Fosmid library construction.

Sediment from 5- to 12-cm depth from the intertidal sand flat of the Wadden Sea was taken for library construction. DNA was extracted as previously described (56) by overnight incubation with proteinase K, gel purified to get rid of humic substances, and additionally purified using the GeneClean Turbo kit (Qbiogene, Carlsbad, CA). A fosmid library was constructed using the EpiFOS fosmid library production kit (Epicenter, Madison, WI) according to the manufacturer's instructions with the following modifications: the concentrated DNA was blunt ended, purified, and concentrated using MICROCON YM-100 columns (Promega, Mannheim, Germany). DNA fragments of appropriate length for cloning were obtained after separation by pulsed-field gel electrophoresis (PFGE). The gel (1.3% low-melting-point agarose) was run on a field contour-clamped homogeneous electric field-PFGE mapper (Bio-Rad, Munich, Germany) at 14°C, 6 V cm⁻¹, for 18 h with 1- to 21-s pulses in 1× Tris-acetate-EDTA buffer. Gel bands containing 40- to 50-kb-long DNA were excised, digested with gelase (Epicenter), purified, and concentrated by application of MICROCON YM-100 columns. Packaging into phage heads and transduction were performed as indicated in the manufacturer's instructions. Three libraries containing approximately 34,000 clones with insert sizes ranging from 32 to 44 kilobases in total were prepared.

The fosmid library from Hydrate Ridge sediments was established as follows. DNA from 1- to 3-cm sediment depth was extracted as described above. The DNA was further purified by anion-exchange chromatography using Genomic-tip 20/G (QIAGEN, Hilden, Germany), embedded in 1% low-melting-point agarose, and dialyzed against Tris-EDTA buffer. The fosmid library was constructed as indicated above with the following modifications. After prior equilibration for 0.5 h at 40°C with end repair mix devoid of enzymes, the end repair reaction was performed in the intact agarose plug. The end repair reaction was performed with doubled nucleotide concentration, and size selection was carried out on a 1% SeaPlaque GTG agarose (FMC BioProducts) PFGE gel (0.5× Tris-borate-EDTA, 1 to 10 s, 14°C, 120°, 11 h). DNA was excised, equilibrated with Tris-EDTA buffer, solubilized with gelase (Epicenter), and concentrated. Ligation, packaging, and transduction were conducted as described above.

Screening for aprA and dsrAB.

In total, 11,000 clones from the intertidal sand flat library were screened for aprA genes. For amplification of the alpha subunit of the adenosine-5′-phosphosulfate reductase gene (aprA), primers APS-1-F and APS-4-RV were used to amplify a fragment of approximately 400 bp. The primer sequences and the PCR protocols will be published elsewhere (B. Meyer and J. Kuever, unpublished data). Screening for dsrAB genes in the fosmid library from Hydrate Ridge sediments was performed similarly. The full-length dsrAB genes were amplified from the fosmid extracts and cells using the previously described primers and PCR conditions (53).

Sequencing, ORF finding, and sequence annotation.

Fosmids were sequenced by a shotgun approach based on plasmid libraries with 1.5- and 3.5-kb inserts. Sequences of small inserts were determined by using Big Dye chemistry (ABI), M13 primers, and ABI3730XL capillary sequencers (ABI) up to a 10-fold coverage. Resulting reads were assembled by Phrap44 and finished in Consed (13). Open reading frames (ORFs) were predicted by the gene prediction program GLIMMER (6), which is integrated into the open source program package GENDB (36). Annotation of the identified ORFs was accomplished on the basis of similarity searches against different databases, such as Pfam, Swissprot, Uniprot, and Interpro. Signal peptides and transmembrane helices were also predicted. All results were evaluated manually. Potential transcription termination sites were predicted using the program TransTerm (11).

Phylogenetic analysis.

