Introduction

Of all the heavy metals, mercury (Hg) is the most toxic and with no biological function (Nies, 1999). Mercury is not abundant in the soil crust, but can be found in ores, the dominant being cinnabar (HgS). Mercury is emitted to the atmosphere by natural sources, for example, soil erosions and volcanic eruptions, and by anthropogenic sources, for example, coal-fueled power plants (Nriagu and Pacyna, 1988). Glacial ice core studies have shown that during the timescale of industrialization in the western world, mercury has increased approximately 20-fold in the ice cores, and anthropogenic activities account for 70% of the last 100 years mercury emissions (Schuster et al., 2002). The atmospheric mercury is transported globally, but deposited locally, and pose a threat to organisms at the top of food webs, due to bioaccumulation of methyl mercury. However, severe mercury contamination of natural environments can often be attributed to inappropriate human activities. Examples include the Minamata Bay in Japan, where the local fish-consuming villagers were exposed and poisoned by high levels of methylmercury in their seafood due to contamination by a nearby acetaldehyde production facility (Eto, 2000; Tomiyasu et al., 2006).

The Y12 nuclear weapons plant facility at Oak Ridge (TN, USA) used mercury for lithium isotope separation during the 1950s and 1960s, and this resulted in an estimated 75–150 tonnes of mercury inadvertently being released to the local environment, including the East Fork Poplar Creek (Han et al., 2006). Examination of the lower East Fork Poplar Creek floodplain (LEFPCF) soil revealed total mercury concentrations ranging from 42 to 2400 μg Hg g−1 dry weight (Barnett et al., 1995). In the LEFPCF studies, the largest fraction of the mercury was estimated to be immobilized as HgS (Revis et al., 1989). Although several studies have been conducted at Oak Ridge dealing with mercury effects on higher trophic level organisms like fish (Burger and Campbell, 2004) and mammals (Stevens et al., 1997), little has been done to elucidate the effects of mercury on the microbial community in the soil.

Three mechanisms for microbial adaptation to environmental stress have been suggested (reviewed in Sørensen et al., 2002): (i) enrichment of populations that carry the required resistance/tolerance traits, (ii) induction of enzymes involved in the detoxification or resistance mechanisms, and (iii) genetic adaptation. The first two are manifested by previously existing sub-populations of resistant strains, while the process of genetic adaptation creates changes in the existing genetic pool of the microbial community and thus the evolution of new capabilities and an increase in the functional diversity of the microbial community. Horizontal gene transfer is a mechanism by which the genetic diversity among bacteria may be enhanced. Thus, it may play an important role in the adaptation and evolution of microbial communities (Sørensen et al., 2005). Metal-resistance genes are often carried on plasmids and other mobile genetic elements (Silver and Phung, 1996) and therefore, exchange of these mobile elements through metal-impacted communities may facilitate rapid alleviation of metal toxicity. This is supported by the higher incidence of plasmid DNA among bacteria isolated from polluted environments (Rasmussen and Sørensen, 1998) and by increased degradation by indigenous bacteria after plasmid transfer with relevant catabolic genes (Top et al., 2002).

Mercury resistance is found in both Bacteria (Barkay et al., 2003) and Archaea (Schelert et al., 2004). Mercury resistance in Bacteria is conferred by the mer operon. Several mer operon-encoded proteins are involved in transport of inorganic oxidized mercury into the cytosol, where the merA-encoded mercuric reductase protein, in a NAD(P)H-dependent manner, reduces Hg2+ (aq) to volatile, less reactive elemental mercury (Hg0 (g)) (mercury in the environment and bacterial resistance is thoroughly reviewed in Barkay et al., 2003). Mercury resistance has been recognized in many different phyla, including Firmicutes, Actinobacteria and Proteobacteria, of both clinical and environmental origin and it is considered to be an ancient resistance mechanism (Osborn et al., 1997). Biochemical and structural studies of the homodimeric mercuric oxidoreductase have shown that there are four cysteine residues in each monomer, that are essential for reducing oxidized mercury to elemental mercury and furthermore that two tyrosine residues are important for optimal functionality (Moore and Walsh, 1989; Distefano et al., 1990; Schiering et al., 1991; Moore et al., 1992). Some mercuric reductase proteins contain a mercury binding N-terminal domain, which gives better protection against oxidized mercury (Ledwidge et al., 2005), but this is not essential, as it is not found in the mercuric reductase enzymes of, for example, Actinobacteria and Archaea.

If there has been long-term selective pressure for mercury-resistant bacteria, all individual bacteria will probably be resistant, and the resistance trait (the merA protein) will probably have evolved differently in different bacteria, thus leading to high functional diversity, closing in on 16S rRNA diversity. Alternatively, if the selective pressure by mercury is of a newer date, only resistant bacteria and those bacteria that could adapt to the new selective pressure by, for example, horizontal gene transfer thus receiving the required resistance traits would proliferate. In the latter case scenario, the community diversity would decrease somewhat, and the functional diversity would be low.

