Genes involved in arsenic transformation and resistance associated with different levels of arsenic-contaminated soils

Background Arsenic is known as a toxic metalloid, which primarily exists in inorganic form [As(III) and As(V)] and can be transformed by microbial redox processes in the natural environment. As(III) is much more toxic and mobile than As(V), hence microbial arsenic redox transformation has a major impact on arsenic toxicity and mobility which can greatly influence the human health. Our main purpose was to investigate the distribution and diversity of microbial arsenite-resistant species in three different arsenic-contaminated soils, and further study the As(III) resistance levels and related functional genes of these species. Results A total of 58 arsenite-resistant bacteria were identified from soils with three different arsenic-contaminated levels. Highly arsenite-resistant bacteria (MIC > 20 mM) were only isolated from the highly arsenic-contaminated site and belonged to Acinetobacter, Agrobacterium, Arthrobacter, Comamonas, Rhodococcus, Stenotrophomonas and Pseudomonas. Five arsenite-oxidizing bacteria that belonged to Achromobacter, Agrobacterium and Pseudomonas were identified and displayed a higher average arsenite resistance level than the non-arsenite oxidizers. 5 aoxB genes encoding arsenite oxidase and 51 arsenite transporter genes [18 arsB, 12 ACR3(1) and 21 ACR3(2)] were successfully amplified from these strains using PCR with degenerate primers. The aoxB genes were specific for the arsenite-oxidizing bacteria. Strains containing both an arsenite oxidase gene (aoxB) and an arsenite transporter gene (ACR3 or arsB) displayed a higher average arsenite resistance level than those possessing an arsenite transporter gene only. Horizontal transfer of ACR3(2) and arsB appeared to have occurred in strains that were primarily isolated from the highly arsenic-contaminated soil. Conclusion Soils with long-term arsenic contamination may result in the evolution of highly diverse arsenite-resistant bacteria and such diversity was probably caused in part by horizontal gene transfer events. Bacteria capable of both arsenite oxidation and arsenite efflux mechanisms had an elevated arsenite resistance level.


Background
The toxin arsenic in soil and aqueous environments is considered as one of the prominent environmental causes of cancer mortality in the World, especially in Bangladesh, India and China. In recent years, chronic intake of groundwater with high levels of arsenic has caused endemic arsenicosis in several provinces of China and new cases of arsenicosis are continuously emerging [1]. Developing efficient and environment-friendly technologies to remove arsenic from soil and water systems is of great importance to many countries including China. Bioremediation of heavy or toxic metal contaminated sites has been often shown to be more efficient than chemical and physical methods, especially when stimulating indigenous microbial communities [2].
Bacteria have developed different strategies to transform arsenic including arsenite oxidation, cytoplasmic arsenate reduction, respiratory arsenate reduction, and arsenite methylation [3]. The primary role of some of these transformations is to cope with arsenic toxicity. Arsenite-oxidizing bacteria oxidize arsenite [As(III)] to arsenate [As(V)] which in many cases is considered primarily a detoxification metabolism since As(V) is much less toxic than As(III). In addition, As(V) is negatively charged and can be easily adsorbed, thus such bacteria have been used in batch reactors together with immobilizing material for removing arsenic from waste water [4,5].
Two families of arsenite transport proteins responsible for As(III) extrusion, ArsB and Acr3p, have been shown to confer arsenic resistance [12,16,17]. The founding member of the ArsB family, ArsB from E. coli, has been extensively characterized and shown to be a 45 kDa, inner membrane protein with 12 transmembrane helices [18,19]. Either ArsB alone or in association with ArsA catalyzes the extrusion of arsenite and antimonite from cells [20]. In most cases, arsB is co-transcribed with arsC encoding an arsenate reductase. It has been suggested that evo-lution and horizontal gene transfer (HGT) of both the ArsB and the ArsC family may have happened simultaneously in microbial evolution [12]. In many cases, As(III) is taken up by aquaglyceroporins [21] and extruded by ArsB [22].
Members of Acr3p transporters showed a function similar to ArsB, but the two proteins have no significant sequence similarity. Even though Acr3p is much less characterized, it has been reported to be present in more phylogenetically distant species than ArsB. Acr3p could be divided into two subfamilies, Acr3(1)p and Acr3(2)p, based on their phylogenetic dissimilarities [16,23]. Acr3p appeared to be more specific and transported only arsenite but not antimonite [24,25], except that Acr3p of Synechocystis was able to transport both arsenite and antimonite [26]. The arsenite transporter gene arsB, ACR3 (1) and ACR3 (2) have been identified in various soil bacteria [16]. However, a correlation between genotype and arsenite resistance level has not been found yet.
The impact of microbial arsenite oxidation and arsenate reduction were reported to influence environmental arsenic cycles [27]. Understanding the diversity and distribution of indigenous bacterial species in arsenic-contaminated sites could be important for improvement of arsenic bioremediation. Microbial species with arsenic biotransforming capabilities had so far not been evaluated in soil systems in China. The objectives of this study were: (1) Study the distribution and diversity of arseniteresistant and arsenite-oxidizing bacteria in soils with different arsenic-contaminated levels; (2) Investigation of the different arsenite oxidase and arsenite transporter genes and attempt to correlate their presence to the arsenic resistance level of these bacteria.

