Computational Analysis of Cysteine and Methionine Metabolism and Its Regulation in Dairy Starter and Related Bacteria

Abstract

Sulfuric volatile compounds derived from cysteine and methionine provide many dairy products with a characteristic odor and taste. To better understand and control the environmental dependencies of sulfuric volatile compound formation by the dairy starter bacteria, we have used the available genome sequence and experimental information to systematically evaluate the presence of the key enzymes and to reconstruct the general modes of transcription regulation for the corresponding genes. The genomic organization of the key genes is suggestive of a subdivision of the reaction network into five modules, where we observed distinct differences in the modular composition between the families Lactobacillaceae, Enterococcaceae, and Leuconostocaceae, on the one hand, and the family Streptococcaceae, on the other. These differences are mirrored by the way in which transcription regulation of the genes is structured in these families. In the Lactobacillaceae, Enterococcaceae, and Leuconostocaceae, the main shared mode of transcription regulation is methionine (Met) T-box-mediated regulation. In addition, the gene metK, encoding S-adenosylmethionine (SAM) synthetase, is controlled via the S_MK box (SAM). The S_MK box is also found upstream of metK in species of the family Streptococcaceae. However, the transcription control of the other modules is mediated via three different LysR-family regulators, MetR/MtaR (methionine), CmbR (O-acetyl[homo]serine), and HomR (O-acetylhomoserine). Redefinition of the associated DNA-binding motifs helped to identify/disentangle the related regulons, which appeared to perfectly match the proposed subdivision of the reaction network.

INTRODUCTION

Many of the characteristic flavors in fermented dairy products such as cheese and yoghurt are the result of metabolic reactions involving sulfur-containing amino acids. The microorganisms applied in these products degrade cysteine and methionine, resulting in the production of flavor components such as methanethiol, dimethyl sulfide (DMS), dimethyl disulfide (DMDS), and dimethyl trisulfide (DMTS). Insight into the regulatory signals and pathways that control the corresponding metabolic fluxes involved in the formation of these flavor compounds and their precursors is essential to rationally control and steer the flavor profiles of said dairy products.

The microorganisms used to produce fermented dairy products belong to the taxonomic order Lactobacillales, which includes the families Enterococcaceae, Lactobacillaceae, Leuconostocaceae, and Streptococcaceae. Many of the respective species are characterized by the fact that they produce lactic acid and are therefore known as the lactic acid bacteria (LAB). The transcription of genes encoding the proteins that are involved in cysteine and methionine metabolism in lactic acid bacteria and other Lactobacillales is controlled by both regulator-binding and RNA structural switches. In various Streptococcaceae, the LysR-family transcription regulators MtaR and CmbR have been shown to be involved in activation as well as repression of genes such as cysD, cysK, metA, metC, metE, and metF (e.g., for Lactococcus lactis [25, 70] and Streptococcus mutans [36, 68]). The transcription regulator HomR was reported to control the expression of metB in S. mutans and Streptococcus thermophilus (69). In addition, three types of RNA structural switches for the regulation of cysteine and methionine metabolism have been reported for low-GC-content Gram-positive bacteria: the T box, the S box, and the S_MK box (14, 22, 59, 77, 80). These sequence elements at the 5′ untranslated region of an mRNA molecule can change conformation, depending on the binding of an effector molecule. The conformational change can terminate transcription (when forming a terminator structure) or allow readthrough (when forming an antiterminator structure) (75, 78). In the case of the T box, a terminator structure is formed shortly after transcription initiation, unless an uncharged tRNA related to a specific amino acid binds to the specifier codon present in the T-box element, whereupon the antiterminator structure is formed (see reference 29). In the case of the S box and the S_MK box, the terminator structure is formed in the presence of S-adenosylmethionine (SAM), whereas in the absence of this molecule, transcription continues (22, 27, 47).

Several studies have described the regulation of sulfur-containing amino acid metabolism for specific LAB and other closely related Gram-positive bacteria. For instance, Hullo et al. (30) reported on the regulatory mechanisms related to cysteine and methionine conversions in Bacillus subtilis. Sperandio et al. described these relations for Lactococcus lactis (70) and Streptococcus mutans (68, 69). Rodionov et al. (59) and Kovaleva and Gelfand (36) performed a comprehensive comparative in silico study for the transcriptional regulators CmbR and MtaR within Gram-positive bacteria. However, the availability of additional experimental and sequence data now allows an overview of the transcriptional control of the key enzymes involved in cysteine and methionine metabolism at a higher resolution. We therefore decided to extend the latter studies and to focus our efforts on the LAB and other Lactobacillales.

In a previous study, we improved the annotation of key enzymes involved in the metabolism of cysteine and methionine in LAB using genome-wide comparative analyses (41). Here, we extend the list of enzymes on the basis of the pathway information present in the KEGG database (32). Redefinition of the binding motifs for CmbR, MetR/MtaR, and HomR in lactococci and streptococci allowed the identification of transcription factor-specific binding sites for these regulators. Also, Met-specific T boxes and S_MK boxes were identified in recently sequenced and published genomes of, e.g., Lactobacillus delbrueckii subsp. bulgaricus, Lactobacillus reuteri, and Lactobacillus casei. The absence of S boxes (SAM-I) in the Lactobacillales observed hitherto was confirmed. Potential structure-forming elements associated with the cysK gene and the hom-thrBC operon in various LAB were revealed, as presented below.

MATERIALS AND METHODS

Genomic information, tools, and data.

Genomic information was retrieved from the ERGO resource (as of December 2009) (53) and from the NCBI microbial genome database (as of September 2011) (http://www.ncbi.nlm.nih.gov/genomes/lproks.cgi). BLAST searches were performed as described elsewhere (1). Multiple-sequence alignments and neighbor-joining (NJ) trees (corrected for multiple substitutions) were generated using the ClustalX program (37). BioEdit software was used to manipulate the alignments and to toggle between translated protein and nucleotide sequences (version 7.0.9; http://www.mbio.ncsu.edu/BioEdit/bioedit.html). Hidden Markov models (HMMs) were made, and genome-wide HMM searches were performed using the HMMER package (http://hmmer.janelia.org/) (13). Genome context was visualized and upstream sequence data were collected using Microbial Genome Viewer 2 (version 1 [34] and version 2 [http://mgv2.cmbi.ru.nl/genome/index.html]; L. Overmars, unpublished data). The original data supporting the analyses presented in this paper can be found at www.cmbi.ru.nl/bamics/supplementary/Liuetal_2012_CysMetregulation.

