Although T-boxes were readily identified, it was more difficult to define their amino acid specificity. To that end, the phylogeny of homologous genes preceded by a T-box from different species was determined and the upstream regions corresponding to each orthologous group were aligned. In all the cases of T-boxes for which the specifier codon had been identified experimentally 3591112141516171821282933, we observed that the specifier codon aligned perfectly within the related orthologous sequences. In fact, this was true for almost all orthologous groups of sequences. Moreover, most of the alignments could easily be clustered by eye into larger groups for which the specifier codon remained directly apparent from the alignment. The resulting alignments and the annotation of the specifier codon can be found at 37. Nevertheless, there remained a few less clear cases. For some (~5%) a secondary structure prediction could be used to provide the additional information required to define the specifier codon in the specifier loop 38. Taken together, a specifier codon could be identified directly for over 90% of the identified T-boxes.

Most amino acids are encoded by multiple codons. Leu for instance, is encoded by six different codons (CUA, CUU, CUG, CUC, UUA and UUG). Remarkably, the T-boxes had a conserved preference for certain codons within as well as between species (Additional file 1). Evaluation of these preferences showed that they complied almost perfectly with the rules observed by Elf et al. for the codon usage by E. coli 39. In an elegant study these authors analyzed the dependence of the charging of various codon-specific tRNAs on the use of various codons in particular proteins. They concluded that: "when codon reading is part of a control loop that regulates synthesis of missing amino acid, the translation rate of the selected codon should be as sensitive as possible to starvation" 39. And, in their paper they showed which codons are the most sensitive in E. coli. We found that for all but one of the most predominantly used specifier codons in T-boxes, the corresponding tRNA is among the highest in sensing shortage of that specific amino acid in E. coli as reported by 39. The only exception was the T-box codon for Ala (GCU). Therefore, assuming the conclusions by Elf et al. are also valid for Gram-positive bacteria, our findings suggest that the codons that are sensitive to depletion are preferentially used in T-box regulation.

As expected, the proposed regulatory role of the T-box elements appeared to be perfectly reflected by the genes under their control. The majority of the T-boxes (62%) were found to precede genes encoding tRNA ligases, while most others were found upstream of genes encoding proteins involved in amino acid transport (12%) or amino acid biosynthesis (18%). The remaining T-boxes (8%; 71 genes in total) were found upstream of genes encoding proteins with unknown function (54 genes), or a function that lacks an apparent relation to amino acid metabolism (17 genes). A complete and species-specific subdivision of T-boxes based on function prediction of the proteins encoded downstream and a list of genes with no apparent relation to amino acid metabolism is provided in the supplementary material, in Additional files 1 and 2.

Many of the genes encoding amino acid transporters were found to be preceded by a T-box, especially in the genomes of the Lactobacilli and Bacilli of the Bacillus cereus-group. The transporters controlled by T-boxes belonged to no less than seven distinct transporter families (MFS, 2.A.1; APC, 2.A.3; NSS, 2.A.22; DAACS, 2.A.23; LIVCS, 2.A.26; NhaC, 2.A.35; and ABC-cassette, 3.A.1; Transporter Classification described by Saier 41). Table 1 gives an overview of the distribution of the transport systems regulated by a T-box over the various Firmicutes species.

Overall, for more than 85% of the T-box regulated transporters the functional annotation (molecular function and/or biological role) could be improved as compared to the entries in the reference database of NCBI. A full list can be found in Additional file 1. We have limited the substrate specificity definition in our annotation to putatively dominant substrates based on the amino acid specificity of the T-box. However, broader substrate specificity is probably more common for transporters. Especially transport systems consisting of only a permease are expected (and have been shown) to display broader substrate specificity (see 42 and 43 for examples), whereas systems that require prior substrate-binding (like in ABC transport) will be more specific. We discuss the T-box based functional annotation of some transport systems in more detail in the following paragraphs and in Additional file 3.

T-box regulation of ABC transport systems was found in most lineages of the Firmicutes but not in the Bacilli. The T-box regulated ABC transporters could be sub-divided into four sub-families, based on the specificity of the substrate-binding protein and the permease. A striking use of extensive T-box regulation in ABC transport was observed in L. plantarum. It appears that in the absence of methionine, L. plantarum uses a single mechanism to switch on not only transport of the amino acid itself, but also of the precursors and co-factors needed for its biosynthesis (see Figure 2 and Additional file 3).

