The transition metal molybdenum is essential for life on earth. It is at the catalytic centre of over 30 enzymes, which are involved in the nitrogen, carbon, and sulphur cycles [1]. Molybdenum is found in the nitrogenase complex, which fixes dinitrogen gas, and in nitrate reductase, which reduces nitrate to nitrite. Other molybdo-proteins include xanthine oxidase, aldehyde oxidase, formate dehydrogenase, sulphite oxidase, nitrite reductase, DMSO reductase, pyridoxal dehydrogenase, xanthine dehydrogenase, and pyrogallol transhydrolase.

Molybdenum is available to organisms in the form of the tetraoxyanion molybdate, which is transported into Escherichia coli cells by an ABC-type transport system encoded by the modABCD operon [2]. A ModE-molybdate complex binds specific DNA target sequences and thus represses or activates transcription of several operons in response to molybdate concentration. The ModE protein from E. coli consists of two domains [3]. At the N terminus is the DNA-binding domain (Pfam accession PF02573, HTH_9) containing a winged helix-turn-helix (HTH) motif. At the C-terminal end is the molybdate-binding domain (Pfam accession PF03459, TOBE), which consists of two sub-domains of OB-folds with exchanging C-terminal strands. In E. coli, ModE bound to molybdate has been shown to contribute to the regulation of modABCD [4], napFDAGHBC [5], moaABCDE [6], and dmsABC operons [7]. These operons encode proteins involved in molybdenum homeostasis and metabolism.

Homologues of ModE are known in γ-Proteobacteria (e.g. E. coli, Haemophilus influenzae) and α-Proteobacteria (e.g. Rhodobacter capsulatus). Now that over 100 complete genome sequences are available, we wished to discover whether this regulatory system is found more widely in organisms other than Proteobacteria and whether we could gain any insights into the evolution of the molybdenum regulon.

Results and Discussion

Taxonomic distribution of ModE homologues and protein domain organisation of their sequences

Probable orthologues of the molybdenum-responsive regulator ModE were found in several Proteobacteria as well as in the green sulphur bacterium Chlorobium tepidum. Sequence analysis using Pfam [9] revealed that these proteins contained the HTH_9 domain at the N terminus and two TOBE sub-domains at the C terminus (Figure 2A). In most cases, only a single ModE homologue was found in each complete genome sequence. However, in the α-proteobacterium Rhodobacter capsulatus, there are two ModE homologues. One is associated with molybdate transport genes in the operon modEABC, whilst the second homologue is encoded by modE2, located upstream of vnfA2, the gene encoding one of two homologues of the nitrogenase regulatory protein, VnfA. The methanogenic Archaea Methanosarcina mazei and M. acetivorans each encode a protein that resembles ModE, but has only one rather than two C-terminal TOBE sub-domains (Figure 2B). No orthologues of ModE were found in the complete genome sequences from other lineages, such as the Gram-positive Bacteria and Cyanobacteria.

Several other Bacteria and Archaea encode proteins that contain the HTH_9 domain characteristic of ModE, but completely lack a molybdate-binding TOBE domain. Examples include Q9RBF7, Q8XXM1 (RSC2092), and Q8ZZY3 (PAE0019) from Alcaligenes eutropha, Ralstonia solanacearum, and Pyrobaculum aerophilum respectively. In these proteins, rather than a TOBE domain, there is a domain at the C terminus that is homologous to the SBP_bac_1 family (Pfam accession PF01547). Members of this family [8] bind diverse solutes such as sugars, peptides, and inorganic ions. In Alcaligenes eutropha, Q9RBF7 (FdsR) has been shown to regulate the fdsGBACD operon, which encodes the soluble NAD+ linked formate dehydrogenase molybdo-enzyme [10]. Based on similarity between the C-terminal domain of FdsR and a domain of formate dehydrogenase, Oh and Bowien [10] proposed that this C-terminal domain binds formate, thus modulating DNA-binding activity.