Collection of genes related to cysteine and methionine metabolism.

The species and strains that were analyzed included all members of the order Lactobacillales with a completed genome published before June 2011 and present within the NCBI database. The KEGG map cysteine and methionine metabolism (map 00270) was used to define a core set of enzyme activities. The set extends the set of enzymes that we previously defined (41). The protein sequences of experimentally verified members of the set (see File S1 in the supplemental material) were used to search orthologs/functional equivalents in other species using BLAST. An orthologous relationship and/or functional equivalency was defined on the basis of our earlier analyses (41), BLAST E values, and, in some cases, multiple-sequence alignments, followed by clustering on the basis of neighbor joining (as described in reference 73). The complete list of enzymes, their function annotation, and the relevant experimental literature are given below. The annotation data were taken from the NCBI (COG numbers below) (57), KEGG (K numbers below) (32), and PFAM (PF numbers below) (18) reference databases. The enzymes have been grouped into five clusters on the basis of the composition of the related operons and shared enzyme nomenclature (EC) numbers.

(i) Enzyme group 1.

Enzyme group 1 consisted of homoserine dehydrogenase (hom; EC 1.1.1.3; COG0460E, K00003, PF03447, and PF00742 [7, 44, 54]), homoserine kinase (thrB; EC 2.7.1.39; COG0083E, K00872, PF08544, and PF00288 [44]), aspartate kinase III (thrA [Bsubtilis_yclM]; EC 2.7.2.4; COG0527E, K00928, PF01842, and PF00696 [7, 35]), and threonine synthase (thrC; EC 4.2.3.1; COG0498E, K01733, and PF00291 [45, 63, 66, 72]).

(ii) Enzyme group 2.

Enzyme group 2 consisted of serine acetyltransferase (cysE; EC 2.3.1.30; COG1045E, K00640, and PF06426 [23, 30, 70]), homoserine O-acetyltransferase (metA; EC 2.3.1.31; COG1897E, K00651, and PF04204 [83]), cysteine synthase A and cysteine synthase-like protein (Bsubtilis_cysK and Bsubtilis _ytkP; EC 2.5.1.47; COG0031E, K01738, and PF00291 [25, 30, 76]), cystathionine gamma-synthase and O-acetylhomoserine (thiol)-lyase (Bsubtilis_yjcL; EC 2.5.1.48; COG0626E, K01739, and PF01053 [3, 33]), O-acetyl-l-homoserine sulfhydrolase and O-acetyl-l-serine sulfhydrolase (cysD; EC 2.5.1.49; COG2873E, K01740, and PF01053); cystathionine beta-synthase for the reverse transsulfurase pathway (Bsubtilis_yrhA; EC 4.2.1.22; COG0031E, K01738, and PF00291 [30]), cystathionine beta/gamma-lyase and homocysteine gamma-lyase (Bsubtilis_yrhB Ecoli_metB; EC 2.5.1.48 and EC 4.4.1.8; COG0626E, K01760, and PF01053 [12, 17, 30, 31]), cystathionine beta/gamma-lyase (Bsubtilis_yjcJ; EC 4.4.1.8 and EC 4.4.1.1; COG0626E, K01760, and PF01053 [3]), and pyridoxal-phosphate (PLP)-dependent C-S lyase (Bsubtilis_patB Llactis_ytjE Ecoli_malY; EC 4.4.1.8 and EC 4.4.1.1; COG1168E, K14155, and PF00155 [2, 31, 46]).

(iii) Enzyme group 3.

Enzyme group 3 consisted of 5,10-methylenetedrahdrofolate reductase (metF; EC 1.5.1.20; ?, K00297, and PF02219 [64]), bifunctional homocysteine S-methyltransferase 5,10-methylenetetrahydrofolate reductase protein (Bsubtilis_yitJ; EC 2.1.1.10 and EC 1.5.1.20; COG0646E [cobalamin dependent], K00547, PF02219, and PF02574 [42]), homocysteine S-methyltransferase (mmuM; EC 2.1.1.10; COG2040E, K00547, and PF02574 [74]), MmuM-associated amino acid permease (mmuP; COG0833E, K03293, and PF00324), methyltransferase (Bsubtilis_yxjG and Bsubtilis_yxjH Llactis_yhcE; EC 2.1.1.14[?]; COG0620E [cobalamin independent], K00548, and PF01717 [9, 38]), 5-methyltetrahydropteroyltriglutamate–homocysteine S-methyltransferase (metE; EC 2.1.1.14; COG0620E [cobalamin independent], K00549, PF08267, and PF01717 [21, 26]), and S-ribosylhomocysteinase (luxS Llactis_ycgE; EC 4.4.1.21; COG1854T, K07173, and PF02664 [38, 58).

(iv) Enzyme group 4.

Enzyme group 4 consisted of C-5 cytosine-specific DNA methylase and SP-beta prophage DNA (cytosine-5-)-methyltransferase (Bsubtilis_ydiO Bsubtilis _ydiP Bsubtilis_mtbP; EC 2.1.1.37; COG0270L, K00558, and PF00145 [51, 81]), and 5′-methylthioadenosine nucleosidase and S-adenosylhomocysteine nucleosidase (mtn Streptococci_pfs; EC 3.2.2.16 and EC 3.2.2.9; COG0775F, K01243, and PF1048 [10, 55]).

(v) Enzyme group 5.

Enzyme group 5 consisted of S-adenosylmethionine synthetase (metK; EC 2.5.1.6; COG0192H, K00789, PF02773, and PF00438 [22, 48]).

