Increased Activity of Cystathionine β-Lyase Suppresses 2-Aminoacrylate Stress in Salmonella enterica

ABSTRACT

Reactive enamine stress caused by intracellular 2-aminoacrylate accumulation leads to pleiotropic growth defects in a variety of organisms. Members of the well-conserved RidA/YER057c/UK114 protein family prevent enamine stress by enhancing the breakdown of 2-aminoacrylate to pyruvate. In Salmonella enterica, disruption of RidA allows 2-aminoacrylate to accumulate and to inactivate a variety of pyridoxal 5′-phosphate-dependent enzymes by generating covalent bonds with the enzyme and/or cofactor. This study was initiated to identify mechanisms that can overcome 2-aminoacrylate stress in the absence of RidA. Multicopy suppressor analysis revealed that overproduction of the methionine biosynthesis enzyme cystathionine β-lyase (MetC) (EC 4.4.1.8) alleviated the pleiotropic consequences of 2-aminoacrylate stress in a ridA mutant strain. Degradation of cystathionine by MetC was not required for suppression of ridA phenotypes. The data support a model in which MetC acts on a noncystathionine substrate to generate a metabolite that reduces 2-aminoacrylate levels, representing a nonenzymatic mechanism of 2-aminoacrylate depletion.

IMPORTANCE RidA proteins are broadly conserved and have been demonstrated to deaminate 2-aminoacrylate and other enamines. 2-Aminoacrylate is generated as an obligatory intermediate in several pyridoxal 5′-phosphate-dependent reactions; if it accumulates, it damages cellular enzymes. This study identified a novel mechanism to eliminate 2-aminoacrylate stress that required the overproduction, but not the canonical activity, of cystathionine β-lyase. The data suggest that a metabolite-metabolite interaction is responsible for quenching 2-aminoacrylate, and they emphasize the need for emerging technologies to probe metabolism in vivo.

INTRODUCTION

Metabolism is guided by the chemical reactivity of molecules in cells. Inherent in this reactivity is the potential for deleterious reactions between molecules in a heterogeneous and crowded intracellular milieu (1). Examples of reactive metabolites spontaneously damaging macromolecules (i.e., DNA or proteins) in cells abound, with well-characterized damage repair systems in place to counteract such damage (2, 3). Examples of metabolite damage resulting from promiscuous enzyme activities or spontaneous reactions between unstable molecules are increasing (4). In many cases, dedicated repair systems for metabolite damage are necessary to prevent reactive metabolites from reaching dangerous levels in cells.

The reactive enamine 2-aminoacrylate (2AA) is produced by pyridoxal 5′-phosphate (PLP)-dependent enzymes associated with amino acid metabolism (5, 6). Due to the short half-life of 2AA in vitro (7), a role for this reactive metabolite in vivo was previously overlooked. However, detailed biochemical and genetic studies demonstrated that 2AA could persist intracellularly and inactivate a variety of target PLP-dependent enzymes by covalently modifying the cofactor in the active site (8–10). The modifications decrease enzyme activity and lead to numerous metabolic deficiencies (11–14). Pioneering studies in Salmonella enterica defined the role of RidA/YER057c/UK114 family proteins in preserving metabolic network integrity by hydrolyzing 2AA and preventing its accumulation inside cells (8, 15) (Fig. 1A). Despite their shared metabolic components, S. enterica and the closely related Gram-negative bacterium Escherichia coli differed in their susceptibility to enamine accumulation and 2AA stress (14, 16). Strains of S. enterica lacking RidA were unable to grow on minimal glucose medium containing serine or on minimal pyruvate medium, due to 2AA stress generated by the biosynthetic serine/threonine dehydratase IlvA (EC 4.3.1.19) (17). In contrast, E. coli strains lacking the three RidA homologs present in the genome experienced a growth defect only when IlvA activity was artificially increased (14). Those studies showed that our understanding of the stress caused by 2AA and its control were incomplete.

FIG 1.