Most of the studies dealing with heavy metal contamination in soil have been conducted with surface soils (Rasmussen and Sørensen, 2001; Muller et al., 2001a; Muller et al., 2001b; Ellis et al., 2003), whereas only few studies have looked at sub-surface soils, and most of these have dealt with other metals, for example, uranium (North et al., 2004; Abulencia et al., 2006) or organic pollutants (Lowe et al., 2002; Zhou et al., 2004).

The purpose of the present study was to examine the diversity of merA genes from surface and sub-surface soil, exposed to mercury for decades, and thus address the role of horizontal gene transfer in bacterial adaptation. It is well known that most soil bacteria are fastidious with regards to cultivation media (Janssen, 2006), and it has been estimated that in 1 g of soil, there might be as many as 106 different bacterial species (Gans et al., 2005). With the obvious limitations of cultivation of soil bacteria, molecular approaches may assist in deciphering bacterial functional diversity. In the present study, we used a cultivation-independent approach, based on PCR amplification, cloning and partial sequencing of the merA gene, enabling determination of merA diversity in soil. Bacterial isolates from LEFPCF (Oregaard et al., 2007) were also examined with regards to the presence of merA genes.

Materials and methods

merA primer design

Mercuric reductase (merA) sequences were obtained by searching the GenBank nucleotide database at NCBI. Putative mercuric reductase genes were found in the following phyla: Firmicutes, Actinobacteria, Deinococcus-Thermus, Bacteroidetes, Proteobacteria (α-, β-, γ- and δ-) and in Crenarchaeota and Euryarcheota, the latter two belonging to Archaea. The DNA sequences were prepared in FASTA format and subjected to alignment by RevTrans (http://www.cbs.dtu.dk/services/RevTrans/; Wernersson and Pedersen, 2003). In short, the sequences were in silico translated to amino acids, aligned and then converted back to DNA sequences, but according to the amino-acid alignment. These multiple DNA alignments were examined manually using clustalX software (Chenna et al., 2003). Primers were manually designed, placing them in conserved regions of the multiple alignments. The merA of the different phyla were too diverse for primers encompassing them all. Primers intended to be specific for the following phyla were designed; α-Proteobacteria, additional β/γ-Proteobacteria primers, Firmicutes and Actinobacteria (Table 1).

Table 1 merA-specific primers for different bacterial phyla
Full size table

The primer pairs were tested in the laboratory on available merA genes from γ-Proteobacteria (merA from Serratia marcessens plasmid pDU1358, cloned into pHG103 (Griffin et al., 1987), kind gift of Dr Tamar Barkay, Rutgers University) and Firmicutes merA genes. Three different merA genes originating from Staphylococcus aureus pI258 (Laddaga et al., 1987), Bacillus sp. RC607 (Wang et al., 1987) and Bacillus megaterium MB1 (Huang et al., 1999), all cloned into Escherichia coli vectors (pRAL2, pYW40 and pGB3A, respectively, kind gift of Dr Tamar Barkay, Rutgers University) were used. Primers targeting Actinobacteria or α-Proteobacteria were tested on mercury-resistant isolates from Oregaard et al. (2007). The plasmids containing merA genes were extracted from their E. coli host cells with QIAprep Spin Miniprep Kit (Qiagen, Hilden, Germany) after overnight growth in LB broth. Genomic DNA of uncharacterized isolates was extracted from colonies grown on dilute media (Oregaard et al., 2007) or in 10% tryptic soy broth supplemented with 10 μ M HgCl2, with High Pure PCR Template Preparation Kit (Roche Diagnostics GmbH, Mannheim, Germany) according to the manual.

Soil

The soil came from LEFPCF, which has a history of mercury contamination. The soil samples were obtained from three depths at the same site: 0–5 cm (surface, soil B), 45–55 cm below surface (soil C) and 90–100 cm below surface (soil D). The soil was stored at 4°C after initial pre-sieving (2 mm mesh).

Soil pre-treatments

The soils were pre-treated with mercury (or water in control treatments) and set up in plastic bags with the equivalent of 20 g dry soil. All pre-treatments were done in triplicate. Pre-treatment with mercury was performed to selectively enrich the soils with mercury-resistant bacteria. A volume of 1 ml autoclaved distilled H2O was added to the control soils (w). The mercury pre-treatments (Hg) included HgCl2, in a final volume of 1 ml. The amount of mercury was adjusted for each soil type to obtain a start mercury concentration of 10 μg g−1 soil–water. The soil microcosms were stored at 25°C in the dark for 7 days, after which the soil was stored at −80°C.

merA clone libraries

Soil DNA was extracted from control and mercury-stimulated soil with Fast DNA Spin for Soil kit BIO 101 Systems (http://www.qbiogene.com).

The β/γ-Proteobacteria PCR reactions were performed with 2 μl soil DNA, primers #91 and #54 (0.4 μ M final concentration) and water in a total volume of 25 μl with PuReTaq Ready-To-Go PCR beads (GE Healthcare, Hillerød, Denmark). Expected amplicon size was 288 bp. The PCR reaction consisted of an initial denaturing step at 95°C/3 min; then 45 cycles with two steps: 95°C/10 s and 60°C/1 min. Final extension was at 72°C for 2 min.

