Resistance Determinants of a Highly Arsenic-Resistant Strain of Leptospirillum ferriphilum Isolated from a Commercial Biooxidation Tank

ABSTRACT Two sets of arsenic resistance genes were isolated from the highly arsenic-resistant Leptospirillum ferriphilum Fairview strain. One set is located on a transposon, TnLfArs, and is related to the previously identified TnAtcArs from Acidithiobacillus caldus isolated from the same arsenopyrite biooxidation tank as L. ferriphilum. TnLfArs conferred resistance to arsenite and arsenate and was transpositionally active in Escherichia coli. TnLfArs and TnAtcArs were sufficiently different for them not to have been transferred from one type of bacterium to the other in the biooxidation tank. The second set of arsenic resistance genes conferred very low levels of resistance in E. coli and appeared to be poorly expressed in both L. ferriphilum and E. coli.

Processes for the biooxidation of gold-bearing arsenopyrite concentrates were developed in the 1980s and are now used in several countries (17). These are among the largest commercial fermentation processes known. Biooxidation is a mineral pretreatment process during which the molecular structure of the arsenopyrite mineral is broken down, exposing the gold and allowing its extraction by cyanide (7). During this process, large quantities of arsenic are released into continuous-flow aeration tanks in which the biooxidation takes place. The majority of arsenopyrite biooxidation processes operate at 40°C and are dominated by a mixture of the sulfur-oxidizing bacterium Acidithiobacillus caldus and the iron-oxidizing bacterium Leptospirillum ferriphilum. During the first few years of operation, the continuous-flow nature of the processes resulted in the selection of highly arsenic-resistant bacteria.
Studies to investigate what genetic changes had taken place that accompanied this high level of arsenic resistance have been carried out with A. caldus. Highly arsenic-resistant strains of A. caldus have been found to contain an unusual Tn21-like ars operon that is not present in less-resistant strains (6,23,24). The 12-kb TnAtcArs is unusual in that the tnpA (transposase) and tnpR (resolvase) genes occur on opposite ends of the transposon, unlike other transposons of the Tn21and Tn3 family, where they form an adjacent unit. These two transposon genes flank a series of genes (arsRCDADA orf7 orf8B) associated with arsenic resistance that are themselves unusual. Genes for an ArsR (negative regulator) (20,27) and ArsC (arsenate reductase) (10,11) are followed by a tandem duplication of the genes for ArsD (a second repressor) (28) and ArsA (an ATPase that associates with ArsB and links arsenite export to ATP hydrolysis) (22). These genes are followed by genes encoding ORF7 (an NADH-like oxidoreductase), ORF8 (a cystathione-␤-synthase [CBS] domain-containing protein), and ArsB, the arsenite efflux pump. Deletion of one copy of arsDA readily occurs, but this deletion does not appear to affect resistance when cloned on a multicopy plasmid in Escherichia coli (24). Similarly, the inactivation or deletion of ORF7 and ORF8 did not affect arsenic resistance in E. coli. Some strains of highly arsenic-resistant A. caldus have a second transposon that is identical to TnAtcArs except that it extends from the inverted repeat adjacent to tnpR to orf7, at which point it is truncated. This truncated transposon lacks an arsB but nevertheless confers low-level arsenic resistance when cloned on its own in E. coli (23).
As the dominant iron-oxidizing bacterium (5), it was anticipated that L. ferriphilum strains from commercial arsenopyrite biooxidation tanks would also possess effective arsenic resistance mechanisms. Whereas A. caldus is a ␥-proteobacterium (9), L. ferriphilum is a member of the nitrospira (12), and we wished to determine whether these two phylogenetically distant but highly arsenic-resistant bacteria isolated from the same biooxidation process would have similar mechanisms for arsenic resistance. More specifically, we wished to determine whether the TnAtcArs transposon found in A. caldus (24) was also present in highly arsenic-resistant strains of L. ferriphilum. We therefore examined the arsenic resistance mechanisms of a strain of L. ferriphilum isolated from a commercial biooxidation tank treating a gold-bearing arsenopyrite concentrate at the Fairview mine (South Africa) and compared these mechanisms with those of other L. ferriphilum isolates.