Identification of putative regulatory elements and their regulons.

cis-Regulatory elements were defined according to the specific footprinting method set out by Francke et al. (20). The method relies on the definition of groups of orthologous functional equivalents (GOOFEs) on the basis of orthology and conserved genomic context. The comparative linear genome maps generated by the Microbial Genome Viewer were used to visualize and inspect the context. For every GOOFE, the upstream regions (normally ∼200 nucleotides) were collected, and conserved sequence elements were searched for by eye from a multiple-sequence alignment and by using the MEME tool (4). The highest-scoring motifs resulting from MEME and the conserved elements in the multiple-sequence alignment were compared, and potential regulatory regions were identified. In case the conserved elements resembled transcription factor binding motifs reported in literature, experimental data on regulators of the same protein family were searched for directly via PubMed (61) or in the reference databases Regulon DB (24) and DBTBS (65). Because members of the same regulator-protein family will, in general, adopt the same fold, the DNA-binding motif should be similar (i.e., have similar compositions and the same size and spacing). Therefore, established binding motifs of regulator-protein family members were taken into account to define the actual binding motif. In addition, we defined the motifs such that they obey general constraints imposed by the molecular nature of the binding process and the helical nature of the DNA molecule. Since most regulator proteins bind to the DNA by virtue of a helix-turn-helix (HTH) domain and as a dimer, a binding site will, in general, be made up of two monomer binding sites and will have to be either palindromic or represent a direct repeat. Moreover, since the DNA is helical, the actual monomer binding site in general has to be shorter than 7 nucleotides and the two sites that make up the dimer binding site have to be interspaced by a fixed number of nucleotides.

The defined motifs (given in File S2 in the supplemental material) were converted to a position frequency matrix, which was used directly to score potential transcription factor binding sites and other regulatory elements of fixed composition and size. In this way, the score of any DNA sequence related directly to its similarity to the input motif. In general, we score the “relative similarity,” which we define as the percentage of the maximally attainable score given the input motif. We have validated and used this approach with success to identify potential binding sites of CcpA and Spo0A in low-GC-content Gram-positive organisms and the sigma 54 promoter in all organisms (19). A cutoff score of >83% relative similarity and a positioning of a maximum of about 200 nucleotides upstream (with some exceptions) of the translation start were used to select potential binding sites for the various regulators. The uniform cutoff score was chosen such that the number of false-positive assignments should be limited, i.e., such that experimentally validated sites were included and that the number of correctly positioned sites was high (position in terms of distance and orientation with respect to translation start of the gene downstream). The identified regulatory elements were related to all genes present in the downstream operon, where an operon was defined as those genes on the same strand that are separated by an intergenic region of less than 250 nucleotides without a termination signal. The locations of (rho-independent) transcription terminators were determined with the Transterm system (15). The analyzed results of the motif searches are given in File S3 in the supplemental material.

Identification of riboswitches and other structural elements.

Hidden Markov models were constructed for the T-box motif and for the S-box motif (SAM-I) on the basis of the available literature (59, 77, 80). Both HMMs were used to scan the selected genomes (cutoff E value, 1 [38]), and the locations of putative boxes were identified. The amino acid specificity of the detected T boxes was established on the basis of the specifier codon, as described by Wels et al. (80) and exemplified in Fig. S1 in the supplemental material. Two characteristic S_MK-box sequences were defined on the basis of information presented elsewhere (22), as given in Fig. 2A, and these were used to scan the selected genomes using the similar motif search procedure (results are presented in File S3 in the supplemental material). Only in case both motifs were found directly upstream of a gene and they were complementary did we consider the site to be a putative S_MK box.

Fig 2.

Binding motifs of various cysteine and methionine metabolism-related regulators in the Lactobacillales. (A) The characteristic SMK box motifs upstream of the metK gene in lactobacilli and streptococci as reported previously (22). (B) Redefined MetR, CmbR, and HomR dimer binding motifs in lactobacilli and streptococci. See the main text for details. (C) CodY binding motif in lactobacilli and streptococci. The motif was created on the basis of the data provided elsewhere (11). The original data can be found in File S2 in the supplemental material. The motifs were created with WebLogo (frequency representation; no correction applied; http://weblogo.berkeley.edu).

RESULTS

Comparative analysis of the enzymes involved in central cysteine and methionine metabolism.

The set of genes related to central cysteine and methionine metabolism was identified on the basis of KEGG map 00270 (32) and earlier work by others (36, 59) and us (41). Orthologs and homologs of these genes were collected from the genomes of all sequenced species/strains of the LAB and other Lactobacillales, as described in Materials and Methods. The related reaction network is given in Fig. 1, and the results of the search and analysis procedure are presented in Tables 1 and 2 and given in File S1 in the supplemental material. Remarkably, the set of genes (and corresponding proteins) can be divided into five separate groups on the basis of genomic organization, the identity of the EC numbers, and the position in the reaction network (Fig. 1). Small differences in operon organization between the families Lactobacillaceae, Enterococcaceae, and Leuconostocaceae (Table 1) and the family Streptococcaceae (Table 2) were observed, implying a difference in pathway modularization between the families.

Fig 1.

Generalized cysteine and methionine metabolism in the Lactobacillales. For most of the studied species, only part of the depicted reactions can take place, as can be concluded from Tables 1 and 2. The map is divided into differently colored boxes on the basis of the operon composition and EC numbers in line with the color scheme in Tables 1 and 2. Abbreviations: DMS, dimethyl sulfide; DMDS, dimethyl disulfide; DMTS, dimethyl trisulfide; CoA, coenzyme A.

Table 1.