The Actinobacteria PCR reactions were done with Phusion High-Fidelity DNA Polymerase (Finnzymes, Espoo, Finland). The PCR reaction mix consisted of 2 μl soil DNA, Act-Fw and Act-Rv primers (0.5 μ M final concentration), 4 μl HF buffer, dNTP (200 μ M final concentration), Phusion DNA polymerase (0.02 U μl) and water in a total volume of 20 μl. Expected amplicon size was 391–397 bp. The PCR reaction consisted of an initial denaturing step at 98°C/2 min; then 35 cycles with three steps: 98°C/10 s, 62°C/10 s and 72°C/20 s. Final extension was at 72°C for 2 min.

The Firmicutes PCR reactions were performed with Phusion polymerase. The PCR reaction mix consisted of 1 μl soil DNA, Fir-Fw and Fir-Rv1892 primers (0.5 μ M final concentration), 2 μl HF buffer, dNTP (200 μ M final concentration), Phusion DNA polymerase (0.02 U μl) and water in a total volume of 10 μl. Expected amplicon size was 455 bp. The PCR reaction consisted of an initial denaturing step at 98°C/2 min; then 35 cycles with three steps: 98°C/10 s, 64°C/10 s and 72°C/20 s. Final extension was at 72°C for 2 min.

The total PCR reaction volume from each reaction was loaded onto 1.2% agarose gels, gel electrophoresed and stained with ethidium bromide. If merA bands of correct size were observed, they were cut out and purified with QIAEX II Gel Extraction Kit (Qiagen). The extracted amplicons obtained with the same primers were pooled before cloning (for example, soil B, stimulated with water, three replicates, all mixed and aliquot of this used for cloning). The merA amplicons were cloned with the TOPO TA cloning kit (Invitrogen, Paisley, UK). A volume of 2–4 μl merA amplicon extracted from gel was mixed with 1 μl salt solution and 1 μl pCR4-TOPO vector, and water to a final volume of 6 μl. The cloning and transformation was done according to the manual, using competent TOP10 E. coli cells. The transformants were grown overnight at 37°C on LB agar with Ampicillin (100 μg ml−1) and re-streaked onto LB agar with Kanamycin (50 μg ml−1).

merA sequencing, analysis and phylogeny

The transformants were grown overnight and the cloning vectors (with merA inserts) were extracted with Qiaprep Spin Miniprep Kit (Qiagen). The plasmid DNA concentration was determined spectrophotometrically (A260/A280). The sequence reaction mix consisted of 4 μl DYEnamic ET Dye Terminator mix (MegaBACE, Amersham, UK), 1 μl M13 forward primer (10 μ M), water and 200–400 ng plasmid DNA in a final volume of 10 μl. The dideoxy nucleotide sequencing (Sanger sequencing) reaction was done in the following way: initial denaturing at 94°C for 3 min followed by 30 cycles consisting of three steps: 94°C/20 s, 50°C/15 s and 60°C/60 s. Subsequently, 30 cycles were done in the same manner but with elongation occurring for 120 s. Finally, there was a single step at 60°C for 5 min. The sequence reaction mix was purified with a Sephadex G-50 plate (96 wells), and the 96-well plates were sequenced on a MegaBACE 1000 Sequenator (Amersham Biotech, Amersham, UK).

The sequences were manually trimmed for vector and primer sequences. Blastn (http://www.ncbi.nlm.nih.gov/BLAST/) (Altschul et al., 1997) was used to detect probable gap positions in the clone sequences, and in silico translation of sequences into functional mercuric reductase protein sequences was done to confirm that the obtained DNA sequence resembled mercuric reductase enzymes. The sequences were subjected to RevTrans alignment as above. The multiply aligned DNA sequences were subjected to DNA distance matrix calculations with the DNADIST of the Phylogeny Inference Package (Phylip, version 3.65, Joseph Felsenstein, http://www.evolution.genetics.washington.edu/phylip.html). The DOTUR software program (Schloss and Handelsman, 2005) was fed with the DNA distance matrices for determination of operational taxonomic units (OTUs) and calculation of different diversity measures, including Shannon–Weaver diversity index and Chao1 richness estimator. Library coverage was calculated as: C=[1−(n/N)]100%, where n is number of OTUs and N the number of sequences.

Sequences

The sequences obtained in this study have been deposited at NCBI. The merA clones of β- and γ-Proteobacteria and Actinobacteria are associated with accession numbers EF460128–EF460310. The merA sequences from the isolates are associated with accession numbers EF455060–EF455079.

Results

Full-length mercuric reductase sequences deposited at GenBank (NCBI) were retrieved and analyzed. The four essential cysteine residues (positions 207, 212, 628 and 629 in Bacillus cereus RC607 merA enzyme, BAB62433) along with the two important tyrosine residues (positions 264 and 605) were used to discriminate mercuric reductase sequences from other similar disulfide oxidoreductases like dihydrolipoamide dehydrogenases. Using these amino acids as criteria for accepting sequences as mercuric reductases, along with general similarity to merA sequences found with blastn/blastp (Altschul et al., 1997; Schaffer et al., 2001), merA of different phyla were identified (Table 2). It was apparent, that many sequences annotated as mercuric reductase encoding genes, were probably wrongly annotated, when using these criteria, since they lacked the two terminal cysteines, Cys628 and Cys629 (data not shown). All mercuric reductase proteins used for primer design by multiple alignments contained all four cysteines and both tyrosines, except one Pseudomonas sequence in which Tyr264 had been substituted with isoleucine (CAC80079). All the Archaea sequences had a phenylalanine residue at position 605, in contrast to tyrosine in bacteria (Table 2). None of the Archaea sequences from the NCBI database used here have been experimentally characterized as conferring mercury resistance to their host. A neighbour joining tree of merA proteins of different phyla is shown in Figure 1. The outgroup sequences were four proteins annotated as mercuric reductases, but without the signature Tyr264, Cys628 and Cys629 amino acids (Table 2).

