Identification of the Leucine-to-2-Methylbutyric Acid Catabolic Pathway of Lactococcus lactis

Abstract

Nutrient starvation and nonculturability in bacteria lead to changes in metabolism not found during the logarithmic phase. Substrates alternate to those used during growth are metabolized in these physiological states, yielding secondary metabolites. In firmicutes and actinobacteria, amino acid catabolic pathways are induced during starvation and nonculturability. Examination of lactococci showed that the population entered a nonculturable state after carbohydrate depletion and was incapable of growth on solid media; however, the cells gained the ability to produce branched-chain fatty acids from amino acids. Gene expression profiling and in silico pathway analysis coupled with nuclear magnetic resonance spectroscopy were used to delineate the leucine catabolic pathway. Lactococci produced acetic and propionic acid during logarithmic growth and starvation. At the onset of nonculturability, 2-methylbutyric acid was produced via hydroxymethyl-glutaryl-coenzyme A (CoA) and acetyl-CoA, along with ATP and oxidation/reduction precursors. Gene expression profiling and genome sequence analysis showed that lactococci contained redundant genes for branched-chain fatty acid production that were regulated by an unknown mechanism linked to carbon metabolism. This work demonstrated the ability of a firmicute to induce new metabolic capabilities in the nonculturable state for producing energy and intermediates needed for transcription and translation. Phylogenetic analyses showed that homologues of these enzymes and their functional motifs were widespread across the domains of life.

Bacteria grow under many nutrient conditions throughout their life cycle. During logarithmic growth, an abundance of nutrients provide carbon and nitrogen to promote cell division and replication, whereas during stationary phase, the carbon and nitrogen sources are often exhausted. While nutrient limitation is one among many conditions that induce bacteria to sporulate (38), this ability is lacking in actinobacteria (e.g., Brevibacterium linens, Mycobacterium, Micrococcus, and Vibrio spp. [8, 18, 38]) and some firmicutes, such as Lactococcus and Listeria spp. These two groups of bacteria enter a metabolically active but quiescent state in response to the local nutrient availability (36).

In this state lactococci metabolize amino acids (36). Normally, lactococci ferment sugars, producing lactic acid. When the cells enter stationary phase a portion of the population retains the ability to maintain an intact cell membrane and begins to use proteins for energy (36). Further incubation results in a nonculturable (NC) state for periods of at least 4 years (15). It is likely that these conditions lead to novel pathways that provide alternate carbon and energy sources and are of primary interest in lactococci.

Catabolism of amino acids such as arginine and branched-chain amino acids (BCAAs) is implicated in long-term survival during starvation of lactococci (4, 14). While addition of arginine extends survival during early post-carbohydrate depletion (4, 36), in the later stages, the consumption of other amino acids, such as BCAAs and α-keto acids, produces branched-chain fatty acids (BCFAs) (13). Known pathways for this catabolism in bacteria yield ATP via substrate-level phosphorylation and reducing equivalents (i.e., NAD) (13). The availability of BCAAs and α-keto acids provides lactococci with sources of nutrients for cellular energy and metabolites that aid cellular maintenance during stress during carbohydrate starvation.

Use of the genome sequence for pathway analysis is providing insight into metabolic capability. The genome sequence of Lactococcus lactis IL1403 is publicly available (3), a draft genome of Lactococcus cremoris SK11 (http://genome.jgi-psf.org/draft_microbes/laccr/laccr.home.html) is in the finishing stages, and the L. lactis MG1363 genome is being annotated (26). The availability of these three closely related genomes offers a new ability to compare metabolic capabilities to determine the details of gene expression due to environmental conditions. Genome-wide expression profiling yields putative targets for the study of metabolic modulation. Combining the gene expression studies with intermediary metabolite analysis will provide insights into new cellular processes that enable new mechanisms of controlling bacterial metabolism. Using the sequences to analyze the presence of homologues also demonstrates the importance of the pathway across multiple organisms.

Previously, we demonstrated that sugar starvation induces the NC state in L. lactis with use of amino acids as an energy source (36). Based on the ability of lactococci to catabolize amino acids to FAs (13) and their ability to survive extended periods of starvation (15), we hypothesized, despite the lack of genomic evidence, that lactococci produce BCFAs from BCAAs during carbohydrate starvation via one of the established lipid metabolic routes in gram-positive bacteria. To prove this hypothesis, we determined the intermediary metabolic products of L. lactis IL1403 from leucine using nuclear magnetic resonance (NMR) spectroscopy while simultaneously measuring the genome expression profile. Specific attention was given to relevant phenotypic milestones (sugar depletion, onset of NC, and BCFA production) to elucidate a putative metabolic route of BCAA catabolism.