Cloning of the ars genes of L. ferriphilum. L. ferriphilum strain Fairview isolated from the biooxidation tanks at the Fairview mine, South Africa (Table 1), was cultured at 30°C in basal medium with FeSO 4 (18), and genomic DNA was isolated as described previously (24). Partially digested 5-to 10-kb Sau3A DNA fragments were used to construct a gene bank in the BglII site of the positive selection vector pEcoR252 (Table 1). E. coli DH5␣ (Table 1) cells were transformed with the ligation mix and plated onto Luria agar plus ampicillin (100 g/ml). Colonies were scraped off and used to prepare the gene bank. The bank was transformed into the E. coli arsenic mutant ACSH50I q , and transformants able to grow on LB plates containing 0.5 mM arsenite were selected. Twelve arsenic-resistant transformants were isolated, their plasmids were mapped, and all were found to have regions in common. One of these, pLfTnArs, was selected for further analysis. The source of insert DNA was confirmed by Southern hybridization using three different fragments as a probe. Total DNA of L. ferrooxidans Fairview and pLfTnArs digested with various restriction enzyme combinations generated fragments that were the same size as the cloned fragments, indicating that the insert DNA originated from L. ferriphilum Fairview (data not shown). Furthermore, the signals obtained suggested that there were two copies of the transposon present on 10-kb and 12-kb BamHI fragments. L. ferriphilum Fairview genomic DNA was digested with BamHI, and fragments in the 8-to 14-kb size range were used to generate a mini gene bank in the same way as described above. This gene bank was screened in E. coli ACSH50I q for arsenite resistance to isolate the 10-kb and 12-kb transposon-containing fragments. Restriction mapping confirmed that the two transposons were identical between the invert repeats but differed in the flanking regions, indicating that they occurred in two different locations on the L. ferriphilum genome. During a metagenome sequencing project for an environmental acid mine drainage microbial community from Iron Mountain (Richmond, Calif.), the near-complete genome sequence of bacteria identified as Leptospirillum group II (i.e., L. ferriphilum) was assembled (25). Analysis of this gapped sequence identified putative arsenic resistance genes, which appeared to be different from those described above. Primers IMArsF and IMArsR (Table 1)  coli ACSH50I q were performed using this fragment as a probe, and a plasmid named pLfArs was isolated.
Sequence analysis of the L. ferriphilum Fairview ars genes. The inserts of plasmids pLfArs and pLfTnArs were sequenced in both directions, and the open reading frames (ORFs) identified are shown in Fig. 1. The arsenic genes present on pLfTnArs were located between divergently transcribed tnpR and tnpA genes in a manner similar to that of the recently reported TnAtcArs transposon from A. caldus (24). Therefore, the name TnLfArs was given to the transposon from L. ferriphilum. The ars genes consist of arsR, arsC, arsD, arsA, and arsB all transcribed in the same direction. Between arsA and arsB is an ORF (157 amino acids [aa]) that encodes a protein with a CBS domain. CBS domains are small domains of unknown function that usually occur in two to four copies per protein and dimerize to form a stable globular structure (2). A similar protein of 158 aa occurs in TnAtcArs that was inactivated without affecting arsenic resistance in E. coli (24). The ars genes are flanked by divergent tnpR (resolvase) and tnpA (transposase) genes (Fig. 1). The amino acid sequences of TnpR, TnpA, and all of the ars gene products were highly conserved, with ArsA being the least conserved (70.5% sequence identity) and TnpR being the most conserved (95.6% sequence identity) ( Table 2). Despite the overall similarities, there were nevertheless some important differences between TnLfArs and TnAtcArs. TnLfArs is only 8,772 bp in length, compared with 12,444 bp for TnAtcArs. The main reason for the size difference is that ORF7 and the arsDA duplication of TnAtcArs were absent from TnLfArs. Furthermore, the amino acid sequences of the two CBS domaincontaining proteins were not as well conserved as the other gene products (52.8% sequence identity) ( Table 2). The TnLfArs 43-bp left and right invert repeats differed from one another by 1 bp and were 3 bp longer than those for TnAtcArs (not shown).