Table 2 Presence of particular amino acids, essential for mercuric reductase activity
Full size table
Figure 1
figure 1

Neighbour-joining tree of mercuric reductase protein sequences. The sequences were obtained from GenBank at NCBI. The boxes reflect phyla with more than one sequence. The gray boxes represent phyla with phenotypic mercury resistance, based on mercuric reductase proteins. The asterisk (*) indicates phyla or isolates in which phenotypic mercury resistance has not been recognized. The horizontal length of a box is equivalent to the maximum length from node to tip of a particular sequence in that phylum. The number at the right-hand corner of the box reflects the number of proteins included in the alignment. The value in parenthesis is the range of GC content of the merA gene of the particular phylum. Black and white circles at nodes reflect bootstrap values higher than 95 and 50%, respectively (out of 1000 bootstraps). The bar at the bottom reflects 5% difference at the amino-acid level. The accession numbers of the proteins used for the neighbour-joining tree were as follows Archaea; NP_376996, NP_147957, CAF18529, NP_560142, NP_344015, YP_256307, NP_110770, YP_023042, ZP_00610004, YP_134400, NP_394797 and ZP_00610602; δ-Proteobacteria: NP_954464, ZP_01140457, ZP_01388909 and YP_383148; Outgroup: ABC45686, NP_952368, YP_628559 and YP_010258; Bacteroidetes: ZP_01060916; Deinococcus-Thermus: YP_144419 and YP_004762. The β/γ-Proteobacteria and Actinobacteria merA protein accession numbers are shown in Figures 2 and 5, respectively. α-Proteobacteria and Firmicutes accession numbers are shown in Figure 4.

Full size image

The multiple alignments were used to design primer pairs targeting merA from Firmicutes, Actinobacteria and α-Proteobacteria, and additional primers for β/γ-Proteobacteria (Table 1). In silico PCR reactions with the primers against the different phyla revealed that they gave amplicons of expected size and were specific to their phyla, except for the Actinobacteria primers, which also seemed to equally well amplify β/γ-Proteobacteria merA.

Three different merA genes originating from S. aureus (Laddaga et al., 1987), Bacillus sp. RC607 (Wang et al., 1987) and B. megaterium (Huang et al., 1999) confirmed that Firmicutes type merA primers, Fi-Fw/Fi-Rv1832 and Fi-Fw/Fi-Rv1892, resulted in amplicons of approximately 395 and 455 bp, respectively (data not shown). The β/γ-type new primers worked with pDU1358 merA (Griffin et al., 1987) (data not shown). To test the applicability of the Actinobacteria primers, genomic DNA from the mercury-resistant Streptomyces isolate (Is-BDOE1) from soil B (Oregaard et al., 2007) was used in PCR reactions. The amplicon was approximately 400 bp, corresponding to the expected 391–397 bp (data not shown).

In the present study, we were interested in describing the diversity of merA genes in surface and sub-surface soils with a history of mercury contamination. Primers targeting β/γ-Proteobacteria (#91 and #54), Actinobacteria (Act-Fw and Act-Rv) and Firmicutes (Fi-Fw and Fi-Rv1892) (Table 1) were used in PCR reactions with DNA extracted from soils pre-treated with mercury or control soils pre-treated with water. The β/γ PCR reactions gave partial merA amplicons in surface and both sub-surface soils, and this was seen for both stimulated and control soils. In contrast, partial Actinobacteria merA genes were only amplified in surface soil (in both control and mercury-stimulated soil). The Firmicutes-type merA was not detected in any soil. The merA amplicons of the β/γ and the Actinobacteria were cloned and transformed. Cloned partial β/γ merA genes, mainly from control soils stimulated with water were sequenced, resulting in 43 sequences from surface soil (B), 49 sequences from sub-surface soil (C) and 51 sequences from deeper sub-surface soil (D). Furthermore, of the 40 sequences amplified with Actinobacteria primers, 20 merA sequences turned out to be β/γ-Proteobacteria in origin (Ba). All β/γ-Proteobacteria merA clone sequences (B, C, D and Ba) were aligned along with 30 reference β/γ merA sequences from GenBank. The multiple alignment was subjected to DOTUR analysis (Schloss and Handelsman, 2005), allowing determination of OTUs and diversity estimates of the different soils. When discriminating bacterial species from one another based on 16S rRNA sequences, a threshold of 3% is typically used, and for genus separation, 5% are used (Schloss and Handelsman, 2005). A threshold of 5% difference was used in the analysis, obtaining more conservative estimates of diversity, compared to 3% threshold. With the 5% threshold for definition of OTUs, the 163 β/γ merA clones fell into 70 OTUs, of which 28 contained more than one sequence. The phylogeny of the β/γ merA sequences from the different soils is shown in Figure 2 along with reference sequences from GenBank. Since the cloned sequences were only around 244 bp in length (excluding the primers), all the reference sequences and plasmids were also truncated, to only contain the fragment obtained with primers #91 and #54 primers. All the cloned sequences showed highest similarity to other β/γ merA genes at GenBank, as determined by blastn or blastp search, except clone C-31, see below. It is apparent from Figure 2 that the β/γ merA gene is highly diverse. Since all partial merA clone sequences were most similar to other merA sequences at GenBank, it is unlikely that the high diversity observed was due to non-specific amplification of genes other than merA. If such non-specific amplification occurred, it would be most probable that the amplicons had different sizes than the merA fragments. Interestingly, all the clones were more than 5% different from the reference sequences, except OTU-03 (B-63 and C-37) grouping with merA from pDU1358, and OTU-07 (Ba-39, Ba-44 and Ba-47) grouping with merA of plasmid pKris-2 and a merA gene from a Pseudomonas isolate (Figure 2).

