Introduction

Polychlorinated biphenyls (PCBs) are a group of molecules which were widely used in a variety of industrial applications prior to their total ban in 1977 [32]. PCBs are chemically and thermally stable and usually recalcitrant to biodegradation. Thus, they tend to accumulate in soils and aquatic sediments, where their toxicity and persistence as pollutants makes them a threat to the ecosystem and to human health [22].

Bioremediation is thought to be a promising approach for restoring PCB-contaminated sites [20]. Anaerobic microbial communities found in sediments can reductively dechlorinate PCBs, such that fewer chlorinated ortho- and ortho- plus para-chlorinated congeners accumulate. Subsequently, aerobic organisms can attack the major congeners resulting from anaerobic dechlorination [13]. With the exception of monochlorinated molecules, chlorinated biphenyls are not known to be growth substrates for bacteria; rather, co-metabolism induced by biphenyl, or other compounds, is required to achieve growth [26]. Indeed, the enzymatic pathways involved in bacterial metabolism of PCBs are also involved in biphenyl metabolism.

Much of our knowledge of biphenyl metabolism in bacteria is derived from studies of Rhodococcus jostii RHA1 and Burkholderia xenovorans LB400. Both of these strains were isolated from sites contaminated with chlorinated hydrocarbons, and the enzymatic pathways underlying biphenyl metabolism in each organism are similar [5, 22, 30]. Briefly, biphenyl is first metabolized to benzoate and 2-hydroxypenta-2,4-dienoate by the sequential action of a 4-component dioxygenase (e.g., BphAa, Ab, Ac, Ad), and the enzymes BphB, BphC, and BphD. Benzoate is then degraded by enzymes encoded by the ben and cat (catechol) genes while 2-hydroxypenta-2,4-dienoate is further metabolized to pyruvate and acetyl-CoA by BphE, BphF, and BphG (reviewed in [22]). Interestingly, in R. jostii RHA1, multiple isozymes exist for each step in this bifurcated pathway; sequence analysis, in some cases supported by experimental data, identified three complete (or partial) 4-component dioxygenases and numerous putative homologs of bphC (13), bphD (eight), bphE, (eight) bphF (seven), and bphG (seven). It has been suggested that this extensive complement of isozymes may contribute to this organism’s aggressive catabolism of complex PCB mixtures [12].

Using microarrays, Goncalves et al. [12] compared transcript abundance in R. jostii RHA1 cells grown in media supplemented with either biphenyl or pyruvate as a sole carbon source. During growth on biphenyl, the bph, ben, and cat genes were significantly upregulated, in some cases by as much as 10,000-fold. However, while some of the predicted isozymes in the biphenyl pathway were upregulated, the majority were not. Notably, some of these apparently unresponsive genes appear to be expressed at constitutively high levels. Similar transcriptomic and proteomic studies of B. xenovorans LB400 have also detected the coordinated upregulation of bph genes in response to biphenyl. However, in addition to the induction of catabolic genes, PCBs and biphenyl were also observed to elicit a stress response, upregulating the chaperones DnaK and GroEL as well as the AhpC [1, 9, 10].

Apart from R. jostii RHA1, other bacteria from the genus Rhodococcus have been found to metabolize a wide variety of environmental pollutants including alkanes, aromatic compounds, and halogenated hydrocarbons [3, 33]. Rhodococcus aetherivorans I24 was isolated from a toluene-contaminated aquifer by Chartrain et al. [7]. It is able to metabolize naphthalene and toluene as sole carbon-energy sources and to convert indene to a variety of indandiols, some of which can be used as precursors of the HIV protease inhibitor indinavir sulfate [23], but it has not been tested yet for its response to PCBs.