In Bacteria and Archaea, transcriptional regulators containing a HTH exhibit a position-function correlation such that repressors and dual function repressors/activators tend to have the HTH-containing domain at the N terminus whilst activators tend to have the HTH at the C terminus [11, 12]. ModE can act as both a repressor and an activator. Consistent with this, ModE and the other proteins described above, the HTH_9 domain occurs at the N terminus. In contrast to this arrangement, Q9PMF6 (Cj1507c) from Campylobacter jejuni has an HTH_9 domain at the N terminus (Figure 2D). This suggests that it might be a transcriptional activator rather than a repressor. Intriguingly the Cj1507c ORF overlaps fdhD encoding the molybdo-enzyme formate dehydrogenase (Cj1508c), strongly suggesting a functional link between this regulatory protein and molybdenum metabolism.

Interestingly several Bacteria and Archaea encode proteins that consist of just the HTH_9 domain, and lack any recognisable molybdate- or other solute-binding domains (Figure 2E). Examples are found in several Archaea and also in Salmonella typhimurium and Agrobacterium tumefaciens. These proteins would certainly not be functional for binding molybdate, but the HTH domain is probably capable of binding DNA and perhaps forming multimeric protein complexes.

Conservation of DNA-recognition elements in ModE homologues

The HTH_9 domain of ModE is responsible for its sequence-specific DNA-binding activity. According to the model proposed by Hall et al. [3] based on the three-dimensional crystal structure of ModE, nine amino acid residues in the HTH directly interact with the target DNA and have a role in sequence-recognition: Ser35, Gln36, Lys39, Ser44, Tyr45, Lys46, Ser47, Trp49, and Asp50 (numbering as in [3]). Interestingly, most of these residues were moderately conserved, even between phylogenetically distant organisms (see alignment in Figure 1). This suggested that the cognate target sequences recognised by these proteins might also be conserved. Therefore we decided to investigate the occurrence of potential ModE-binding sites in a range of prokaryotic genome sequences.

Figure 1
figure 1

Alignment of the helix-turn-helix (HTH) domains homologous to the molybdenum-responsive transcriptional regulator ModE. Amino acid residues predicted to directly interact with the target DNA are marked with asterisks (*). The alignment was taken from the Pfam database

Figure 2
figure 2

Domain architectures of some example members of the HTH_9 family as defined by Pfam (Bateman et al., 2002). The domains are described in the text. A. Q9I631 (Pseudomonas aeruginosa), Q88AA0 (Ps. syringae), Q88QX4 (Ps. putida), Q9CMR6 (Pasteurella multocida), MODE_HAEIN (Haemophilus influenzae), Q8Z8A6 (Salmonella Typhi), Q8ZQR8 (Salmonella Typhimurium), MODE_ECOLI (E. coli), MODE_YERPE (Yersinia pestis), Q8EAN6 (Shewanella oneidensis), Q8XQR8 (Ralstonia solanacearum), Q8KC82 (Chlorobium tepidum), MODE_AZOVI (Azotobacter vinelandii), Q9ABA4 (Caulobater crescentus), MOPA_RHOCA (Rhodobacter capsulatus), MOPA_RHOCA (Rhodobacter capsulatus), Q9F4K4 (Herbaspirillum seropedicae). B. Q8PWN4 (Methanosarcina mazei), and Q8TTZ2 (M. acetivorans). C. Q9RBF7 (Alcaligenes eutrophus), Q8XXM1 (Ralstonia solanacearum), and Q8ZZY3 (Pyrobaculum aerophilum). D. Q9PMF6 (Campylobacter jejuni). E. Q97Z66 (Sulfolobus solfataricus), Q97ET9 (Sulfolobus tokodaii), Q8ZYE6 (P. aerophilum), O29240 (Archaeglobus fulgidus), Q8TVF9 (Methanopyrus kandleri), and Q98KI4 (Rhizobium loti).

Identification of ModE-binding sites in genomic DNA sequence

The consensus ModE-recognition sequence has dyad symmetry and can be approximately represented as atCGcTATATAN6TATATAaCGat [5]. As a first step to fully characterising the set of target sequences for this family of transcription factors, we constructed an alignment (Figure 3) of several known and probable ModE-binding sites from γ-Proteobacteria, also illustrated as a sequence logo in Figure 4. Using a position-dependent weight matrix derived from this alignment (Figure 5), we scanned the protein-coding and non-coding regions of a range of complete genomes to find matches to the canonical ModE-binding DNA sequence motif. The results of these searches are summarised in Tables 1 [see Additional File 3], 2, and 3. It is important to remember that the mere presence of DNA sequence similarity to a consensus is not sufficient to demonstrate a biologically significant protein-binding site; in other words this approach of scanning DNA sequence against a weight matrix model will inevitably yield some false-positives. For example, several high-scoring matches were found within protein-coding regions in most of the genomes that we examined; these intragenic sites are unlikely to be functional regulatory sites. Furthermore, many of the intergenic matches fell upstream of genes not obviously involved in molybdenum metabolism and therefore probably not subject to regulation by ModE.