MATERIALS AND METHODS

Cultures and media.

L. lactis IL1403 was obtained from the Utah State University culture collection and grown in M17 broth (Difco Laboratories, Detroit, MI) with 2% glucose added (M17G broth). Stock cultures were prepared by growing the cells twice in 10 ml of M17G broth at 30°C for 24 h. Cultures were frozen at −70°C in 10% nonfat dry milk containing 30% glycerol. Before each experiment, a frozen stock was thawed and subcultured twice at 30°C for 24 h in 10 ml of M17G broth. The culture was harvested by centrifugation (3,800 × g for 15 min at 4°C), washed twice with and resuspended in sterile saline, and inoculated (1%) into sterile chemically defined basal medium (22). The medium contained 0.2% glucose and 200 mg/liter of BCAAs, as opposed to the chemically defined medium used in other studies (35). The pH was adjusted to 7.2 and maintained with 0.19 M 3-(N-morpholino)propanesulfonic acid (Sigma-Aldrich, St. Louis, MO). The medium was filter sterilized with a 0.2-μm-pore-size sterile filter (Corning Inc., Corning, NY).

Culturability and viability.

Culturability was estimated on M17 agar containing 0.5% glucose. Viability of the culture was simultaneously determined using the Baclight Live-Dead viability kit (Molecular Probes Inc., Eugene, OR) as previously described (36). Intracellular ATP concentration was estimated by chemiluminescence as previously described (36).

Glucose determination.

Glucose was quantified with the spectrophotometric assay of Dubois et al. (7). Concentrations were estimated from a standard curve that was linear over the detection range with an R² value of ≥0.98.

Gas chromatography of FAs.

FAs were estimated from culture supernatants at various time points during incubation. Gas chromatography was used to determine FAs as previously described (13), with a slight modification to the column (see Materials and Methods in the supplemental material).

α-Keto acid analysis.

α-Keto acids were estimated from cell extracts as previously described (6). Corrected peak areas were used to calculate concentrations of individual α-keto acids from standard curves that were linear over the detection range (R² ≥ 0.98) (data not shown).

Metabolic assays.

Substrate utilization was determined at the time of sugar exhaustion and NC as described previously (13). The substrate for BCAA catabolism was 20 mM [2-¹³C]leucine in 190 mM 2-(N-morpholino)propanesulfonic acid-sodium salt buffer (pH 7.2). Leucine assays were done in the presence and absence of 20 mM pyruvate. The cells were incubated with the substrates at 30°C for 3 h. After incubation, the filter-sterilized assay buffer was analyzed for products from [2-¹³C]leucine by NMR spectroscopy as described below, FAs were analyzed by gas chromatography, and α-keto acids were analyzed by capillary electrophoresis. Cell extracts were prepared from cell pellets for NMR spectroscopy as described previously (6). Products from the assay buffer were considered extracellular, while ¹³C-labeled products from the cell extract were considered intracellular.

NMR spectroscopy.

The products from ¹³C-labeled leucine (Isotec Inc., Miamisburg, OH) were identified using NMR spectroscopy as previously described (14) (see Materials and Methods in the supplemental material).

Pathway identification.

Based on the observed products from NMR spectroscopy, Pathcomp in KEGG was used to determine a possible pathway (23, 24). The ¹³C labels of the identified products are shown in Table 1.

TABLE 1.

Locations of ¹³C labels in products of samples analyzed by NMR spectroscopy and their putative candidate genes and enzymes

^aProduct was found both inside and outside the cell.

^bProducts were found only inside the cell.

^cGray circles indicate ¹³C label locations.

Protein homology.

KEGG maps for leucine metabolism were produced using the IL1403 genome (see Fig. SA-1 in the supplemental material), the finished genome sequence for L. cremoris SK11 (see Fig. SA-3 in the supplemental material), and other organisms for comparison (see Fig. SA-2 and SA-4 in the supplemental material) (23, 24). If a functionally annotated gene for a particular enzyme was absent, a list of genes that were capable of catalyzing the reaction was identified using the putative EC number and cluster of orthologous groups classification. This list was examined further for gene candidates by comparing the domains of the primary protein structure to other known protein domains using the conserved domain database (version 2.03) (29). Significance of structural homology was assigned at an E value of 10⁻⁶. Gene expression data for identified protein homologues were mapped to the pathway reactions to monitor the expression profiles of the possible enzymes involved in the pathway.

Gene expression profiles.

Gene expression was determined in the NMR assay by collecting RNA for indirect labeling as described by Yi et al. (39) with slight modifications (see Materials and Methods in the supplemental material). NimbleGen Systems hybridized the labeled cDNA. The raw data were normalized using the R program by the robust multichip average method (21). Further details and the annotations of genes described in this study are available (see Tables SA-1 to SA-3 in the supplemental material).