A common feature of the studies cited above is that they focus on measuring changes in gene expression brought about by specific substrate molecule (e.g., biphenyl) under laboratory conditions (i.e., in culture media). To date, high throughput techniques have not been applied to assess the effect of PCBs on bacteria in complex environments, such as polluted soils or sediments. Several challenges associated with the presence of humic acids, organic contaminants, and metals, which may interfere with the hybridization [36], and with the difficulty of identifying which factors caused the expression of specific genes, must be addressed; in this context, studies employing pure cultures, polluted and unpolluted soil, or sediment matrices can be very useful. Similarly, there have been few (if any) investigations of how “naïve” organisms such as R. aetherivorans I24—those which have not been extensively exposed to PCBs—respond to these pollutants.

This study employs microarrays to assess gene expression in R. aetherivorans I24 during growth in a variety of sediment slurries and culture media. The primary objective of this work is to explore gene expression in contaminated sediments and contrast this with what is observed in defined medium supplemented with PCB mixtures. In addition, as R. aetherivorans I24 was not isolated from a site contaminated with polychlorinated hydrocarbons, its response to PCB/biphenyl exposure may offer insight into the evolution of xenobiotic degrading bacteria.

Methods

Experimental Conditions

R. aetherivorans I24 was grown overnight in 10 ml LB before inoculating 1 L of minimal medium (MM) [19] supplemented with glucose (3 g L−1) and biphenyl (0.69 mg L−1, i.e., 4.5 μM). Cells were incubated overnight at 30°C on a rotary shaker before repeating biphenyl supplementation and incubating for a further 3 h. Prior to treatment, the cells were washed twice in 1 L PBS and then resuspended in MM containing 3 g L−1 glucose (OD600 = 0.5). For treatments, 60 ml aliquots of this culture were then dispensed into 500-ml flasks. The treatments were as follows:

  1. 1.

    glucose only (GLU)

  2. 2.

    0.69 mg L-1 of BP

  3. 3.

    5 mg L-1 of the PCB mixture Aroclor 1254 (PCB)

  4. 4.

    a contaminated sediment (C)

  5. 5.

    a non-contaminated sediment (NC)

All samples were incubated for 48 h at 30°C on a rotary shaker (200 rpm). RNA was then isolated from 20 ml volumes which were harvested by centrifugation. For all conditions, three biological replicates were performed, each of which was hybridized in duplicate. Since the variation among these hybridization replicates was minimal (data not shown), the average of the two was taken for all data analyses.

For the sediment treatments, 9 g of contaminated or non-contaminated sediment was added to each culture flask. These sediment samples were collected in 2000 from the Wijnhaven area of Rotterdam harbor [11]. This site has 30 years history of contamination by a variety of pollutants, PCBs included. The amounts of PCBs, heavy metals, polycyclic aromatic hydrocarbons, and pesticides in both sediment samples are presented in Table 1. In a previous work, it was shown that in the contaminated sediment, the average bioavailability of PCBs congeners was 42% of the total content, with a maximum of 67% for PCB 52 and a minimum of 23% for PCB138 [24]. All sediment samples were autoclaved for 30 min at 120°C prior to the experiments.

Table 1 Total content of PCBs, PAHs, DDT, and its breakdown products DDD and DDE in the non-contaminated and contaminated sediments
Full size table

Aroclor 1254 (ULTRA Scientific, Inc) is the commercially available PCB mixture which most closely resembles the congener pattern found in the contaminated sediment. MM was spiked with PCBs to a concentration of 5 mg L−1 using acetone as a carrier solvent. The acetone was allowed to evaporate completely prior to beginning the experiment.

Preliminary experiments (data not shown) indicated that the conditions described and the presence of glucose in the minimal medium allowed in all treatments the growth of the strain and sufficient RNA yields for microarray analyses.

RNA Isolation

RNA was extracted from sediment treatments with the QBIOgen FastRNA Pro Soil Direct Kit as per the manufacturer’s recommendations (Morgan Irvine, CA). For nonsediment treatments, the QBIOgen Pro Blue Kit was used. Following isolation, RNA samples were subjected to a double DNAse treatment with Qiagen’s RNase-free DNase and RNeasy mini kit (Valencia, CA). All RNA samples were stored at −80°C prior to microarray analyses, and their quality was confirmed by Bioanalyzer (Agilent, Santa Clara, CA).

