Previous sequence comparisons using BLAST analyses indicated that one mal operon (the ancestral mal1 operon, mal1_An, along with the ancestral malK2_An gene in Figure 1) transferred from an ancestor of Tc. litoralis to an ancestor of P. furiosus, likely involving a transposition mediated by the insertion sequences now found in P. furiosus 8. Based on a simple comparison of the genes encoding ABPs in the mal and mdx operons in Tc. litoralis with their homologs in P. furiosus, a copy of an ancestral mdxK gene (malK2_An) was postulated to have recombined downstream of the mal1_An operon in an ancestor of Tc. litoralis (Fig. 1), where it now encodes the ABP for the mal-encoded transporter 18. This ancestral mdxK could have originated from gene duplication or horizontal acquisition. We performed phylogenetic analysis of all the proteins encoded by the mal and mdx operons to test these hypotheses and to determine the origins of these evolutionarily mobile genes.

The MdxE_Pf (PF1938) protein sequence was used as a query for a BLAST search of the non-redundant protein database at NCBI. The top 250 hits were aligned using ClustalX and then used to construct neighbor joining trees. Using these preliminary relationships as a guide, sequences were chosen from archaea, Thermotogales species, and bacteria closely related to the archaeal sequences to represent a broad spectrum of taxa. These selected sequences were then placed in a new database (approximately 60 sequences for each gene). Associated MalF and MalG protein sequences were concatenated with each MalE-like sequence only if these three ORFs were the only ORFs in an apparent ABC transporter operon. These concatenated sequences were aligned and their phylogenetic relationships were examined by maximum likelihood (ML) and Bayesian analyses. The tree constructed from this ML analysis is shown in Figure 2 with ML bootstrap values and Bayesian posterior probabilities for each analysis placed thereon.

The most obvious feature of this tree in regard to the Thermococcales sequences is the independent evolution of the archaeal mal and mdx operons. The distant evolutionary relationship of these two operons is well supported. The Mdx sequences cluster with strong support with those from other archaea while the Mal sequences cluster separate from other archaeal sequences. The latter cluster with sequences from a variety of bacteria including cyanobacteria and members of the Thermotogales. These data support the independent evolution of these two sets of genes and suggest that the mal genes found in Tc. litoralis and P. furiosus were acquired from a bacterium by an ancestral member of the Thermococcales (Fig. 1). We shall discuss the relationship of these Thermotogales operons to the Thermococcales mal operons after considering the history of the Thermococcales malK genes.

The archaeal malK genes have undergone a very different evolutionary history than those in their adjoining mal operons. We examined the relationships among malK homologs using both MrBayes and PHYML and both gave similar, though not identical trees, each with overall relatively weak support. However, within those trees, clusters of strong support were found and these provide reliable information about the evolutionary relationships within those clusters. As shown in the PHYML-derived phylogeny in Figure 3, all the Thermococcales malK and mdxK homologs group together in a well-supported clade. This supports the hypothesis that an ancestral malK2_An homolog either was duplicated in an ancestral Thermococcales or was horizontally acquired by that ancestor from another Thermococcales. The malK2_An was later transferred along with the mal1_An operon to a P. furiosus ancestor (Fig. 1) 18. In these genomes, these malK genes are near, but not immediately adjacent to, the three-gene mal/mdx ABC operons, each encoding two MSPs and one SBP. An amylopullulanase gene typically separates the malK homolog from the other genes, but some of these clusters have more genes in the intervening region, perhaps indicative of other HGT or deletion events in some species.

Since the Tt. maritima mal operons do not contain adjacent malK homologs, we sought in this tree any indication of the identity of a possible Thermotoga malK homolog, perhaps derived from the archaeal malK or mdxK genes. Unfortunately, only two Tt. maritima ORFs appear in Fig. 3 among archaeal sequences and all are distantly related to the Thermococcales malK/mdxK cluster. The functions of these archaeal transporters are unknown and since none of their ABP or MSP homologs appeared in the BLAST results using the P. furiosus MalE, MalF or MalG sequences as queries, they unlikely to be mal-related ABPs. A function of TM0421, one of the Tt. maritima ORFs found in the vicinity of these archaeal sequences (though with low bootstrap support), has been suggested. Data has indicated that it is a myo-inositol transporter's ABP (InoK) 4. The function of the other ORF, TM1232, is unknown, though its expression was reported to be upregulated when Tt. maritima was grown in co-culture with Methanocaldococcus jannaschii 23.

A relatively weakly supported cluster of Thermotogales ORFs apparent in Fig. 3 is unrelated to any archaeal ORFs. This cluster includes the Tt. maritima ORF TM1276 which, based upon gene expression data, has been suggested to encode a maltose transporter ABP 2425. Unfortunately this analysis does not provide additional support for that observation. It does not appear that any archaeal malK/mdxK homolog was acquired by ancestors of the currently sequenced species of Thermotogales.