A BLAST search using the L. ferriphilum ars genes indicated that they were most similar to transposon-related ars genes recently identified in Methylobacillus flagellatus (GenBank accession number NZ_AADX01000013) and Alcaligenes faecalis (accession number AY297781), although studies of these ars transposons have not been reported. A comparison of the operon structures of the four ars transposons identified thus far is shown in Fig. 1. In general, the operon structures of TnLfArs, TnAtcArs, and the Alcaligenes faecalis transposons are most closely related, with TnAtcArs having an additional NADH oxidoreductase-like ORF7 and the Alcaligenes faecalis transposon appearing to have undergone a deletion of tnpR and an inverted repeat. The structure of the M. flagellatus transposon is different in that the tnpA and tnpR genes are adjacent to each other as for most other Tn21 subfamily (and Tn3 family) transposons where they form a functional unit (13). In addition, the M. flagellatus arsR and arsC genes appear to have been fused. Interestingly, the last 39 amino acids of TnpA are repeated downstream of M. flagellatus arsB, reading in the same orientation as arsB (Fig. 1). This remnant could be the link between the arrangement of the tnpR and tnpA genes in the TnAtcArs, TnLfArs, and Alcaligenes faecalis operons compared to the gene layout in the M. flagellatus operon.
A comparison of the sequence identities of all of the ars transposon-associated gene products with those of TnLfArs is made in Table 2. Unexpectedly, although TnLfArs and TnAtcArs origi-nated from bacteria from the same environment, the amino acid sequences of their gene products were not necessarily the most closely related. TnpR of TnLfArs was most related to that of TnAtcArs, while ArsR and ArsC were equally related to both TnAtcArs and the transposon of Alcaligenes faecalis. However, the predicted ArsD, ArsA, ArsB, and TnpA proteins of TnLfArs were much more related to those of the Alcaligenes faecalis and M. flagellatus transposons than to those of TnAtcArs.
At first glance, the finding of two similar ars transposons in the two types of acidophilic bacteria is not surprising, as transposons, and particularly those of the Tn21 family, are very effective at distributing the genes they carry (13). This finding is consistent with the prediction, reported previously by Woese (26), that "cosmopolitan genes," of which arsenic resistance genes are a typical example, are more likely to be a characteristic of particular environments than of particular organismal lineages. If cosmopolitan genes are not restricted to certain lineages, they might be expected to pass between different types of bacteria within an environment that selects for them. It is highly unlikely that the ars transposons have been passed from A. caldus to L. ferriphilum or vice versa in the biooxidation tank environment. The differences in DNA sequences between the genes and the amino acid sequences of their products are far too substantial for them to have arisen in the 10 to 15 years or less since the biooxidation plants began operation and the time of sample collection. This suggests that although the two types of bacteria have arrived at similar solutions to their need for arsenic resistance, they have acquired the ars transposons independently of each other.
Restriction mapping (Fig. 1) and partial sequence analysis of the chromosomal ars genes present on pLfArs revealed three ORFs with sequence identity to arsR, arsC, and arsB genes as found in the L. ferriphilum acid mine drainage isolate whose genome was partially sequenced (25). Surprisingly, in spite of the acid mine drainage sample having originated from California and the Fairview isolate having originated from South Africa, the nucleotide sequences of these ars genes were identical. The arsRC genes are atypical in that they form one continuous open reading frame and seem to be translated as a fusion protein.
The predicted amino acid sequences of these gene products and the equivalent proteins of TnLfArs were not highly conserved (Table 2).
Resistance conferred by the L. ferriphilum Fairview ars genes in E. coli. The ability of the constructs pLfTnArs and pLfArs to confer resistance to arsenite and arsenate was tested in E. coli ACSH50I q . Cultures grown overnight were diluted 100-fold into fresh medium (Luria broth for arsenite assays and lowphosphate medium [15] supplemented with 2 mM K 2 HPO 4 for arsenate assays) containing various concentrations of sodium arsenite/arsenate and incubated at 37°C for 5 h, and the absorbance at 600 nm was determined. Cells containing the transposon-located ars genes (pLfTnArs) were more resistant to both arsenite and arsenate but were considerably more resistant to arsenite than the control cells (pEcoBlunt) (Fig. 2). The chromosomal ars genes (pLfArs) conferred only slightly more arsenic resistance than the control. The very weak resistance conferred by pLfArs could be due to weak promoter expression in E. coli. Gene expression analyses were therefore performed to investigate this further.