DNA Microarrays

Array manufacturing, cDNA labeling, hybridization, and scanning were carried out by Roche NimbleGen, Inc (Madison, WI). A total of 3,524 genes were represented on the arrays; of these, 41 and 176 are found on the R. aetherivorans I24 plasmids pRA2 and pRA3, respectively. Based on the current annotation of the R. aetherivorans I24 genome, this accounts for 64% of the chromosomal genes, 55% from pRA2, and 54% from pRA3 (Cahill et al. in preparation). On average, there were ∼3 probes per gene with a range of 1–43.

Raw probe-level intensities were initially processed using Nimblescan (Roche NimbleGen, Inc, Madison, WI). Data were subjected to quantile normalization and gene expression summary values were then calculated using the robust multiarray average algorithm [4, 15]. Statistical analysis of gene expression data was carried out with Gene Spring v.10 (Agilent, Santa Clara CA). Data were log-transformed and percentile filtered with a lower cutoff of 20 (3,366 of 3,524 genes passed filtering). Significant differences in gene expression between each treatment and the glucose-only reference were identified using volcano plot comparisons (T test unpaired, Benjamin–Hochberg multiple testing correction, p value cutoff 0.05, fold change cutoff 2.0). Fold changes were expressed as the ratio (treatment/reference) of averaged normalized expression values. All array data generated in this study have been deposited in the NCBI’s Gene Expression Omnibus (GEO; http://www.ncbi.nlm.nih.gov/geo/) under the accession number GSE17033.

Sequence Analysis

To assign genes to cluster of orthologous groups of proteins (COG) functional categories, BLASTP was used to identify the reciprocal best hit for each R. aetherivorans I24 gene among the sequences in eggNOG (database downloaded Mar. 25, 2009) [16]. R. aetherivorans I24 genes which could be assigned to COGs by this method were then mapped to functional categories using the legend provided by the COG database (http://www.ncbi.nlm.nih.gov/COG/grace/uni.html) [31].

Similarly, reciprocal best hits were used to identify putative R. aetherivorans I24 orthologs of isozymes from the biphenyl pathway of R. jostii RHA1 (as identified by Goncalves et al. [12]). The R. jostii RHA1 protein sequences were downloaded from http://www.rhodococcus.ca and are accessible in the NCBI under the accession numbers NC_008268.1, NC_008269.1, NC_008270.1, and NC_008271.1.

Results

Patterns in Gene Expression

Comparison of each of the four experimental treatments to the glucose-only condition identified a total of 626 genes which were significantly up- or downregulated (A full list of upregulated genes is included as part of the Supplementary material; data are summarized in Table 2). The greatest number of differentially expressed genes was found in the contaminated sediment treatment (383 total) whereas expression in non-contaminated sediment was most similar to that of the glucose-only reference (111 total). Among the sediment treatments, there was considerable overlap of the genes which were upregulated. Specifically, 37 of the 44 genes upregulated in the non-contaminated sediment were also upregulated in the contaminated sediment. Similarly, PCBs, regardless of whether in medium or contaminated sediments, elicited the upregulation of 100 genes. These common genes constitute a large proportion of the total number of genes upregulated in each PCB-containing condition (129 and 207 genes, respectively). In contrast, there were very few upregulated genes in common between the biphenyl and PCB treatments in media (12 of 184 and 129, respectively).

Table 2 Summary of up- and downregulated genes in each condition, relative to the glucose-only treatment
Full size table

An examination of upregulated genes in the context of their membership in orthologous groups (i.e., COGs) did not reveal any obvious biases (Table 3). Under all conditions, similar percentages of upregulated genes fell into the categories of “Cellular processes and signaling,” “Metabolism,” or are “Poorly characterized.” The only apparent trend was that in both sediment and media among the genes upregulated by PCBs, a greater proportion was involved in “Information storage and processing” than in the other categories. More specifically, PCBs elicited the upregulation of a large number of genes involved in “Translation, and ribosomal structure and biogenesis.”

