|Home | About | Journals | Submit | Contact Us | Français|
Desulfitobacterium strains have the ability to dechlorinate halogenated compounds under anaerobic conditions by dehalorespiration. The complete genome of the tetrachloroethene (PCE)-dechlorinating strain Desulfitobacterium hafniense Y51 is a 5,727,534-bp circular chromosome harboring 5,060 predicted protein coding sequences. This genome contains only two reductive dehalogenase genes, a lower number than reported in most other dehalorespiring strains. More than 50 members of the dimethyl sulfoxide reductase superfamily and 30 paralogs of the flavoprotein subunit of the fumarate reductase are encoded as well. A remarkable feature of the genome is the large number of O-demethylase paralogs, which allow utilization of lignin-derived phenyl methyl ethers as electron donors. The large genome reveals a more versatile microorganism that can utilize a larger set of specialized electron donors and acceptors than previously thought. This is in sharp contrast to the PCE-dechlorinating strain Dehalococcoides ethenogenes 195, which has a relatively small genome with a narrow metabolic repertoire. A genomic comparison of these two very different strains allowed us to narrow down the potential candidates implicated in the dechlorination process. Our results provide further impetus to the use of desulfitobacteria as tools for bioremediation.
Halogenated organic compounds are released into the environment from natural and anthropogenic sources. Many anthropogenic halogenated chemicals, like chlorinated haloalkenes (7, 10, 46), benzenes (1), and dioxins (5), are of particular concern due to their toxicity to humans and other forms of life. This toxicity is often paired with high recalcitrance to degradation, especially in anaerobic environments, leading to persistent contamination.
Anaerobic environments are frequently characterized by limited availability of electron acceptors. Theoretical calculations have shown that coupling the reduction of many halogenated organic compounds to the oxidation of suitable substrates is a way to harness energy (46). As determined two decades ago, this source of energy is utilized by the microbial community. The oxidation of available electron donors coupled to the reduction of halogenated organic compounds while energy is conserved is called dehalorespiration (7, 10, 46). Dehalorespiring strains have been isolated independently from contaminated sites around the world. The two most prominent genera resulting from these isolation efforts are Dehalococcoides (29) and Desulfitobacterium (51), and various strains of these genera are used as model systems to study dehalorespiration (8, 11, 51).
Dehalococcoides ethenogenes 195 is one of the few strains isolated to date which can dechlorinate tetrachloroethene (PCE) to ethene (29). D. ethenogenes 195 can use only hydrogen as an electron donor and chlorinated compounds as electron acceptors (29).
Desulfitobacterium strains are also known to dechlorinate a wide variety of substrates, including halophenolic compounds and chloroalkenes (7, 10, 46). Although several strains can use PCE or trichloroethene (TCE) as an electron acceptor, no Desulfitobacterium strain isolated so far completely dechlorinates these compounds to ethene (7, 14, 48). In contrast to Dehalococcoides strains, Desulfitobacterium strains can utilize electron acceptors other than chlorinated compounds. Several strains that are capable of deiodination (21) and reduction of As(V), Fe(III), Se(VI), Mn(IV), and a variety of oxidized sulfur species (37) have been isolated, although currently little is known about how widespread these capabilities are in this genus.
Since Desulfitobacterium and Dehalococcoides strains are frequently encountered at contaminated sites, these genera have attracted considerable attention for use as bioremediation agents. The use of these strains in real life, however, is hampered by the lack of information about how the dehalogenation process is embedded in the general metabolism of the organisms and the conditions that allow these microorganisms to proliferate in the environment.
Here we report the first complete genomic sequence of thegenus Desulfitobacterium. Desulfitobacterium hafniense Y51 (formerly Desulfitobacterium sp. strain Y51) was isolated from a contaminated site in Japan based on its ability to efficiently dechlorinate PCE even at its highest water solubility (48). The recent publication of the D. ethenogenes 195 genomic sequence (43) allowed us to compare the two sequences and highlight the similarities and differences between the organisms.