Statistical analysis and data visualization.

All experiments were conducted in two biological replicates. The robust multichip average-normalized data were averaged across replicates before visualization. Expression maps were drawn by using Hierarchical Clustering Explorer version 3.0 (34, 35). Two levels of expression changes were set to a 2.5-fold change apart from each other (25). The heat map for each class of enzymes was drawn separately to generate expression maps. Fold changes for each gene at the end of assay were calculated with data from the start of assay as the control. Statistical significance for differential gene expression was determined using the SAM (Significance Analysis for Microarrays) package (39). The false discovery rate was assigned to be 20% (see Materials and Methods in the supplemental material).

RESULTS

Culturability, sugar exhaustion, and viability.

L. lactis IL1403 utilized 0.2% glucose during exponential growth and attained a maximum cell number of ∼10⁹ CFU/ml within 24 h; however, IL1403 did not reach the NC state until day 21 of incubation. During further incubation the cells remained intact and at a constant level without colony formation on M17-glucose agar plates that were incubated for 48 h (Fig. 1). The intracellular ATP concentration increased to and remained constant at 200 pM during the 1-year incubation period of this study (see Fig. SA-6 in the supplemental material).

FIG. 1.

Cell counts (▪) and fluorescent counts of live (⧫) and dead cells (•) from L. lactis IL1403 during incubation. RFU, relative fluorescence units.

BCFA production.

During logarithmic growth, IL1403 produced acetic and propionic acid. Production of 0.1 mM isovaleric acid was identified in culture at the onset of NC (i.e., day 21). The time of induction for BCFA production was unusually short in comparison to that of the other strains tested (ML3 and SK11) (15), which required a longer time in the NC state for production of BCFAs.

Presuming that all lactococci use pathways similar to that of brevibacteria for production of isovaleric acid (14, 19, 20), we examined the genome of IL1403 for the presence of related genes for BCFA production. While this was successful for the aminotransferases (ATases), unfortunately we did not find homologues for the catabolic pathways for leucine in the genomes of L. lactis IL1403 or L. cremoris SK11. Both lactococcal genomes contained seven ATase homologues, which are used in the first step in amino acid catabolism (17, 28, 33, 41). Abolishing the activity of the branched-chain ATase (ilvE) did not reduce BCFA production but rather changed the FA profile (16), nor did it change the ability to reach NC. While IL1403 produced BCFAs from amino acids, the genes responsible for the known catabolic routes were not identified using in silico genome reconstruction tools. These observations indicate that lactococci may use an unknown pathway for the production of BCFA products or that the annotations in IL1403 are not complete, thereby causing the in silico reconstruction effort to fail in assigning the correct gene to the correct metabolic step. Therefore, we determined the metabolic pathway for comparison with the genome annotations using NMR spectroscopy and gene expression profiles to determine the metabolic intermediates and the associated genes. This resulted in construction of a complete metabolic route for leucine catabolism in lactococci that is different from the pathway in brevibacteria.

NMR spectroscopy.

Whole cells were assayed with [2-¹³C]leucine in buffer to identify metabolites during the NC state (Table 1). The expected transamination product, α-ketoisocaproate, with the ¹³C label on the α-keto group, was detected intracellularly. The primary BCFA product was 2-methylbutyric acid and contained ¹³C labels on the alkyl side chains of the C-3 and C-4 carbons (Table 1), which were detected intra- and extracellularly. The 3-hydroxy-3-methylglutaric acid (HMGA) end product was derived from intracellular 3-hydroxy-3-methylglutaryl (HMG)-coenzyme A (CoA). The addition of pyruvate did not alter the end products of leucine catabolism (data not shown). Interestingly, IL1403 shuffled the ¹³C label to produce 2-methylbutyric acid rather than isovaleric acid (3-methylbutyric acid) during starvation as confirmed by gas chromatographic analysis, indicating that the cell used a complex set of metabolic interconversions between these intermediates.

The other intracellular products labeled with ¹³C were α-ketoisovalerate, citrate, and glutamate (Table 1). The presence of these intermediates indicated that lactococci utilized leucine by different pathways than brevibacteria (14). To delineate the potential metabolic route, we used in silico analysis with the Pathcomp feature in KEGG to compute the biological routes to catabolize leucine to 2-methylbutyric acid, citrate, glutamate, and α-ketoisovalerate (23, 24). This estimation matched two intermediates that were related to products identified by NMR to produce 2-methylbutyric acid. While α-ketoisocaproate, the expected transaminated product of leucine (Table 1), was found, HMGA, the product of HMG-CoA hydrolysis, was also found (Table 1). The pathway from leucine to citrate shared common intermediates until the formation of acetyl-CoA. Pyruvate derived from acetyl-CoA subsequently yielded oxaloacetate and citrate. The computed pathway for glutamate intersected with citrate production, which was further catabolized to glutamate in this state. Leucine was catabolized to α-ketoisovalerate via the generation of pyruvate with three further steps. The in silico analysis and observed intermediates from NMR were used to identify the putative enzymes and their gene homologues in IL1403 and SK11.