RidA paradigm and MetC reaction scheme. (A) In Salmonella enterica, 2AA is generated as an intermediate by the pyridoxal 5′-phosphate-dependent serine/threonine dehydratase IlvA acting on the alternative substrate serine. Following release from IlvA, 2AA can be hydrated to pyruvate by solvent water or by RidA. In the absence of RidA, some 2AA persists long enough to damage target enzymes and to cause metabolic stress. (B) Cystathionine β-lyase, encoded by metC, is a pyridoxal 5′-phosphate-dependent enzyme that converts cystathionine to homocysteine, pyruvate, and ammonia in the penultimate step of methionine biosynthesis. MetC additionally acts on a diverse set of substrates (24).

The present study was carried out to identify non-RidA mechanisms to diminish 2AA stress in S. enterica. The data showed that overexpression of metC in strains lacking RidA allowed wild-type growth under 2AA stress conditions. MetC is a PLP-dependent cystathionine β-lyase enzyme that catalyzes the penultimate step in methionine biosynthesis (Fig. 1B). The cystathionase activity of MetC was not required for suppression of ridA mutant phenotypes, and the MetC protein did not act on 2AA enzymatically. The data support a model in which MetC generates a metabolite that is able to react with 2AA.

RESULTS AND DISCUSSION

Expression of metC in trans reverses growth defects of a ridA mutant strain.

Strains of S. enterica lacking RidA are unable to grow on medium with exogenous serine or cysteine, due to the detrimental effects of the accumulated 2AA (8, 12). A plasmid library containing Salmonella enterica LT2 genomic DNA was transduced into a ridA mutant strain (DM3480). Transductants were screened for growth on minimal glucose medium containing serine (5 mM). Plasmids allowing growth were moved into a naive parental strain to confirm the suppression phenotype, and a representative plasmid (pSR43) was chosen for additional characterization. The presence of pSR43 restored growth to a ridA mutant strain in minimal glucose medium containing serine (Fig. 2A). Sequence analysis showed that pSR43 contained a fragment of chromosomal DNA consisting of seven open reading frames (ORFs), including four genes of unknown function, two genes encoding components of a TonB-dependent transport system, and the metC gene, encoding a PLP-dependent cystathionine β-lyase that is required for methionine biosynthesis (Fig. 2B). The association of ridA mutant phenotypes with PLP-dependent enzymes suggested that metC could be involved in the suppression of serine sensitivity. Plasmid pSR43 was engineered to contain two sequential stop codons in metC, generating plasmid pMC10. pMC10 (DM9407) failed to restore the growth of a ridA mutant in the presence of serine (Table 1), showing that the MetC protein was required for suppression by pSR43. The metC ORF alone was individually cloned into a pSU19 vector and a pBAD vector (yielding pSW1 and pDM1480, respectively). Unexpectedly, neither of these constructs restored growth to a ridA mutant on medium with serine or medium with pyruvate as the sole carbon source (Table 1).

FIG 2.

MetC in trans suppression of growth defects of ridA mutants. (A) Growth of wild-type (Wt) and ridA mutant strains containing the empty pBR328 vector (vector-only control [VOC]) or pSR43 in minimal glucose plus serine (5 mM) medium is shown. Data reflect the averages and standard deviations of experiments performed in triplicate in a 96-well plate format. (B) The genomic organization around metC is depicted. The inserts in the relevant plasmids are indicated. Site-directed mutagenesis generated a metC-null version of pSR43, pMC10. The metC ORF was cloned into the multiple cloning site of pSU19 or pBAD24 to generate pSW1 or pDM1480, respectively.

TABLE 1.

Alleviation of ridA strain growth defects by construct-dependent expression of metC

Strain	Genotype	Growth yield (OD650) at 24 ha
		Glucoseb	Glucose plus serineb	Pyruvatec
DM7576	ridA pSW1	1.14	0.07	0.56
DM7575	ridA pSR43	1.15	1.29	1.07
DM9407	ridA pMC10	1.07	0.14	0.37
DM15539d	ridA pDM1480	1.37	0.11	NT

^aAverages from three independent cultures, with errors of less than 0.2 absorbance unit, are shown.

