|Home | About | Journals | Submit | Contact Us | Français|
Comparative genomics approaches are broadly used for analysis of transcriptional regulation in bacterial genomes. In this work, we identified binding sites and reconstructed regulons for 33 orthologous groups of transcription factors (TFs) in 196 reference genomes from 21 taxonomic groups of Proteobacteria. Overall, we predict over 10 600 TF binding sites and identified more than 15 600 target genes for 1896 TFs constituting the studied orthologous groups of regulators. These include a set of orthologues for 21 metabolism-associated TFs from Escherichia coli and/or Shewanella that are conserved in five or more taxonomic groups and several additional TFs that represent non-orthologous substitutions of the metabolic regulators in some lineages of Proteobacteria. By comparing gene contents of the reconstructed regulons, we identified the core, taxonomy-specific and genome-specific TF regulon members and classified them by their metabolic functions. Detailed analysis of ArgR, TyrR, TrpR, HutC, HypR and other amino-acid-specific regulons demonstrated remarkable differences in regulatory strategies used by various lineages of Proteobacteria. The obtained genomic collection of in silico reconstructed TF regulons contains a large number of new regulatory interactions that await future experimental validation. The collection provides a framework for future evolutionary studies of transcriptional regulatory networks in Bacteria. It can be also used for functional annotation of putative metabolic transporters and enzymes that are abundant in the reconstructed regulons.
Bacteria in most ecological niches are constantly exposed to variations in many factors, including nutrient availability. Changes in gene expression using transcription factors (TFs) allow bacteria to adapt to these variations. Knowledge of transcriptional regulatory networks is essential for understanding cellular processes. Comparative genomics is the analysis and comparison of genomes from different species. Thousands of sequenced bacterial genomes provide an opportunity to reconstruct transcriptional regulatory networks using comparative genomics. Despite the importance of transcriptional regulation of the central metabolism for systems-level metabolic modelling of Bacteria, our understanding of the respective transcription factor regulons is limited for the majority of sequenced bacteria. In this study, we have applied the comparative genomics approach to describe regulatory networks of genes involved in central metabolism in four major classes of Proteobacteria. The reconstructed regulatory networks involve 33 groups of orthologous TFs with different DNA recognition motifs. Large-scale phylogenomic analysis of the reconstructed TF regulons reveals and classifies various evolutionary processes that shape regulatory networks in Bacteria. The reconstructed regulon contents suggest numerous novel functional associations between both known and uncharacterized genes encoding enzymes and transporters, thus providing testable hypotheses for future experimental studies. This study demonstrates the power of comparative genomics for the reconstruction of TF regulons in bacteria.
Regulation of gene expression is an important mechanism for fast adaptation of prokaryotic metabolism to changing environmental conditions. Transcription factors (TFs) repress or activate gene transcription via specific binding to TF binding sites (TFBSs) in regulatory gene regions. The binding ability of many bacterial TFs depends on the presence or absence of an effector, such as intracellular metabolites, inorganic chemicals or physical stimuli (Browning & Busby, 2004). A set of genes directly controlled by a single TF is called a regulon. Global TF regulons in bacteria contain large sets of genes (operons) that share similar TFBSs in their promoter regions, while local TFs control one or several operons that are often co-localized with a TF gene (Rodionov, 2007).
Experimental studies have built a foundation for understanding the mechanisms behind transcription regulation (Minchin & Busby, 2009). However, even with high-throughput technologies as ChIP-Seq or RNA-Seq, these approaches still consume a lot of time and resources and therefore are restricted to a few model organisms (Grainger et al., 2009). At this point, comparative genomics studies of a growing number of sequenced bacterial genomes provide a successful approach to extend our knowledge of known TF regulons to a wide range of bacterial lineages, as well as to perform ab initio prediction of novel TF regulons (Rodionov, 2007). Comparative genomics-based regulon reconstruction combines identification of conserved cis-acting TFBSs, and their genomic and metabolic context analysis in a set of closely related genomes. Finally, it results in determination of a regulog, i.e. a set of genes/operons co-regulated by orthologous TFs in closely related organisms. Implementation of this approach in the RegPredict web tool (Novichkov et al., 2013) produced numerous computational reconstructions of TF regulogs across a wide range of bacterial taxa (Ravcheev et al., 2011, 2013; Rodionov et al., 2011, 2013; Leyn et al., 2013). The substantial amount of data on regulon reconstructions captured in the RegPrecise database (Novichkov et al., 2013) provides the basis for description of most common types of events associated with the evolution of TF regulons in bacteria, such as duplications and losses of TFs and their TFBSs that result in expansions, shrinkages, mergers and split-ups of regulons (Gelfand, 2006; Rodionov et al., 2006; Ravcheev et al., 2014). New non-orthologous TFs could be introduced to control equivalent pathways or, vice versa, orthologous TFs could control distinct pathways in related taxonomic groups of bacteria (Yang et al., 2006; Rodionov et al., 2008; Kazakov et al., 2009; Leyn et al., 2014).
We recently conducted a comprehensive comparative genomics analysis of regulatory systems for methionine metabolism in nearly 200 representative genomes from 22 taxonomic groups from the phylum Proteobacteria (Leyn et al., 2014). In he Gammaproteobacteria, two TFs, MetJ and MetR, are implicated in the control of methionine metabolism, whereas this function is taken by other TFs (SahR and SamR) or RNA regulatory systems (e.g. SAH and SAM riboswitches) in other lineages of Proteobacteria. The core of MetJ regulons includes a large number of genes that are highly conserved in most lineages of gammaproteobacteria. In contrast, the core of MetR regulons includes only two genes, metE and metR, whereas regulatory interactions between MetR and other target genes are mostly lineage-specific. Regulatory system replacement and lineage-specific regulon expansions in Proteobacteria were also observed in the comparative genomics analyses of TF regulons involved in fatty acid degradation (FadR, PsrA, FadP), branched-chain amino acid utilization (LiuR, LiuQ), N-acetylglucosamine utilization (NagC, NagR, NagQ) (Yang et al., 2006; Kazakov et al., 2009), biotin biosynthesis (BirA, BioR) (Rodionov & Gelfand, 2006) and central carbohydrate metabolism (HexR) (Leyn et al., 2011).