Table 3 Functional classification of up-regulated genes based on COG membership
Full size table

Sediment Gene Expression

Depicting the upregulated genes in a Venn diagram consisting of four sets allowed us to refine our pairwise comparisons (Fig. 1). While a common set of 37 genes were upregulated in both sediment conditions, only eight of these were found exclusively in these conditions. In fact, the majority of the 37 common genes, 23 in total, were also upregulated by PCBs in medium (Table 4). This latter set included two paralogs of the chaperone groEL and a universal stress protein. In addition, a thioredoxin as well as the alkyl hydroperoxide reductase protein C (ahpC) were upregulated in both sediment conditions and the PCB-containing medium. Both of these proteins are known to be involved in the response to oxidative stress [25, 29].

Figure 1
figure 1

A Venn diagram depicting the relationships (i.e., overlaps) between the sets of genes upregulated in each treatment, relative to the glucose-only reference. (BP biphenyl in medium; PCB PCBs in medium; C contaminated sediment, NC non-contaminated sediment). Total numbers of upregulated genes per treatment are reported in parenthesis

Full size image
Table 4 Fold changes of selected genes
Full size table

PCB Gene Expression

The 100 upregulated genes which were common to the PCB-containing medium and contaminated sediment conditions comprise two major groups. The minority was the 23 genes which were also upregulated in non-contaminated sediment, as discussed in the preceding section. A second set of 67 upregulated genes was specific to PCBs (Table 4). This group included genes encoding additional proteins related to oxidative stress response, namely a superoxide dismutase (sodA), a catalase/peroxidase (katG), and a thioredoxin reductase. This PCB-specific set also included several ribosomal proteins (six total) and a ribosomal recycling factor.

A relatively small proportion of the genes upregulated in PCB-containing medium were specific to that condition (26 of 129), and the predicted functions of these genes are diverse. In contrast, almost one half of the genes upregulated in contaminated sediment were specific to this treatment (95 of 207). Among this set were numerous additional ribosomal proteins (12 total), alkyl hydroperoxide reductase protein F (ahpF), and several catabolic genes (e.g., Raet_200057—a putative catechol 2,3-dioxygenase).

Biphenyl Gene Expression

R. aetherivorans I24 can use biphenyl as its sole carbon and energy source. To account for this, the complete genome sequence of R. aetherivorans I24 was examined to identify orthologs of enzymes in the biphenyl pathway of the closely related bacterium R. jostii RHA1 (Table 5). R. aetherivorans I24 was found to possess all of the enzymes required for a complete biphenyl pathway, as well as a complete collection of ben genes. Specifically, the linear plasmid pRA3 encodes orthologs of bphAaAbAcAd, bphB1, and bphC1 in the same arrangement as the corresponding genes in R. jostii RHA1. In addition, orthologs of etbD1 and bphE3, bphG3, and bphF3 are also present on pRA3; the arrangement of the latter three is syntenic with their orthologs in R. jostii RHA1. While R. aetherivorans I24 does possess a core set of enzymes required for biphenyl metabolism, it lacks orthologs of many of the isozymes identified by Goncalves et al. [12].

Table 5 Isozymes from the biphenyl pathway of R. jostii RHA1 from Goncalves et al. [12] and their putative orthologs in R. aetherivorans I24.
Full size table

Of each of the treatments in this study, biphenyl elicited the most distinctive response. Of 184 genes upregulated in the biphenyl treatment, 166 were specifically induced by biphenyl. However, none of the orthologs of key enzymes in the biphenyl pathway, for example bphAaAbAcAd, were featured in this set, nor were they upregulated under any of the other conditions. In fact, only two orthologs of isoyzmes noted by Goncalves et al. [12] were upregulated by the biphenyl treatment (Raet_101733 and Raet_100698). In addition, biphenyl upregulated components of numerous transporters as well as genes with putative functions that suggest a role in aromatic catabolism. These include a ring hydroxylating dioxygenase (Raet_101087), a carveol dehydrogenase (Raet_101089), and a phenol hydrolase (Raet_101551).