^bCells were grown with glucose (11 mM) as the sole carbon source, with serine (5 mM) added as indicated.

^cPyruvate (50 mM) was used as the sole carbon source. NT, not tested.

^dArabinose (0.1%) was added to all cultures for this strain.

One interpretation of the results described above was that other genes on the pSR43 construct were required for suppression of ridA phenotypes by MetC. ExbB and ExbD form a complex with TonB to transduce energy to the outer membrane, for use by TonB-dependent transporters (18). The possibility that TonB was required for pSR43 suppression of ridA phenotypes was examined. After 24 h, a ridA tonB double mutant grew in the presence of 5 mM serine when pSR43 was provided (optical density at 650 nm [OD₆₅₀] of 1.12 ± 0.035) but not when a plasmid with no insert was present (OD₆₅₀ of 0.064 ± 0.063). These data indicated that the ExbB-ExbD-TonB complex was not required for suppression of a ridA mutant. The remaining genes of unknown function flanking metC (yghA, yhjG, yghB, and yqhC) were similarly dispensable for pSR43-mediated suppression of ridA phenotypes (data not shown).

In the absence of evidence for a role of additional proteins in suppression by pSR43, strains were assayed for cystathionase activity. Strains carrying the pSR43 construct had >10-fold higher cystathionase activity than did strains with plasmids carrying only metC (Table 2). These data provided an explanation for the different behavior of the metC-containing plasmids, since cystathionase activity positively correlated with suppression of ridA strain growth defects. The significantly higher levels of MetC activity in strains with pSR43 suggested a context-dependent effect on metC expression, which was not pursued here.

TABLE 2.

Correlation of cystathionine β-lyase activity with growth of ridA mutants

Strain	Genotype	β-Lyase activity (ΔA412/mg/min)a
DM7576	ridA pSW1	0.2 ± 0.3
DM7575	ridA pSR43	11.6 ± 1.9
DM9407	ridA pMC10	0.3 ± 0.05
DM15539b	ridA pDM1480	1.3 ± 0.8

^aCells were grown in nutrient broth, crude extracts were generated, and cystathionine β-lyase (MetC) activity was assayed as described in the text. The data are reported as the averages and standard deviations of three independent experiments.

^bArabinose (0.1%) was added to cultures containing DM15539.

Cystathionine degradation by MetC is not required to suppress growth defects of a ridA strain.

The positive correlation between cystathionase activity and suppression of ridA strain phenotypes indicated that MetC was required, but it did not implicate this activity per se in the phenotype. Methionine limits flux through the methionine biosynthesis pathway through feedback inhibition of MetA (19), thus diminishing the synthesis of the MetC substrate cystathionine. Adding methionine to cultures of a ridA strain containing pSR43 that was grown on minimal pyruvate medium or minimal glucose plus serine medium did not affect MetC suppression of ridA strain growth defects (data not shown). Furthermore, disruption of the metB gene, encoding cystathionine synthase, did not prevent pSR43 from suppressing the growth defect of a ridA mutant strain grown in minimal glucose medium containing serine and methionine (data not shown). Thus, cystathionine was not required for MetC-dependent suppression of ridA strain phenotypes. Furthermore, one of the products of MetC-catalyzed cystathionine α,β elimination, homocysteine, was found to inhibit a ridA strain, implying that increased degradation of cystathionine would be harmful rather than helpful (Fig. 3). This inhibition by homocysteine was attributed to increased accumulation of 2AA, since (i) homocysteine increases IlvA activity in E. coli (20) and (ii) growth inhibition was reversed by isoleucine or glycine, conditions that prevent or bypass, respectively, damage from 2AA accumulation (21). In total, the results described above indicated that the MetC protein, when highly expressed, but not its canonical cystathionine β-lyase activity suppressed the growth defects of a ridA strain.