Here, we extended these observations toward large-scale regulon reconstructions for 21 known TFs that have orthologues in a wide phylogenetic range of Proteobacteria. By comparing the metabolic context of the reconstructed TF regulons, we identified the core, taxonomy-specific and genome-specific members of regulons, and proposed evolutionary scenarios for the regulation of several pathways involved in the metabolism of amino and fatty acids, nucleotides and co-factors in Proteobacteria. Additionally, we predicted novel regulators of aromatic amino acid metabolism replacing the TyrR/PhrR and HmgR regulons in Alteromonadales and Pseudomonadales (named HmgS and HmgQ), and a novel regulator of NAD metabolism in betaproteobacteria and alphaproteobacteria, named NadQ. The obtained regulatory reconstructions for both known and new TF regulons across 196 reference genomes of Proteobacteria will be useful for development of theoretical models for the evolution of microbial regulatory networks.
For regulon reconstruction, we selected 196 reference genomes of alpha-, beta-, gamma and deltaproteobacteria and subdivided them into 21 sets of evolutionarily related genomes (Table S1, available in the online Supplementary Material). Closely related strains and species were excluded from the analysis because they skew the TFBS training set and thus decrease the sensitivity of the TFBS recognition rule. Genomes and the phylogenetic species tree were downloaded from MicrobesOnline database (Dehal et al., 2010). Each taxonomic group includes 4–16 genomes of bacteria. Orthologues of TFs in the selected genomes were identified as bidirectional best hits using protein blast searches (Altschul et al., 1997) and were additionally confirmed via phylogenetic trees using precomputed protein trees in MicrobesOnline. The genomes of epsilonproteobacteria were not analysed because of the absence of orthologues for target TFs. Zetaproteobacteria, which is represented by a single genome in the MicrobesOnline database, was not suitable for the comparative genomic analysis.
Genes in the reconstructed regulons were considered orthologues if they were classified as specific tree-based orthologues in MicrobesOnline. Conservancy of the genomic context through related genomes was considered as an additional support for gene orthology. Comparative analysis of conserved gene neighborhoods was conducted in MicrobesOnline. Biological functions of genes were predicted by blast searches against the SwissProt/Uniprot database (UniProt C, 2014) , by domain architecture analysis in the Pfam database (Finn et al., 2014) and by using gene function assignments in the PubSEED database (Overbeek et al., 2005). Known metabolic pathways were taken from KEGG (Kanehisa & Goto, 2000) and EcoCyc (Karp et al., 2014). Sequence logos for TF binding sites were drawn using the WebLogo package (Crooks et al., 2004).
For regulon reconstruction we used an established comparative genomics approach implemented in the RegPredict interactive tool (Novichkov et al., 2013). This approach is based on construction of positional weight matrices (PWMs) for TFBS motifs, and further genomic searches for additional regulon members on the basis of predicted TFBSs in upstream gene regions (Rodionov, 2007). The bioinformatics workflow used for regulon reconstruction is described in Fig. S1. Two main workflows were applied for regulon reconstructions: (i) propagation and expansion of known TF regulons that were previously experimentally studied in model organisms (Table S2) and/or computationally reconstructed in Shewanella species (Rodionov et al., 2011); and (ii) ab initio prediction of novel TF regulons for sets of potential target genes involved in the same metabolic pathway. To find conserved TFBS motifs for the known TFs in each taxonomic group where their orthologues are present, we used initial training sets of genes that are orthologous to previously established regulon members in model species, and then updated each set by potential regulon members confirmed by the comparative genomics checks. For novel TF regulons, the original training sets included genes from the respective metabolic pathways and/or conservative chromosomal gene neighbourhoods around analysed TFs.
A simple iterative procedure implemented in the Discover Profile tool in RegPredict was used for identification of conserved palindromic DNA motifs and construction of PWMs. For most of the analysed TFs, their DNA motifs have palindromic structure and length between 15 and 25 nt, whereas the TFBS motifs of NagQ and BirA represent tandem and inverted repeats, respectively. The obtained PMWs (both known and ab initio predicted) were further used for identification of additional candidate sites in upstream gene regions as previously described (Leyn et al., 2014; Ravcheev et al., 2014). Each predicted regulatory interaction was analysed for conservation within the analysed groups of genomes using the Clusters of co-Regulated Orthologous operoNs (CRONs) approach in RegPredict. Further analysis of functional and genomic context and curation of each CRON resulted in the final TF regulon model. All reconstructed TF regulons including TFBS motifs and sets of TF-regulated genes/operons with their functional annotations are accessible in the latest release of the RegPrecise database (Novichkov et al., 2013) (Data Citation 1). Each TF regulon in RegPrecise belong to two types of regulon collections classified by either taxonomy of studied bacteria, or by the name of TFs.