Figure 2
figure 2figure 2

Neighbour-joining tree of merA DNA sequences. The sequences were obtained from β/γ-type merA clone libraries (B, C, D and Ba) or from exogenously isolated plasmids (pKris). The boxes reflect OTUs with more than one sequence. The horizontal length of a box is equivalent to the maximum length from node to tip of the particular OTU. The OTUs were defined by DOTUR. Black circles at nodes reflect bootstrap values higher than 90% (out of 1000 bootstraps). Letters in parenthesis reflect the soil type. The number inside the box, or at the upper left-hand corner of the box reflects the number of sequences within the particular OTU. Archaea merA gene from Haloarcula marismortui was used as outgroup. The bar at the bottom reflects 5% difference.

Full size image

Exogenous plasmid isolation had been conducted on soils B, C and D, resulting in four different plasmids (pKris-1 to -4), as determined by plasmid restriction enzyme analysis (Lipthay et al., 2007). All plasmids conferred mercury resistance to their host (E. coli or Pseudomonas putida). In the present study, the merA gene was amplified with β/γ primers BG-Fw349 and #54 and sequenced, obtaining over 1100 nucleotides from each merA gene. The phylogenetic relationship of these merA genes is as mentioned above, shown in Figure 2. When looking at the 244 bp fragment obtained with primers #91 and #54 (excluding primers), the four plasmids fell into three different OTUs (⩾95% identity) (Figure 2).

Estimates of merA diversity in the three different soil microcosms, based on the sequenced clones, are shown in Table 3. The diversity estimates were similar for all three soils. Although all three soils were very similar in merA diversity, the actual cloned sequences were generally different as is shown in the Venn diagram in Figure 3. The circles represent different soil clones and plasmids from these soils, as indicated by letters on their side, and reference sequences from GenBank, with the numbers inside indicating the number of OTUs. The smaller interconnecting circles show how many OTUs were shared between the connected circles (environments). Only two OTUs contained sequences from all three soils (OTU-01 and OTU-02, in total 12 sequences) (Figure 2). The OTUs with representatives of different soils seemed fairly equally distributed between B/C (four OTUs), B/D (three OTUs) and C/D (eight OTUs) (Figure 3). Note that only three OTUs were connected to GenBank sequences.

Table 3 Diversity of cloned β/γ-Proteobacteria merA sequences
Full size table
Figure 3
figure 3

A Venn diagram representing shared and unique β/γ-type merA OTUs. The OTUs were defined as being no less than 95% identical at the DNA level. Shared OTUs are interconnected between the relevant circles. The area of the circles reflects the number of OTUs belonging to these. A total of 198 sequences were included in the analysis, of which 30 came from NCBI (27 OTUs), four sequences from exogenously isolated plasmids (three OTUs), 20 Ba sequences (six OTUs), 43 B sequences (21 OTUs), 49 C sequences (29 OTUs) and 51 D sequences (25 OTUs). OTUs, operational taxonomic units.

Full size image

The Actinobacteria merA PCR and cloning resulted in 20 sequences showing highest similarity to Actinobacteria merA proteins. One of these was discarded, as it seemed chimeric, by high N-terminal similarity to β/γ merA, whereas the C-terminal protein region was most similar to Actinobacteria. The remaining 19 sequences were aligned against nine Actinobacteria merA references and a DNA Neighbour-Joining tree was calculated (Figure 4). The 19 sequences fell into 10 OTUs (5% threshold), of which seven OTUs contained one sequence. It is noteworthy that most of the cloned sequences are quite divergent from the sequences deposited at GenBank.

Figure 4
figure 4

Neighbour-joining tree of Actinobacteria merA DNA clone sequences, isolate Is-BDOE1 and reference sequences from GenBank. The boxes reflect OTUs with more than one sequence. The horizontal length of a box is equivalent to the maximum length from node to tip of the particular OTU. The OTUs were defined by DOTUR (⩾95% similar). Black circles at nodes reflect bootstrap values higher than 95% (out of 1000 bootstraps). Values in parenthesis reflect the number of sequences in the particular OTU. Archaea merA gene from Haloarcula marismortui was used as outgroup. The bar at the bottom reflects 5% difference. OTU, operational taxonomic unit.