D. hafniense Y51 was cultured as described previously (48). The genome was sequenced using the whole-genome shotgun method (12). Genomic DNA was isolated using a standard phenol-chloroform extraction-based protocol and was mechanically sheared. Two genomic DNA libraries with average insert sizes of 2 kb and 8 kb were constructed in the pUC118 vector (53). Sequencing was performed using an ABI Prism ABI3730 DNA analyzer (Applied Biosystems). The sequences were base called and assembled using Phred/Phrap/Consed (11, 15). Gaps were closed by primer walking for gap-spanning plasmid clones, direct sequencing of PCR products, and nested PCR-assisted contig extension. Misassemblies and frameshifts were corrected by verifying the positions of repeated DNA regions (rRNA gene, repetitive sequences) or ambiguous DNA regions using PCR. The final genome sequence is based on 98,319 reads. The error rate is 0.04 base per 10 kb as calculated using Consed.
rRNA-encoding genomic regions were located by a BLASTN homology search against the 16S rRNA sequence of D. hafniense Y51 and the 23S and 5S rRNA sequences of Thermoanaerobacter tengcongensis (2). tRNA-encoding regions were predicted by tRNA scan SE (25).
Protein coding sequences (CDS) were predicted by glimmer (39) trained on the whole genome sequence using an open reading frame cutoff value of 240 bp. In order to identify false-positive hits, we compared all glimmer predictions with entries in the Swiss-Prot database and with all coding sequences of completely sequenced organisms (as of 9 July 2005) using BLASTP (e-value, <1e-10). Conflicting coding sequences were removed from the coding sequence list. The remainder of the genome was screened for the presence of CDSs by a BLASTX homology search against CDSs of Clostridium acetobutylicum ATCC 824, Bacillus subtilis subsp. subtilis 168, and Escherichia coli K-12. This second step allowed us to identify CDSs missed by glimmer, either because they were shorter than 240 bp or because the signature was not recognized as a coding sequence. The homologous regions identified were extended to CDSs. The start codon of each CDS was manually revised when it was necessary.
Functional annotation of the proteome was carried out by a BLASTP homology search against the NCBI Clusters of Orthologous Groups (COG) database (ftp://ftp.ncbi.nih.gov/pub/COG/old/) (50). Subcellular localization of the coding sequences was predicted by using PSORTb (13).
The homolog of each D. hafniense Y51 coding sequence that was most similar to any coding sequence of a completely sequenced genome (as of July 2005) was determined by a BLASTP search using a cutoff value of 1e-4. A small self-written Perl script was used to extract the metadata containing the strain information associated with the highest-similarity hits.
The predicted coding sequences of D. hafniense Y51 and D. ethenogenes 195 were compared to each other by BLASTP using a cutoff value of 1e-4. Reciprocal highest levels of similarity were used to identify a set of 751 orthologous coding sequences. The 751 D. hafniense Y51 coding sequences were compared to a sample containing all coding sequences of completely sequenced organisms (as of July 2005), including that of D. ethenogenes 195. Conversely, the 751 D. ethenogenes 195 coding sequences were compared to a sample containing all coding sequences of completely sequenced organisms (as of July 2005), including that of D. hafniense Y51 but not that of D. ethenogenes 195. A small self-written Perl script was used to extract the D. hafniense Y51 and D. ethenogenes 195 coding sequences (and the metadata associated with them) that exhibited the highest levels of similarity to D. ethenogenes 195 and D. hafniense Y51 coding sequences, respectively.
The complete D. hafniense Y51 genome sequence has been deposited in the DDBJ database under accession no. AP008230.
The D. hafniense Y51 genome is a single circular 5,727,534-bp chromosome with 5,060 predicted CDSs (Table (Table11 and Fig. Fig.1).1). This strain harbors no plasmids. The replication origin of the chromosome was defined using the position of the transition point of GC skew (Fig. (Fig.2)2) (24, 30) and the presence of the characteristic replication protein encoded by dnaA. The GC skew analysis also clearly identified the chromosomal arms. In most prokaryotic organisms the sizes of the two chromosomal arms are usually similar, but in D. hafninense Y51 one arm is approximately twice as long as the other. To our knowledge, this is the most extreme case in any completely sequenced microorganism with a circular chromosome to date. The G+C content is 47.4%, and the overall variation of the G+C content in the genome is low (Fig. (Fig.2).2). Local changes in the coding density and the clustered presence of phage-related genes were identified, suggesting that multiple prophages in various states of decay are present in the genome (data not shown).