FIG 3.

Toxicity of the homocysteine product of cystathionine β elimination to ridA mutants. Wild-type (Wt) and ridA strains were grown in minimal glucose medium containing 0.25 mM homocysteine. No supplement, isoleucine (1 mM), or glycine (0.67 mM) was added to the growth medium containing homocysteine. Addition of isoleucine or glycine had no effect on the growth of the wild-type strain. The data represent the averages and standard deviations of three independent cultures.

MetC overproduction reduces 2AA accumulation.

Growth defects associated with a ridA mutation are caused by the accumulation of 2AA, which is generated primarily by the serine/threonine dehydratase IlvA (8, 21). 2AA inhibits several enzymes, most critically the serine hydroxymethyltransferase GlyA (21). Suppression of the growth defect can be accomplished by (i) decreasing the accumulated 2AA (8) or (ii) compensating for a 2AA-damaged target by rerouting metabolic flux (21). These two possibilities can be distinguished by assessing the activity of transaminase B (IlvE) (EC 2.6.1.42), a PLP-dependent enzyme that is inactivated by free 2AA (8). Thus, IlvE activity serves as a proxy for 2AA accumulation and is reduced in ridA strains due to 2AA accumulation (8, 22). IlvE activity was significantly reduced in a ridA strain carrying pMC10 (55 nmol 2-keto-3-methylvalerate [2KMV]/min/mg), relative to the wild-type control (120 nmol 2KMV/min/mg) (Fig. 4). In contrast, the IlvE activity of a ridA strain carrying pSR43 (90 nmol 2KMV/min/mg) was not significantly different from that of the wild-type control (122 nmol 2KMV/min/mg) (Fig. 4). These data indicated that the mechanism of MetC suppression reduced 2AA levels.

FIG 4.

Restoration of IlvE activity by MetC in trans in ridA mutants. Transaminase B (IlvE) activity was measured in crude extracts of wild-type (Wt) or ridA strains containing pSR43 or pMC10, grown to late log phase in minimal glucose medium. IlvE activity was normalized to the crude protein concentration. Data were analyzed by one-way ANOVA with Tukey's multiple-comparison test (α = 0.01), and the averages and standard deviations of three biological replicates are shown. Asterisks denote statistically significant (P < 0.01) variation between samples.

The serine sensitivity of a ridA mutant is suppressed by isoleucine (17) due to its allosteric inhibition of IlvA, preventing 2AA generation (23). It was formally possible that MetC overproduction diminished 2AA levels through allosteric inhibition of IlvA. An IlvA variant (encoded by ilvA219) that is insensitive to allosteric regulation (22) was introduced into a ridA strain carrying pSR43. The presence of pSR43 maintained growth of the ridA mutant in the presence of serine (or with pyruvate) even when IlvA activity could not be allosterically regulated (Table 3). Taken together, these data confirmed that MetC reduced 2AA accumulation in vivo and did not depend on regulating IlvA for this effect. In total, these results supported a model in which MetC acted downstream of IlvA to sequester and/or to degrade 2AA and thus prevent the inactivation of target enzymes.

TABLE 3.

Absence of requirement for allosteric IlvA inhibition for MetC-mediated suppression

Strain	Genotype	Growth yield (OD650) after 12 ha
		Glucoseb	Glucose plus serineb	Pyruvatec
DM3480	ridA	1.26	0.16	0.32
DM5970	ridA ilvA219	0.51	0.1	0.13
DM9384	ridA ilvA219 pSR43	1.29	1.16	1.24
DM9385	ridA ilvA219 pMC10	0.36	0.07	0.12

^aAverages from three independent cultures, with errors of less than 0.2 absorbance unit, are shown.

^bCells were grown with glucose (11 mM) as the sole carbon source, with serine (5 mM) added as indicated.

^cPyruvate (50 mM) was used as the sole carbon source.