Full size image

The C-31 clone sequence mentioned above and obtained with β/γ-type primers did not show similarity to other merA genes by blastn. When the C-31 clone sequence was virtually translated into 81 amino acids, the protein sequence showed similarity to other putative merA proteins from α-Proteobacteria, for example, Xanthobacter autotrophicus (59 identical amino acids (72%)) (Figure 5). The neighbour-joining tree in Figure 5 includes nine partial merA sequences obtained from anoxic river sediments, contaminated with mercury and obtained from PCR reactions using similar primers as in the present study (Ni Chadhain et al., 2006). Note how two of the sequences (SBC-049 and SBC-183) are most closely related to other α-Proteobacteria merA sequences. The other seven SBC sequences all seem closer related to α-Proteobacteria than to the Gram-positive Firmicutes-type merA genes.

Figure 5
figure 5

Neighbour-joining tree of α-Proteobacteria merA clone sequences obtained in the present study (C-31) and in a study of anoxic sediments (SBC-prefix), see Discussion for details. All the reference sequences from α-Proteobacteria are putative merA genes. Firmicutes-type merA DNA is also included. I and II are clades defined in the original paper describing these clones, obtained with similar primers as used for the β/γ-type merA clone library of the present study. The values to the left at some of the nodes reflect the bootstrap values (out of 1000). Archaea merA gene from Haloarcula marismortui was used as outgroup. The bar at the bottom reflects 5% difference.

Full size image

Several isolates obtained from the same soils as used in this study (Oregaard et al., 2007) were subjected to merA-specific PCR with the primers shown in Table 1, and some were partially sequenced (Table 4). The Streptomyces (Is-BDOE1) isolate seemed to contain a merA sequence diverging from its closest relative by approximately 27% at the amino-acid level (Table 4). The merA gene of Is-BDOE1 was amplified on two occasions and sequenced with both primers (Act-Fw and Act-Rv) but only the forward primer resulted in acceptable sequence data. The sequences obtained with the forward primer were identical. The Dyella-like isolates had merA genes that differed somewhat from the closest match, pSB102, a full-sequenced plasmid from Alfalfa rhizosphere (Schneiker et al., 2001).

Table 4 LEFPCF soil isolates from which partial merA sequences were obtained
Full size table

Partial α-Proteobacteria merA DNA sequences of several different isolates were subjected to both blastn and blastp search. Blastn did not reveal any relatedness to other merA genes and did not result in any meaningful hits. Blastp search showed similarity to all the putative α-Proteobacteria merA protein sequences used in the phylogenetic tree in Figure 5. The α-Proteobacteria merA fragments of the Bradyrhizobium, Rhizobium and Sphingomonas isolates (Table 4) were identical. The merA of the Rhizobiales-like isolate was somewhat different. The partial merA sequence of the Rhizobiales-like isolates showed similarity to other putative merA genes of α-Proteobacteria, ranging from 62% to 49% identical amino acids. Mercuric reductases from Firmicutes were second in similarity to the Rhizobiales-like merA gene, the highest at 45% identity (YP_148949, Geobacillus kaustophilus). These are the first (partial) merA sequences of α-Proteobacteria isolates, which have a mercury-resistant phenotype.

Discussion

The diversity of mercury resistance traits—merA—was high in all soil samples tested in this study. There were no significant differences in diversity between surface and sub-surface samples. The merA diversity estimators were nevertheless lower than the similar estimators calculated based on 16S rRNA gene clone libraries from the same soil samples (unpublished results, Oregaard, de Lipthay and Sørensen). This difference could indicate recent horizontal spread of the mercury-resistance trait among the soil bacteria in response to the metal contamination. This is corroborated by the findings of identical merA genes in different genera of α-Proteobacteria, isolated from different depths (see below), and that conjugative plasmids conferring resistance to mercury have been isolated from LEFPCF soils (Lipthay et al., 2007). The adaptation of the bacterial community in the soil has probably not been solely due to horizontal gene transfer of few broad host-range mercury-resistance plasmids, since many different merA genes are present in the soil. A combination of horizontal gene transfer and selection of resistant subpopulations seems like a plausible explanation for the observations in the present study.

Even though the soils were very similar in merA diversity, the composition of merA genes was very different, and most OTUs were unique to the depth at which they were found. It has been argued that environmental factors exerting selective pressure on particular traits will probably not be effective at the phylum level, but more so at genus or species level (Janssen, 2006), and in a recent study of approximately 1900 sequenced 16S rRNA gene clones from three agricultural soils experiencing different management treatments, highly similar diversity values were found, but the bacterial community compositions were different in the three soils (Hartmann and Widmer, 2006). The authors argue that to understand the effects of a particular environmental stimulus, community composition analysis is required, since diversity indices are often insensitive to changes inferred by the stimulus (Hartmann and Widmer, 2006). In the present study, merA composition at the examined depths seemed very distinct, and less than 20% (11/65; BCD) of the OTUs were shared between different soil compartments, indicating that these vertically adjacent bacterial soil communities only experience limited mixing. This would mean that horizontal gene transfer would only have limited effect on dissemination of adaptive traits between the vertically separated soil communities.

