20018 was constructed by P1vir transduction (25) of E. coli LF20001 into E. coli 40 and selection on LB plates that contained ampicillin. The resultant Ap r clone (E. coli LF20018) was tested for arsenic induction of ␤-galactosidase, and the location of the MudI prophage was determined by Southern blotting and hybridization, with the lacZ gene as the probe.
Isolation of the arsenic-inducible operon. To isolate the proximal portion of the arsenate-inducible operon, a lac operon-Mu attR-E. coli DNA fragment was cloned from strain LF20001 via isolation of total cellular DNA (13), cleavage with BglII, ligation into BamHI-cleaved pBR322 DNA, and transformation into E. coli BU5029. One colony, which developed a blue color on LB agar with ampicillin and X-Gal (because of amplification of the lac operon), was selected, and its plasmid was designated pJS29 (Fig. 1). The cloned chromosomal DNA adjacent to the right end of the MudI insertion was isolated and used as a probe to identify a 15-kb PstI fragment from the chromosome of E. coli 40. The 15-kb fragment was cloned into pBR322 (digested with PstI) to yield plasmid pJC076 ( Fig. 1) by standard procedures (31). A 3-kb NsiI-BglI fragment that encom-passed the site of MudI insertion in strain LF20001 was subcloned (plasmid pJC103) and completely sequenced.
␤-Galactosidase assays. ␤-Galactosidase assays were performed as described by Miller (25) by the chloroform-sodium dodecyl sulfate cell lysis procedure. Cells were grown to an A 550 of 0.4 in LB broth at 32ЊC and exposed to various arsenic and antimony compounds, and aliquots were removed for ␤-galactosidase assays after 30 min.
Arsenic sensitivity tests. The sensitivities of E. coli strains to trivalent and pentavalent arsenic ions were determined by preparing petri plates that contained LB agar and various concentrations of sodium arsenate and sodium arsenite. Overnight cultures of E. coli strains grown in LB broth were diluted in fresh LB broth and grown at 32ЊC to an A 550 of approximately 0.4. Then, cells were diluted 10 5 -fold in LB broth, and 0.1 ml of these dilutions was spread (in triplicate) on different agar plates. Petri dishes were incubated at 32ЊC overnight, and then colonies were counted.
Nucleotide sequence accession number. The nucleotide sequence reported here has been submitted to the EMBL database, and it has been assigned accession number X80057.

Discovery of a chromosomal ars operon homolog.
A collection of lacZ transcriptional gene fusions was prepared by using E. coli 40 and the MudI bacteriophage (6). In order to identify any gene whose transcription is specifically affected by arsenic salts, this collection of clones was replicated on petri dishes in the absence and presence of various concentrations of sodium arsenate and the ␤-galactosidase indicator substrate X-Gal. A single clone which formed blue colonies on petri plates that contained sodium arsenate and white colonies in its absence was identified. This clone, designated strain LF20001, was isolated for further study. A P1vir transductant of the MudI prophage region was also prepared in E. coli 40 and designated E. coli LF20018. The DNA adjacent to the MudI prophage was mapped by Southern blotting (with the lac operon as the probe), cloned, and used as a probe to map and isolate the DNA sequences that flank the MudI insertion site from the chromosome of E. coli 40 (Fig. 1). By hybridization of the  (22), it was determined that the arsenate-inducible gene was located at 77.5 min on the E. coli genetic map. A total of 2.973 kb of DNA was sequenced (EMBL accession number X80057) ( Fig. 2A) from plasmid pJC103 and used to scan databases with the University of Wisconsin Genetics Computer Group sequence analysis software package. It was found that this region of the chromosome is highly homologous to the arsenic-inducible ars operons of plasmid isolates from E. coli (7,32,48), Staphylococcus aureus (17), and Staphylococcus xylosus (30) (Fig. 2B); thus, because of its homology and arsenate inducibility, it was designated ars. This chromosomal ars operon was found to consist of three cistrons, which we have named arsR, arsB, and arsC because of strong homology to the plasmid-borne ars operons. The arsR cistron is 77.0% homologous (at the protein level) to the same cistron in the ars operon isolated from plasmid R773 of E. coli, while the arsB and arsC cistrons are 90.7 and 94.3% homologous, respectively, with this operon. Weaker, though still significantly homologous, are the plasmid-encoded ArsR, ArsB, and ArsC proteins of both S. aureus and S. xylosus (Fig. 2B). The location of the MudI prophage in E. coli LF20001 was found to be in the arsB gene, 799 bp downstream from the ATG start codon. Transcription of the inserted promoterless lac operon would