A T-box was identified in front of an APC-family protein encoding gene in all the studied Bacillus genomes. In B. subtilis and B. licheniformis the gene ybvW is preceded by a Leu T-box, whereas such a box is lacking upstream of the orthologous genes, which are found in E. faecalis, G. kaustophilus and in L. lactis (co-orthologs: yibG and ysjA) (Figure 3). The Leu T-box suggests that the YbvW protein is a Leucine transporter, in line with the general functionality of transporters of the APC family (family characteristics described in 4445). Surprisingly, in the members of the Bacillus cereus-group another APC family gene is preceded by a T-box, specific for threonine. Although an orthologous gene is present in most of the Firmicutes genomes, e.g. ykbA in B. subtilis, it is regulated by a T-box only in the species of the Bacillus cereus-group (Figure 3). The protein encoded by ykbA in B. subtilis has recently been shown to be a Ser/Thr exchanger and was consequently renamed SteT 45. A similar functionality of the protein ortholog in the members of the Bacillus cereus group is supported by the codon identification of the T-box.

The Bacilli of the Bacillus cereus group, the Lactobacilli and the Clostridia contain several branched-chain amino acid cation symporters of the LIVCS-family 46, some of which are T-box regulated (Figure 3). Since the three branched-chain amino acids share very similar molecular properties (e.g. size and hydrophobicity) we expect that these transporters are not highly specific despite their proposed amino acid specific control, but merely that expression of the "multi-specific" system has been brought under the control of the individual amino acids. Indeed, the orthologous transporters that have been characterized in L. delbrueckii (BrnQ; 43), C. glutamicum (BrnQ; 4748) and P. aeruginosa (BraZ; 49) displayed transport of all three branched-chain amino acids.

Finally, the LIVCS family branched-chain amino acid transporter BraZ of P. aeruginosa, was shown to have a clear preference for isoleucine and valine over leucine 49. In this respect it is noteworthy that in the Bacilli of the Bacillus cereus group the expression of one of the homologs of the NSS family (neurotransmitter:sodium symport) is controlled by a Leu T-box, whereas such a box is lacking for the LIVCS homologs in those species (Figure 4). Besides the Leu T-box regulated NSS transporter, the Bacilli of the Bacillus cereus group contain three other homologs of the same family, two of which are controlled by a Trp and a Phe T-box, respectively (Figure 3). These two amino acids agree well with the experimentally determined tryptophan transport functionality of the NSS homolog TnaT in S. thermophilum 50. The presence of a regulatory T-box ranging from Leu to Trp, and Phe suggest that the members of the NSS transporter family may display a rather broad amino acid specificity.

Another class of proteins to which significant functional information could be added (when compared to the NCBI-annotation) using the specificity of the detected T-box is that of the so-called hypothetical proteins or unknown function proteins (data accumulated in Table 2). Obviously, when orthologous proteins in related species were also of unknown function, specifier codon information clearly improved the annotation. Examples of new annotations related to amino acid biosynthesis or transport are enzymes (methionine synthase, cystathionine gamma synthase, chorismate mutase, anthranilate synthase), transporters (Leu-, Lys- and His-specific permeases), tRNA-ligase related functions, and regulation (anti-TRAP protein).

It appeared (Figure 4) that regulation by T-boxes is conserved in almost all (at least 29 out of 34) Firmicutes for several tRNA ligases (ileS, a laS, serS and thrS), whereas some tRNA ligases (lysS, asnS and gltX; the latter gene encodes a Glu-tRNA ligase (see also the legend of Figure 4)) appeared to be controlled by a T-box in only a few species. The genes encoding the tRNA ligases for cysteine and asparagine were often found as the second gene in a putative operon that was T-box regulated. In several organisms, multiple copies of amino-acid specific tRNA-ligase encoding genes are found and in more than half of the cases (58%) only one of them is subject to T-box regulation.

A clear phylogenetic effect is observed when the four major orders within the Firmicutes (Bacillales, Clostridia, Lactobacillales and Mollicutes) are compared. This is true for both the number and the type of tRNA-ligase encoding genes regulated by a T-box. Most T-box regulated tRNA-ligase encoding genes are found in the Bacillales and especially in species of the Bacillus cereus group. Within the Lactobacillales, it appears that Streptococci have far less tRNA- ligase encoding genes regulated by T-boxes than Lactobacillus species. The lowest number of tRNA ligases regulated by T-boxes was found for S. thermophilus, S. pneumoniae and some strains of S. pyogenes. The relatively low amount of T-box regulation in these species could be the result of regressive evolution, a process that was suggested to be the underlying mechanism for the large loss of functionally active genes in S. thermophilus 54.