Distribution and diversity of arsenite-resistant bacteria in soils with different levels of arsenic
Analysis of microbial species and diversity of arseniteresistant bacteria were performed in 4 soil samples with high (TS), intermediate (SY) and low (LY and YC) levels of arsenic contamination. A total of 230 arsenite-resistant bacteria were obtained and 14 of them showed arsenite oxidizing abilities. Based on analyses of colony morphologies and 16S rDNA-RFLP, a total of 58 strains were obtained including 5 arsenite-oxidizing bacteria.

Arsenite resistance levels of arsenite-resistant bacteria vary greatly
The MIC range for arsenite of the 58 strains was from 2 mM to 34 mM (Fig. 1). The numbers of the strains with MIC values ≥ 2 mM, ≥ 5 mM, ≥10 mM, ≥15 mM, ≥ 20 mM,

Identification and distribution of the arsenite oxidase gene aoxB and arsenite transporter genes arsB, ACR3(1) and ACR3(2)
A total of 5 arsenite oxidase genes and 51 arsenite transporter genes were successfully amplified from 38 strains using PCR with degenerate primers. The ACR3(1) and ACR3 (2) were amplified in strains of the high and intermediate arsenic-contaminated soils only. In contrast, aoxB and arsB were found in strains isolated from all three kinds of arsenic-contaminated environments ( The aoxB sequences were amplified from all of the five arsenite oxidizers. No aoxB sequences were amplified from any of the 53 non-arsenite oxidizers. The deduced amino acid sequence of aoxB from Achromobacter sp. SY8 showed 95% identity to AoxB from Achromobacter sp. NT-10 (GenBank accession no. ABD72610) [15]. The deduced AoxBs from Agrobacterium spp. TS43, TS45, and LY4 displayed 95%, 96%, and 95% identities to AoxB of Agrobacterium tumefaciens 5A (GenBank accession no. ABB51928) [31], and 94%, 96%, and 95% identities to AoxB from Rhizobium sp. NT26 (GenBank accession no. AAR05656) [14] respectively. The identity of AoxBs between Pseudomonas sp. TS44 and Alcaligenes faecalis NCIB 8687 was only 61% (Fig. 2). Comparison between 16S rRNA phylogenetic tree, MICs, and related genes 16S rDNA and deduced AoxB phylogenetic trees indicated that their evolutionary relationship was similar.
TS33 were shown to belong to β-Proteobacteria, but were assigned to the γ-Proteobacteria clade using Acr3(2)p sequences; (iv) LY4 belonged to α-Proteobacteria based on the 16S rRNA gene, but its ArsB was in γ-Proteobacteria clade (Fig. 4). The phylogenetic discrepancies exhibited that these 9 arsenite transporter genes were probably acquired by horizontal gene transfer (HGT). Furthermore, 6 of these horizontally transferred ACR3(2) genes were from the strains isolated from the highly arsenic-contaminated TS soil.