The genome is predicted to include six rRNA operons and 59 tRNA genes. There are several codons which are not represented by cognate tRNAs, suggesting that the codon recognition by the tRNA is wobbly in this organism. Eighty percent of the 5,060 predicted CDSs are transcribed in the same direction as DNA replication. Preferential use of the leading strand for transcription is also found in Clostridium perfringens (44) and Clostridium tetani.
D. hafniense Y51 belongs to the clostridia based on rRNA sequence comparison-based taxonomy. Consistent with this, the CDS homology search revealed that most D. hafniense Y51 CDSs exhibited the highest levels of similarity to CDSs of clostridia, including T. tengcongensis, a gram-negative, anaerobic, thiosulfate- and sulfur-reducing organism (2), and various Clostridium strains (Fig. (Fig.3).3). The next most prevalent group was the bacilli, which are known to be closely related to clostridia (Fig. (Fig.3).3). A large proportion of the CDSs, however, had no obvious orthologs or paralogs in clostridia or bacilli and exhibited the highest levels of similarity to CDSs of phylogenetically distant strains, especially members of the γ-Proteobacteria and Archaea, suggesting that the D. hafniense Y51 genome may contain many genes acquired by horizontal transfer at some stage of its evolution.
Of the of 5,060 predicted CDSs, over 75% had BLASTP hits to the COG database (50) with an e-value less than 1e-4. Functional classification of the predicted proteome revealed 430 CDSs related to energy production and conversion (functional classification group C) (Table (Table22).
From the viewpoint of dehalorespiration the most noteworthy group of respiratory enzymes is the corrinoid-containing reductive dehalogenases (Fig. (Fig.4).4). The PCE dehalogenase encoded by pceA has been purified and characterized. It contains an Fe4S4 cluster binding motif and forms a complex with a membrane anchor subunit, PceB (49). A putative regulatory protein, PceC, and a trigger protein-like folding chaperone, PceT, are also encoded by the operon. A similar pceABCT cluster has also been reported in Dehalobacter restrictus and D. hafniense TCE1 (28). The cluster is sandwiched between the genes encoding two transposases in D. hafninese Y51, suggesting that it was acquired by horizontal transfer. PceA contains a Tat (twin arginine translocation) signal peptide (49) and is predicted to be transported through the cell membrane into the periplasmic space by the bacterial Tat-dependent type II secretion system as a prefolded complex (41). Four tatA-like genes and a tatC-like gene are present in the genome, but no tatB gene is present. This is unlike the situation in Escherichia coli (16), in which the type II secretion system was originally described, but it is just like the situation in Bacillus subtilis (18). In these microorganisms the TatA protein probably has a dual role and is also responsible for the TatB function. The other dehalogenase gene neither occurs in a cluster nor is surrounded by genes encoding transposases. The dehalogenase is very similar to the ortho-chlorophenol reductive dehalogenase of Desulfitobacterium frappieri PCP-1, which exhibits dechlorinating activity for several polychlorophenols (4). It is currently not known whether D. hafniense Y51 dechlorinates polychlorophenols. The finding that only two dehalogenase genes are present is a surprise considering that there are 19 such genes in the D. ethenogenes 195 genome (43) and nine such genes have been found in the partially sequenced strain D. hafniense DCB-2 (D. hafniense DCB-2 whole-genome shotgun project; GenBank accession number AAAW00000000). D. hafniense DCB-2 does not dechlorinate PCE and TCE, which may be explained by the presence of a different set of dehalogenases in this strain.
In D. hafniense Y51 the CDSs that form the largest paralogous group are the CDSs that encode dimethyl sulfoxide (DMSO) reductase A subunits (dmsA) (3), most of which are accompanied by a CDS encoding small DmsB-like Fe-S cluster-containing accessory proteins (Tables (Tables33 and and4).4). Many of the complexes are encoded byoperons that also contain the genes for a DMSO reductase anchor subunit (dmsC) (54) or polysulfide reductase (nrfD) (17), two types of membrane subunits which are thought to participate in the electron transfer process (Table (Table4).4). These complexes are known to catalyze the reduction of DMSO, trimethylamine-N-oxide (40), arsenate (45), and a variety of other compounds, although the substrate specificities of most paralogs are not known. Indeed, these compounds can be utilized by this strain (data not shown).