Presence and regulation of genes encoding central cysteine and methionine metabolism in Enterococcus, Lactobacillus, and Leuconostoc genomes^a

^aThe genes present in the same operon are indicated by a similar coloring of the cells and/or by †. In case more than one closely related sequence was present, the total number is given. The number is in parentheses in case not all are present in the same operon and/or preceded by a similar putative binding site. In case the gene might be present but was not called, it is indicated by “0?.” The genes have been grouped into five clusters on the basis of the composition of the operons and in case of shared EC numbers. Putative regulator-binding sites upstream of the indicated gene are given in capital letters in superscript: C, CmbR motif; M, MetR motif; N, CodY motif; S, S_MK box; T, T box. Low-scoring putative binding sites are indicated by lowercase letters and a question mark. In several cases we found putative binding sites for two regulators. Species name abbreviations: Lb., Lactobacillus; L., Leuconostoc; P., Pediococcus. The related data and NCBI gi codes can be found in File S1 in the supplemental material, and the analysis of upstream regions can be found in File S3 in the supplemental material. The original data are provided at www.cmbi.ru.nl/bamics/supplementary/Liuetal_2012_CysMetregulation. *, for most enzymes, the related gene names are provided. In most cases these represent names common in all lactobacilli. In cases with little uniformity, the names are derived from the orthologs found in the Bacillus subtilis genome (also see Materials and Methods). Abbreviations in enzyme names: methylenethf, methylenetedrahdrofolate; synth. rev. pathway, synthase for reverse pathway. The E. faecalis V583 genome has a copy of the cmbR gene (indicated by #), and the Lactobacillus delbrueckii subsp. bulgaricus and Lactobacillus plantarum genomes have a copy of the metR gene (indicated by ##). In Lactobacillus plantarum, the second copy of hom is associated with metA and cysD in an operon.

Table 2.

Presence and regulation of the genes encoding central cysteine and methionine metabolism in streptococcal genomes^a

^aThe genes present in the same operon are indicated by a similar coloring of the cells. In case more than one closely related sequence was present, the total number is given. The number is in parentheses in case not all are present in the same operon and/or preceded by a similar putative binding site. In case the gene might be present but was not called, it is indicated by “0?.” The genes have been grouped into five clusters on the basis of the composition of the operons and in case of shared EC numbers. Putative regulator-binding sites upstream of the indicated gene are given in capital letters in superscript: C, CmbR motif; M, MetR motif; N, CodY motif; S, S_MK box; T, T box. Low-scoring putative binding sites are indicated by lowercase letters and a question mark. In several cases, we found putative binding sites for two regulators. Species name abbreviation: Lc., Lactococcus. The related data and NCBI gi codes can be found in File S1 in the supplemental material, and the analysis of upstream regions can be found in File S3 in the supplemental material. The original data are provided at www.cmbi.ru.nl/bamics/supplementary/Liuetal_2012_CysMetregulation. *, for most enzymes the related gene names are provided. In most cases these represent names common in all lactobacilli. In cases with little uniformity, the names are derived from the orthologs found in the Bacillus subtilis genome (also see Materials and Methods). Abbreviations in enzyme names: methylenethf, methylenetedrahdrofolate; synth. rev. pathway, synthase for reverse pathway.

The presence-absence list of genes is in agreement with the results obtained in earlier analyses (36, 41, 59). Nevertheless, we observed a few differences. For instance, we found that O-acetylhomoserine sulfhydrolase (gene cysD) is absent in various Streptococcus pneumoniae strains, including strain TIGR4, and that vitamin B₁₂-dependent methionine synthase (gene yxjH) is present in Streptococcus gordonii. We also found that the enzyme cystathionine beta/gamma-lyase of Streptococcus pyogenes is more similar to that of B. subtilis (gene yjcJ) than to that of Lactococcus lactis (gene ytjE). A multiple-sequence alignment of MetA showed that the protein in all Lactobacillales carries the specific Glu residue at position 111 that renders the MetA protein of Bacillus cereus an acetyltransferase instead of a succinyltransferase (as shown elsewhere [83]) (see File S4 in the supplemental material).

Because we could include a large number of species and strains (i.e., 96 genomes), the general trends in the presence or absence of certain enzymes became more apparent than in the earlier comparative analyses (36, 41, 59). Owing to the larger number of genomes, cases of putative horizontal gene transfer could be recognized. As an example of the latter, orthologs of cystathionine synthase and cystathionine beta/gamma-lyase, represented by genes yrhA and yrhB, respectively, in B. subtilis, were not found among the streptococci, except for the sequenced S. thermophilus strains. We observed that the upstream region associated with the yrhA gene in S. thermophilus was identical (besides a few single nucleotide polymorphisms) to that of yrhA in Lactobacillus helveticus and L. delbrueckii subsp. bulgaricus and that of yrhA on plasmid pLC1 of Lactobacillus rhamnosus Lc 705w (see Fig. S2 in the supplemental material), suggesting a recent plasmid-mediated transfer of the genes and upstream sequences to all the different strains individually or an extreme stability of this particular sequence.

We found the enzymes 5′-methylthioadenosine nucleosidase/S-adenosylhomocysteine nucleosidase (the corresponding gene is designated mtn in B. subtilis and Escherichia coli or pfs in Streptococcaceae) and S-adenosylmethionine synthetase (gene metK) to be present in all analyzed genomes and thus potentially essential. Indeed, the metK gene product was shown to be essential to the growth of E. coli K-12 (79), and the mtn (i.e., pfs) gene product was recently put forward as a critical enzyme for bacterial metabolism (55). The enzyme S-ribosylhomocysteinase (gene luxS) was found to be absent only in Lactobacillus sakei. Besides, serine acetyltransferase (gene cysE) and cysteine synthase A (gene cysK) are present in all Streptococcaceae and N5-methyltetrahydrofolate methyltransferase (gene yxjH or metE2) is present in all Lactobacillaceae except for L. sakei and L. helveticus. The remaining makeup of cysteine and methionine metabolism appeared to be more variable between species. L. sakei (Lactobacillaceae) and Streptococcus equi (Streptococcaceae) have the least extensive enzyme repertoire, with 2 and 7 enzymes, respectively.

Identification of riboswitches.

Recently, three comprehensive studies described the occurrence and evolution of T boxes among prokaryotes (29, 77, 80). We have used the T-box HMMs described previously (80) to search recently acquired genome sequences of, e.g., L. delbrueckii subsp. bulgaricus, Lactobacillus brevis, and L. reuteri. We identified many new T boxes associated with genes/operons involved in cysteine and methionine metabolism in these genomes. As a control, we also scanned the other genomes included in previous studies (77, 80). The conservation of the ATG specifier codon in the multiple-sequence alignment of the newly recovered T boxes (see Fig. S1 in the supplemental material) implies that they all respond to the absence of methionine. We found that genes/operons metB (BS_yjcL), metE-metH (BS_metE-yitJ), and hom1-metA-cysD from Lactobacillus plantarum WCFS1 and genes/operons LEUM_1806-LEUM_1803 (BS_metA-yjcL-yjcJ-yxjG), LEUM_1802 (luxS), and LEUM_1795-LEUM_1794 (metE-metF) from Leuconostoc mesenteroides are regulated by a Met T box, in agreement with previously published findings (77, 80). cysE, encoding a serine acetyltransferase, was also found to be preceded by a Met T box in B. subtilis, as reported previously (56). The association with a Met T box appeared to be almost fully conserved within the Lactobacillus species for the genes that encode the proteins responsible for the synthesis of methionine from homocysteine, i.e., metEF, yxjH-yxjG, and luxS, and for the genes encoding a methionine ABC import system, where metQ encodes the substrate binding protein.