It was formally possible that increased MetC competed with IlvA for the PLP cofactor, resulting in decreased PLP-dependent activity of IlvA and less generation of 2AA. Data failed to support this model, as the (i) overexpression of other PLP-dependent enzymes failed to suppress the ridA phenotypes, (ii) overexpression of MetC did not decrease the activity of transaminase B (Fig. 4), and (iii) addition of exogenous pyridoxal failed to prevent the suppression by MetC (data not shown). In the absence of support for a model invoking competition for PLP, we pursued a model in which activity of MetC led to the quenching of 2AA.

MetC does not sequester or quench 2AA in vitro.

Two scenarios were considered to explain how MetC overproduction decreased 2AA accumulation in vivo. First, MetC could act directly on 2AA, by a mechanism similar to that of RidA. Second, MetC could generate a metabolic product that reacts with the unstable 2AA to generate a stable adduct. In vitro efforts to distinguish these scenarios were pursued. IlvA acts on serine to generate 2AA, which is hydrated to pyruvate by water. RidA activity is detected as an increased rate of pyruvate formation, when RidA deaminates 2AA, in these assays (7). In contrast, adding MetC to a reaction mixture containing IlvA and serine had no effect on the rate of pyruvate formation (data not shown). These data indicated that MetC did not hydrolyze 2AA and did not bind and sequester this metabolite, making the first scenario unlikely.

Working model for metabolic quenching of 2AA.

Considering the requirement for the MetC protein for suppression and the absence of evidence for MetC interacting with 2AA, a metabolite quenching model was developed (Fig. 5). In this scenario, a reactive metabolite generated by MetC would react with, and thus sequester, 2AA to minimize the metabolic stress in the absence of RidA. MetC is reported to have a broad substrate range (24) and is capable of generating reactive persulfide species (25). We considered a scenario in which MetC generates a product that can result in persulfide-mediated 2AA sequestration. Efforts to query a specific model with cystine as a source of MetC-generated cysteine persulfide were complicated by the inability to provide the relevant reactive intermediates directly. Instead, 2AA and cysteine persulfide were generated in situ, in the reaction mixture, by IlvA and MetC, respectively. No unique product that would indicate a reaction of 2AA and cysteine persulfide was detected (data not shown). Similarly, strategies to generate cysteine persulfide from cystine with sodium sulfide inhibited IlvA activity, which was needed to produce 2AA. There are a number of potential explanations for the failure of the in vitro efforts to identify the proposed interactions. First, MetC has a broad substrate range, and cystine might not be the relevant metabolite. Second, the intracellular conditions might not be adequately replicated by an aqueous solution.

FIG 5.

Working model of 2AA sequestration by MetC-generated persulfides. 2AA is generated by serine/threonine dehydratases (i.e., IlvA). MetC can catalyze a variety of elimination reactions using sulfur-containing amino acid substrates. In the scenario depicted, the elimination of pyruvate from cystine yields cysteine persulfide (thiocysteine), which can react with and sequester 2AA by one of two mechanisms. The reaction generates a stable adduct, effectively removing 2AA stress in the cell in the absence of RidA.

Conclusions.