occur from an upstream promoter ( Fig. 2A), presumably located in a position similar to that of the ars operon in plasmid R773 (47). Effects of various arsenic and antimony compounds on ars gene expression. The plasmid-borne ars operons are inducible by various toxic arsenic and antimony compounds (38). In order to measure induction of expression of the chromosomal ars operon by these compounds, E. coli LF20001 was grown in LB broth and exposed to various compounds, and ␤-galactosidase activity was quantified. Maximal levels of ␤-galactosidase activity were reached at approximately 60 min after exposure of strain LF20001 to sodium arsenate at final concentrations that ranged from 1 to 10 g/ml (data not shown). When strain LF20001 was exposed to increasing concentrations of sodium arsenite, induction of gene expression was detectable at 30 min postexposure and 5 g/liter, with maximal induction observed at 1 g/ml (Fig. 3). Sodium arsenate, the pentavalent (and less toxic) form of arsenic, did not induce ars operon expression at 30 min postexposure and 5 g/liter. However, higher concentrations (100 and 1,000 g/liter) were able to induce expression of the arsB::lacZ fusion. Antimony (as antimony oxide), located just below arsenic in the periodic table, was also found to induce ars operon expression, as it does for plasmid-borne ars operons (34). However, cacodylic acid, a relatively nontoxic pesticide which contains arsenic in an organic formulation (3), was unable to induce expression of the arsB::lacZ fusion, even when added at arsenic concentrations as high as 1 g/ml (Fig. 3). The observed threshold and concentration-dependent induction of the chromosomal arsB:: lacZ fusion are very similar to those observed for plasmidborne ars operons (8). No induction of the arsB::lacZ fusion was observed with other (Pb, Zn, and Cu) metal ions at concentrations that ranged from 0.01 to 10 g/ml, nor was ␤-galactosidase activity from wild-type E. coli induced or affected by arsenic and antimony compounds at the concentrations used here (data not shown).
A functional role for the ars operon in protection from arsenic toxicity. In order to determine if the chromosomal ars operon plays a functional role in arsenic detoxification, clones with (E. coli 40) and without (E. coli LF20001 and LF20018) a functional ars operon were tested in identical genetic backgrounds for their growth in arsenic-containing LB media. Increasing concentrations of sodium arsenite or sodium arsenate were added to LB agar on petri dishes, and the colony-forming capacities of various strains were examined after overnight growth. The 50% lethal concentrations of sodium arsenate and sodium arsenite for E. coli 40 were found to be between 200 and 2,000 g/ml (Fig. 4), and similar results were obtained for E. coli MG1655 (data not shown). Disruption of the chromosomal ars operon by MudI insertion was found to increase the sensitivity of E. coli 40 to sodium arsenite (Fig. 4A) and sodium arsenate (Fig. 4B) by approximately 10-to 100-fold (Fig. 4A). However, when the complete ars operon, cloned on a multicopy plasmid (pJC103), was introduced into wild-type E. coli 40, resistance to sodium arsenite (Fig. 4A) and sodium arsenate (Fig. 4B) increased by at least 3-to 10-fold, though the absolute levels of resistance were somewhat lower than those observed for E. coli that contained the arsRDABC operon of plasmid R773 (34). The ability of E. coli arsB mutant strains (LF20001 and LF20018) to survive at arsenic levels which induce ars operon expression (0.01 to 1 g/ml) may be due to other cellular detoxification mechanisms, such as those provided by glutathione and thioredoxin (11,15,16). Sequences homologous to the chromosomal ars operon are highly conserved. Because of the high degree of homology between the protein products of the E. coli chromosomal ars operon and those found on plasmids from both gramnegative and gram-positive bacteria, we sought to determine if the chromosomal operon was conserved at the DNA level (and thus, possibly the progenitor of plasmid-based arsenic resistance determinants). DNA was isolated from a number of plasmid-free, gram-negative bacterial species (1), hydrolyzed with restriction enzymes, and Southern blotted after agarose gel electrophoresis. After hybridization with an E. coli ars-specific probe, sequences that were homologous to the E. coli chromosomal ars operon were found in all of the enterobacterial species examined. Moreover, homologous sequences to the ars operon were detected in the nonenterobacterial species P. aeruginosa (Fig. 5). This high degree of evolutionary conservation at the DNA level strongly reinforces notions that the chromosomal ars operon is functionally important and that its chromosomal presence is not of recent origin.