S-Adenosylmethionine-sensitive SAM-I riboswitches are often found upstream of genes involved in sulfur metabolism and transport in bacilli and clostridia (3, 59), but they have not been reported in LAB. In accordance, we did not detect SAM-I riboswitches upstream of genes involved in cysteine and methionine metabolism in the genomes that were analyzed. However, another SAM-responsive element was reported (22) upstream of the metK gene in lactobacilli and streptococci and was named the S_MK box. We used two conserved structures forming stretches of about 6 nucleotides from the reported box (given in Fig. 2A) to search for potential S_MK boxes. We found the two motifs in the correct order upstream of the metK gene in all analyzed genomes but no additional hits (see File S3 in the supplemental material). For all Lactobacillaceae and many Streptococcaceae, the distance was about 50 nucleotides, whereas in other streptococci, like Streptococcus dysgalactiae, Streptococcus gallolyticus, Streptococcus mitis, S. mutans, Streptococcus pasteurianus, and S. pyogenes, this distance was much larger, at about 350 nucleotides. This huge variability in spacing that was also observed previously (22) raises interesting questions related to the way in which sequences of such different lengths can form similar three-dimensional structures to accommodate SAM binding.

LysR-family regulators related to cysteine and methionine metabolism.

MetR from S. mutans, MtaR from Streptococcus agalactiae, CmbR from L. lactis, and HomR from S. mutans have been identified as being important regulators of the genes related to cysteine and methionine metabolism in streptococci (25, 68, 69). They belong to the LysR family of transcription regulators. A neighbor-joining tree of LysR-family proteins was constructed on the basis of a comprehensive BLAST search for homologs (cutoff E value, 1e5). There was a clear division of protein sequences in three subclusters, corresponding to MetR/MtaR, CmbR, and HomR, within the resulting NJ tree, as was also observed previously (36). The tree (Fig. 3) clearly shows that MetR of S. mutans and MtaR of S. agalactiae are orthologous and that CmbR and HomR are their closest relatives. The phylogenetic analysis further implies that CysR from S. mutans is orthologous to CmbR from L. lactis, although CysR was proposed to be a new regulator separated from the CmbR cluster (69). Our assessment of orthology between the two transcription factors is supported by the similarity of their cognate DNA-binding motif (see below). For the other species, only Lactobacillus delbrueckii subsp. bulgaricus, L. plantarum, and Enterococcus faecalis possessed a homolog. The first two are clearly orthologous to MetR/MtaR, whereas the enterococcal regulator seems to be the most related to CmbR.

Fig 3.

Bootstrapped (n = 1,000) partial neighbor-joining tree of LysR-family transcription regulators in representative members of the Lactobacillales. Only the branches of the CmbR, MetR/MtaR, and HomR subfamilies are shown. NCBI reference sequence (RefSeq) gi codes precede the species/strain names (synchronized with Tables 1 and 2). Experimentally validated regulatory proteins are indicated by dots. Abbreviations: L., Lactococcus; Lb., Lactobacillus.

The LysR-family proteins have a domain architecture that is common for transcriptional regulators in prokaryotes. The member proteins consist of a signal molecule binding domain, followed by an HTH DNA-binding domain. The family contains a number of well-studied proteins, including AlsR, CcpC, CitR, GltC, YwfK, and CmbR (43). We have compared the results of DNA-binding studies for LysR-family members for a number of Firmicutes (see the data in Table S1 in the supplemental material). A straightforward alignment and comparison of the reported binding sites reveal a common motif structure of the cis elements, namely, ATNNNN---NNNNAT. The motif displays a clear dyad symmetry and a conserved spacing of 3 nucleotides. This architecture agrees well with the observation made by Schell (62) that LysR-family members generally recognize a box with a conserved sequence, T-N₁₁-A, located 50 to 80 bp upstream of the transcriptional start site. In fact, LysR-family members in general assemble as tetramers (dimer of dimers), and thus, their binding locus is most often composed of two adjacent dimer binding sites where the spacing between the two sites may vary (43). The motifs for LysR binding that have been defined more recently in some cases deviate from this family consensus, and we have therefore redefined them, as described below, taking the mechanistic/molecular characteristics of the binding into account. The motifs were then used to search the lactobacillus genomes for putative binding sites, and the results are given in Tables 1, 2, and 3.

Table 3.

Presence and regulation of the cysteine and methionine ABC transport systems in streptococcal genomes^a

^aThe genes present in the same operon are indicated by similar shading of the cells. In case more than one closely related sequence was present, the total number is given. In case the gene might be present but was not called, it is indicated by “0?.” The number is in parentheses in case not all are present in the same operon and/or preceded by a similar putative binding site. Putative regulator-binding sites upstream of the indicated gene are given in capital letters in superscript: C, CmbR motif; M, MetR motif. Low-scoring putative binding sites are indicated by lowercase letters and a question mark. In several cases we found putative binding sites for two regulators. Species name abbreviation: Lc., Lactococcus. The related data and NCBI gi codes can be found in File S1 in the supplemental material, and the analysis of upstream regions can be found in File S3 in the supplemental material. The original data are provided at www.cmbi.ru.nl/bamics/supplementary/Liuetal_2012_CysMetregulation.

MetR/MtaR.