A set of 196 representative genomes of gamma-, beta-, alpha- and deltaproteobacteria selected from the MicrobesOnline database was classified into 21 taxonomic groups by analysing the phylogenetic species tree (Table S1). For the analysis of evolution of transcriptional regulation, we selected a set of 21 transcriptional regulators of central metabolism that are present either in E. coli and/or Shewanella species and that are conserved in five or more taxonomic groups of Proteobacteria (Table 1; Fig. S1). The selected TFs include the previously known regulators that control biosynthesis/utilization of amino acids (ArgR, HutC, HypR, LiuR, MetJ, MetR, TrpR, TyrR), fatty acids (FabR, FadR, PsrA), nucleotides (NrdR, RutR) and vitamins (BirA, NrtR), as well as nitrogen and carbon metabolism (HexR, GlcC, LldR, NagC, NtrC, PdhR). Sixteen of these TFs are present in E. coli, of which 12 regulators are also shared by Shewanella species, whereas the remaining five TFs (LiuR, HutC, HypR, NrtR, PsrA) are unique for Shewanella species. We also studied 12 additional TFs that appear to substitute for some of the above TFs in the control of specific metabolic pathways, and thus can be assumed to benon-orthologous TF replacements. These include known and predicted regulators that control metabolism of amino acids (HmgQ, HmgR, HmgS, LiuQ, SahR, SamR) and fatty acids (FadP), vitamin biosynthesis (BioR, NadR, NadQ) and N-acetyl-glucosamine metabolism (NagQ, NagR) (marked with an asterisk in Table 1).
Application of the comparative genomics procedure to 33 analysed groups of orthologous TFs resulted in reconstruction of 283 regulogs containing 1896 regulons that are unevenly distributed across 21 taxonomic groups of Proteobacteria (Fig. S2). Each regulon includes a set of target genes/operons that are co-regulated by the same TF in a particular genome. A regulog represents a set of regulons under control of orthologous TFs in a specific taxonomic group of Proteobacteria. The most widespread orthologous groups of analysed TFs are NrdR (186 regulons, 20 regulogs), NtrC (169 regulons, 19 regulogs), MetR (117 regulons, 14 regulogs), HutC (113 regulons, 18 regulogs), LiuR (104 regulons, 16 regulogs), HexR (95 regulons, 13 regulogs) and BirA (94 regulons, 11 regulogs).
The taxonomic distribution of analysed TF regulogs across four subdivisions from the phylum Proteobacteria is summarized in Fig. 1. Overall, 30 out of 33 analysed TFs are present in gammaproteobacteria, and 14 of these regulators do not have orthologues in other classes of Proteobacteria. Alphaproteobacteria possess, in total, 15 studied TFs, including one regulator (BioR), which is unique for this class. Among 17 studied TFs in betaproteobacteria, two regulators (LiuQ, FadP) are unique for this class. Deltaproteobacteria, which represent the most taxonomically diverse subdivision of Proteobacteria, have orthologues for only five studied TFs. Several TFs (such as LldR, GlcC, RutR) that are present in several classes of Proteobacteria show a mosaic distribution across the analysed genomes and taxa, while other TFs (such as ArgR, FabR, FadR, MetJ, TrpR and TyrR) are highly conserved in many taxonomic groups of gammaproteobacteria but are absent in other classes. The diverse distribution of TFs suggests different evolutionary pathways for the studied metabolic regulons.
The detailed descriptions of reconstructed regulons and regulogs are captured in the RegPrecise database (Data Citation 1), whereas the complete list of regulatory interactions between the studied TFs and their target genes is provided in Table S3. Overall, the obtained regulons included 10 663 candidate TFBSs and 15 690 target genes (Table 1). The largest average number of target genes per genome (more than 10 genes per genome) was observed for regulators of amino acid metabolism (ArgR, LiuR, MetJ, TyrR), the fatty acid degradation regulators FadP and PsrA, the carbohydrate metabolism regulators HexR, NagC and NagR, and the pyrimidine utilization regulator RutR.
For most of the studied TFs, their cognate DNA binding motifs are generally conserved across the analysed taxonomic groups (see the RegPrecise database for detailed lists of taxonomy-specific TFBS motifs, Data Citation 1). However, for several TFs including FabR, HypR, NrtR, RutR, SahR and TrpR, we observed taxon-specific substitutions in their cognate DNA motifs, whereas the GlcC-binding DNA motifs in the alpha- and beta-/gamma-subdivisions of Proteobacteria are characterized by different length of the spacer between the conserved palindromic half-sites (Fig. 2). Finally, the HexR and NagQ motifs in several taxonomic groups of gammaproteobacteria have different consensus sequences and structures (Yang et al., 2006; Leyn et al., 2011).
To analyse conservation of regulatory interactions in the reconstructed regulogs, we calculated the conservation score as the number of gene occurrences in a regulog divided by the number of regulons in a regulog. The mean of these taxonomy-specific conservation scores was calculated for all orthologous groups of target genes across analysed lineages of Proteobacteria. For each group of orthologous TFs, we plotted the average conservation score of a target gene against the number of taxonomic groups, in which this gene is regulated. The obtained plots visualize average conservation of regulatory interactions and thus help to determine the core, taxonomy-specific and genome-specific target genes within the reconstructed TF regulons (Fig. S3).
The core regulon members determined by this approach represent regulatory interactions with high average conservation scores that are conserved in more than half of reconstructed TF regulogs. The core members of most of the analysed regulons are consistent with major biological functions and molecular effectors of their cognate TFs (Table S4). For instance, the arginine repressor regulon ArgR in gammaproteobacteria has a conserved core that includes genes involved in arginine biosynthesis (argABCEFGH, carAB), transport (artPIQM) and degradation (astAD), as well as the argR gene itself (Fig. 3). The cores of most other reconstructed TF regulons include their cognate TF genes. Exceptions to this observation include the FabR, FadR and NadQ regulons that include their cognate TF genes only in some taxonomic groups of Proteobacteria; the biotin repressor BirA, which is autoregulated only in Desulfovibrionales; and the deoxyribonucleotide reductase regulator NrdR, which was never found under autoregulation. Negative autoregulation of a TF gene is a common feature of bacterial regulatory networks. Here, we demonstrate that this type of regulatory interaction is highly conserved in the evolution of regulatory networks of Proteobacteria.