Fumarate is the electron acceptor that leads to the fastest growth (48). It is predicted to be reduced by the three-subunit fumarate reductase encoded by frdABC. Interestingly, the genome encodes 30 paralogs of the flavoprotein subunit (frdA) (Tables (Tables33 and and5)5) This group of coding sequences is also expanded in Shewanella oneidensis (Table (Table3).3). Nevertheless, the function of the flavoprotein subunits in fumarate reduction or other processes has not been established yet.
The genus Desulfitobacterium was originally described as a taxon containing organisms that reduce elemental sulfur and sulfite but not sulfate (51). D. hafniense Y51, however, has been reported to be capable of reducing sulfate (48). Indeed, the genome encodes sulfate reductases in addition to sulfite reductases (Table (Table5).5). D. hafniense Y51 also encodes a nitrate reductase, as well as two periplasmic nitrite reductase complexes (6) composed of a cytochrome c catalytic subunit, NrfA, and a cytochrome c membrane-anchoring subunit, NrfH (Table (Table55).
D. hafniense strains have been shown to be capable of utilizing metal ions as electron acceptors. The D. hafnienese Y51 genome encodes at least six c-type cytochromes, far fewer than the 111 and 42 paralogs found in metal ion-reducing strains of Geobacter sulfurreducens and S. oneidensis, respectively (31). Furthermore, tetraheme cytochrome c (cymA) required for S. oneidensis metal ion-dependent respiration (35, 42) is not present in D. hafniense Y51. This not only shows that the use of metal ions as electron acceptors may be rather limited but also hints that c-type cytochromes do not play a role in dehalorespiration.
D. hafniense Y51 cannot grow on mono- or oligosaccharides used as electron donors. We attribute this to the lack of suitable transport systems in this strain to import these compounds from the environment (functional classification G) (Table (Table22).
Both pyruvate and lactate have been reported to be used as electron donors by D. hafniense Y51 (48). Pyruvate is converted to acetate in the presence of PCE or TCE via a series of reactions (Fig. (Fig.5),5), as is the case in Desulfitobacterium dehalogenans (52). The D. hafniense Y51 genome encodes three pyruvate formate lyases and two pyruvate ferredoxin oxidoreductases that may mediate the conversion of pyruvate to acetyl coenzyme A (acetyl-CoA) (Fig. (Fig.5).5). Lactate may be oxidized to pyruvate by a broad-specificity malate dehydrogenase (27). The reducing equivalents formed during the conversion of pyruvate are channeled to the electron acceptors, including halogenated compounds (32). Growth on pyruvate or lactate is fast, since phosphate acetyltransferase and acetate kinases catalyze the conversion of acetyl-CoA to CoA and acetate coupled to substrate-level ATP synthesis; thus, not only energy is harnessed from the respiration process.
Formate supports slower growth than pyruvate or lactate, suggesting that there is no additional help from substrate-level phosphorylation when this electron donor is used. Formate might be metabolized by a selenocysteine-containing formate dehydrogenase (Table (Table5).5). Alternatively, formate might be split into hydrogen and carbon dioxide by the formate hydrogen lyase (Table (Table5),5), and the resulting hydrogen might be oxidized by a membrane-associated hydrogenase (Fig. (Fig.4).4). In contrast to the situation in D. ethenogenes 195, the use of hydrogen as an electron donor has not been proven to support D. hafniense Y51 growth. Nevertheless, three Hup-type Ni-Fe periplasmic hydrogenases and a putative Fe periplasmic hydrogenase, each with a cytochrome b subunit, and an Ni-Fe hydrogenase that is a neighbor of a putative cytochrome c similar to the split-soret cytochrome c of the sulfite-reducing organism Desulfovibrio desulfuricans (9) are encoded by D. hafniense Y51 (Table (Table5).Transposon-based5).Transposon-based mutagenesis of D. dehalogenans provided evidence forthe involvement of a Hup-type hydrogenase and the formate hydrogen lyase-like complex in dehalorespiration (47); hence, the orthologous genes are likely to be essential components of the genes encoding the dehalorespiration pathway. The genome also harbors the coding sequences for HypA to HypF implicated in hydrogenase maturation (26). The complex is encoded by the hypAB and hypCDEF operons in two different regions of the D. hafniense Y51 chromosome.