Discussion
The first goal of this study was to determine the distribution and diversity of arsenite-resistant bacteria from soils with different levels of arsenic contamination. In addition, the ability to oxidize arsenite was further analyzed. Since the soils were collected from the surface and subsurface zones, only aerobic conditions were used in bacterial isolation. Thus, only aerobic/facultative aerobic bacteria were obtained in this study. We identified Acinetobacter, Comamonas, Pseudomonas, Stenotrophomonas, Delftia, Agrobacterium, and Bacillus as the major genera of arseniteresistant bacteria in the four arsenic-contaminated soils. Among them, Acinetobacter, Agrobacterium, Bacillus, and Pseudomonas species were commonly found at other arsenic-contaminated sites [16,29,30,[32][33][34][35]. To our knowledge, Janibacter, Micrococcus, Thauera, and Williamsia were novel arsenite-resistant bacteria isolated in this study.
We found that the high arsenic TS site revealed a much higher diversity of arsenite-resistant bacteria and the resistance levels observed were also much higher than in isolates found in the intermediate and low arsenic-contaminated sites. It is a limitation that only one medium (CDM) was used for bacterial isolation which could result in the observed differences between sites. The 12 strains with arsenite MICs > 20 mM were all obtained from the high arsenic soil. Generally, it has been proposed that high arsenic contamination is likely to exert a strong selective pressure leading to low microbial diversity [16,32]. However, the TS site used in our study had several hundred years of smelting history [36] which may result in the evolution of more bacterial species that were already well adapted at elevated arsenic concentrations. Moreover, Pennanen et al. [37] reported that at long-term field sites, soil microbial communities have had time to adapt to metal and/or metalloid stress. Turpeinen et al. [33] also found that the diversity of arsenic-resistant bacteria in higher arsenic-, chromium-and copper-contaminated soil was higher than that in less contaminated soil. These results suggested that microorganisms had been adapted to high arsenic stress and maintained their diversity in TS site after a long-term exposure to arsenic.
The aoxB genes were detected in all of the five arsenite oxidizers but not in the non-arsenite oxidizers. This indicates that aoxB may be specific for most of the aerobic arseniteoxidizing bacteria and useful for detecting arsenite-oxidiz-ing microorganisms in the environment. Inskeep et al. [15] reported that arsenite oxidase genes are widely present in different arsenite oxidizers and widespread in soil-water systems. We have enriched pristine soils with arsenite to isolate arsenite-oxidizing bacteria from noncontaminated soils but without success. To our knowledge, all of the cultured arsenite oxidizers obtained so far were isolated from arsenic-contaminated sites. Inskeep et al. [15] detected aoxB-like sequences from arsenic-contaminated environments but not from pristine soils indicating that arsenite oxidation is a major process in arseniccontaminated environments. The expression level of aoxB could probably be applied to monitor environmental arsenic-contaminated levels.
A phylogenetic analysis of the 5 arsenite oxidizers based on the 16S rRNA genes and the aoxB genes showed a similar phylogeny indicating genomic stability of the aoxB genes. The MICs of the arsenite oxidizers displayed a maximal arsenite resistance level in both Pseudomonas and Agrobacterium. Furthermore, strains containing both the arsenite oxidase and any type of transporter gene showed a higher arsenite resistance level. These results suggest that bacteria capable of both arsenite oxidation and arsenite efflux mechanisms have an elevated arsenite resistance level. We also found that arsenite can be fully oxidized even at concentrations close to the MIC in arsenite oxidizers SY8 and TS44 (data not shown). Recently, we have amplified and sequenced the arsC/ACR3 operon (arsC 1 -arsR-arsC 2 -ACR3-arsH) in the adjacent downstream region of aoxB in Pseudomonas. sp. TS44 (data not shown; Gen-Bank, EU311944). Kashyap et al. [31] found that in Agrobacterium tumefaciens strain 5A, disruption of aoxR caused a loss in the ability to oxidize arsenite and furthermore resulted in an apparent reducing phenotype probably due to the action of cytosolic ArsC and subsequent pumping Phylogenetic evidence of potential HGT of arsB/ACR3(2) Figure 4 Phylogenetic evidence of potential HGT of arsB/ACR3 (2). Phylogenetic comparison between 16S rRNA genes (A) and potential horizontally transferred arsB/ACR3(2) genes (B). All sequences used in A's and B's construction are subsets of Fig. 1 and Fig. 3 respectively.