A 17-bp palindromic conserved sequence, TATAGTTTNAAACTATA, was identified upstream of metY, metA, metQ, metI, and the metEF operon in streptococci and upstream of the metEF operon in L. lactis (36, 59). However, at that time, the regulatory protein, which should bind to the so-called MET box, had not yet been identified. Rodionov et al. (59) initially proposed that the transcriptional regulator MtaR, known to be involved in methionine uptake in S. agalactiae, was a good candidate. Indeed, it was shown soon after that MetR is the regulator protein that binds to the MET box in S. mutans (68). In fact, MetR and MtaR can be inferred to fulfill the same role, as they are orthologs (Fig. 3). The MET-box motif was identified in the upstream regions of atmB, metE, cysD, metA, and smu.1487 in S. mutans, and binding of MetR to these MET boxes was confirmed using gel mobility shift assays and base substitutions in the MET boxes (68).

We have analyzed the upstream regions of the genes orthologous to metE and metH (mmuM) in the selected genomes for the presence of a sequence similar to that of the MET box. Like others, we observed two MetR/MtaR-binding sites in the upstream region of the metH (mmuM) and metE genes in the L. lactis and S. thermophilus genomes. The first site is often located about 60 to 70 bp from the transcriptional start and displays an activating role (68). The second putative binding site, which is closer to the transcriptional start, is far less conserved. The observed organization of the binding sites appears to be general for LysR-family members and relates to a mechanism that requires tetramer formation of the transcription factor (43). It has clear implications for the dynamics of LysR-family-mediated transcriptional regulation, as described for the enterobacterial nitrogen assimilation control protein Nac (60).

The dissected sites were used to redefine the binding motif for MetR/MtaR in both lactococcal and streptococcal strains (Fig. 2B; see File S2 in the supplemental material). We used the experimentally identified binding sites of all LysR-family regulators to guide the definition (see above and Materials and Methods). The putative binding sites of MetR/MtaR showed an overall palindromic structure, ATA-N₉-TAT, which is typical for the LysR family. The most conserved element, ATAGTT, is located upstream, whereas the downstream element, N₃-XXCTAT, shows somewhat less conservation. The complete MetR/MtaR dimer-binding motif that we thus define is ATAGTT-N₃-XXCTAT. The newly defined MetR/MtaR-binding motif is completely covered by the earlier defined MET box, yet it is 2 nucleotides shorter so that it agrees with the LysR-family characteristic. The recovery of putative binding sites in genome-wide searches is very sensitive with respect to the precise composition of the search motif, and therefore, we used both the newly defined motif and the extended motif (1 nucleotide at each side) reported previously (59) to identify MetR/MtaR-binding sites. We observed only a few differences between the two searches (not shown), although the latter search seemed to be more discriminative, and the results given in Tables 2 and 3 and File S3 in the supplemental material therefore relate to the second search. The observed positive effect of the addition of the flanking nucleotides in the search and their conservation could well be related to potential effects of the flanking nucleotides on the molecular structure of the binding site (where the physical binding occurs) and therefore on binding affinity.

A clear MetR/MtaR-binding motif was discerned in the upstream region of metA, yjcL patB, metEF or metE yitJ, mmuM (metH), and yxjH and the operon related to ABC-mediated transport of methionine in most streptococcal genomes (Tables 2 and 3; see File S3 in the supplemental material), as reported earlier (59). These regulatory relations seem to be conserved (with a few exceptions, as described below); i.e., when the genes are present, the MetR/MtaR-binding site is also present. However, some sites score relatively low, which could be indicative of lost function but could also be very well related to slightly altered binding preferences between species. In the case of S. gordonii, Streptococcus parasanguinis, and Streptococcus sanguinis, the S_MK box upstream of metK is preceded by a MetR/MtaR-binding site. As S. equi, S. dysgalactiae, Streptococcus parauberis, and Streptococcus uberis lack the related genes, MetR/MtaR-mediated regulation seems to be restricted to methionine import via the ABC transport system. Another regulatory connection that was conserved in at least three species was observed in S. mitis, Streptococcus oralis, and S. pneumoniae for the genes fhs and folD, encoding formate-tetrahydrofolate ligase and a bifunctional methylenetetrahydrofolate dehydrogenase (NADP⁺)/methenyltetrahydrofolate cyclohydrolase, respectively, which are important in the biosynthesis of the cofactor tetrahydrofolate.

In L. lactis, MetR/MtaR-mediated regulation seems to be restricted to metEF, although most of the other genes are present in the L. lactis genomes. In the lactobacilli L. delbrueckii subsp. bulgaricus and L. plantarum, only one gene seems to be connected to MetR/MtaR. In L. delbrueckii subsp. bulgaricus, the gene metE2 is preceded by an obvious binding site. Remarkably, in other lactobacilli, the gene is preceded by a T box. As the gene is adjacent to the gene encoding the regulator, it could well be that this unit has been acquired from some Streptococcus. In fact, the genes metE and metR are also neighbors on the L. plantarum genome. Surprisingly, in L. plantarum the MetR/MtaR-binding site appears to have been lost. However, a binding site was found upstream of trxB1 (lp_0761), a gene encoding a thioredoxin reductase, more remotely connected to sulfur metabolism. Similarly, a binding site was found upstream of coaE in L. lactis, which encodes dephospho-coenzyme A kinase, the enzyme that catalyzes the final step in the biosynthesis of the thiol compound coenzyme A.

CmbR/CysR.

CmbR positively regulates the metC-cysK (i.e., yrhB-cysK1) operon in L. lactis (25). In fact, the expression of 18 genes is affected by a cmbR knockout in this species (70). The genes, which are grouped in several transcriptional units, include cysD, cysM (i.e., cysK2), yhcE (i.e., yxjG or yxjH), metC-cysK (i.e., yrhB-cysK1), and metA-metB1-ytjE (i.e., metA-yjcL-patB) and the methionine (plpABCD ydcBD) and cystine (yjgCDE) ABC transport systems (70). It was shown that binding of CmbR to the metB2-cysK (i.e., yrhB-cysK1) promoter is stimulated by O-acetyl-l-serine and by low cysteine and methionine concentrations in L. lactis (16, 25). In vitro promoter binding studies confirmed strong binding of CmbR to the metB2-cysK (i.e., yrhB-cysK1) promoter and weaker binding to the promoter of cysD and cysM and of the methionine (plpA) and cystine (yjgC) ABC transport system operons. Hardly any binding was observed for metA-metB1-ytjE (i.e., metA-yjcL-patB) and yhcE (i.e., yxjG or yxjH) (see Fig. 2 in reference 70). In S. mutans, cysK, tcyD (i.e., tcyJ or tcyK), metBC (i.e., yjcL-patB), and homR expression was shown to be affected upon deletion of cmbR (69).