Figure 3
figure 3

Alignment of known and strongly suspected ModE-binding sites used to generate the weight matrix.

Figure 4
figure 4

Sequence logo representation of the alignment of known and strongly suspected ModE-binding in Figure 3. A graphic representation of an aligned set of binding sites. The relative heights of the letters are proportional to the frequencies of bases at each position. The degree of sequence conservation is measured in bits of information and is indicated by the total height of a stack of letters. The vertical scale is in bits, with a maximum of 2 bits possible at each position. The logo was generated using WebLogo [27].

Figure 5
figure 5

Position-specific weight matrix derived from the alignment of known and strongly suspected ModE-binding sites used to generate the weight matrix in Figure 1.

Table 2 Potential ModE-binding sites identified by scanning archaeal genome sequences against the weight matrix derived from alignment in Figure 3. Sites that fall immediately upstream of genes implicated in molybdenum metabolism are indicated in bold. Scores are Kullback-Leibler distances that have been normalised such that the maximum possible score is 100. Essentially, the higher the score is, the greater the magnitude of the theoretical binding energy [26]. All sites scoring more than 75 are listed. Distances are given (number of bases) between the downstream end of the putative ModE-binding site and the predicted translational start codon.
Table 3 Frequencies of matches to the ModE-binding consensus in several complete genomes. The number of sites scoring greater than 75 for similarity to the ModE-binding consensus was counted for each of the complete genome sequences. Note that the consensus sequence is a nearly palindromic, so only sites on on one strand were counted. Matches were designated intragenic if the center of the 28 base site was located within a protein-coding region and intergenic otherwise.

However, notwithstanding the occurrence of some false-positives, we are confident that our method had some discriminatory value in identifying bona fide ModE-binding sites because of the high proportion of the matches that did fall upstream of molybdenum-linked genes. Despite the importance of molybdenum-dependent processes, relatively few bacterial genes are directly involved in molybdenum metabolism: certainly no more than a few percent of the genome. Nevertheless, a disproportionately high number of our identified matches were located immediately upstream of genes implicated in molybdenum metabolism. For example in E. coli K12, we found a total of 26 high-scoring matches to the ModE-binding consensus (Table 3). Of these 26 sites, 17 occurred in non-coding DNA. Five of these 17 sites were located upstream of genes or operons associated with molybdenum metabolism or molybdo-enzymes (Table 1 [see Additional File 3]). In some cases, these ratios were more impressive: for example in Agrobacterium tumefaciens, Bradyrhizobium japonicum, Chlorobium tepidum, Mesorhizobium loti and Ralstonia solanacearum, at least 50% of the intergenic matches were located immediately upstream of molybdenum-associated genes (Table 3). The frequent occurrence of these ModE operator-like sequences upstream of molybdenum-associated genes is consistent with their being biologically significant rather than mere coincidence.

As expected, we found good matches to the proposed consensus ModE binding sequence immediately upstream of genes implicated in molybdenum metabolism in the γ-proteobacterial genomes. For example, high-scoring sites were detected upstream of homologues of modA, moeA, dmsA, and napF in E. coli and Salmonella species, which correspond to ModE-binding sites. Additional good candidate sites were found upstream of yecK in both E. coli and Haemophilus influenzae. The yecKbisZ operon in E. coli encodes a molybdo-enzyme that forms part of a respiratory system [13]. Therefore it is not surprising that the yecKbisZ operon might be regulated by ModE.

We observed some differences in the repertoires of predicted ModE-binding sites between the closely related strains of enteric bacteria (E. coli, Yersinia pestis, Shigella flexneri, and Salmonella enterica). There appears to be a core ModE-dependent regulon consisting of the dmsABC, moaABCDE, modABCD and napFDAGHBC operons, the only exception being that there is no apparent ModE binding site associated with dmsABC in Y. pestis. Additionally, yecK also has a putative upstream ModE binding site in E. coli and S. flexneri; yecK appears to be absent from S. enterica and Y. pestis. Furthermore, in S. enterica a second dms operon also has sites, whilst Y. pestis has additional sites upstream of pflA and narP. These differences in the predicted ModE-dependent regulon between closely related species suggest some evolutionary plasticity in the regulon. We also observed clear differences between the close relatives Haemophilus influenzae and H. ducreyi, the latter lacking a ModE homologue and lacking conserved ModE operator-like sites upstream of its mod operon.