Full-length dsrAB sequences of the three fosmid clones were translated into proteins and phylogenetically analyzed using the ARB program package (31). Maximum likelihood trees were constructed using JTT amino acid substitution matrix for evolutionary distance along with the ProML program of the Phylip program package integrated in ARB. Distance matrix trees were calculated using the neighbor-joining function of ARB. Deletions and insertions were not considered in the DsrAB treeing methods. The amino acid sequences of QmoB and Sat were aligned using the ClustalW program package. For phylogeny analysis, calculations were performed without and with a 50% positional conservation filter. Subsequently, consensus trees were generated from the results of all treeing methods. Protein length heterogeneities were considered, and terminal sequence stretches were excluded from calculations. For QmoB 899 amino acid positions and for Sat 433 amino acid positions were included in the calculations.

Sulfur energy metabolism.

All three fosmids encoded several proteins implicated in dissimilatory sulfate reduction. The majority significantly resembled proteins from sulfate/sulfite-reducing prokaryotes and sulfur-oxidizing bacteria (Table ​(Table11 and Tables S2 and S3 in the supplemental material). The following analysis is focused on ORFs found in fosmid ws39f7, since it provided the largest gene set involved in DSR (Table ​(Table1).1). To improve clarity, “ORF” refers to the predicted gene whereas “Orf” refers to the deduced protein. The arrangement of the genes in the fosmids and of selected homologs in other organisms is depicted in Fig. ​Fig.11.

Open in a separate window

FIG. 1.

Genomic organization of the dsrAB locus in fosmids ws39f7, ws7f8, and hr42c9, sequenced SRP genomes, Desulfitobacterium hafniense, Chlorobium tepidum, and Allochromatium vinosum. ORF numbers match those in Table ​Table1.1. Numbers above ORFs refer to the homolog ORF in fosmid ws39f7.

As noted above, fosmids ws39f7 and ws7f8 included the gene for adenosine-5′-phosphosulfate reductase (aprAB, ORF20 and ORF21). Besides, a gene coding for the ATP sulfurylase (sat, ORF19) was identified directly upstream of aprAB (Fig. ​(Fig.1).1). These enzymes catalyze the activation of sulfate and the subsequent reduction of adenosine-5′-phosphosulfate (APS) to sulfite. In contrast, the sat-aprAB cluster was not found on fosmid hr42c9.

Fosmids ws39f7 and ws7f8 carried genes encoding four proposed subunits of the cytoplasmic dissimilatory sulfite reductase (DsrABDC, ORFs 3, 4, 5, and 8), accounting for the terminal steps in the reduction of sulfite to hydrogen sulfide. The putative dissimilatory sulfite reductase protein (ORF3 and 4) contains the characteristic siroheme binding sites and is phylogenetically closely related to DsrAB from known SRP (see the end of Results) (Fig. ​(Fig.2).2). ORF5 is predicted to be functionally similar to DsrD despite the absence of significant database hits. The predicted protein size of approximately 9 kDa and the position downstream of dsrB are in good accordance to DsrD in other SRP (16, 23, 44) and the sulfite-reducing bacterium Desulfitobacterium hafniense (Fig. ​(Fig.1).1). The DsrD amino acid sequence is highly variable. Nevertheless, in all homologs DNA-binding motifs are conserved. Its function is still not fully resolved (18), but it has been postulated that DsrD might play a role in DNA binding (37) or interact with DsrAB (28) to bind sulfite. It has never been identified in sulfur-oxidizing organisms; thus, it appears to be specifically involved in the dissimilatory reduction of sulfite (5, 23). The gene encoding DsrC (ORF8) was displayed in fosmids ws39f7 and ws7f8. DsrC was proposed to further reduce oxidized sulfur intermediates, such as thiosulfate (48, 50). Both DsrC and DsrD are not essentially tightly associated with DsrAB (18, 40). Thus, they also may act as independent proteins rather than as subunits of the Dsr complex.

Open in a separate window

FIG. 2.

DsrAB-based phylogenetic reconstruction. The scale bar corresponds to 10% estimated sequence divergence. The tree was inferred using the maximum likelihood and neighbor-joining methods. Multifurcations were introduced when branching orders were not supported by all phylogenetic methods used.