Gene expression profiles of candidate enzymes. (i) Identifying gene candidates.

Use of gene expression profiles in combination with NMR led to a putative set of genes and enzymes involved in the metabolic route of BCFAs. A number of steps in the putative pathway used enzymes that were functionally characterized in lactococci, while the remaining steps were catalyzed by enzymes that have not been characterized. The large number of genes is prohibitive for classical gene deletion strategies. Therefore, a metabolomics and bioinformatics approach was used to reconstruct the set of putative genes for further discovery.

As previously established, the first step in the catabolic route used ATases, which are extensively characterized in lactic acid bacteria (32, 41). For the unidentified enzymes needed for the BCFA production pathway, protein domain analysis based on the genome sequence was used to determine a putative identification for misannotated, unknown, or uncharacterized genes. Subsequent studies focused on identifying the genes for catabolism of BCAAs to BCFAs on the basis of NMR studies and protein domain analysis.

Domain homology searches revealed considerable differences in gene annotations between the genomes of SK11 and IL1403 (see Fig. SA-1 to SA-4 in the supplemental material). In some cases this analysis led to genes having the same name but different functional descriptions, despite having significant domain homologies (E value, <10⁻⁶ to 10⁻¹⁰⁰). It also led to genes in other bacteria that encoded enzymes for leucine degradation that were not annotated in lactococci. These observations supported the assertion that the IL1403 genome required additional annotations of genes associated with leucine catabolism for a predictive estimation of the metabolic route by in silico analysis in KEGG.

The role of these domain homologues was investigated using gene expression studies to explore the complex metabolic web of pathways of leucine degradation. Identification of the candidate genes was included in subsequent pathway reconstruction. The pathway contained ATases, dehydrogenases, acyltransferases, carboxylases, transacylases, and acylkinases to convert leucine to 2-methylbutyric acid (see the supplemental material). Homologues of these genes, as determined by CD-Search analysis and domain presence, were identified in over 100 unique genera of bacteria and a limited number of plants, fungi, and eukaryotes (Fig. 2; see the supplemental material).

FIG. 2.

Putative pathway for catabolism of leucine to 2-methyl butyric acid, citrate, glutamate, and α-ketoisovalerate and its gene expression. Compounds were identified by NMR (gray) and in silico analysis (black). Intermediates are shown for production of 2-methylbutyric acid only. The gray circles on the adjacent compound structures reflect peaks that were identified by NMR. Reactions with unique features (ATP production by substrate-level phosphorylation and carbon fixation) are depicted in gray. Regular arrows for reactions show single-step reactions. Dotted arrows depict products from multistep pathways. Based on the products identified by NMR, three forms of labeled acetyl-CoA are potentially present inside the cell, as shown. Color changes within the expression maps are as follows: green to black, 2.5-fold increase in gene expression; black to red, an additional ≥2.5-fold increase in gene expression and vice versa. Lanes: 1, 0 h of IL1403 at sugar exhaustion; 2, 3 h after incubation of IL1403 at sugar starvation in [2-¹³C]Leu; 3, 0 h of IL1403 at NC; 4, 3 h after incubation of IL1403 at NC in [2-¹³C]Leu.

(ii) Gene expression profile.

Whole-genome expression analysis of leucine metabolism during intermediate determination by NMR revealed that none of the genes involved in FA metabolism changed significantly (q value, ≤0.2) at sugar exhaustion, suggesting that they were already expressed above the mean level. At the onset of NC, expression of four unknown genes related to FA production, all of which were designated to be dehydrogenases based on protein domain similarities, significantly decreased (q value, ≤0.2) as follows: ycfD, 2.5-fold; yteC, 2-fold; yuiC, 3-fold; ynjF, 2.5-fold. This indicates that the cells were primed to produce FAs from amino acids and did not need to regulate gene expression to initiate FA production. The limited list of significantly different genes was due to the stringent level of significance testing used to minimize the false discovery of genes that were not associated with this metabolic route. Therefore, use of heat maps for visualization of gene expression patterns at key metabolic events allowed conclusions to be made regarding the relative expression level and direction of expression, which was verified by biological experimentation.

(iii) Aminotransferases.