Regulation of expression of the L. ferriphilum Fairview ars genes.
To determine whether the ars genes on pLfArs were expressed in E. coli, mRNA was analyzed using reverse transcription (RT)-PCR. Total RNA was isolated (24) from 50 ml of mid-exponential-phase cultures of E. coli ACSH50I q carrying pLfArs grown in LB containing 25 M arsenate, 25 M arsenite, or no arsenic and with antibiotic selection. For isolation from L. ferriphilum Fairview, cells were first grown without arsenic and then diluted 100-fold into fresh medium containing 0.1 mM arsenite/arsenate or no arsenic. RNA was isolated using the same method as that used for E. coli. A first-strand cDNA synthesis kit (AMV; Roche) was used for cDNA synthesis from mRNA. A PCR containing 2 l of the 20-l (total volume) reverse transcriptase reaction mixture was performed. The LfArsBrtR primer was used for cDNA synthesis from arsB mRNA (Table 1) and was used in combination with the LfArsBrtF primer for the PCR. A 330-bp arsB-arsRC product was obtained, indicating that arsRCB expression was obtained in E. coli and that arsB was cotranscribed with arsRC ( Fig. 3). However, as this determination was not quantitative, the level of expression was unknown.
Further analyses were performed on the same mRNA described above using slot blot hybridization. Total mRNA was transferred onto a Hybond-N ϩ nylon membrane (Amersham) using a slot blot apparatus (Hoeffer Scientific) and hybridized according to the manufacturer's instructions by using digoxigenin-labeled DNA probes. Probes specific for the chromosomal arsB transcript (0.6-kb PvuI-NcoI) and the 16S rRNA gene (as an internal control) were used. Signal intensities were measured with a Macbeth TD 109 transmission densitometer. In both E. coli and L. ferriphilum Fairview, weak induction (less than twofold) of arsB expression was detected from arsenateand arsenite-induced samples, and this varied substantially between experiments (not shown). This confirmed that the ars genes were expressed in E. coli, presumably from a promoter upstream of arsRC.
To further test this, a translational lacZ fusion of the putative promoter region for arsRCB in pLfArs was made, and ␤-galactosidase expression was determined in E. coli ACSH50I q . A region of the ArsR protein that included 344 bp upstream of the ATG start codon was amplified by PCR using the LfArsRLacZF/ LfArsRLacZR primer pair (Table 1). The PCR products were digested with BamHI-EcoRI and ligated into the promoterless lacZ reporter gene of pMC1403, resulting in the construct pLfArsRLacZ. Cultures of E. coli ACSH50I q cells harboring the various constructs grown overnight were diluted 1:200 into fresh medium containing the appropriate antibiotics, sodium arsenate or sodium arsenite (25 M) when indicated, and 0.4 mM IPTG (isopropyl-␤-D-thiogalactopyranoside) and incubated at 30°C to an optical density at 600 nm of 0.5. The ␤-galactosidase activities were measured using a method described previously by Miller (14). Constructs in which ArsRC (ptacLfArsRC) ( Table 1) was expressed from a tac promoter as well as ArsR (ptacLfArsR) (with an artificial translational stop codon inserted between ArsR and ArsC) ( Table 1) were added in trans. Similar ␤-galactosidase activities of approximately 120 units were obtained for the arsR-lacZ construct in the absence or presence of 25 M arsenate or arsenite, ArsR, and ArsRC (not shown). Taken together with the arsenic growth studies using E. coli, it was concluded that the L. ferriphilum ars genes present on pLfArs were poorly expressed and regulated in both E. coli and L. ferriphilum. One could speculate that this provided the selection pressure for L. fer- riphilum Fairview to acquire TnLfArs to enhance its ability to survive in an arsenic-rich environment, like the biooxidation tanks from which it was isolated. Expression and regulation analyses of the transposon ars genes were not performed, as a similar study was conducted for the closely related TnAtcArs genes identified in A. caldus in which 25 M arsenate or arsenite resulted in a fourfold induction (24), and both of these transposons conferred similar resistance to arsenic in E. coli (not shown).