DISCUSSION
We have discovered a functional, arsenic-inducible operon in the chromosome of E. coli, with homologous sequences detectable in many other gram-negative bacterial species. This operon displays strong homology, both in protein sequence and genetic organization, with plasmid-borne arsenic detoxification operons. During the later stages of this work, continued sequencing of the E. coli genome also uncovered this chromosomal ars operon homolog, though no functional studies were performed (40). The names arsE, arsF, and arsG were given to these three homologous cistrons, but arsR, arsB, and arsC more accurately reflect their evolutionary relatedness and probable function(s). Thus, it is likely that the chromosomal ars operon is organized as a single transcription unit that is regulated by the arsenic-and antimony-inducible ArsR repressor. Moreover, the structural genes of the chromosomal ars operon appear to encode an arsenate reductase (arsC) and an arsenitespecific efflux system (arsB). The apparent strong evolutionary conservation of chromosomal ars determinants also suggests that this operon may be the progenitor of plasmid-borne ars operons. The origins of many plasmid-borne resistance determinants have not yet been elucidated. However, it is known that ␤-lactamases are also highly conserved, whether their lo- cations are chromosomal or on plasmids (26,33). In addition, hemolysins show similarly strong evolutionary conservation (2,10). More recently, however, a chromosomal homolog of a plasmid-borne copper resistance operon has been found in Pseudomonas syringae (23). To our knowledge, its evolutionary conservation has not been determined. It has been proposed that the structure of the plasmidborne, ATP-driven arsenic efflux pump, made up of the ArsA and ArsB proteins, may be structurally related to the multiple drug resistance ATP-driven efflux pump found amplified in mammalian cancer cells (36,37,49). During chemotherapy of cancer patients, cells become resistant to anti-cancer chemotherapeutic agents by amplification of the number of copies of the multiple drug resistance gene and thus overexpression of the multiple drug resistance pump (28). In analogous fashion, amplification of the chromosomal ars operon, by its presence on multicopy plasmids, should allow increased resistance to cellular exposure to toxic arsenic salts. In this regard, we found that the presence of this operon in pBR322 (pJC103), under its own regulation, conferred at least a 3-to 10-fold increase in the resistance of E. coli to arsenate or arsenite exposure.
A lacZ fusion to the chromosomal ars operon was found to be induced by arsenic compounds at concentrations that reflected their relative toxicities (arsenate Ͻ arsenite). Moreover, antimony oxide also induced expression of the arsB::lacZ fusion. These results are consistent with those observed for the R773 ars operon of E. coli (47). We also observed a threshold of ars operon expression; induction was not observable at a concentration of less than 1 g of arsenite per liter. A similar threshold effect on induction has been shown for the mer operon by its inducer, mercury (27). In addition, cacodylate did not induce the arsB::lacZ fusion, even at elevated concentrations. Thus, the induction of ars gene expression appears to reflect the relative toxicities of arsenic and antimony compounds. The LF20001 (lacZYA inserted in arsB) clone may therefore prove useful in determining the potential cytotoxicities of arsenic compounds (for enteric bacteria), as the assay for ␤-galactosidase is both rapid and quantitative (25).
In contrast to the E. coli plasmid-borne ars operon, the chromosomal ars operon contains neither the arsD nor arsA cistron. Low-stringency Southern blotting (41) and hybridization with cloned arsR, arsB, and arsC cistrons from plasmid R773 enabled these cistrons to be detected in the chromosome of E. coli, and the pattern of fragments was consistent with restriction enzyme mapping of the chromosomal ars region (data not shown). However, no signal was detected with the arsD and arsA cistrons as probes, even after long exposure times. These results imply that the arsD and arsA cistrons are not present at another location on the E. coli chromosome. Whether (and from where) plasmid-borne ars operons have gained the arsD and arsA cistrons or the E. coli ars operon has lost them is not clear at present. In this regard, it is interesting that the chromosomal ars operon organization more closely resembles that of gram-positive bacteria in lacking arsD and arsA, even though it is more homologous to that of gramnegative bacterial species.
The lack of an arsA-encoded ATPase subunit in the chromosomal ars operon is striking. Nonetheless, the ArsB subunit may function as an arsenite-specific efflux system which uses membrane potential instead of ATP hydrolysis by the arsA cistron. This mechanism has been observed for the plasmidborne ArsB protein from Staphylococcus spp. (4) and in tetracycline efflux for the Tn10-derived tetracycline resistance determinant (19). Determinations of the structure(s) and function(s) of other chromosomal ars operons and the origin(s) of the arsD and arsA genes are currently in progress.