The remaining members of reconstructed regulons were classified into the taxon- and genome-specific groups depending on their average conservation scores. The taxonomy-specific regulon members are characterized by strong conservation of regulatory interactions restricted to 50 % or less of taxonomic groups containing an orthologous TF. Thus, taxon-specific regulon members were defined as genes that are regulated in more than 65% of genomes in at least one taxonomic group. In contrast, the genome-specific regulon members are characterized by low conservation of regulatory interactions, when in each taxonomic group conservation of a regulatory interaction is less than 65%. The groups of taxonomy-and genome-specific TF regulon members with assigned metabolic pathways are often involved in the same major biological process as the core regulon members (Table S4). However, in some TF regulons these categories also involve genes that participate in other metabolic pathways or biological processes. For example, the taxon-specific members of ArgR regulons include genes involved in arginine biosynthesis (argD) and transport (artJ, argW, omp), arginine degradation (astBC), as well as genes from glutamate (gltBD) and branched-chain amino acid (ilvMGDA) biosynthesis, putrescine transport (potFGHI) and pyruvate metabolism (oadABG) (Fig. 3). The genome-specific ArgR regulon members include arginine degradation genes (arcABCD, astE), proline (proVWX) and histidine (hisJMPQ) transporters, and putrescine metabolism genes (speF, potE). Likewise, in our previous analysis of the methionine-specific regulons MetJ, MetR and SahR in Proteobacteria, we identified the core-, taxonomy- and genome-specific members of regulons and demonstrated their involvement in different aspects of the methionine metabolism. Other amino-acid-specific TF regulons analysed in the current work are described in more detail in the following sections.
The aromatic amino acids tryptophan, tyrosine and phenylalanine are synthesized in Proteobacteria by the common pathway leading from erythrose 4-phosphate to 2-dehydro-3-deoxy-d-arabinoheptonate-7-phosphate (DAHP), and shikimate to chorismate (Fig. 4). After chorismate, the pathway divides into the three terminal biosynthetic pathways that are specific for each aromatic amino acid. E. coli has three DAHP synthase isoenzymes, AroF, AroG and AroH, which are feedback inhibited by tyrosine, phenylalanine and tryptophan, respectively. The biosynthesis of aromatic amino acids is regulated at both the DNA and the RNA level. The DNA-binding transcription factors TyrR and TrpR jointly control the expression of genes involved in aromatic amino acid metabolism in E. coli (Pittard & Yang, 2008). At the RNA level, the trpEDCBA operon encoding the tryptophan biosynthesis enzymes and the phenylalanine biosynthesis gene pheA are regulated by translational attenuation in E. coli and other gammaproteobacteria (Panina et al., 2001). The tryptophan-responsive regulator TrpR in E. coli acts as a repressor of the trpEDCBA operon, the tryptophan transporter gene mtr and the regulatory gene trpR (Czernik et al., 1994; Jeeves et al., 1999). In addition, TrpR negatively regulates the expression of the shikimate kinase aroL and the DAHP synthase aroH that are involved in chorismate biosynthesis. The tyrosine-responsive regulator TyrR in E. coli negatively controls the tyrosine biosynthesis genes tyrB, aroF-tyrA and aroLM,the aromatic amino acid transporter aroP and the tyrR gene itself. In addition, TyrR activates the tyrosine- and tryptophan-specific transporters tyrP and mtr and the folate biosynthesis gene folA in the presence of tyrosine or phenylalanine (Yang et al., 2004; Pittard et al., 2005). The TyrR regulon was also partially studied in two other Enterobacteria. In Citrobacter freundii, it activates the tyrosine degradation gene tpl (Smith & Somerville, 1997). In Enterobacter cloacae, TyrR activates the ipdC gene involved in the synthesis of indole acetate from tryptophan and represses a divergently transcribed gene, akr, encoding a putative aldo-keto reductase (Coulson & Patten, 2015). An orthologue of TyrR in Pseudomonas putida, known as PhhR, is responsible for the activation of genes essential for phenylalanine degradation and phenylalanine homeostasis (Herrera et al., 2010). In Pseudomonas aeruginosa, PhhR directly controls the phhABC, hpd and dhcA transcriptional units involved in the catabolism of phenylalanine and tyrosine (Palmer et al., 2010).
The comparative genomics approach was applied to analyse regulons controlled by transcription factors homologous to TrpR and TyrR/PhhR and to predict novel regulons for aromatic amino acid metabolism (Table S5). The TrpR-family regulons reconstructed in eight lineages of gammaproteobacteria control genes for tryptophan biosynthesis, uptake and catabolism (Fig. 4). In most of the analysed taxonomic groups, TrpR regulates the trpR and trpE genes that form the conserved regulon core, whereas other genes from the tryptophan biosynthesis pathway and the mtr transporter were classified as taxon-specific members of the regulon. Two other known targets of TrpR in E. coli, the chorismate biosynthesis genes aroLM and aroH, represent regulatory interactions that are conserved in a small number of closely related Enterobacteriales genomes but not in other lineages. New predicted members of the TrpR regulons include various aromatic amino acid biosynthesis genes such as aroG in Pasteurellales, aroFtyrA in Shewanellaceae, aroF2 and aroA in Vibrionales, as well as the tryptophan degradation genes tnaAB and a predicted tryptophan transporter from the COG0733 family in Vibrionales. Interestingly, the TrpR regulon in Shewanella species does not include the tryptophan biosynthesis operon, which is regulated by a translational attenuator at the RNA level (Panina et al., 2001).