Of the 62 OTUs obtained with the β/γ-Proteobacteria primers (Figure 3), only one is ⩾95% similar to a GenBank reference sequence. This highlights that the actual functional diversity in natural environments is vast compared to what is recognized in culture collections. The group by Dr Janssen has demonstrated elegantly that successful cultivation of hard to culture recalcitrant soil bacteria can be achieved by long-term incubation on solid media with complex carbon sources (Janssen et al., 2002; Davis et al., 2005). In a parallel study of the soils used in the present study, several mercury-resistant bacteria were obtained by long-term incubation (Oregaard et al., 2007). Although most of the isolates are similar to previously cultivated bacteria, many of the merA genes are novel, and when virtually translating them to protein sequences and comparing with GenBank merA proteins, they are less than 95% identical at the amino-acid level (Table 4).

To the best of our knowledge, this is the first time that merA genes have been shown in mercury-resistant α-Proteobacteria (soil) isolates. The first known mercury-resistant α-Proteobacteria was a marine Rhizobiales-like isolate from a hydrothermal vent plume in the Pacific ocean, at approximately 2.5 km depth (Vetriani et al., 2005). The authors showed that it grew on artificial seawater medium supplemented with 2 μ M HgCl2 and that it volatilized oxidized Hg(II) to elemental Hg(0), but did not confirm the presence of a merA gene in this isolate. In a recent study with the same soils as used in the present study (Oregaard et al., 2007), bacterial isolates showing high 16S rRNA gene similarity to Azospirillum, Bradyrhizobium, Rhizobium, Sphingomonas and Rhizobiales were isolated on dilute agar media amended with HgCl2 and could subsequently be cultured in liquid 10% tryptic soy broth (TSB) supplemented with 20–50 μ M HgCl2, except the Rhizobiales-like isolate, which resisted growth in 10% TSB. In the present study, manually designed primers were obtained by examination of multiple alignments of α-Proteobacteria genomic merA sequences. merA PCR amplicons (approximately 812 bp) were obtained from all five different isolates mentioned above. The partial merA sequences were identical in the Bradyrhizobium, Rhizobium and Sphingomonas isolates, whereas the Rhizobiales-like merA gene was somewhat different (the Azospirillum merA fragment was not sequenced). The three different isolates, Rhizobium, Sphingomonas and Bradyrhizobium, were isolated from surface, intermediate and deep sub-surface soils. We think this is a strong indication that horizontal gene transfer plays a role in dissemination of adaptive traits between related bacteria. The genome-sequenced α-Proteobacteria, containing putative merA, are all originating from marine environments, except the Xanthobacter autotrophicus Py2, originating from sludge. None of these bacteria have physiologically been characterized with regards to mercury resistance. The deep branching of the merA of the isolates compared to the reference sequences (data not shown) can be due to very different environments, from which they originate.

In the present study, 62 different OTUs were found in soils B, C and D (Figure 3) after sequencing 143 clones obtained with β/γ-specific merA primers. In a study by Barkays group, mercury-contaminated anoxic sediments from Berry's Creek, a tributary of the Hackensack river (NJ, USA), were analyzed for the presence of merA of the β/γ type (Ni Chadhain et al., 2006). The reverse primer was the same as used in the present study (#54), whereas the forward primer was shifted three nucleotides downstream, resulting in 285 bp long amplicons. The authors found a high diversity, and with a 5% discrimination threshold between phylotypes, as used in this study, the Chao1 richness estimator gives around 92 phylotypes. The 68 OTUs in the present study were obtained from three different soils depths. If these are all considered as originating from the same site, that is 163 merA sequences found within 1-m depth from surface, the Chao1 richness estimator gives 143 phylotypes with a 5% threshold (Table 3), indicating that these soils are more diverse with regards to merA than the anoxic river sediments at Berry's Creek.

When comparing the merA sequences found in the present study, with the sequences from Berry's Creek sediment, and using a 5% discrimination threshold, only three OTUs contain similar sequences (OTU-01 and WBC-054; OTU3 and SBC-204+WBC-007; OTU10 and SBC-116). This very low similarity between these sites could be due to the anoxic nature of the freshwater sediments, with selection of different merA genes, but this is mere speculation and more studies are required to elucidate this difference.

Interestingly, three of the clades (I–III) found in the study cited above, grouped closely to Firmicutes-type merA of S. aureus pI258 and Bacillus RC607 (Ni Chadhain et al., 2006) (Figure 5). We find it probable that these sequences are of α-Proteobacteria origin, due to high blastp similarity to α-type merA and by high bootstrap values (95%) at the branch point of α-type and Firmicutes-type merA in the neighbour-joining tree (Figure 5). Clades I–III (Ni Chadhain et al., 2006) and the genomic α merA seem like a monophyletic group, distinct from the Firmicutes. The phylogenetic relationship between α-type SBC clones and α-type isolates sequenced in this study cannot be determined, since their sequences do not overlap. An attempt to amplify merA of the α isolates in the present study with primers Al-Fw and #54 was unsuccessful.