As CmbR belongs to the LysR family of transcription regulators, one would expect its binding site to have a common LysR-family motif. However, all studies dedicated to CmbR binding to date have resulted in a distinctly different CmbR-binding motif definition. In the first study, a deletion analysis showed that a direct repeat of ATAAAAAAA is required for metC activation by CmbR (25). In the second study, the upstream regions of seven transcriptional units found to be regulated by CmbR in L. lactis were analyzed. A first consensus binding sequence, TWAAAAATTNNTA, centered 46 to 53 bp upstream of the transcriptional start, with a second consensus, TWAAAWANNTNNA, located 8 to 10 bp upstream, was proposed (70). The approximate location of CmbR binding in the metB2 and cysD upstream regions was determined by gel-shift experiments (25, 70). A more recent publication by Kovaleva and Gelfand (36) describes an in silico analysis of upstream regions of cysteine biosynthesis genes in streptococcal species and, to increase the recognition power, defines the CmbR-binding motif to be TGATA-N₉-TATCA-N_2-4-TGATA.

To resolve the discrepancies resulting from the previous analyses, we reanalyzed the reported CmbR-binding sites in L. lactis and Streptococcus species. We identified the consensus binding sequences in the upstream region of the CmbR-regulated cysK gene and its paralogs/orthologs from lactococcal and streptococcal genomes (Fig. 2B; see File S2 in the supplemental material). Since two subclusters with respect to the cysK gene can be distinguished by phylogeny and gene context in the L. lactis strains, at first, a putative CmbR-binding site was identified for each subcluster. In case of the so-called cysM1 cluster, CmbR-binding sites from other streptococci such as S. agalactiae and S. gordonii were also taken into account. Nevertheless, all the putative CmbR-binding sites in both Lactococcus and Streptococcus could be summarized by a motif of dyad symmetry ATA-N₉-TAT, with a preference for a long stretch of adenines following the starting ATA. This putative CmbR-binding motif has the general LysR-family signature, and, in addition, all the previous experimental work on CmbR binding supports this assignment. Moreover, like in the case of MetR/MtaR, a second more degenerate site is found directly downstream. Similarly, we observed a positive effect of the addition of the flanking nucleotides in the search for putative CmbR-binding sites, and we have thus used the extended motif in our searches (Fig. 2B).

In the NJ tree, the CmbR subfamily is most closely related to the MetR/MtaR subfamily (Fig. 2). Therefore, one might expect that their binding sites resemble each other. Indeed, the motifs are remarkably similar. However, there were some small differences, especially in the conserved flanking residues. As in the case of MetR/MtaR, we have thus used the extended CmbR motif to search for additional binding sites (in this case, 3 nucleotides on each side). The results are depicted in Tables 1 to 3 (see File S3 in the supplemental material). We retrieved the known binding sites in the region upstream of metB2-cysK in the Lactococcus strains and cysK in the Streptococcus strains. We also identified a novel site upstream of cysD in the S. thermophilus strains, whereas in, for example, S. mutans, the regulation of cysD seems to be mediated by MetR/MtaR. Another conserved relation of CmbR was found with homologs of the gene encoding the ABC transport-related cystine substrate binding protein (6, 52). In many streptococci, two CmbR-regulated copies of the gene are present; the first is related to tcyA or yxeM of B. subtilis, and the second is related to tcyJ and tcyK of B. subtilis. The first is associated with the MetR/MtaR-regulated methionine ABC transport-related operon (49, 82), and the second is associated with genes encoding the permease and ATP-binding subunit of a separate cystine ABC transport system (related to tcyLMN in B. subtilis [6]).

Remarkably, the scoring suggests that in S. suis the regulation of metBC, metEF, and the methionine ABC transport-related operon is also CmbR dependent instead of MetR/MtaR dependent, like in the other streptococci. For S. mutans, our observations fit the regulon derived on the basis of the cmbR-knockout experiments (69). In contrast, the number of putative CmbR-binding sites detected in L. lactis was low compared to the number of reported putative sites (70). Only the highest-affinity promoter was retrieved. Apparently, the defined motif is not very discriminative in the analysis of the L. lactis genomes. Other regulatory connections that were conserved in at least three species were observed with genes encoding pyruvate formate lyase, dihydrofolate reductase, a glutamine amidotransferase, a beta-lactamase, glyceraldehyde 3-P dehydrogenase, and a specific RNA methyltransferase in various species (Table S2 in the supplemental material).

HomR.

HomR is the third transcriptional regulator of the LysR family that is closely related to MetR/MtaR and CmbR and which was found only in S. gallolyticus, S. mutans, S. pasteurianus, and S. thermophilus. The expression of S. mutans metBC, encoding cysteine biosynthesis genes, and the tcyDEFGH (i.e., tcyJKLMN) cluster, encoding a cysteine transport system, is specifically affected in a homR-knockout mutant (69). We examined the upstream region of metB in the HomR-containing streptococci and could identify a clear HomR-binding motif according to the common structure of LysR-family DNA elements (as given in Fig. 2B). The binding motif resembles that of MetR/MtaR, with the most obvious differences being located in the flanking nucleotides. Similar to MetR/MtaR and CmbR, HomR exerts control over metB expression via a second binding site located downstream with lower similarity. We have used the extended HomR-related motif to search for additional binding sites (see File S3 in the supplemental material). Besides the conserved relation with regulation of the metBC (i.e., yjcL-patB) operon, we found a HomR-specific site upstream of folD in S. pasteurianus and panE (encoding a 2-dehydropantoate 2-reductase) in S. thermophilus. In addition, the relative scoring is suggestive of some overlap with the regulation via MetR/MtaR for the metE yitJ operon in S. gallolyticus, S. mutans, and S. pasteurianus and for metA in S. thermophilus.

CodY.