Discussion

To the best of our knowledge, this is the first study in which microarrays have been applied to assess the global gene expression of a bacterial strain in PCB-contaminated sediments. To distinguish sediment-induced effects from those specific to PCBs, we included two additional treatments: non-contaminated sediment and defined medium supplemented with PCBs. Our aim was to assess how a bacterial strain of ecological importance such as R. aetherivorans I24 reacts to pollutants in the sediment matrix and to identify genes of potential ecological interest, e.g., for the development of biosensors. After some preliminary experiments (data not shown), we decided to pre-induce the cultures to be used in the experiments with biphenyl (0.69 mg L−1), and resuspend them in MM added with glucose (3 g L−1). The choice of biphenyl was based on the fact that co-metabolism or pre-induction with biphenyl is a necessary condition for PCBs metabolism in other degrading strains such as R. jostii RHA1 B. xenovorans LB400 [8, 10]. On the other hand, the addition of glucose in all treatments ensured an active growth of cells and at the same time a sufficient RNA yield for microarray analyses. We think this choice is also representative of the ecology of soils and sediments, where easily available carbon sources such as glucose are usually present [27].

As part of this study, we also assessed gene expression in defined medium supplemented with biphenyl. This simplified treatment was intended to aid in the interpretation of the more complex treatments involving sediments and/or PCB mixtures. As well, we hoped to gain insight into the catabolism of this substrate by a bacterium which had not been extensively exposed to polychlorinated hydrocarbons. Our analysis of the complete genome sequence of R. aetherivorans I24 showed that while this species does possess the core enzymes required for biphenyl metabolism, it lacks orthologs of the majority of the isozymes in the biphenyl pathway of R. jostii RHA1 (Table 5). Given that multiple isozymes are thought to be advantageous for the catabolism of mixtures of related compounds (e.g., PCBs) [2], it may be that the extensive collection of isozymes in this latter species reflects a selective pressure imposed by the environment from which it was first isolated.

Even though R. aetherivorans I24 possesses all of the core enzymes of the biphenyl pathway, most of which were represented on the arrays (19 of 24), very few of these genes were upregulated in any of the treatments in this study. Of the R. aetherivorans I24 genes which are likely to be orthologs of the isozymes identified by Goncalves et al. [12], only isozymes of bphC and bphD were upregulated in the biphenyl treatment. In addition to these, we also observed upregulation of several genes which, while not obvious orthologs of bph genes, are predicted to be involved in aromatic catabolism. Thus, while some of the genes upregulated in biphenyl imply that the cells are responding to this substrate, we did not observe the coordinated upregulation of all of the core enzymes in the pathway, along with numerous isozymes, as seen under comparable conditions in R. jostii RHA1. One explanation for this may be the presence of glucose in all of the treatments used here. Taken together, the observations of Goncalves et al. [12] and Sakai et al. [28] suggest that in R. jostii RHA1, the expression of the bphE1F1G1 operon is repressed in rich medium. If the same were true for all of the bph genes in R. aetherivorans I24, then we might not expect significant upregulation of these genes in any of the treatments. An alternative possibility is that the response observed in R. jostii RHA1 is, like its extensive complement of isozymes, a product of selective pressures operating in the environment from which it was isolated. This scenario implies that in a common ancestor of the rhodococci, enzymes in the bph pathway were not primarily involved in biphenyl metabolism; rather, selective pressures specific to the R. jostii RHA1 lineage led to the recruitment of these enzymes for the metabolism of polychlorinated hydrocarbons.

Sequence analysis, taken together with our microarray observations, thus revealed that although R. aetherivorans I24 encodes all of the necessary genes for biphenyl/PCB metabolism, these genes do not appear to be subject to the same coordinated regulation that is characteristic of bacteria which have been extensively exposed to polychlorinated hydrocarbons.