The Actinobacteria merA clones showed that the functional diversity was not unique to the β/γ-Proteobacteria, but also occurring within the Actinobacteria phylum. With only 19 sequences, 10 different OTUs were found, and all of these were more than 5% different at the DNA level to all reference merA genes retrieved from GenBank. Seven of the OTUs were most closely related to Arthrobacter-like genes. The functional importance of these bacteria in the soils with regards to mercury is unclear, since it is thinkable that the primers are biased. When using the merA primers targeting Actinobacteria, we found that amplification of merA from Is-BDOE3, an Arthrobacter-like strain from Oregaard et al. (2007), resulted in a slightly larger merA fragment than obtained from Is-BDOE1. Sequencing revealed that the amplicon was most similar to a ferrochelatase-encoding gene (data not shown). A second PCR and sequencing confirmed this result. The ferrochelatase gene was quite different from all the Actinobacteria merA clones. We find it most probably that isolate Is-BDOE3 contains a merA gene, since it grows well on 10% tryptic soy broth agar supplemented with 25 μ M HgCl2. The primers targeting merA of Actinobacteria are less than optimal due to false-negative amplification (inability to amplify merA of isolate Is-BDOE3) and false-positive amplification (ability to amplify merA of β/γ-Proteobacteria, the Ba clones; see Figures 2 and 3). The fact that the Arthrobacter-like isolate did not result in a proper merA amplicon also hints that the seven OTUs clustering around the Arthrobacter merA genes might be of different origin than Arthrobacter.

Firmicutes-type merA was not amplified from control soil nor mercury stimulated soil. Initial testing of the Firmicutes merA-specific primers on three different merA genes of Firmicutes origin gave proper size amplicons (data not shown). It is therefore most likely that potential Firmicutes bacteria of these soils were not mercury-resistant, or that their DNA was extracted insufficiently, thus leading to no amplification. Insufficient DNA extraction could be due to either low abundance of Firmicutes-type bacteria or due to inefficient lysis/extraction.

Searching the NCBI databases for putative mercuric reductase proteins led to the definition of eight (monophyletic) groups, consisting of Firmicutes, Actinobacteria, α-, β/γ- and δ-Proteobacteria, Bacteroidetes, Deinococcus-Thermus and Archaea (Figure 1). The criterion for accepting putative merA genes as likely mercuric reductase proteins was the occurrence of essential cysteine and tyrosine residues (Table 2). Apart from differences in the actual sequence, the merA genes vary in length (according to whether they contain zero, one or two Hg-binding domains at the N terminus) and GC content. The putative mercuric reductase proteins of α- and δ-Proteobacteria, along with the Deinococcus-Thermus all lack the heavy metal associated (HMA)-domain, found in β/γ and Firmicutes-type merA. The HMA domain is not essential for resistance to mercury (Moore and Walsh, 1989; Ledwidge et al., 2005), and both Actinobacteria and Archaea isolates resistant to mercury do not have this region.

The genome-sequenced Salinibacter ruber DSM 13855 (Mongodin et al., 2005) belonging to the Bacteroidetes phylum has two genes annotated as mercuric reductases (YP_446491 and YP_444230), but both lack the C-terminal cysteines, and also the tyrosine at position 264 and are thus most unlikely merA proteins. The outgroup in Figure 1 consisted of wrongly annotated Salinibacter merA gene YP_446491 along with three δ-Proteobacteria sequences (NP_952368, YP_628559 and YP_010258) annotated as merA, but without the required amino acids described above.

Interestingly, the newly genome-sequenced marine isolate Leeuwenhoekiella blandensis (Pinhassi et al., 2006) belonging to the Bacteroidetes seems to contain a merA gene, although the particular gene is not annotated as such (ZP_01060916). The Leeuwenhoekiella putative merA also contains a heavy metal binding domain at the N-termini, an important feature of the mercury-resistance mechanism (Ledwidge et al., 2005). The Leeuwenhoekiella-putative merA gene is to the best of our knowledge the first likely mercuric reductase, found in the Bacteroidetes phylum. We have found that the L. blandensis isolate grows on marine broth agar supplemented with 30 μ M HgCl2, but not with 40 μ M HgCl2 (unpublished results, Oregaard and Sørensen) corroborating that this marine Bacteroidetes isolate is mercury-resistant.

In conclusion, we have shown that merA genes from mercury-contaminated soil environments are very diverse. The three soil communities seemed to share only few OTUs, and most clones were less than 95% identical at the DNA level to merA sequences deposited at GenBank. Several merA genes of isolates obtained from the soils used in the present study were partially sequenced, and many were less than 95% identical at the amino-acid level to GenBank protein sequences. The α-Proteobacteria merA genes are the first evidence of a similar resistance mechanism in this sub-phylum, as observed in β/γ-Proteobacteria. However, nothing is known about the mer operon structure of the α-Proteobacteria isolates, and whether these genes are chromosomally or plasmid encoded. Future work will focus on these issues. With many new genomes added to the NCBI database every month, analysis of the genomes may allow design of primers targeting merA of hard to culture bacteria from phyla with only few cultivable representatives, thus allowing assessment of their importance in mercury contaminated environments.