The content of reconstructed TyrR (PhhR) regulons is highly variable across the analysed six lineages of gammaproteobacteria (Table S5). The most conserved members of these regulons are the tyrR gene itself, the tyrosine biosynthesis and transport genes aroFtyrA and tyrP, as well as the phenylalanine and tyrosine degradation genes phhAB, hmgABC and hpd (Fig. 4). The aromatic amino acid transporters mtr and aroP, as well as the chorismate biosynthesis genes aroLM, which were previously known as TyrR-regulated genes in E. coli,belong to the TyrR regulons only in Enterobacteriales. The folA gene represents another previously known member of the TyrR regulon in E. coli, although we were unable to find conserved TyrR-binding sites upstream of folA orthologues in other Enterobacteriales.
A novel predicted tyrosine transporter from the COG2814 family was found under TyrR regulation in seven genomes of Enterobacteriales. The reconstructed TyrR regulons in Vibrionales are extended to include a predicted novel tyrosine transporter from the COG0733 family and the aroG gene. The most significant shifts in the regulon content were identified in Shewanellaceae, where TyrR controls the degradation pathways for various amino acids including phenylalanine (phhAB), tyrosine (hmgCB), tryptophan (tdo-kyn), branched chain amino acids (ldh, brnQ, liu, ivd and bkd operons), proline (putA), methionine (mdeA) and oligopeptides (various peptidase genes). In addition, the conserved part of the TyrR regulons in Shewanellaceae includes the tyrosine/phenylalanine biosynthesis genes aroA and tyrB, as well as the aceBA and acsA genes from central carbon metabolism. Finally, the tpl gene encoding an alternative pathway of tyrosine degradation belongs to the TyrR regulons in Citrobacter koseri and two Pasteurellales, whereas the indole acetate synthesis gene ipdC is regulated by TyrR in some Enterobacteriales and Shewanellaceae genomes.
The homogentisate pathway of tyrosine degradation encoded by the hmgABC operon in P. putida is regulated by the IclR-family repressor HmgR and homogentisate as anti-repressor (Arias-Barrau et al., 2004). Similar HmgR regulons were reconstructed in four other Pseudomonas species, whereas the hmgABC genes in two other Pseudomonas species belong to the TyrR-family PhrR regulons (Table S5). In a closely related bacterium from the family Pseudomonadaceae, Azotobacter vinelandii, which lacks both PhrR and HmgR regulons, we identified a novel LysR-family regulon for the homogentisate pathway genes, which we termed HmgQ. Orthologous HmgQ regulators in the Shewanellaceae family are predicted to control the hmgA–hpd genes, whereas the hmgCB genes of the homogentisate pathway belong to the extended TyrR regulon in Shewanella species. Another novel regulator from the MarR family (termed HmgS) was identified in several Alteromonadales and Pseudoalteromonadales species, where it is predicted to control the hmgAB genes.
In conclusion, the transcriptional regulation of aromatic amino acid metabolism is highly variable among major lineages of gammaproteobacteria (Table S5). The TyrR-family regulators control the aromatic amino acid biosynthesis, uptake and/or catabolic pathways in most of the analysed taxa. In contrast, the reconstructed regulons in the family Shewanellaceae predict a global regulatory role of TyrR for genes that are involved in catabolism of various amino acids and in central carbon metabolism. The mode of TyrR action on its predicted novel targets in Shewanella is to be determined experimentally. Preliminary comparative analysis of positions of the TyrR-binding sites in the promoter gene regions suggest that TyrR probably acts as an activator for most of the amino acid degradation operons in Shewanella species (data not shown). The homogentisate pathway in gammaproteobacteria is controlled either by TyrR/PhhR or by non-orthologous local regulators from the IclR, LysR, and MarR protein families. We also observed interchangeability between the TyrR and TrpR regulons: the aroF–tyrA genes are controlled by TyrR in most of the analysed lineages, whereas in the family Shewanellaceae this operon is predicted to be controlled by TrpR. Overall, the major biological role of TyrR in Enterobacteriales and Pasteurellales is the regulation of aromatic amino acid biosynthesis and transport genes, although in Pseudomonas and Shewanella it mostly controls the amino acid degradation pathways, whereas in other Alteromonadales species, as well as in Vibrionales and Aeromonadales, it is implicated in the control of both biosynthetic and catabolic pathways.
l-Proline, the only proteinogenic imino acid, is used by many bacteria as a source of energy and a precursor for the synthesis of other amino acids. 4-Hydroxy-l-proline, which can be synthesized post-translationally from l-proline, is also a rich source of carbon and nitrogen for many micro-organisms. In Sinorhizobium meliloti, the hydroxyproline transport and utilization genes (hyp) are negatively regulated by the GntR-family regulator HypR, with 4-hydroxy-l-proline as an inducer (White et al., 2012). Orthologues of HypR were identified in 13 taxonomic groups that mostly belong to gammaproteobacteria but also include three alphaproteobacterial and two betaproteobacterial taxa (Fig. S2). The core of reconstructed HypR regulons includes hypR, which is autoregulated in 32 out of 46 studied genomes, and hypE, hypO, hypD and hypH encoding enzymes involved in the conversion of hydroxyproline to α-ketoglutarate (Fig. 5). Operon organization of the hyp genes varies among the studied bacteria. Taxonomy-specific regulon members include an alternative 4-hydroxyproline epimerase (hypY) and two putative enzymes from the TCA cycle, malate dehydrogenase (hypS) and citrate isomerase (hypX), that are potentially involved in α-ketoglutarate utilization (Fig. 5). PutA, the main enzyme of the proline catabolic pathway that provides proline oxidation to 1-pyrroline-5-carboxylate, is the taxonomy-specific member of the HypR regulons in the Shewanellaceae and Aeromonas species. Ornithine can be converted to proline via ornithine cyclodeaminase (COG2423), which is predicted to be a part of the HypR regulon in Vibrio parahaemolyticus, Paracoccus denitrificans and several Shewanella species. Thus, HypR function probably expands onto ornithine degradation.