Screening of Leptospirillum isolates for ars genes. Genomic DNA from three L. ferrooxidans strains (ATCC 49879, DSM2705, and Chil-Lf2) and three L. ferriphilum strains (ATCC 49881, Fairview, and Warwick) was isolated as described previously (24) and was analyzed for the presence of the transposon and chromosomal ars genes by using Southern hybridization. When probing was performed with a fragment specific for TnLfArs, only L. ferriphilum Fairview gave a signal (data not shown). This suggested that this organism likely acquired the ars transposon in the biomining environment via horizontal gene transfer. When hybridization was performed with a 1-kb PvuI-StuI fragment from pLfArs containing the arsRCB genes, all the L. ferriphilum strains gave a hybridization signal, but none of the L. ferrooxidans strains gave a hybridization signal, even under lower-stringency conditions (data not shown). The L. ferriphilum ATCC 49881 and Fairview chromosomal ars genes were located on the same 4-kb EcoRV fragment, indicating that their ars genes and flanking regions are likely identical. A hybridization signal was only obtained from L. ferrooxidans ATCC 49879 when it was probed with an A. caldus chromosomal arsB-specific fragment and under lowstringency conditions (data not shown), suggesting that the isolates of L. ferrooxidans must have a set of ars genes that are different from those of the L. ferriphilum group.
As only the L. ferriphilum strain originally isolated from the Fairview mine had the TnLfArs, we tested whether this strain was more resistant to growth in arsenic than the L. ferriphilum type strain ATCC 49881 and two L. ferrooxidans strains, DSM2705 and ATCC 49879. L. ferriphilum Fairview was capable of growth and iron oxidation in up to 60 mM As(V) as well as As(III) (data not shown). At higher concentrations, it was difficult to keep the arsenic in the solution, and the resistance of this strain may therefore be higher. L. ferriphilum ATCC 49881 and L. ferrooxidans DSM2705 grew up to approximately 40 mM As(V) and As(III), while L. ferrooxidans ATCC 49879 grew up to 30 mM As(III) and 20 mM As(V). Therefore, L. ferriphilum Fairview had maintained its ability to grow in higher concentrations of arsenic than other strains of Leptospirillum even though it had not been exposed to arsenic during 8 to 10 years of subculture in the laboratory.
TnLfArs is capable of transposition in E. coli. To test whether TnLfArs was transpositionally active in E. coli, a similar strategy used previously was used here (24). Briefly, cells containing a nonmobilizable plasmid, pLfTnpUC1 (containing the L. ferriphilum ars transposon genes) (Table 1), and the conjugative plasmid pSa were mated with an E. coli ACSH50I q -Rif recipient and plated onto selective medium. Plasmid DNA was isolated from two As r Km r Ap s transconjugants (pLfTn1 and pLfTn2), digested with BamHI-HindIII restriction enzymes, and analyzed by a trans-alternating field electrophoresis gel (Fig. 4A) and Southern hybridization (Fig. 4B) using an arsDcontaining fragment as a probe. Hybridization signals were obtained for both transconjugant plasmids but not for pSa alone, suggesting that the transposon had jumped to pSa and was conjugated into the recipient cell. Furthermore, the size of the largest pSa fragment was increased (from 23 kb to approximately 36 kb), confirming the insertion of the transposon (Fig.  4A, compare lanes 2 and 3).
As both the L. ferriphilum and A. caldus ars transposons have been found to be capable of transposition in E. coli, it is likely that TnAtcArs is functional in A. caldus and that TnLfArs is functional in L. ferriphilum. The observation that one type of bacterium is unlikely to have acquired the transposon from the other might indicate that horizontal gene transfer does not easily occur between these two types of bacteria or that the biooxidation tanks are not a suitable environment for this to occur.
Exactly how these transposons may have evolved relative to each other is difficult to determine given this limited set of data. However, the discovery of four arsenic resistance-associated transposons in widely different geographical locations in bacteria that are physiologically and phylogenetically diverse suggests that other related transposons are likely to be found in other arsenic-rich environments. An understanding of the evolutionary development of these transposons awaits their discovery.
Nucleotide sequence accession number. The sequence of inserts of plasmid pLfTnArs was deposited in the GenBank database under accession number DQ057986.