Fosmid ws39f7 displayed two genes that are likely responsible for the synthesis of siroheme, the presumed prosthetic group of DsrAB. ORF16 encodes a subunit of siroheme synthase (CysGA). It is followed by a gene (ORF17) coding for a bifunctional protein that represents a fusion of siroheme synthase (CysGB) and uroporphyrinogen III synthase (HmeD), an arrangement very similar to the dsr locus in Chlorobium tepidum (Fig. ​(Fig.1).1). Downstream of dsrD, a gene (ORF6), which is related to dsrN in SRP (16, 23, 26, 27, 44), D. hafniense, and sulfur-oxidizing bacteria, was predicted (5, 10). The derived protein resembled cobyrinic acid a,c-diamide synthase, which is part of the vitamin B₁₂ biosynthetic pathway and catalyzes the amidation of cobyrinic acid. However, a distinct function of this protein in DSR and also in sulfur oxidation was proposed (5, 26). Generally, siroheme is considered to be the prosthetic group of DsrAB. In contrast, in Desulfovibrio spp. an amidated siroheme was identified, and a DsrN-like protein was accounted for the amidation step (26, 35). In D. desulfuricans, Desulfovibrio vulgaris, D. hafniense, Thermodesulforhabdus norvegica (27), Desulfobacter vibrioformis (26), and our fosmids, dsrN homologs are found in the direct neighborhood of dsrAB (Fig. ​(Fig.1),1), suggesting siroheme amide as the actual prosthetic group of DsrAB. Interestingly, in all fosmids dsrN is juxtaposed to a dsrL-like gene (ORF7). The amino acid sequence of Orf7 showed high similarities to DsrL in phototrophic sulfur-oxidizing bacteria in both N- and C-terminal sequences and to a subunit of glutamate synthase (GltD). Like in DsrL of Allochromatium vinosum, Orf7 has two flavin adenine dinucleotide (FAD)- and two Fe₄S₄-binding sites as well as one NAD(P)H-binding motif. For the reverse pathway of sulfur oxidation, Dahl et al. (5) proposed a function in electron transfer from NAD(P)H or FAD to an acceptor protein to other proteins of the dsr locus rather than in glutamate synthesis. Related proteins are also present in known genomes of bacterial SRP and sulfite-respiring D. hafniense and Moorella thermoacetica.

Membrane proteins of electron transport.

On three fosmids investigated in this study, genes were found in proximity to dsrAB that are most probably involved in the electron transfer to AprAB and DsrAB (Fig. ​(Fig.1).1). Orf10 to Orf14 of fosmid ws39f7 showed high similarity to the heterodisulfide reductase-like menaquinol-oxidizing enzyme complex (HmeABCDE) in Archaeoglobus fulgidus and to the DsrMKJOP complex in A. vinosum. The identified complex is distinct to other respiratory complexes, such as the high-molecular-weight cytochrome (Hmc) complex in D. vulgaris (45), with respect to gene arrangement and gene composition. The amino acid sequences of Orf10 to Orf14 revealed specific characteristics supporting a function similar to that predicted for the Hme complex of A. fulgidus (32). Heme b- and heme c-binding sequence motifs were identified in ORF10 and 12. Cysteine motifs for Fe-S binding were found in ORF11 and 13. ORF10 encodes a transmembrane heme b-type protein and likely is the homolog of HmeC/DsrM. Orf11 is an Fe-S protein related to a DsrK/HmeD. The deduced protein from ORF12 is predicted as a periplasmic triheme with a typical sec-signal peptide that may serve as a membrane anchor on the periplasmic side, similar to DsrJ and HmeE. Just as DsrO/HmeA, Orf13 contains an Fe-S binding site and an arginine leader sequence indicative for the Tat transport system across membranes (1). Orf14 is related to DsrP/HmeB and concordantly showed 10 membrane-spanning helices and similarities to polysulfide reductase motifs. This protein family participates in the electron transfer from quinones to terminal acceptors.