The results presented here indicate that MetC overproduction suppresses the multiple growth defects displayed by a ridA mutant. The data indicate that suppression by MetC does not depend on cystathionine β-lyase activity or IlvA regulation and MetC does not bind 2AA directly. Despite the negative results with in vitro efforts, we think that the model depicted in Fig. 5 remains the most plausible explanation of our results. A number of chemical precedents support the thinking that led to this model. Cystathionine β-lyase enzymes are able to decompose a variety of sulfur-containing amino acids of potential biological significance, including cystine, djenkolic acid, lanthionine, and others (24). There is precedence for cysteine being able to react with 2AA generated by cysteine desulfhydrase (CdsH) (EC 4.4.1.1) in semipure enzyme assays, leading to the accumulation of a cyclized end product (26). These data confirmed that the nucleophilic cysteine thiol could attack 2AA before spontaneous tautomerization and hydrolysis could take place. Cysteine persulfide (thiocysteine) is produced by a variety of promiscuous cystine lyases, including cystathionine lyases, and affects the cellular thiol landscape in many organisms (27). In a general model, thiocysteine could act either as a strong nucleophile capable of attacking 2AA or as an electrophile susceptible to attack by 2AA, depending on the sulfur atom considered (Fig. 5). A variety of persulfide species beyond thiocysteine, derived from precursors other than cystine, play important roles in sulfur trafficking, protein persulfidation, and other cellular processes (28). These considerations emphasize that an unbiased in vivo approach is needed. A viable approach would not be limited to exploring the impact of a single MetC substrate (e.g., cystine) on 2AA levels but would detect unknown adducts derived from serine when MetC is expressed in S. enterica. In vitro experiments may fail to reconstitute conditions of the intracellular space, including molecular crowding and a lack of abundant free water, conditions that may be required for the proposed interaction between 2AA and a MetC-generated metabolite. The need for new approaches to in vivo metabolic questions was reinforced by a parallel study to identify enzymes that control 2AA in the RidA-deficient archaeon Methanococcus maripaludis. An aspartate/glutamate racemase (MMP0739) (EC 5.1.1.13) that suppressed S. enterica ridA phenotypes was found. Similar to the conclusions reached herein, the data implicated the enzyme (i.e., MMP0739), but not the known substrates, in the suppression effect (29). Creative application of in vivo NMR metabolomic approaches, or other technological advances, could provide insights into the metabolic reactions occurring in cells when reactive metabolites accumulate.

MATERIALS AND METHODS

Bacterial strains, media, and chemicals.

Strains used in this study were derived from Salmonella enterica serovar Typhimurium LT2 and are listed, with their genotypes, in Table 4. Tn10d(Tc) is the transposition-defective mini-Tn10 (Tn10Δ16Δ17) described by Way et al. (30). MudJ refers to the transposition-defective Mu element Mud1734 described previously (31).

TABLE 4.

Strains, plasmids, and primers

Designation	Description
Strains
DM3480	ridA3::MudJ
DM5970	ivlA219 ridA3::MudJ
DM7575	ridA3::MudJ pSR43
DM7576	ridA3::MudJ pSW1
DM9384	ilvA219 ridA3::MudJ pSR43
DM9385	ilvA219 ridA3::MudJ pMC10
DM9386	metC::Tn10d(Tc) pSR43
DM9387	metC::Tn10d(Tc) pMC10
DM9404	Wild-type (isogenic to DM3480)
DM9405	DM9404 pSR43
DM9407	ridA3::MudJ pMC10
DM9567	ridA3::MudJ metB::Tn10d(Tc)
DM13225	ridA3::MudJ tonB256::Tn10d(Tc) pSU19
DM13226	ridA3::MudJ tonB256::Tn10d(Tc) pSR43
DM14277	ridA3::MudJ pBAD24
DM14370	DM9404 pBAD24
DM15535	DM9404 pSU19
DM15536	DM9404 pDM1480
DM15538	ridA3::MudJ pSU19
DM15539	ridA3::MudJ pDM1480
DM16218	DM9404 pBR328
DM16219	ridA3::MudJ pBR328
Plasmids
pBR328	40
pBAD24	41
pSU19	37
pSW1	metC cloned into pSU19 multiple cloning site
pSR43	metC and flanking genomic DNA cloned into pBR328 BamHI cloning site
pMC10	pSR43 with two stop codons inserted in metC ORF
pDM1480	metC cloned into pBAD24 multiple cloning site
Primers
MetC-SDMfor	CAGGAAACGCAACATGACGGATTAATAGTTGGATACCAAACTGG
MetC-SMDrev	GCGTTTACCAGTTTGGTATCCAACTATTAATCCGTCTATGTTGCG
MetC_For_NdeI	CATATGACGGATAAACAGTTGGATACCAAACTGGTAAACGC
MetC_Rev_XhoI	CCCCGGCACTCGAGCACAATTC
MetC_For_KpnI	GAGAGGTACCAACGGATAAACAGTTGGATAC
MetC_Rev_HindIII	GAGATCTAGATTACACAATTCTGGCGAAGC