The first step in catabolizing amino acids is an exchange of the amino group with an α-keto acid, preferably α-ketoglutarate (40). The genome of IL1403 contains nine annotated ATases (EC 2.6.1.42), six of which are biologically uncharacterized. IL1403 also contains an amidase (ybgE) that catalyzes amino group removal, which also is characterized to a limited extent.

The expression of three ATase genes (yeiG, nifZ, and aspB) was regulated by at least 2.5-fold at sugar starvation and NC. One of these genes (yeiG) was repressed by 3.5-fold during the incubation at sugar exhaustion and was not induced during NC, but it was expressed above baseline levels compared to other genes (Fig. 2), while aspB was expressed at the same levels at sugar exhaustion and induced 3-fold during NC. Expression of ybgE, araT, and bcaT did not increase in relation to leucine catabolism, and these genes were not expressed above baseline levels. ATases possess multiple substrate specificities (1, 16, 32). Presumably the production of BCFAs from BCAAs can be routed through alternate ATases that were responsive to starvation conditions, such as yeiG, nifZ, or aspB, which are the primary candidates to catalyze this reaction in these conditions.

(iv) Dehydrogenases.

The conversion of the α-keto acid to an acyl-CoA is catabolized in two different steps by dehydrogenases (EC 1.2.1.25 and 1.3.99.3). The L. lactis IL1403 genome contains 38 uncharacterized dehydrogenases (3). Any biological activity that catalyzes short chain molecules would be favorable for four reactions in the proposed pathway—the α-keto acid dehydrogenase (EC 1.2.1.25) reaction and three different acyl-CoA dehydrogenase (EC 1.3.99.3, 1.1.1.35, and 1.3.1.44) reactions (Fig. 2). Comparing across the assays, four genes from this subset—yphC, ygcA, ypjF, and yugB—were repressed by twofold but their expression remained higher than the baseline, while six genes (yteC, ysjB, ypaI, ypjH, ycgD, and menD) were induced 2- to 2.5-fold during the incubation time (Fig. 2). Presumably, any of these genes may be involved in the catalysis of these dehydrogenase reactions. The gene ywjF is annotated as 3-hydoxyisobutyrate dehydrogenase (2) and was initially a candidate to be the primary enzyme to catalyze this reaction, but the gene expression was near baseline levels and did not change during the incubation time. This indicates that additional dehydrogenases other than ywjF that were induced at NC were more likely to be involved in BCFA production than the remaining dehydrogenases in the genome (Fig. 3).

FIG. 3.

Comparison of pathway motifs of the BCAA catabolic pathway to those of fatty acid oxidation and glycolysis. The preparatory and payoff phases of each pathway are indicated by dotted lines and dashed lines, respectively. Reactions that produce or utilize ATP by substrate-level phosphorylation and those of carbon fixation are depicted. Note that all pathways have acetyl-CoA as an end product.

The two-step dehydrogenase reaction is also catalyzed via α-keto acid dehydrogenases that are structurally very similar to the pyruvate dehydrogenase complex of L. lactis IL1403 (see Tables SA-2 and SA-3 in the supplemental material). The lactococcal genes for the four subunits, except for pdhC at NC, were induced in both assay conditions by 1.5-fold (Fig. 2). This allows one to speculate that the lactococcal pyruvate dehydrogenase complex may be involved in catabolism of other α-keto acids such as α-ketoglutarate, α-ketoisovalerate, and α-ketoisocaproate, assuming that the substrate specificity is this broad, but that the complex may later be disabled, thereby directing the pathway to specific dehydrogenases that were induced at NC and beyond.

(v) Acyl transferases.

An acyl transferase catalyzes the reaction of EC 1.2.1.25 entirely by itself, which involves the removal of the C1 carbon from the α-keto acid to produce acyl-CoA (Fig. 2). This class of enzymes is also involved in the addition of acetyl groups onto the propionyl-CoA backbone (EC 2.3.1.16) (Fig. 2). Among the 12 potential acetyl transferases found in the IL1403 genome, the expression of three genes (yhjG, ycjC, and ycjD) increased by 1.5- to 3-fold over the incubation time (Fig. 2). These three enzymes are likely to be involved in production of 2-methylbutyric acid but are uncharacterized.

(vi) Carboxylases.

Multiple reactions in the proposed pathway result in fixation or removal of carbon via carboxylases. These enzymes are involved in the carboxylase reaction (Fig. 2). There are four homologous genes (accA, ipd, pycA, and pdc) that may catalyze carbon fixation and subsequent carboxylation (EC 6.4.1.4) of 3-methylcrotonyl-CoA. All of these genes were expressed above the baseline or induced by 1.5- to 3-fold across the entire incubation time. Hence, all four known carboxylases are candidates to catalyze this reaction step.

(vii) Hydratases.