Many vanillate-specific O-demethylase corrinoid protein (odmA) (19) homologs support growth on lignin-derived compounds abundant in forest soil because phenyl methyl ethers are components of lignin in plants (20). D. hafniense PCE-S is known to utilize phenyl methyl ethers, including vanillate and syringate, as electron donors (36). D. hafniense Y51 contains 15 homologs of odmA (Table (Table3)3) and two genes similar to the vanillate:corrinoid protein methyltransferase gene (odmB) (19), suggesting that the use of phenyl methyl ethers is widespread in this species. Although not known to be autotrophic, D. hafniense Y51 contains the Wood-Ljungdahl pathway (22, 23, 34). This pathway might be used to channel methyl groups from phenyl methyl ethers to the central metabolism (Fig. (Fig.5).5). The presence of multiple paralogs suggests that they are important in Desulfitobacterium biology.
D. hafniense Y51 encodes a functional Embden-Meyerhof-Parnas pathway. Like many strict anaerobes, D. hafniense Y51 lacks 2-oxoglutarate dehydrogenase of the tricarboxylic acid cycle. This organism is predicted to be self-sufficient for nucleotides and amino acids, although apparently it cannot efficiently degrade them (data not shown). The genome encodes complete pathways for synthesis of the cofactors flavin adenine dinucleotide, NAD, menaquinone, heme, and cobalamin (Table (Table5).5). The presence of the cobalamin synthesis pathway is especially noteworthy, since it is required by both the phenyl methyl ether-utilizing O-demethylases and the reductive deghalogenases (Fig. (Fig.44).
The predicted nitrogenase complex of D. hafniense Y51 exhibited the highest levels of similarity to the complexes found in some methanogens and photosynthetic nitrogen-fixing bacteria (38). Experimental verification of the nitrogen-fixing ability of this strain is required for a more thorough understanding of these genes.
Desulfitobacteria have been characterized as organisms that grow only in strictly anaerobic conditions. The genomes of these bacteria encode five putative catalases, two superoxide dismutases, and several rubrerythrin-rubredoxin systems with four rubrerythrin and two rubredoxin paralogs. D. hafniense Y51 also has a cytochrome bd oxidase operon composed of four genes, the structure of which is similar to that of Moorella thermoacetica (8). The presence of these CDSs contributes to the relatively high tolerance of this strain to dioxygen.
It is very interesting that D. hafniense Y51 and D. ethenogenes 195 both have dechlorinating ability despite the fact that they are phylogenetically distantly related and have very different genomic features (Table (Table1).1). Since closely related Desulfitobacterium and Dehalococcoides strains contain vastly different numbers and kinds of dehalogenases, it is tempting to speculate that the genes are horizontally acquired due to anthropogenic environmental pressure (47).
To compare D. hafniense Y51 and D. ethenogenes 195, we identified an orthologous subset consisting of 751 genes in the two strains (Fig. (Fig.6),6), only 54 of which are related to energy production and conversion (Table (Table2).2). We argue that this set of 751 coding sequences contains two classes, sequences that are responsible for dehalorespiration and sequences needed for other functions. While the former class is likely to be horizontally transferred (47), the latter class is predominantly vertically inherited and thus predicted to exhibit higher levels of homology to the orthologs of closely related strains than to the orthologs of strains which are more distantly related. By enriching for possible horizontally transferred genes within this subset, we should also enrich for coding sequences that may have a role in dehalorespiration.
To this end we identified the closest homologs of each of the 751 CDSs in the orthologous group to a CDS present in 1 of more than 200 completely sequenced organisms. Thirty-eight of the 751 D. hafniense Y51 coding sequences exhibited the highest levels of similarity to their D. ethenogenes 195 orthologs, and 72 of the 751 D. ethenogenes 195 coding sequences exhibited the highest levels of similarity to their D. hafniense Y51 orthologs (Fig. (Fig.6).6). For these two groups of coding sequences, 18 were found to be reciprocal best hits, meaning that the D. hafniense Y51 and D. ethenogenes 195 orthologs showed more homology to each other than they showed to any other paralogous sequence from another strain (Table (Table66 and Fig. Fig.66.).