In S. meliloti, hydroxyproline is imported into the cell via the ABC-family transporter HypMNPQ, which is predicted to be a part of HypR regulons only in three other genomes and thus was classified as a taxonomy-specific regulon member. Two other predicted hydroxyproline transporters, namely HypT from the MFS family and HypABC from the TRAP family, were identified as genome-specific members of the HypR regulons. Furthermore, HypR-regulated genes encoding a putative TonB-dependent outer membrane transporter in several Alteromonadales and a COG531-family permease in Erwinia amylovora may also be involved in hydroxyproline/proline transport.
Large amounts of proline and hydroxyproline are found in the abundant protein collagen (Phang et al., 2015). We identified three secreted collagenases (colA1, colA2 and colA3) within the reconstructed HypR regulons in the Alteromonadales (mostly in Shewanella species), suggesting that the HypR regulons evolved in these species include the upstream metabolic steps in the proline/hydroxyproline utilization pathway (Fig. 5). Moreover, the reconstructed HypR regulons in the Alteromonadales and Aeromonas taxa of gammaproteobacteria include various proline (di)peptidases (PrdP, AmpP and YpdF) and proline dipeptide/tripeptide permease (PdtP), suggesting these species utilize another upstream source of proline for the catabolic pathway (Fig. 5).
Histidine is a well-known source of carbon, nitrogen and energy for many bacteria. The histidine degradation pathway was studied in Klebsiella aerogenes and Salmonella typhimurium and involves four reactions catalysed by HutH, HutU, HutI and HutG (re-named HutG2 in this work, belonging to the COG0010 family), whereas in Pseudomonas species the pathway involves an alternative HutG enzyme from the COG3741 family, as well as an additional reaction catalysed by HutF (Fig. 6) (Goldberg & Magasanik, 1975; Zhang & Rainey, 2007). The histidine utilization genes are regulated by orthologous HutC repressors in the above three gammaproteobacteria. Orthologues of HutC were identified in 113 genomes from all studied taxa of alpha-, beta- and gammaproteobacteria except the Pasteurellales (Fig. S2). The core of reconstructed HutC regulons includes all known histidine catabolic enzymes, including both alternative HutG enzymes and the HutD protein, which has an as yet unknown function in the pathway. Thus, the HutC regulons in the Pseudomonadaceae, Enterobacteriales, Burkholderiales, Rhizobiales and Vibrionales often include multiple paralogues of the histidine ammonia-lyase HutH. In addition to the hut genes, HutC was predicted to co-regulate the histidine biosynthesis operon in Colwellia psychrerythraea and Idiomarina loihiensis (Fig. 6).
The taxonomy- and genome-specific members of the reconstructed HutC regulons are represented by several known or putative transporters involved in histidine uptake (Fig. 6). These include three known histidine uptake systems: the HisT permease in the Burkholderiales, Moraxellaceae and Pseudomonadaceae, and two distinct ABC-family transporters, HisJMPQ in Burkholderia and HutXYZ in Pseudomonadas and some Rhizobiales. A novel histidine transporter from the ABC family (named HisXYZ) was predicted in the Comamonadaceae, Ralstonia and Rhizobiales. Moreover, the reconstructed regulons include a novel TonB-dependent outer membrane transporter in the Caulobacterales and Sphingomonadales, a COG2814-family permease in Ralstonia species and a COG3314-family transmembrane protein in Aeromonas, Psychromonas and Marinomonas species that are potentially involved in histidine uptake. Additional putative transporter from the COG1457 family was identified in the HutC regulons of Burkholderia, Pseudomonas, Klebsiella and Acinetobacter species, although all these species already possess the HisT permease. Therefore, there is not enough information to support assignment of histidine specificity to these novel COG1457-family transporters, which are homologous to purine and allantoine transporters.
In addition to the above-described five amino-acid-specific regulons and also the previously described methionine regulons (Leyn et al., 2014), we observed many variations in reconstructed regulons for orthologous TFs in different lineages of Proteobacteria, including regulon expansion and contraction and many cases of recruiting non-orthologous TFs to control equivalent pathways. The most interesting and novel examples of the observed evolutionary changes in the reconstructed TF regulons are briefly described below.
The FabR repressor, which was previously known to control the fatty acid biosynthesis genes fabAB in E. coli and other Enterobacteriales, co-regulates the unsaturated fatty acid biosynthesis genes desABC in several lineages of gammaproteobacteria. In addition, the FabR regulon in six Shewanella species is expanded by the pfaRABCD operon encoding polyunsaturated fatty acid synthase. The fatty acid degradation pathway in Shewanellaceae and other gammaproteobacteria is regulated by PsrA, whereas in Enterobacteriales the analogous pathway is regulated by FadR. The FadR regulon in Shewanellaceae is contracted and retains only two operons shared with the orthologous regulon in Enterobacteriales and Vibrionales (fadIJ and fadL). The PsrA regulon in Shewanellaceae is expanded by several operons involved in the TCA cycle (aceBA, sdh, gltA). The biological role of PsrA regulons in two lineages of betaproteobacteria, Ralstonia and Burkholderia, is shifted to control the fatty acid biosynthesis genes, whereas the fatty acid degradation genes are predicted to be co-regulated by a novel TetR-family regulator, termed FadP, in the above two lineages, as well as in Comamonadaceae, which lack PsrA or FadR regulons (Kazakov et al., 2009).