The four purified proteins HmeACDE in A. fulgidus were proposed to play a critical role in electron transfer from the menaquinol to sulfur intermediates, facilitating the reduction of sulfate to sulfide (32, 33). In Archaeoglobus profundus, HmeC and HmeD homologs are involved in the oxidation of H₂, the only known electron source for this organism (33). In A. vinosum, DsrMKJOP also copurified with DsrAB and were shown to be indispensable for the oxidation of sulfur intermediates (5, 42). The proximity of hme to dsrAB in all of our investigated fosmids, in D. hafniense, and in sulfur-oxidizing C. tepidum and A. vinosum supports an interaction of the gene products (Fig. ​(Fig.1).1). Genome comparison revealed that genes of Hme-related membrane complexes consistently show a conserved structure, which contains at least five genes. In contrast to A. fulgidus, the other investigated genomes contain an additional homolog gene with unknown function linked to this operon (ORF9) (Fig. ​(Fig.1).1). We repeatedly found homologs to ORF9 that are linked to genes encoding the b-type cytochromes in the DsrMKJOP/Hme operon of SRB genomes, in D. hafniense, and in C. tepidum but not in A. fulgidus (Fig. ​(Fig.1).1). The predicted protein has a calculated mass of 21 kDa and displays highest sequence similarity to a conserved hypothetical protein in D. desulfuricans.

On fosmid ws39f7, the two genes upstream of dsrAB (ORF1 and 2) probably form an interactive unit with ORF18. These three genes encode a membrane complex homolog to the quinone-interacting membrane-bound oxidoreductase (QmoAB) complex in D. desulfuricans (ORF1, 2, and 18), a novel membrane-bound respiratory complex (41). It also showed high similarity to heterodisulfide reductase in methanogenic archaea. The deduced proteins of ORF1 and ORF2 code for soluble proteins each containing Fe-S and FAD for electron and proton transport according to QmoA and QmoB in D. desulfuricans. No signal peptides or transmembrane helices were identified, indicating a cytoplasmic location. The third gene, ORF18, encoded six membrane-spanning helices and thus likely is a transmembrane protein as QmoC. It harbored both a hydrophobic domain with homology to the heme b protein HdrE and a hydrophilic domain with homology to the Fe-S protein HdrC in A. fulgidus. Based on comparison of our data with the detailed analysis of sequences and proteins in D. desulfuricans, ORF1, 2, and 18 proteins likely are functionally similar to the Qmo complex. The derived proteins are predicted to interact physically and to catalyze the transfer of electrons to AprAB for the reduction of APS. The qmo operon is juxtaposed next to aprAB in Desulfotalea psychrophila, D. vulgaris, and C. tepidum and partially in fosmid ws39f7, which also confirms a direct functional linkage to AprAB (Fig. ​(Fig.1).1). This is supported by the absence of qmoABC, sat, and aprAB in the sulfite-reducing D. hafniense and M. thermoacetica.

On all three fosmids, additional genes that are potentially involved in sulfur-based energy metabolism were identified (Table ​(Table11 and Tables S2 and S3 in the supplemental material). The derived protein of ORF28 (ws39f7) showed a weak similarity to high-molecular-weight cytochrome complex subunit E (HmcE) of D. vulgaris (45). It is probably physically associated with an Fe-S protein encoded by ORF29 that in turn resembles the HmcF subunit. The latter seems to form a fused transcriptional unit with a response regulator receiver domain (ORF29) and may be functionally linked with the adjacent two-component regulator sensor (ORF30). Similarly, Orf25 displayed the best hit to an Fe-S protein of A. fulgidus. In D. hafniense, the homologs are in close proximity to dsrAB, which supports a functional linkage to DsrAB (Fig. ​(Fig.1).1). Similarly to ORF29, a close link of ORF25 to regulatory elements (ORF27 and 28) was found. A dedicated function could not be assigned to both two-protein units; however, the close association with regulatory elements suggests a role in transcriptional control. We further found a predicted NADH oxidase (Orf15) that was homolog to NoxA-3 of A. fulgidus and showed highest similarity to a predicted gene product in D. hafniense. Related homologs were also found in known genomes of SRP and in C. tepidum, indicating a specific relation to sulfur-based energy metabolism. NADH oxidases might also translocate electrons directly from NADH to APS, as was proposed for D. vulgaris (4). More generally, NADH oxidases play a role in protection of proteins under oxygen stress (34).

Novel genes probably involved in DSR.