Our data show that probable ModE-binding sites were found upstream of modA homologues not only in enteric bacteria and H. influenzae as has been previously established, but also in A. tumefaciens, Helicobacter hepaticus, M. loti, Pasteurella multocida, several Pseudomonas strains, R. solanacearum and Shewanellla oneidensis. This indicates that ModE-dependent regulation of the molybdate ABC transporter is widespread amongst proteobacteria.

Interestingly, we found strong potential ModE-binding sites upstream of modA (involved in molybdate transport) and napA (encoding a molybdenum-containing nitrate reductase) in the ε-proteobacterium Helicobacter hepaticus (but not in its close relatives H. pylori and Campylobacter jejuni). The complete genome sequence of H. hepaticus does not encode a full-length ModE homologue; it does encode a 135 amino acid protein (HH0653), consisting of a single HTH_9 domain, which is therefore probably able to specifically bind to the identified sites. However, since it lacks a molybdate-binding TOBE domain, it cannot directly use molybdenum availability to modulate DNA-binding activity. Interestingly, a potential ModE-binding site is also observed upstream of HH0158, predicted to encode a molydenum-containing periplasmic nitrate reductase in H. hepaticus.

A similar scenario is observed in Agrobacterium tumefaciens. A ModE-binding site is found upstream of modA but A. tumefaciens has no full-length ModE homologue, only a 131 amino acid protein (ATU2654, Q8UCD4) consisting of a HTH_9 domain. It is possible that the ModE-dependent regulatory system is in a process of degeneration in these organisms. However, it is equally possible that these proteins are functional and that DNA-binding is modulated by protein-protein interaction with some unknown factor or factors.

Strikingly, we also found a good candidate ModE-binding sites associated with molybdenum metabolism in some more distantly related organisms. One striking example was a site upstream of an open reading frame (CT1544) annotated as a probable periplasmic molybdenum-binding protein component of a molybdenum ABC transporter complex in the green sulphur bacterium C. tepidum [14]. This suggested that the full-length orthologue of ModE (CT1543, Q8KC82) functions in regulation of molybdenum transport in C. tepidum in a similar manner to ModE in E. coli.

The genome of C. tepidum [14] has at least three loci that contain genes implicated in molybdenum homeostasis and metabolism. The first of these loci includes a mod transport system and genes involved in synthesis of the molybdopterin genes cofactor, moaCBmoeEmobBAmoaA. A second includes CT1765, encoding a homologue of the ModG molybindin of Azotobacter vinelandii that binds molybdate. CT1765 has been misleadingly annotated as a molybdopterin-binding protein, as have most other molybdate-binding proteins in the public databases. It is probably involved in molybdate homeostasis. The third locus includes moaCB, moeE, mobBA, moaA and the nif genes that encode nitrogenase. Structural genes for the nitrogenase system in C. tepidum most closely resemble those of the Archaea rather than those of other nitrogen-fixing bacteria [14]. This raises the question of horizontal transfer between Bacteria and Archaea. However, a strong candidate σ54-dependent promoter is located upstream of C. tepidum nifH [15], which encodes the nitrogenase. σ54-dependent transcription of nitrogenase genes is common among Proteobacteria, but σ54 is not found in Archaea. Taken together, the σ54-dependent promoter and the ModE binding site centred just upstream of CT1544 suggest that the regulation of nitrogenase in C. tepidum shares much in common with that in Proteobacteria such as A. vinelandii. Does this tell us anything about the direction of horizontal transfer of nitrogen-fixation genes between Archaea and green sulphur bacteria? If the green sulphur bacteria acquired nitrogen-fixing genes from an archaeon, then this would imply that the recruitment of σ54 to the regulation of nitrogen fixation has occurred at least twice independently. The alternative, and more likely, scenario is that the Archaea acquired nitrogen-fixation genes from a relative of the green sulphur bacteria.