Hydratases catalyze the addition of a water molecule to or its removal from short chain molecules. In production of 2-methylbutyric acid they were involved in the hydrolyzing of coenzyme A from HMG-CoA (Fig. 2) and in the reaction of EC 4.2.1.17 (Fig. 2). While no specific hydratases are known for these reactions the IL1403 genome contains nine known hydratases that have similar action (2) and a homologue to the Bacillus subtilis genome (menB). The increase in the gene expression profiles was over twofold for fabZ1 and aroD, while enoB was expressed above baseline throughout the incubation, even after threefold repression at sugar exhaustion (Fig. 2). There was a twofold reduction in the expression of menB in both sugar exhaustion and NC but no change for ysiB, suggesting that these genes were not involved in this catabolic step. Hence, enoB, fabZ1, and aroD are likely candidates to catalyze this reaction.

(viii) HMG-CoA synthase.

This enzyme bidirectionally catalyzes or degrades HMG-CoA to or from acetyl-CoA and acetoacetyl-CoA. The gene hmcM was repressed by twofold under both assay conditions (Fig. 2). This suggests that this gene may not be involved or may be repressed by product accumulation. Alternatively, an unidentified homologue of this gene may be involved in this reaction.

(ix) Phosphotransacylase and acyl kinases.

The action of phosphotransacylases (EC 2.3.1.19) on acyl-CoA intermediates combined with acyl kinases (EC 2.7.2.7) yields FAs via substrate-level phosphorylation (Fig. 2). This is one mechanism for the production of ATP from acyl-CoA, yielding FA end products. The lactococcal genome also contains genes for enzymes that hydrolyze the phosphate molecule from short chain acyl phosphates without generating ATP (apl, yfjC) (2). The expression levels of the transacylase (pta) and acyl kinases (ackA1 and ackA2) were higher than those of the lesser-known phosphatases (Fig. 2). This was in spite of a 2.5-fold reduction in the pta expression level at sugar exhaustion (Fig. 2). The expression of the phosphatase genes was not induced during incubation at starvation or NC. This points to the role of pta in the production of BCFAs.

The identification of an alternate metabolic pathway raised the question of its regulation. Typically in bacteria, activator or repressor proteins bind upstream of the promoter, which are induced by abiotic factors, such as stress. In addition, genes of a particular pathway are commonly organized in operons and are collectively regulated by the same effector. Analysis of the gene organization and binding motifs upstream of the genes for regulator binding sites for this pathway found no common or known regulatory motifs (see Results in the supplemental material).

Determination of the genetic regulatory mechanism using classical isogenic deletion strategies was not feasible for the entire pathway that contains over 80 putative genes. Comparison of the entire pathway structure was used to explore features common to extensively characterized catabolic pathways. For example, acetyl-CoA is a common intermediate from many catabolic pathways that includes glycolysis and FA oxidation.

Many pathway motifs of BCAA catabolism, FA oxidation, and glycolysis were strikingly common (Fig. 3). All three pathways involved a preparatory phase of activation of the substrate via a highly electronegative molecule (i.e., CoA or phosphate). The energy-rich intermediates were catabolized during the “payoff phase” to yield energy (i.e., ATP) from FA oxidation via the Krebs cycle. While some reactions were homologous across BCAA catabolism and FA oxidation, further similarities among BCAA catabolism and glycolysis included intermediates that led to other biosynthetic pathways. For example, α-ketoisocaproate and acetoacetyl-CoA may be diverted to biosynthesis of other amino acids. Fructose-6-phosphate is used for N-amino sugar biosynthesis, while phosphoenolpyruvate and 3-phosphoglycerate are amino acid precursors. Notably, both pathways used nine intermediates containing CoA adducts prior to ATP generation. The motif most common between BCAA catabolism and glycolysis was energy production that results in two net ATP molecules via substrate-level phosphorylation. Glycolysis uses two ATP molecules to activate glucose and produces four ATP molecules via substrate-level phosphorylation. BCAA catabolism uses one ATP to generate three acetyl-CoA molecules. Each acetyl-CoA reacts with a propionyl-CoA to form 2-methyl-acetoacetyl-CoA and, subsequently, 2-methylbutyric acid with one ATP. Acetoacetyl-CoA splits to yield two acetyl-CoAs, which can each lead to 2-methylbutyric acid. Effectively, three acetyl-CoAs are produced from HMG-CoA, leading to three ATP molecules from each leucine molecule via substrate-level phosphorylation (Fig. 3).