Using the P. furiosus MalE sequence as a query, homologs were revealed in the genome sequences from F. nodosum, Ts. melanesiensis, Tt. petrophila, and Tt. neapolitana but not Tt. maritima (Fig. 2). Each of these genomes contains complete malF and malG genes downstream from this malE homolog in the order malEFG. For Tt. petrophila and Tt. neapolitana this reveals a third apparent mal operon. Ts. melanesiensis and F. nodosum have only one other mal operon (see below), so this is a second for them. This third operon is closely more related to the mal operons of Tc. litoralis and P. furiosus than are the mal1 and mal2 operons (Fig. 2). Although in that figure they may appear to be specific relatives of one another, that relationship is not strongly supported by either maximum likelihood or Bayesian analyses.

To examine the possibility of HGT of these mal3 genes between the Archaea and the Thermotogales, we aligned and analyzed sequences from several additional bacteria in the clade that includes mal3. That analysis, shown in Figure 4, does not show a specific association of the Thermococcales Mal sequences with these Thermotogales sequences, discounting the possibility of HGT of these genes between the Thermococcales and Thermotogales.

Interestingly, the additional mal operon in Tt. lettingae (depicted here as "Mal3_Ttl" though it only has two mal operons) is not closely related to its homologs in the other members of the Thermotogales. It is also not a specific relative of the archaeal mal genes. It was likely acquired from a bacterial group different than the donor to the other Thermotogales.

Although the Tt. maritima Mal paralogs are associated with sequences from enteric bacteria in the P. furiosus mdxE-derived phylogeny (Fig. 2), the resolution of that analysis was too low to determine the detailed relationships among the bacterial orthologs. To examine the Tt. maritima mal1/mal2 evolutionary history in detail and to determine the relationships among the mal1 and mal2 genes in the Thermotogales, we used the MalE1_Ttm (TM1204) sequence as a query and selected from that dataset sequences from those genes that are arranged in three-gene operons as described previously. Those sequences were concatenated with their MalF and MalG partners. The concatenated sequences from the Mdx_Pf operon was included as an outgroup and all these concatenates were aligned using ClustalX. Trees were constructed using ML and Bayesian analyses as above.

The resulting phylogenetic tree (Fig. 5) grouped Thermotogales sequences together and the relative arrangement of species in this group reflects their branching order as derived using 16S rRNA gene sequence comparisons 326. Tt. lettingae, F. nodosum, and Ts. melanesiensis have only one of this type of mal operon while the remaining three Thermotoga species each have two of this kind (Fig. 5). These mal operons in F. nodosum and Ts. melanesiensis are related to one another to the exclusion of those from the Thermotoga species. The single Tt. lettingae mal operon is most closely related to the mal1 orthologs. The mal2 operon appears to have arisen within the Thermotoga lineage either through a duplication of an ancestral mal1 or through a horizontal acquisition from an unknown close relative.

An examination of the genomic contexts of these mal operons suggests that a simple loss of one of the mal operons does not best explain this evolutionary history. As shown in Table 1, the mal operons in F. nodosum and Ts. melanesiensis are not located in an area of the genome that has synteny with any of the other species considered here. This genomic context analysis also supports the observation that this mal operon in Tt. lettingae belongs to the mal1 cluster and that mal1 was likely found in the ancestor of all these Thermotoga species. We found no region in the Tt. lettingae genome sequence that contains a cluster of homologs of the genes surrounding the mal2 operon of the other species. Consequently, a simple loss of mal2 from the Tt. lettingae lineage is not evident. A duplication of mal1 after the divergence of Tt. lettingae would cause the Mal1 and Mal2 lineages to be exclusively related to one another, but that is not apparent in Fig. 5. Consequently, mal2 does not appear to have arisen by a duplication of mal1. Rather the data indicate that mal2 was acquired by an ancestral Thermotoga species from an unknown the Thermotoga species after the divergence of Tt. lettingae.

With the exception of Tt. maritima all species examined here have intact malF2 genes. Tt. maritima has a malF2 pseudogene. Our analyses cannot determine whether the mutations in malF2_Ttm occurred during laboratory cultivation of this strain or prior to its isolation from nature.

Sequence comparisons showed that the Thermotogales cluster of Mal1 and Mal2 proteins are specifically related to those from enteric bacteria, so we sought other evidence to support this unusual evolutionary association of proteins from such different kinds of bacteria. We examined putative structural features of the MalF and MalG membrane proteins by hydropathy analyses using the TMpred program 27.

The MalF_Ec is a member of the CUT1 (carbohydrate uptake) family 28 and is one of the most extensively studied membrane permeases 29. Unlike other membrane permeases, MalF_Ec is unusual in that it consists of eight transmembrane helices instead of six and has a large periplasmic domain of 180 amino acids between transmembrane helices 3 and 4 (Fig. 6) 30. This peculiar membrane topology is reportedly conserved in all MalF homologs from bacteria closely related to E. coli 30. The function of this domain is unknown, but mutational alterations in this loop affect the localization of MalK_Ec to the membrane bound MalF_Ec 29.