out of As(III). It is noteworthy to point out that there are two processes of As(V) reduction in the environment. One is the use of As(V) as a terminal electron acceptor under anaerobic conditions. The other is the intracellular reduction of As(V) to As(III) under aerobic conditions due to the ArsC-dependent cytoplasmic arsenate reduction as part of the arsenic resistance system (ars operon). Since As(III) is the species being pumped out of cell (by arsB or ACR3), the presence of As(III) in the environments can also be detected under aerobic condition.
One of the main purposes in this research was to determine the correlation among the bacterial arsenite resistance level, bacterial distribution in the environment and the different types of arsenite transporter gene families. We found that the ACR3 genotypes were predominant over arsB (33 ACR3 vs. 18 arsB) in our samples which was in agreement with a report by Achour et al. [16]. In addition, we found any two types of arsenite transporter genes can coexist in the same strain [arsB and ACR3(1), arsB and ACR3(2), ACR3(1) and ACR3 (2)]. Related reports also found the presence of multiple sets of arsenic resistance genes and operons in one strain, especially the arsenite transporter genes. Pseudomonas putida KT2440 contains two operon clusters (arsRBCH) for arsenic resistance [38]. Acidithiobacillus caldus has three sets of arsenic resistance determinants, one located on the chromosome and the other two exist on the transposon [39,40]. Corynebacterium glutamicum has two typical arsenic-resistant operons and additional arsB and arsC genes, of which two arsenite transporter genes belonged to the ACR3(1) group [41]. The genome of Herminiimonas arsenicoxydans revealed the presence of four arsenic resistance operons including two arsB genes and one ACR3 [42]. Multiple sets of arsenic resistance determinants were also reported in B. subtilis [18] and Desulfovibrio desulfuricans G20 [43]. These findings indicate that microorganisms have acquired multiple resistance determinants via chromosomal duplication or horizontal gene transfer to cope with arsenic toxicity. In this study, ACR3(1) and ACR3(2) appeared to be mostly associated with high arsenite resistance since they were only identified from the high and intermediate arseniccontaminated sites, while arsB was found in all three sites. One explanation is that ACR3 may have a higher affinity and velocity to extrude arsenite than arsB and thus seems to be more effective.
Heavy metal contaminated environments were shown to provide a strong selective pressure for transfer of related resistance genes within soil systems [44]. In this study, aoxB and ACR3(1) appeared to be more stable than ACR3 (2) and arsB since phylogenetic discrepancies between 16S rRNA genes and ACR3(2)/arsB were found which supported HGT events of ACR3 (2) and arsB. Most of the HGT occurred in strains identified from the highly arsenic-contaminated TS soil [6 ACR3(2)]. This indicates that arsenite transporter genes may be horizontally transferred and increasingly present in a microbial population under conditions of long-term elevated arsenic stress. It is important to note that HGT occurred in somewhat closely related species in this study, however, this does not detract from the suggestion of HGT and it is likely that the HGT events occurred between these closely related species. Martinez et al. [45] reported that P IB -Type ATPases (pbrA/ cadA/zntA) were broadly transferred in Arthrobacter and Bacillus in radionuclide and metal contaminated soils. Jackson and Dugas [46] also suggested that horizontally transferred arsC resulted in the diversities and complexities of arsenate reductase during its evolution. Excluding arsC, other genes related to arsenic resistance (e.g. arsA, arsB/ACR3) had not been reported as being transferred by HGT. To our knowledge, this is the first study to report widespread horizontal transfer of arsenite transporter genes. The HGT event and subsequent maintenance may have occurred increasingly under the high arsenic pressure [47] and resulted in plastic changes in microbial diversity.