Besides ORF9 linked to the Hme complex, several other ORFs could not be assigned to functionally characterized proteins. Homologs to Orf23 were identified in all SRP genomes, in Desulfobacula toluolica, in C. tepidum, and in Thiobacillus denitrificans but not in the draft sequences of sulfite-respiring D. hafniense and M. thermoacetica, indicating a role in electron transfer to AprAB. Orf23 displayed high motif similarities to a protein family of unknown function (UPF0153). These proteins contain eight conserved cysteine residues that may form a metal-binding site. Their function is still unknown but might be a part of an oxidoreductase complex. The deduced protein has a calculated mass of 34 kDa. In both fosmid ws39f7 and fosmid ws7f8, the respective gene was coupled to a gene encoding a conserved hypothetical protein (ORF22) and to aprAB (Fig. ​(Fig.1).1). The hypothetical protein derived from ORF22 is related to a gene in C. tepidum, which is also adjacent to aprAB (Fig. ​(Fig.1).1). Another probable protein was encoded by ORF31, which showed weak but significant hits to hypothetical proteins in known SRP genomes and D. hafniense. The gene context does not allow designation of a function, but it is apparently linked to regulatory elements on fosmid ws39f7, in D. desulfuricans, and in D. hafniense. Furthermore, a gene encoding a putative transmembrane transport protein (ORF33) that is associated with regulatory elements was identified. The deduced protein belongs to a family of di- and tricarboxylate transporters and to sodium/sulfate symporter proteins. It showed similarity to an uncharacterized transport protein in D. desulfuricans and might be involved in sulfate transport into the cytoplasm.

Genome comparison of gene arrangement.

When the gene arrangement on fosmid ws39f7 is compared to that in the genomes of A. fulgidus, D. psychrophila, D. vulgaris (16, 23, 44), and Desulfobacterium autotrophicum (unpublished data), it is shown that the homolog genes are dispersed similarly throughout the genomes (Fig. ​(Fig.3).3). Clustering of genes involved in DSR was also observed to a lower extent in fosmids ws7f8 and hr42c9 (Fig. ​(Fig.1).1). In sulfite-respiring D. hafniense, 11 homolog genes were found to cluster with dsrAB. Phototrophic sulfur-oxidizing A. vinosum and in particular C. tepidum also show a high degree of gene clustering similar to fosmid ws39f7. In C. tepidum, 17 homolog genes are located in the neighborhood of two physically separated dsrAB copies (Fig. ​(Fig.1),1), with a gene order partially identical to that of fosmid ws39f7.

Open in a separate window

FIG. 3.

Comparison of positions of ortholog genes on fosmid ws39f7 and complete genomes of Archaeoglobus fulgidus and Desulfovibrio vulgaris. Ortholog genes are connected by black lines. Mb, megabases.

Comprehensive model of dissimilatory sulfate reduction.

Based on our gene context analyses, a more comprehensive model for dissimilatory sulfate reduction in prokaryotes can be proposed (Fig. ​(Fig.6).6). To indicate the level of gene clustering, we refer to ORFs in fosmid ws39f7. According to this model, sulfate is transported into the cell either by a hypothetical sodium/sulfate symporter protein (Orf32) or by general sulfate transporters not encoded on the fosmid. The cytoplasmic steps of DSR are already well known (39). In the cytoplasm, sulfate is activated by an ATP sulfurylase (Orf19, EC 2.7.7.4) to form APS. An electron translocation to APS is catalyzed by AprAB (Orf20 and 21, EC 1.8.99.2), resulting in sulfite and adenosine monophosphate. DsrABDC subunits (Orf3, 4, 5, and 8, EC 1.8.99.3) reduce sulfite to intermediates and finally to hydrogen sulfide. Furthermore, a complete gene set required for the synthesis of the prosthetic group of DsrAB, which is generally regarded to be siroheme (24), was identified. Recent experimental results and the analysis of the genetic context of the dsr locus provide evidence that siroheme-amide may serve as the actual prosthetic group rather than siroheme (26, 35) (Table ​(Table11 and Fig. ​Fig.1).1). The physical proximity of dsrN and dsrAB in our fosmids, D. hafniense, and phototrophic sulfur oxidizers (5) strongly supports this hypothesis. The juxtaposition of dsrL to dsrN points at the DsrL protein to provide glutamine for amidation of siroheme. In contrast, Dahl et al. (5) proposed that DsrL functions as a NADPH:acceptor oxidoreductase.