PceB and PceT have no obvious orthologs in D. ethenogenes 195, providing some circumstantial evidence that the membrane-anchoring mechanism is not conserved or dispensable and that the PceT trigger factor-like folding chaperone might not be essential or may be complemented with a nonhomologous protein. We, however, expected to find PceA and PceC among the reciprocal best hits. Surprisingly, PceA was not included in this group. Although our assumption was that D. hafninese Y51 and D. ethenogenes 195 were the only dehalorespirers among the microbial strains used for our comparative study, the genome of a recently sequenced marine microorganism, Silicibacter pomeroyi DSS-3 (33), contained a reductive dehalogenase gene. Nothing is known about the dehalorespiring capability of this microorganism, but the presence of a dehalogenase in yet another group of microorganisms provides additional evidence that the dehalogenases are frequently horizontally transferred. Exclusion of the coding sequences of this strain from the comparison did not modify any of our other results except that it added PceA as the 19th member of the reciprocal best-hit group (data not shown). As expected, the PceC-like putative transcription regulator was found in the group of 18 best hits, suggesting that there might be some similarity in the transcriptional regulation of the dehalogenases in the two organisms. Since there are many paralogs of this putative transcriptional factor in both organisms, our assumptions need to be corroborated experimentally.
Included in the group of 18 reciprocal best hits were the large subunit and the maturation factor of a Hup-type Ni-Fe hydrogenase. Since it has been experimentally proven that Hup-type hydrogenases are necessary for dehalorespiration (47), the high level of similarity of the D. hafniense Y51 and D. ethenogenes 195 orthologs tempted us to speculate about the existence of a dehalorespiration-specific Ni-Fe hydrogenase. Both strains possess multiple copies of various corrinoid transport systems, multiple subunits of which exhibit unusually high levels of similarity. This clearly highlights the importance of scavenging corrinoid cofactors from the environment (Fig. (Fig.4).4). This is particularly important for D. ethenogenes 195, which, unlike D. hafniense Y51, does not encode the complete de novo corrinoid synthesis pathway (Table (Table3).3). Although carbon monoxide dehydrogenase activity is not known to be involved in dehalorespiration, the CDSs encoding the putative carbon monoxide dehydrogenase/acetyl-CoA synthase (Fig. (Fig.5)5) are conserved to a great extent in the two strains, as are several uncharacterized coding sequences which to date have not been implicated in dehalorespiration or other processes.
Although the similarity of D. hafniense Y51 and D. ethenogenes 195 is interesting from the viewpoint of dehalorespiration, the differences are also noteworthy. D. hafniense Y51 contains an unprecedented number and variety of respiration-related genes, most of which are not present in D. ethenogenes 195 (Tables (Tables33 and and5).5). This should be one reason why D. ethenogenes 195 utilizes only hydrogen as an electron donor and chlorinated organic compounds as electron acceptors (29). D. hafniense Y51 is known to possess a flagellum and is highly motile (48). Indeed, the genome encodes multiple copies of methyl-accepting chemotaxis proteins (Table (Table3)3) and contains a large cluster of motility genes. It should be interesting to study whether chlorinated compounds act as chemoattractants for this strain. In contrast, D. ethenogenes 195 is a coccoid organism whose genome encodes no motility. This is a disadvantage in bioremediation studies, since this species might not be as efficient in locating and approaching the target to be degraded.
The comparison of the genomes of D. hafniense Y51 and D. ethenogenes 195 showed that two superficially similar organisms, both of which were isolated based on their PCE-reducing abilities, are very different. Although excelling in the variety of chlorinated compounds that it can use as electron acceptors, D.ethenogenes 195 is a true dechlorination specialist; its limited metabolic repertoire and its apparent inability to disperse efficiently in the environment probably mean that this organism must be used as part of a bacterial community in bioremediation. D. hafniense Y51, on the other hand, is a generalist. It exhibits very high and still unexplored flexibility and uses a wide variety of electron donors and acceptors, which broadens the scope of its biotechnological applications. It is motile and largely self-sufficient for factors needed for reductive dehalogenation. The two genomes not only establish a firm background for research on dehalorespiration but pave the way for metabolic engineering of these strains to better suit the purposes of bioremediation.
This research was supported by New Energy and Industrial Technology Development Organization (NEDO), Japan.
We thank R. H. Doi (University of California, Davis) and C. Omumasaba for critical reviews of the manuscript.