Conclusion
This work investigates the distribution and diversity of microbial arsenite-resistant species in soils representing three different levels of arsenic contamination, and further studies the arsenite resistance and arsenic transforming genes of these species. Our research provides valuable information of microbial species and genes responsible for arsenite oxidation and resistance, and increases knowledge of the diversity and distribution of the indigenous bacteria that may be stimulated for successful bioremediation of arsenic contamination.

Site description and soil sample collection
Four soil samples representing high (TS), intermediate (SY) and low (LY/YC) levels of arsenic contamination were used in this study. The TS soil was collected in Tieshan District, a highly arsenic-contaminated region, which is located in Huangshi City, Hubei Province, central China. There are several kinds of metal (gold, copper, iron) mines that have been smelted for many years. Concentrations of arsenic and heavy metals are extremely high in water, soil and sediments of this area [36]. The SY soil was collected from a pig farm, Shayang County, Jingmen City, Hubei Province where there are certain levels of arsenic in the soil due to the long-term usage of arsenic in feed material to resist disease and stimulate pig growth. The other two samples, LY and YC, were collected from the low arsenic-contaminated soils near the Yellow Sea of Lianyungang and Yancheng Cities Jiangsu Province, eastern China, respectively.

Isolation and identification of arsenite-resistant and arsenite-oxidizing bacteria
One hundred grams of each soil sample was amended with NaAsO 2 to a final concentration of 500 mg kg -1 and incubated at 28°C for a week. During incubation sterilized H 2 O was added to the jars to reach the original moisture value. Isolation of arsenite-resistant bacteria was performed by adding 10 g (triplicates) of each soil to 90 mL 0.85% NaCl in a 250 mL Erlenmeyer flask and shaken at 160 rpm for 30 min. 1 mL of the above mixture was added to 9 mL 0.85% NaCl for serial dilution and plated on chemically defined medium (CDM) plates [9] with a final concentration of 800 μM NaAsO 2 and incubated at 28°C for another week. Single colonies were picked and restreaked several times to obtain pure isolates.
The obtained arsenite-resistant bacteria were tested for their abilities to oxidize As(III) (NaAsO 2 ) using a qualitative KMnO 4 screening method [10]. Each arsenite-resistant bacterium was inoculated in CDM broth with a final concentration of 800 μM NaAsO 2 and then shaken at 160 rpm for 5 days at 28°C. For each isolate 1 mL culture was added to a 1.5 mL centrifuge tube containing 30 μL of 0.01 M KMnO 4 and the color change of KMnO 4 was monitored. A pink color of the mixture indicated a positive arsenite oxidation reaction [formation of As(V)]. The sterile CDM medium containing the same amount of NaAsO 2 was used as an abiotic control. The arsenite oxidizing phenotype was also detected using the molybdene blue method with a spectrophotometer (DU800, BeckMan, CA, USA) [48].
Total DNA of each strain was extracted using standard molecular genetic methods. Nearly full-length 16S rDNA of the bacteria was amplified by PCR using universal primers Uni-27F and Uni-1492R (Table 1)  Colony morphologies and 16S rDNA PCR-RFLP technique were used to remove the repeated isolates for each sample. PCR-RFLP was performed by enzyme digestion at 37°C for 3 hrs in a 20 μL volume containing 2 μL of 10 × enzyme buffer, 2.5 units of HaeIII or MspI and 5-10 μL of the 16S rDNA PCR products, amending ddH 2 O to a final volume of 20 μL. The digested DNA fragments were separated in 2% agarose gels and the digestion patterns were grouped by DNA fingerprinting profiles.

Identification of the aoxB gene encoding the arsenite oxidase Mo-pterin subunit and arsB, ACR3(1) and ACR3(2) genes encoding different arsenite transport proteins
The PCR amplification of aoxB was performed using degenerate primers (Primers #1F and #1R) ( Table 1) and following the PCR conditions as described by Inskeep et al. [15]. The amplification of arsB, ACR3(1) and ACR3 (2) genes were performed using three pairs of degenerate primers [darsB1F and darsB1R for arsB, dacr1F and dacr1R for ACR3(1), dacr5F and dacr4R for ACR3 (2)] (Table 1) as  described by Achour et al. [16]. The PCR products were purified using the Gel Extraction Kit (SBS Genetech, Shanghai, China). The purified PCR products were ligated into pGEM-T (Promega, Madison, WI, USA) and the ligation products were used to transform E. coli DH5α competent cells by electroporation. The transformants were grown on LB agar containing ampicillin, X-Gal and IPTG at 37°C for 16 hrs according to the manufacturer's recommendations.

DNA sequencing and phylogenetic analysis
The PCR products were purified using the UltraPure™ PCR Kit (SBS Genetech). DNA sequencing analysis was performed using ABI 3730XL DNA analyzer by Sunbiotech company (Beijing, China). All sequences were analyzed by BlastN (for 16S rRNA gene) and BlastX (for deduced AoxB and ArsB/Acr3p) searching tools [50]. All sequences were checked manually and edited for the same lengths using ClustalX 1.83 software [51]. MEGA 3.1 software [52] was used to construct phylogenetic trees by the neighbor-joining distance method [53] and reliability of inferred trees was tested with 1000 bootstrap replicates. Some reference sequences from the GenBank were used in constructing phylogenetic trees for clarification.

Determination of the minimal inhibitory concentrations (MICs) of arsenite
The MIC, defined as the lowest concentration of arsenite that inhibited growth in CDM broth, was performed with all arsenite-resistant bacteria. Triplicate samples of each single colony were inoculated in 3 mL CDM broth supplemented with increasing concentrations of NaAsO 2 , incubated with shaking at 28°C for one week and the OD 600 values were determined. The initial screening for MICs was performed with 5 mM, 10 mM, 15 mM, and 20 mM of NaAsO 2 . Subsequent determinations were performed with 1 mM NaAsO 2 intervals over the appropriate range. The sensitivity of MIC detection was 1 mM.
Publish with Bio Med Central and every scientist can read your work free of charge