BCAA catabolism differed from glycolysis in the use of HS-CoA as opposed to phosphate as the high-energy leaving group that activated catabolism of the substrate. Another distinct feature of the BCAA pathway is that carbon fixation occurred in two steps of the pathway, an unusual feature in catabolic pathways. BCAA catabolism would consequently be of great utility for carbon salvage and storage during carbon starvation (Fig. 3). Notably, phylogenetic analysis showed that many bacterial genera contain one or more homologous genes that belong to the different classes of enzymes of this pathway (see Fig. SA-5 in the supplemental material), suggesting that this pathway may be prevalent across the domains of life.

DISCUSSION

Secondary metabolism during NC is important to lactococci in food and the environment. For organisms like lactococci that are of plant origin and have been domesticated for fermentation processes, the ability to survive nutrient starvation has been largely unexamined due to the belief that the majority of the cell population die and lyse, which is not the case for the entire population (36). A major proportion of the 10⁹-CFU/g population either in medium or cheese presumably remains viable but NC while ∼0.001% of the cell population is made up of lysed or dead cells from the growth phase (12). With this small proportion of lysis relative to the entire population, we examined a metabolic observation that is important to firmicutes that lack extensive genetic redundancy (27). The catabolism of BCAAs in lactococci, a group of firmicutes that primarily uses substrate-level phosphorylation for ATP production during carbohydrate starvation, is of particular interest since it also lacks sigma factors used by other firmicutes for regulation. This pathway identifies a portion of the core genes needed for survival during nutrient starvation (10, 13). While the use of lactococci is common in fermented foods, the exact biochemical mechanisms of the pathways are only being characterized now. A number of pathways have been proposed to generate FAs, but none are verified to be involved in production of straight-chain FAs or BCFAs (11, 37), and none have been shown to link carbon starvation to ATP production.

An interesting feature of leucine conversion to 2-methylbutyric acid is the molecular efficiency in production and use of the intermediates and cofactors resulting in energy and reducing equivalents. The products from earlier steps are subsequently consumed in the later steps or vice versa, as with many biochemical pathways. For example, NADH generated earlier by dehydrogenases is used in subsequent dehydrogenase reactions (Fig. 2). The CO₂ released by the dehydrogenase reenters the pathway via the carboxylase reaction (Fig. 2). Coenzyme A used in earlier steps is also regenerated. The splitting of HMG-CoA into three molecules of acetyl-CoA generates three ATP molecules from the substrate-level phosphorylation reaction with the acyl kinase (Fig. 2). This suggests that during substrate starvation and NC the cells exist in a homeostatic condition where the cell reroutes and reuses the cofactors for carbon fixation. This results in survival during cellular stress from carbon limitation.

This study found straight- and branched-chain FA products (α-ketoisocaproate, HMGA, 2-methylbutyric acid, glutamate, citrate, and α-ketoisovalerate). A shift in the metabolic end products from short, straight-chain FAs to BCFAs occurred only at the onset of the NC state. NMR studies revealed unstable intermediates and accumulation of multiple end products, while gene expression profiles provided insight into the genes involved in the biotransformation. This combination led to a putative mechanism for the utilization of leucine that has not been previously observed in lactic acid bacteria (5), despite being found in Pseudomonas and Clostridium spp. (see Fig. SA-2 and SA-4 in the supplemental material) (9, 30).

Previously postulated mechanisms for the involved pathways have primarily relied on a single or limited combination of reactions, such as an ATase to derive the α-keto acid, dehydrogenases to reduce it to acyl-CoA, and a decarboxylase to produce BCFAs (11). The putative pathway in this study proceeds via a longer metabolic route than previously speculated. Determination of the functionality of dehydrogenases based on structural homology (42) was extended in this study to putatively identify 38 in the IL1403 genome. The activities of dehydrogenases, acyltransferases, carboxylases, and dehydratases produce 2-methylbutyryl-CoA, which is acted on by transacylases and acyl kinases to produce 2-methylbutyric acid. These coupled reactions produce ATP, which provides the cell a competitive advantage during NC. At least one gene for each of these enzymes was expressed (Fig. 2), suggesting that these enzymes are involved in leucine degradation but have extensive redundancy. This is unusual for an organism that has a genome size of 2.3 Mb, perhaps pointing to the critical nature of pathways that involve substrate-level phosphorylation during starvation.

The 2-methylbutyric acid molecule was labeled on both carbon moieties (Table 1). This suggests that acetyl-CoA was labeled at both of its carbons because it forms the aceto branch of 2-methylaceto-acetyl-CoA (Table 1). This was further reduced by a series of steps to the alkyl group of 2-methylbutyric acid (Fig. 2). As none of the subsequent steps involve addition of a carbon moiety to introduce a label, the metabolism of HMG-CoA becomes of interest as it determines the BCFA produced.