HexR in Enterobacteriales is a local regulator of glucose 6-phosphate dehydrogenase (zwf), whereas in other lineages of gamma proteobacteria it co-regulates zwf–pgl with genes from the Entner-Doudoroff pathway (edd, eda), glucokinase (glk) and pyruvate kinase (pykA). The HexR regulons in Shewanellaceae and Vibrionales are significantly expanded to include various other genes from the central glycolytic and fermentation pathways, glucose transport, mannitol utilization, nucleoside metabolism, glutamate biosynthesis and the glycine cleavage system (Leyn et al., 2011). The pyruvate-responsive regulator PdhR, which solely controls the pyruvate dehydrogenase operon aceEF–lpdA in Enterobacteriales, Vibrionales and several other lineages, undergoes radical expansion to include genes involved in the TCA cycle (sdhCDAB, gltA, aceAB, oadGAB) and fermentation (pflBA, focA, lldP–dld) in Shewanellaceae.
NtrC protein is a regulator of nitrogen assimilation described in E. coli and Rhodobacter capsulatus (Reitzer, 2003; Masepohl & Hallenbeck, 2010). Orthologues of the NtrC protein were found in 19 analysed taxa. A core part of the regulon contains genes for glutamine synthetase (glnA), ammonium transporter (amtB) and nitrogen assimilation regulatory genes (glnBK and ntrB, ntrC). On the other hand, the NtrC regulon demonstrates taxon-specific expansions to other nitrogen metabolism genes. NtrC-dependent regulation of glutamate dehydrogenase (gdhA) is a specific feature of the Moraxellaceae, Rhodobacterales and deltaproteobacteria. Genes encoding glutamate synthase (gltBD) are regulated by NtrC in the Moraxellaceae and Shewanellaceae, as well as in some beta-, alpha- and deltaproteobacteria. In a number of taxa, the NtrC regulon is expanded to the genes encoding hydrogenases that act on carbon–nitrogen bonds. Thus, allophanate hydrolase (atzF) and agmatinase (speB) are regulated in the Rhizobiales and betaproteobacteria, respectively, whereas urease (ureABC) and urea ABC transporters (uctABC and urtABCDEF) are regulated in the Alteromonadales, Oceanospirillales, Rhizobiales and Rhodobacterales. In alpha- and betaproteobacteria, the NtrC regulon is expanded to genes involved in nitrogen oxide uptake and reduction, such as assimilatory reductases of nitrate (nasAB) and nitrite (nasDE and nirA), nitrate–nitrite antiporter (narK) and nitrate ABC transporter (nrtABC). Together, the core of the NtrC regulon includes genes necessary for inclusion of ammonia into organic compounds through glutamine synthesis, whereas the taxon-specific regulon members are necessary for generation of ammonia via metabolism of various nitrogen compounds.
Three different TFs were previously found in Proteobacteria to control the N-acetylglucosamine catabolic pathway, namely NagC, NagQ and NagR (Yang et al., 2006). All three regulons have similar cores consisting of two central enzymes from the N-acetylglucosamine pathway (nagA and nagB) and a PTS-family transporter that is involved in uptake and phosphorylation of N-acetylglucosamine. The taxonomic distribution of these three regulatory systems is not uniform: NagC and NagR were found in gammaproteobacteria, whereas NagQ was identified in alpha- and betaproteobacteria, as well as in some lineages of gammaproteobacteria. Interestingly, Reinekea sp. MED297 has two distinct regulators, where the NagQ regulon contains genes involved in the sugar catabolic pathway (nagKAB) and chitin degradation (cdxA, chiA), whereas NagC controls genes encoding an N-acetylglucosamine-specific PTS transporter and chemotaxis proteins. Furthermore, in two Xanthomonas species, we found both NagQ and NagR, where NagQ regulates the N-acetylglucosamine catabolic and transport genes (nagAB and nagP), while NagR regulons include the N-acetylglucosamine kinase and TonB-dependent outer membrane receptors that might be involved in sugar transport across the outer memberane. Overall, all three N-acetylglucosamine-specific TF regulons are expanded in many genomes to include various chitin utilization genes. Also, a significant expansion of the NagC regulon was observed in the Vibrionales, where it includes the central glycolytic (gapA, gapB, fbaA, pgk, eno) and glycogen biosynthesis (glgAC) genes.
Two distinct TFs, BirA and BioQ, control the biotin/vitamin B7 biosynthesis pathway in Proteobacteria. BirA was previously studied in detail in E. coli (Beckett, 2005). It is a protein that functions both as a transcriptional repressor and as a biotin-protein ligase, which covalently links biotin to biotin-dependent enzymes. BirA enzymes are ubiquitous in micro-organisms, while the N-terminal DNA-binding domain can be only found in a subset of BirA proteins from a broad number of Proteobacteria, Firmicutes and several other lineages of Bacteria and Archaea. Among the studied Proteobacteria, BirA repressors and regulons were found in the gamm- and delta subdivisions. The most conserved part of BirA regulons consists of the biotin biosynthesis genes (bioABCDF). The BirA regulons in Desulfovibrionales are expanded to include the fatty acid biosynthesis genes (fabF, fabH, acpP), which are involved in the same pathway with the biotin-dependent acetyl-CoA carboxylase. The BirA proteins from beta- and alphaproteobacteria have lost their DNA-binding domains, and apparently the BirA regulons do not exist in these species.