T-box regulation of genes related to amino acid biosynthesis has been described previously for various amino acids 32324. Like in the case of the t-RNA ligases, T-box control of amino acid biosynthesis displayed clear phylogenetic patterns (Figure 5). For instance, the biosynthesis of Branched Chain Amino Acids (BCA: isoleucine, leucine, valine) was found to be T-box regulated in Bacillales and Clostridia, whereas several families within the Bacillales (e.g. Staphylococci and Listeria) as well as several Streptococci consistently lack T-box control of BCA biosynthesis. Similarly, we found that the species of the B. cereus group contain a T-box in the upstream region of the tyrosine biosynthesis operon consistent with the experimental data that showed that tyrosine biosynthesis from shikimate is T-box regulated in B. anthracis 24. The B. cereus group representatives are the only organisms in our study that encode a phenylalanine-4-hydroxylase ortholog (converting phenylalanine into tyrosine). We found this gene also to be T-box regulated in all members of the B. cereus group.

An important observation related to T-box evolution was made by Grundy et al. 912. They showed that a single nucleotide change in the specifier codon of the Tyr T-box of tyrS in B. subtilis was enough to change the amino acid specificity. In addition, while analyzing the distribution of T-boxes over the various transporter families we were struck by the fact that although some of these families are very large, there was only one (or a few) family-member(s) found to be regulated by a T-box and only in a restricted number of species (see Table 1 and e.g. Figure 3). In our opinion, the only likely scenario to explain the phylogenetically limited occurrence of the transporter T-box associations that would not imply massive loss of the T-box regulation was that of acquisition of the regulatory element by the transporter encoding gene in a specific lineage. Moreover, the results of Grundy et al. 912 imply that in principle the T-Box can change specificity easily.

To analyze this further, the T-boxes preceding the genes that encode the t-RNA-ligases for the branched-chain amino acids (ileS, leuS and valS) were examined. These were chosen because T-box regulation of these genes is most wide-spread (Figure 4) and because the proximal genes themselves, ileS, leuS and valS form a separate tRNA-ligase sub-family with a very clear evolutionary lineage (Figure 6A). In sharp contrast, the NJ-tree of the aligned complete T-boxes (approximately 200 – 300 nt) appeared extremely unreliable (Figure 6B). This finding implies that amino acid specificity of a T-box is not prominent on the overall sequence level and that the apparent overall sequence variability in time of the T-box is relatively high. Simultaneously, in case T-box sequences do cluster in reliable clusters (high bootstrap support) in a NJ-tree they must therefore be closely related in time.

To limit possible obscuring effects of comparing sequences between species, we collected and compared the T-box sequences within species and for clarity restricted the comparison to the T-boxes that accompany transport systems and the related tRNA ligases. For all three analyzed species: B. anthracis, L. acidophilus and L. plantarum. similar phenomena were observed. The multiple sequence alignment of the analyzed T-box sequences and the associated NJ-tree (depicted in Figure 6C for B. anthracis) are strongly suggestive of a close evolutionary relationship between several of the T-boxes. For example in B. anthracis, the Thr T-boxes found in front of BA4970 (transport systems of LIVCS-type) and the Thr-tRNA ligase (BA4820) were highly similar and the same was observed for the Ile T-boxes found in front of another LIVCS homolog and the Ile-tRNA ligase. As the LIVCS homologs appear closely related in time (apparent duplication in the B. cereus group ancestor, see Figure 3), the data imply that the regulatory T-box was not inherited in a similar way but 'acquired' independently. In fact, this explanation fits the observed scattered appearance of the T-boxes for the various transporter families perfectly.

The results presented in Figure 6 are also suggestive of another way in which the T-boxes have evolved. The NJ-tree relates the Phe T-box found in front of one of the NSS family transporters to the Tyr T-box associated with the Tyr-tRNA ligase (Figure 6C). It thus seems that the Tyr T-box of the tRNA ligase was duplicated -as this T-box is present in various Firmicutes species- and has diverged/adapted to control a Phe transporter in the Bacilli of the B. cereus group. In fact, the similarity between the T-box upstream of Tyr-tRNA ligase and the Phe T-box in front of the transporter is higher than between the Tyr T-box and the Tyr T-box preceding BA4353 (NhaC family transporter). This is consistent with the fact that: the Tyr T-box acquisition of the NhaC ortholog should have occurred earlier in history, as the Tyr T-box control of the NhaC ortholog is present in several Firmicutes, and the sequences thus had more time to diverge.