Our measurements of gene expression in PCB-containing medium broadly agree with observations of the proteome of B. xenovorans LB400 under similar conditions. In common with Agullo et al. [1], we note that exposure to PCBs upregulates chaperones (e.g., groEL) and enolase. However, while we observed that PCBs induced numerous genes involved in the response to oxidative stress, Agullo et al. [1] only identified a single gene (ahpC), and it was induced in response to biphenyl, rather to PCBs. In contrast, our observations have little in common with what was found in a microarray-based investigation of PCB-induced changes in gene expression in B. xenovorans LB400 [21].

One mechanism by which PCBs are toxic to bacteria is through the production of reactive oxygen species [8]. Accordingly, PCBs, in both sediment and medium, increased the expression of sodA, katG, ahpC, the chaperone groEL, thioredoxin, and thioredoxin reductase. All of these genes are known to be induced in response to oxidative stress and/or PCBs [1, 8, 14, 34, 35]. These findings suggest that the response of R. aetherivorans I24 to PCBs, regardless of whether in sediment or medium, is directed at ameliorating the effect of reactive oxygen species, rather than catabolism. Furthermore, the fact that PCBs induced a very similar transcriptional response of the strain in both sediment and medium suggest that the interference of sediment and soil components such as humic acids on the gene expression patterns and hybridization process [35] is minimal. This needs to be confirmed but sounds promising for future studies aimed at comparing the transcriptional response of bacterial strains in media of different complexity.

Exposure of cells to contaminated sediments or PCBs in pure cultures induced a significantly higher ribosome synthesis. No previous findings on this induction in bacteria by PCB are available; only one study found a higher induction of these genes by a microarray study of gene expression in mammalian cells exposed to PCBs [6].

Of the 408 genes which were upregulated among all treatments, only eight were restricted to the sediment treatments. Among the 37 genes common to C and NC treatments, 23 are also found in PCB-containing medium. This common set includes chaperones and a universal stress protein. Thus, it may be that even in non-contaminated sediment, most of the upregulated genes are part of a general stress response that is also activated by PCBs, regardless of whether they are in sediment or medium.

The genes upregulated in contaminated sediment include those common to the non-contaminated sediment and PCB-containing medium (discussed above), as well as those specific to the conditions which included PCBs. Among this set are numerous genes that may play a role in the cell’s response to the oxidative stress caused by the PCBs. There were also a considerable number of genes (95) which were upregulated only in the contaminated sediment. The sediment samples used in this study are known to be contaminated with other pollutants (Table 1) in addition to PCBs. These condition-specific genes may reflect the cell’s response to these additional substrates.

Our assessment of global gene expression in R. aetherivorans I24 revealed that response in pure culture can be to a certain extent a predictor of the ecological behavior of the strain in presence of contaminated sediments. The high induction of stress genes and, on the other hand, the lower or absent induction of genes related to biphenyl metabolism in cells exposed to PCBs in pure culture or in contaminated sediment may indicate that the strain, even if pre-induced with biphenyl and containing PCBs degradation genes, is actually more involved in counterbalancing the stress caused by the presence of pollutants rather than degrading them, at least in the 48 h incubation used here and even in the presence of glucose. This can hold important consequences for biosensor-based risk assessment and for the bacterial-aided bioremediation of contaminated sites. Monitoring of the induction of stress genes in contaminated samples may in fact represent a way to identify potential constraints to soil bioremediation practices, while the in situ application of biosensors specific for pollutants such as PCBs in soils and sediments may be limited by a low or absent induction of genes involved in the degradation if those genes are used as promoters [17, 18]. On the other hand, some of the genes identified here could be used as indicators not of the effects of a single class of specific molecules such as PCBs but rather to abiotic stresses caused by chemical pollution of different origins. Finally, specific experiments aimed at understanding the optimal exposure time and conditions for the expression of catabolic genes would be very useful for bioremediation purposes.