Previous comparative genomics analysis of biotin pathway genes has identified a novel GntR-family TF in alphaproteobacteria from the Rhizobiales and Rhodobacteriales lineages, termed BioR, which was predicted to control the biotin metabolism genes (Rodionov & Gelfand, 2006). The reconstructed BioR regulons include the biotin biosynthesis genes (bioABDF, bioCG, bioZ), as well as a novel ECF-family transporter for biotin (bioYMN) (Hebbeln et al., 2007). The BioR regulon was later experimentally validated in Brucella melitensis and Paracoccus denitrificans (Feng et al., 2013, Feng et al., 2015).
In the Enterobacteriales, the NAD cofactor metabolism genes are controlled by the NadR regulator, which is absent in all other lineages of Proteobacteria. The most conserved part of the NadR regulon is the nadA–pnuC operon that encodes a de novo NAD biosynthesis enzyme and a ribosyl nicotinamide transporter. In several enterobacterial genomes, NadR controls additional NAD biosynthesis and nicotinate/vitamin B3 salvage genes (nadB, pncB, niaP, nadR).
In alphaproteobacteria and several species of beta- and gammaproteobacteria we found a novel TF, termed NadQ, which presumably controls the NAD biosynthesis genes. Proteins from the NadQ orthologous group belong to an as yet undescribed protein family (COG4111) that has a characteristic C-terminal DNA-binding domain with a helix-turn-helix (HTH) motif, which is homologous to another regulator of NAD metabolism, NrtR (Rodionov et al., 2008). However, the N-terminal effector binding domain of NadQ is unique as it is not similar to the ADP ribose-binding domain of NrtR. Palindromic DNA binding motifs for NadQ identified in seven lineages of Proteobacteria are characterized by the common consensus sequence ttATRCTCannntGAGYATaa, where R and Y stand for purines or pyrimidines, respectively. The nadQ genes are often clustered on the chromosome with the de novo NAD biosynthesis genes (nadABC). Thus, the core of reconstructed NadQ regulons in proteobacteria includes these central NAD biosynthesis genes. In the Rhodobacterales and Caulobacterales, the NadQ regulon is expanded to include the lower NAD biosynthesis pathway genes (nadE and nadD). The effector molecule for the novel NadQ regulator has yet to be determined experimentally but the regulon content suggests that it may be either NAD itself or one of the pathway intermediates.
We used the comparative genomics approach for reconstruction of regulatory networks for amino acid and other central metabolic pathways that are controlled by specific groups of TFs. The results of this study demonstrate considerable variability of the TF regulons for the central metabolic pathways in Gram-negative bacteria from the phylum Proteobacteria. The core members of the characterized TF regulons are involved in arginine biosynthesis (ArgR), biotin biosynthesis (BirA), fatty acid biosynthesis (FabR) and degradation (FadR, FadP, PsrA), glycolate, lactate and pyruvate utilization (GlcC, LldR, PdhR), central carbohydrate metabolism (HexR), histidine and hydroxyproline/proline utilization (HutC, HypR), branched‐chain amino acid degradation (LiuR, LiuQ), methionine metabolism (MetJ, MetR, SahR), nitrogen assimilation (NtrC), deoxyribonucleotide biosynthesis (NrdR), N-acetylglucosamine utilization (NagC, NagQ, NagR), pyrimidine degradation (RutR), tyrosine and phenylalanine metabolism (TyrR), and tryptophan biosynthesis (TrpR). Large-scale phylogenomic and comparative genomics analyses of these TFs reveal numerous examples of various evolutionary processes for regulators and their regulons at the levels both of a taxonomic group/class of bacteria and of an individual genome. These predicted evolutionary processes can be classified into three distinct types: (i) 'regulon expansion' in certain taxa compared with other lineages that can range from additions of several regulon members to large-scale shifts in the regulated metabolic pathways (e.g. PdhR, TyrR and TrpR regulons in Shewanella species); (ii) 'fuzzy regulons' when a regulon contains a small conserved core and a large periphery of taxon- and genome-specific genes (e.g. ArgR, HexR and NtrC regulons); and (iii) 'regulon loss or acquisition' when an entire regulon (including a TF and all its TFBSs and target genes) is absent or present only in specific genomes within a taxonomic group of bacteria (e.g. GlcC and HypR, which are present in individual species of Enterobacteria, or NagR, which was found in all but one Shewanella species). The most conserved regulatory interactions were identified within TF regulons that are involved in the control of certain essential biosynthetic pathways (e.g. BirA, NrdR and FabR). Overall, this study provides new insights into the evolutionary mechanisms that shape transcriptional regulatory networks in Bacteria.
This work was supported by the Russian Foundation for Basic Research (14-04-00870) and by the Russian Academy of Sciences via the programme 'Molecular and Cellular Biology'. Additional funding was provided by the Genomic Science Program (GSP), Office of Biological and Environmental Research (OBER) and U.S. Department of Energy (DOE), and this is a contribution of the Pacific Northwest National Laboratory (PNNL) Foundational Scientific Focus Area. Regulon reconstruction in sulfate-reducing deltaproteobacteria, data accumulation and visualization in the RegPrecise database, and project maintenance in the RegPredict web server by ENIGMA – Ecosystems and Networks Integrated with Genes and Molecular Assemblies (http://enigma.lbl.gov), a Scientific Focus Area Program at Lawrence Berkeley National Laboratory, are based upon work supported by the U.S. Department of Energy, Office of Science, Office of Biological & Environmental Research under contract number DE-AC02-05CH11231.
Supplementary File 1