Even more striking was our finding that good candidate ModE-binding sites occur immediately upstream of several genes implicated in molybdenum metabolism in the methanogenic Archaea (Table 2). For example, we found high-scoring matches to ModE-binding sites in Methanosarcina acetivorans upstream of modA and fmdE, which encodes a molybdenum-containing hydrogenase. We also found sites upstream of vnfH and nifI1, implicated in nitrogen-fixation. MA1213 (vnfH) is annotated as an iron-containing nitrogenase, VnfH [16]. However, examination of this sequence suggests that it is more likely to be a molybdenum-containing nitrogenase [17]. The predicted ModE-binding site upstream of fmdE is also conserved in Methanosarcina mazei. Therefore it is possible that the ModE-like proteins Q8PWN4 and Q8TTZ2 from Methanosarcina mazei and Methanosarcina acetivorans respectively are functional for molybdate-responsive regulation of these molybdo-enzymes.


Transcriptional regulation of molybdo-enzymes, and genes involved in molybdenum metabolism and homeostasis, is performed by the molybdate-responsive transcription factor ModE in E. coli and related Proteobacteria. We found that homologues of ModE are also found in the green sulphur bacterium C. tepidum, and in methanogenic Archaea. Moreover, we found that its cognate DNA recognition element is also highly conserved even between Bacteria and Archaea. As far as we are aware this is only the third report of a regulatory DNA element whose sequence appears to be conserved in bacteria and Archaea. The other two examples are the regulator of biotin metabolism, BirA [18], and the metal dependent regulator MDR1/DtxR [24].

Although the basal transcription apparatus in Archaea is similar to eukaryotic RNAPII rather than to bacterial RNAP, the emerging view is that Archaea use a varied repertoire of regulatory mechanisms that includes both eukaryal and bacterial paradigms (e.g. [1922]). Archaea contain large numbers of HTH-containing domains, which are more similar to bacterial HTH domains rather than to eukaryotic proteins [23]. However, most of the Archaeal HTH-containing proteins form Archaea-specific families [23] and so the archaeal and bacterial HTH domains may share very ancient common origins [11]. The high-affinity molybdate transport system, which has a small size and simple organisation, may have a very ancient origin. The fact that the ModE regulon is conserved across large phylogenetic distances also suggests an ancient common origin followed by loss in multiple lineages rather than multiple transfer events. This may be further supported by the observation of partially degenerated ModE regulatory systems in bacteria such as A. tumefaciens and H. hepaticus.


Domain architectures were inferred from protein sequences using Pfam [9] release 10.0.

Matches to ModE-binding sites were identified in non-coding regions of complete genome sequences using a standard position-specific weight matrix scoring approach. A frequency matrix was generated from the alignment of known binding site sequences using the script [see Additional File 2]. This matrix contained the relative frequencies of each base being found at each position in the alignment (see Figure 5). The matrix can be considered to be a statistical model of the DNA sequence motif. Another way of visualising the DNA sequence motif is as a sequence logo [25] as in Figure 4.

Matches to this DNA motif were detected in a genome sequence by the following method. The entire genome sequence was scanned such that every window of 28 bases was assigned a score using the matrix and the script [see Additional File 1]. This score, known as the Kullback-Leibler distance, reflects the theoretical binding energy of the DNA protein interaction [26] and is calculated using the formula in Figure 6. The maximum possible score for a window, i.e. that given by a perfect match to the consensus, differs slightly from genome to genome according to percentage G+C content. Therefore, all scores were normalised such that 100 is the highest possible score for a 28 base window, and the scores were rounded to the nearest integer. Incidentally, we would expect a site scoring 100 only once in about 7.2 × 1016 bases of random DNA sequence (assuming 50 G+C content), so it is perhaps not surprising that no sites were found to have a score of 100.

Figure 6
figure 6

Formula used to calculate I seq , the Kullback-Leibler distance [26], where i is the position within the site, p b is the frequency of that base in the genome, and f b,i is the observed frequency of each base at that position (from the weight matrix). Values for p b were calculated from the percentage G+C content of the genome sequence.

We chose 75 as a cut-off threshold. That is, we only considered sequences that scored greater than 75. This threshold value was chosen as all of the known (and strongly suspected) ModE binding sites scored greater than 75.

The Perl scripts used for the analysis are available as Additional Files 1 and 2.