A nutritional regulator involved in the global control of amino acid metabolism in Gram-positive bacteria is CodY (67, 71). It has been extensively studied in L. lactis (11, 28) and other streptococci (8, 39, 40) and was reported to control the expression of hom, thrA, thrB, thrC, and cysD in these organisms. We defined a CodY binding motif on the basis of the data provided elsewhere (11). The motif is given in Fig. 2C and is almost identical to the experimentally defined motif of B. subtilis (5). The motif was used to search for putative CodY binding sites upstream of the genes related to cysteine and methionine metabolism. The results of the search are listed in Tables 1 and 2 and File S3 in the supplemental material. We found similar motifs upstream of thrA, thrC, and/or the hom-thrB operon in many of the streptococcal genomes. However, the upstream regions are relatively dissimilar between the species, and as a result, for various other streptococci, it is not particularly clear whether the CodY binding site is conserved (not shown). We found a potential site upstream of the metA-metB1-ytjE (i.e., metA-yjcL-patB) operon in L. lactis and the metB1-ytjE (i.e., yjcL-patB) operon in S. thermophilus.

Novel elements upstream of cbs and thrB genes.

A comparative analysis (see Materials and Methods) of the upstream regions of the other genes involved in cysteine and methionine metabolism did yield three additional putative regulatory elements. The yrhA-yrhB operon in various lactobacillus strains, such as Lactobacillus salivarius, L. plantarum, L. delbrueckii subsp. bulgaricus, and Lactobacillus acidophilus, as well as in Oenococcus oeni and Leuconostoc, is preceded by a conserved palindromic motif, AAAGGGCGCGAA-N_(11-18)-TTCGCGCCTTTT (Fig. 4A). As described earlier, all S. thermophilus strains carry an orthologous operon with a completely conserved upstream region. We could not detect identical sites elsewhere on the genome. Considering the variable spacing between the complementary stretches, the motif probably represents a structural element. Similarly, a multiple-sequence alignment of the upstream regions of thrB from L. acidophilus and Lactobacillus gasseri revealed another previously unidentified conserved motif (Fig. 4B). The putative motif consists of inverted repeats separated by 15 bp (ATTGTAAC-N₁₅-GTTACAAT). Interestingly, each part of the inverted repeat complements itself (e.g., in the first part, ATTG is immediately followed by its complement sequence, TAAC). Again, no similar sites were found elsewhere on the genome. Another potential structural motif was recovered upstream of thrB in L. lactis (Fig. 4C). The 12-nucleotide sequence is found more than 100 times in a seemingly random distribution in all L. lactis genomes and less than 2 times in all other analyzed genomes (not shown). The motif was discovered before and previously called the highly repetitive motif (50). The absolute conservation and occurrence imply functional importance to L. lactis, although it has yet to be discovered what the precise regulatory role might be.

Fig 4.

Potential regulatory motifs related to the control of Cys and Met metabolism. The motifs were identified upstream of the yrhAB operon in LAB (A), in thrB in LAB (B), and in L. lactis (C). The motifs are palindromic and therefore could relate to structure-forming elements. The matching parts of the sequence have been underlined in different shades. The motifs were created with WebLogo (frequency representation; no correction applied; http://weblogo.berkeley.edu).

DISCUSSION

We have performed a genome-wide in silico analysis to reveal the transcription regulatory interactions that control the expression of the genes encoding various key enzymes involved in cysteine and methionine metabolism in all sequenced species of the order Lactobacillales. The associated regulators that we could identify were the known regulatory proteins, such as CmbR, MetR/MtaR, HomR, and CodY, as well as known RNA riboswitches. In addition, we found two potential regulatory structure-forming elements. We redefined specific binding motifs for CmbR, MetR/MtaR, and HomR in lactococci and streptococci. All motifs follow the characteristic LysR-family signature (ATA-N₉-TAT) and are substantiated by the available experimental data. The motifs allowed a computational separation of the various binding sites within the streptococci. The identified regulator-gene associations overlapped well with the subdivision of the reaction network that was made on the basis of the operon composition and on the basis of EC numbers (compare Fig. 1 with Tables 1 and 2).

A clear diversity in the transcriptional regulation network between the different families was observed. In various Lactobacillaceae like L. plantarum, the biosynthesis of methionine, as well as its precursors, i.e., homocysteine and cystathionine, is responsive to low methionine levels through T-box-mediated regulation. Similarly, the ABC transport of methionine is regulated via a T-box riboswitch. We could not identify new general sites related to the synthesis of cysteine from homocysteine and the degradation of both cysteine and methionine in the Lactobacillaceae. This may indicate that most of the relevant regulators are known or that potential other elements might be less conserved or species specific. In streptococci, the gene-regulator associations nicely fit the established inducer substrates. MetR/MtaR controls the expression of the genes that encode biosynthesis and transport of methionine, whereas CmbR relates to the control of cysteine biosynthesis and cystine transport. HomR appears to be specifically dedicated to the control of yjcL-patB expression in four streptococci. The latter relationship makes sense, as the genes encode the enzymes responsible for the conversion of acetylhomoserine to cystathionine and further. In S. thermophilus, an additional gene controlled by HomR encodes the protein that catalyzes the production of acetylhomoserine. The intracellular levels of homoserine seem to be controlled by the global regulator CodY.

The biosynthesis of flavor compounds, e.g., H₂S, methanethiol, DMS, DMDS, and DMTS, is catalyzed directly by cystathionine beta/gamma-lyase, which is encoded in most streptococci by a single gene, the expression of which is controlled by MetR/MtaR (methionine). Of course, changing the levels of the flavor precursors methionine, cystathionine, and cysteine will also affect the rate of synthesis of flavor compounds. These levels are determined by the activity of cysteine synthase (cysK) and cystathionine synthase (yrhA), homocysteine S-methyltransferase (yitJ and mmuM), and S-adenosylmethionine synthetase (metK) and thus controlled by CmbR (i.e., O-acetyl[homo]serine), MetR/MtaR (i.e., methionine), and an S_MK box (i.e., S-adenosylmethionine), respectively. Inclusion of the above-described reaction network and the regulatory interconnections into a quantitative metabolic network model may help to rationalize new strategies for controlling the sulfuric flavor formation in various LAB-fermented food products.