Open in a separate window

FIG. 6.

Tentative schematic model of dissimilatory sulfate reduction based on the deduced proteins potentially encoded on fosmid ws39f7. The illustrations of Hme and Qmo complexes were modified after Mander et al. (32) and Pires et al. (41). indicates heme.

It is still unclear which proteins universally present in SRP transfer electrons to the cytoplasmic enzymes. A few membrane complexes have been described for Desulfovibrio spp. and Archaeoglobus spp., such as the DvH Hmc (45), Dd27k 9Hc (46), DvH TpIIc3 (52), Hme (32, 33), and Qmo (41) complexes. These complexes are related, and all but Qmo contain a periplasmic cytochrome c subunit. However, they differ substantially in their gene order, protein structure, and number of heme-binding sites. Here it is shown that the HmeA-E/DsrMKJOP complex to date appears to be the only cytochrome c-transmembrane complex of which all conserved homologs are universally present among known SRP and also in phototrophic sulfur-oxidizing bacteria (5, 23).

Similarly, the qmoABC homologs along with the gene arrangement appear to be unique in published genomes of sulfate-respiring prokaryotes, D. autotrophicum (data not shown), C. tepidum, and T. denitrificans. This complex does not show periplasmic components; therefore, it was speculated to account for the electron transfer directly to AprAB for the reduction of adenosine-5′-phosphosulfate (41). This observation is also confirmed by the capability of Desulfotomaculum aeronauticum to utilize sulfate only by amendment of the quinone precursor menadione. Apparently here electrons are transferred from the menaquinone pool to APS rather than from periplasmic electron sources (17).

Haveman et al. (15) found a downregulation of dsrMKJOP/hme, qmoABC, and ATP synthase genes in D. vulgaris upon inhibition of DSR by nitrite amendment. They concluded that both membrane complexes donate electrons to AprAB and DsrAB. Here it is shown that these genes are more universally distributed among SRP and most likely are of general importance in electron transfer of DSR. The proteins homologous to HmeA-E/DsrMKJOP and QmoABC (32, 41) may function as universal menaquinone oxidoreductases and transmembrane electron/proton transfer proteins in SRP. Such an obligate interaction would provide for the missing link between the menaquinone pool/periplasm and the cytoplasmic enzymes. The ubiquitous presence of the dsrMKJOP/hme and qmoABC genes in sulfate-reducing prokaryotes and their presence in some sulfite-reducing and phototrophic sulfur-oxidizing bacteria emphasize their relevance for general sulfur-based energy metabolism.

In this model, the previously supposed cytoplasmic electron shuttles (39, 43) from the membrane donors to cytoplasmic DsrAB and AprAB are not illustrated. In contrast, there is evidence for a direct interaction of DsrABDC with the membrane components. In Desulfovibrio spp. (49) and also in A. vinosum (5), a membrane-bound fraction of DsrAB was observed.

Several conserved hypothetical proteins that were often closely linked to known genes, such as ORF9 and ORF23, were identified. Thus, they might play a so far unknown role in sulfur-based energy metabolism. Further experiments should elucidate their role in these pathways.

Not all components important for sulfate respiration may necessarily be encoded on this subgenomic fragment. Pyrophosphatase, additional sulfate transporters, quinones, ferredoxins, flavodoxins, and oxidoreductases are widespread and could be easily supplied by general metabolism.

We thank the staff of R/V Sonne, Katja Nauhaus, and Antje Boetius for sampling at the Hydrate Ridge and Katja Bosselmann for sampling in the Wadden Sea. Melissa Duhaime, Jakob Pernthaler, and Friedrich Widdel are acknowledged for helpful suggestions to improve the manuscript. Alexander Loy, University of Vienna, helped with the phylogenetic analysis of DsrAB sequences. High-quality sequencing was done in the unit of the Max Planck Institute for Molecular Genetics headed by Richard Reinhardt. Andreas Ellrott provided technical support in replicating and handling of libraries.

This work was funded by the Max Planck Society, the European Commission (NOE Marine Genomics Europe, GOCE-CT-2004-505403), and the “Research Group BioGeoChemistry of Tidal Flats” funded by the German Science Foundation (DFG).