The presence of HMGA also suggests that it may act as a metabolic shunt that accumulates carbon during BCAA catabolism. HMG-CoA is a branch point to pathways other than that of FA production, as evidenced by the production of ¹³C-labeled glutamate. Precursors of sterol biosynthesis (HMGA and HMG-CoA), which is incomplete in lactococci, were also observed. HMGA may not be formed unless an excess of HMG-CoA occurs. This may be attributed to product inhibition by either 2-methylbutyric acid or acetyl-CoA and acetoacetyl-CoA (Fig. 2). This enzyme is also inhibited by acetate, a possible product of acetyl-CoA catabolism, and glutamate, produced by ATases (30), and was formed during log growth. The HMG-CoA synthase reaction is reversible and thereby able to be driven by substrate or energy requirements. The reversibility of the reaction may also explain the observation that HMGA carried labels at alternate positions (Fig. 2). Acetyl-CoA labeled at C-2 may combine with acetoacetyl-CoA labeled at C-3 to form HMG-CoA with alternate labels. Acetoacetyl-CoA may be labeled from two molecules of acetyl-CoA with one having a ¹³C-label at the C-2 position.

Gene deletion for a pathway with multiple gene homologues was not feasible as part of this study without identification of a particular critical gene. None of the genes involved in leucine catabolism were located in an operon. Attempts to identify a regulatory mechanism using known sequence motifs with an intergenic sequence analysis failed. Considering that BCAA catabolism was linked to carbon starvation by pathway comparison and phenotypic measurements, it was also surprising that none of the genes contained upstream binding sites for CcpA, a known carbon flow regulator. The lack of CodY-binding sites was expected, considering that this motif was also not identified in Bacillus subtilis upstream of BCAA catabolic genes (31). This suggests that in silico binding motif analysis alone lacks the power to identify mechanisms of gene regulation. Given that the genes for the main catabolic steps were not organized in a single operon and that over 80 genes were regulated, a detailed strategy to determine the expression regulation is needed but not obvious considering the size of the problem.

The BCAA catabolic pathway shared common motifs with known exergonic pathways (Fig. 3). While FA oxidation shared similarities in biochemical mechanisms, glycolysis and BCAA catabolism shared similar purposes. For example, the BCAA pathway and glycolysis catabolize a six-carbon molecule to produce ATP via substrate-level phosphorylation. While their primary role remains production of ATP, glycolysis also functions as an anabolic pathway by channeling phosphorylated sugar-derived intermediates into products essential for cell structure and function. The BCAA pathway accomplishes the same purpose via keto acids and acetyl-CoA.

The distinctions between BCAA catabolism and glycolysis highlighted features that may be used to control BCAA regulation in firmicutes. The use of coenzyme A as the leaving group in BCAA catabolism, as opposed to phosphate in glycolysis, suggests the existence of a regulatory mechanism linked to coenzyme A metabolism in amino acid catabolism. The inability to use phosphate may also be due to phosphate limitation during starvation. Carbon fixation during BCAA catabolism allows the cell the ability to store carbon in various forms such as FAs, HMGA, and other CoA derivatives during nutrient limitation. The ability to fix carbon suggests that unidentified regulators to control carbon metabolism may be involved in activation of BCAA catabolism in firmicutes. While ATP was produced by BCAA catabolism, the cells remained unable to divide, suggesting that energy production from BCAA catabolism did not facilitate cell division during starvation but allowed transport and metabolic capabilities.

The diverse presence of this pathway across various domains of life suggests that this pathway may support life under carbohydrate deprivation. Organisms undergo nutrient starvation at many points during their life cycle. In many cases, when sugar reserves are depleted, alternate sources of energy are required to maintain life and metabolic activity. This includes the need for carbon sources in various forms for biosynthetic and catabolic activities as well as the messenger molecules for the pathways and gene products involved. BCAA catabolism serves these purposes in a bacterium, using homologous enzymes that were present across various domains of life, suggesting that the switch to amino acid catabolism from sugar catabolism may be a widespread trait in living organisms.

Conclusions.

Lactococcus lactis IL1403 consumed leucine to produce 2-methylbutyric acid, α-ketoisocaproate, α-ketoisovalerate, HMGA, glutamate, and citrate during carbohydrate starvation and NC. Gene expression profiles coupled with intermediate metabolite analysis and in silico analysis resulted in identification of a catabolic pathway previously uncharacterized in firmicutes. Gene lists of possible enzymes were narrowed to a few candidate enzymes for all reactions, except those involving dehydrogenases due to extensive redundancy. This work demonstrated that lactococci do not completely lyse upon nutrient limitation but rather enter an NC state that induces unique metabolic capabilities to produce BCFAs. This pathway produced three ATP molecules per leucine molecule. The cells maintained their ability to transport and metabolize substrates in the NC state.