To determine if the MalF topology found in E. coli is conserved in the MalF homologs from its close relatives, including members of the Thermotogales, transmembrane topological analyses of these sequences were performed. Representative data from those TMpred analyses are shown in Figure 6 and Additional file 1. Both MalF1 and MalF2 homologs from all members of the Thermotogales show the eight transmembrane helices, large transmembrane loop, and ABP-interaction motif found in MalF_Ec. These unusual features support the close relationship between the Thermotogales MalF1 and MalF2 homologs and those from E. coli relatives. The phylogenetic analyses reported above are not influenced by the presence of this large loop since tree topologies are not changed when these loops were removed from the sequences prior to alignment (data not shown).

The MalF homolog from Propionibacterium acnes, also related to the E. coli MalF, has a smaller, but still evident, loop (see Additional file 1). The conservation of this unique large loop in the MalF permeases in these distantly related organisms is consistent with their shared evolutionary history. Interestingly the MalF from Deinococcus radiodurans (and Thermus thermophilus, not shown) has a smaller loop while Thermoanaerobacter tengcongensis MalF (and those closely related to it) show no evidence of a loop (see Additional file 1). Whatever its function, this loop is more pronounced in close relatives of the enteric form of MalF and was apparently lost from the largely gram positive clade of organisms.

The MalF3 homologs found in some members of the Thermotogales and in P. furiosus and Tc. litoralis do not contain this loop feature (Fig. 6). This observation supports the large evolutionary distance between the mal3 genes and the mal1-mal2 genes.

The MalG1 and MalG2 homologs in the Thermotogales are unusually large as compared to their homologs from other bacteria including E. coli (Fig. 7). The alignments of MalG homologs showed that the C-terminal 300 amino acids of MalG1_Ttm and MalG2_Ttm are similar to the MalG homologs from E. coli and related bacteria, but the N-terminal 525 amino acids showed no significant sequence similarity to any known proteins. TMpred analyses revealed a large hydrophilic region of about 500 amino acids between transmembrane helices one and two of these MalG1 and MalG2 sequences (Fig. 7). The MalG3 homologs lack this region, supporting their phylogenetic placement relative to Mal1/Mal2 (Fig. 7).

Protein sequences were retrieved by BLASTP searches of the nonredundant protein database at NCBI 34 using as queries amino acid sequences of Tt. maritima, P. furiosus, or Tc. litoralis MalE, MalF, or MalG. The P. furiosus MalK2 (PF1933) was used as the query to search for MalK homologs among bacteria and archaea. The top 250 sequences retrieved by BLASTP searches were assembled into datasets and repeated sequence entries and closely related sequences (determined using neighbor-joining trees of aligned sequences) were removed. Those sequences from three-gene operons (malE, malF and malG) were retained and the cognate genes were concatenated. Alignments of the amino acid sequences of these concatenates were prepared using ClustalX v1.83.1 35.

TM1837, the malF2_Ttm ORF, contains an authentic frameshift mutation. To make comparisons between MalF homologs, the pseudogene was corrected by removing nucleotides at positions 254 (A) and 298 (G) (identified using GENIO/frame 36). Corrections at these positions restored the reading frame without introducing new stop codons. malF2_Tm is preceded by a small ORF, TM1838. The putative 54 amino acid peptide product of TM1838 is 53% similar to the amino terminus of MalF1_Ttm. By adding a nucleotide between positions 138 and 139 in the TM1838 sequence and combining it with the corrected malF2_Ttm, a single polypeptide is obtained that has a high similarity to malF1_Tm as measured by PRSS (P value 1.6 e-21) 37. It appears that a series of point mutations in an ancestral malF2 created TM1838 and the translationally truncated malF2_Ttm. Consequently, there is no functional MalF2_Ttm protein produced by Tt. maritima.

Aligned sequences from the concatenate datasets were used to construct consensus trees using MrBayes v3.1 38 and PHYML v2.4.4. The MrBayes analyses were performed using the default GTR model and gamma distributed rate variation with two runs, each with four chains, for 1,000,000 generations (780,000 for the tree shown in Fig. 2) and taking a consensus tree after a burn-in of either 100,000 or 65,000 generations. These analyses were performed at the Bioinformatics Facility of the University of Connecticut Biotechnology Bioservices Center. Maximum likelihood analyses of these alignments were performed in PHYML with 1000 bootstrap resamplings, the JTT substitution model, a fixed proportion of invariable sites, one category of substitution rate, and the BIONJ input tree.

The analyses above were confirmed with alternative alignments using MUSCLE 39 and T-COFFEE 40 each using their default parameters. These analyses produced phylogenies nearly identical to those shown here when used with PHYML (WAG model, estimated pinvar, estimated gamma distribution) (not shown). Transmembrane topological analyses were performed using TMpred 4127.