Rich growth medium consisted of Difco nutrient broth (NB) (8 g/liter) and sodium chloride (5 g/liter). Minimal medium was composed of no-carbon E salts (NCE) with 1 mM magnesium sulfate (32), trace elements (33), and either glucose (11 mM) or pyruvate (50 mM) as the sole carbon source. Difco BiTek agar (15 g/liter) was added to make solid growth medium. Amino acids, i.e., serine (5 mM), methionine (0.3 mM), homocysteine (0.25 mM), isoleucine (1 mM), and glycine (0.67 mM), were added to minimal medium as indicated. Antibiotics were added to rich and minimal medium at the following respective final concentrations: chloramphenicol, 20 and 5 mg/liter; kanamycin, 50 and 12.5 mg/liter; tetracycline, 20 and 10 mg/liter. Amino acids, including l-cystine dihydrochloride, and antibiotics were purchased from Sigma-Aldrich (St. Louis, MO).

Genetic techniques and growth analyses.

Strains were constructed by transductional crosses using the high-frequency general transducing mutant of bacteriophage P22 (HT105/1, int-201) (34). Techniques used to perform transductions, to eliminate phage contamination from cells, and to identify phage-free recombinants were described previously (35). Strains were freshly streaked on solid NB medium prior to determination of growth curves. Individual colonies were inoculated into 2 ml of NB broth and grown overnight at 37°C, at 200 rpm. Cell growth was measured in liquid medium, as described previously (36); 5 μl of an NB culture grown overnight was used to inoculate 195 μl of growth medium contained within each well of a 96-well microtiter plate (Corning). Microtiter plates were incubated at 37°C with shaking, and growth was monitored as the change in OD₆₅₀, using a BioTek Elx808 plate reader. All growth experiments were performed in triplicate, and the resulting data were plotted using GraphPad Prism 7.0, generating curves in log₁₀ format that displayed the averages and standard deviations of the replicates. When final ODs were reported, growth analyses were performed by inoculating 100 μl of NB overnight into 5-ml cultures contained within 30-ml culture tubes and incubated at 37°C and 200 rpm.

Molecular methods and library construction.

A plasmid library carrying chromosomal DNA from Salmonella enterica partially digested with Sau3A and ligated into the BamHI site of pBR328 was provided by C. Miller. Plasmid pSR43 from this library conferred the phenotypes described to a S. enterica ridA mutant strain. Two sequential stop codons were engineered into the beginning of the metC sequence by site-directed mutagenesis using primers MetC-SDMfor and MetC-SMDrev. The DNA changes A10T and C13A, corresponding to amino acids K4 (ochre) and Q5 (amber), respectively, were introduced into the plasmid pSR43 using a standard QuikChange protocol (Stratagene), to generate the plasmid pMC10. Plasmid pDM1480 was generated by cloning the metC gene into pBAD24. The metC ORF was amplified using primers MetC_For_KpnI and MetC_Rev_HindIII, digested with KpnI and HindIII, and ligated into pBAD24 digested with the same restriction enzymes. Plasmid pSW1 contained the metC coding sequence in pSU19, where it was constitutively expressed from the lac promoter (37).

Crude cystathionase assays.

MetC activity was assessed as the cystathionine-dependent formation of free thiols in cell extracts, according to a published protocol (38). Briefly, 25-ml cultures were grown to full density in NB with 20 μg/ml chloramphenicol at 37°C, with shaking. Cells were pelleted by centrifugation (7,000 × g) and resuspended in 1 ml of 0.85% NaCl. Cells were lysed by sonication at 4°C, and cell debris was removed by centrifugation (17,000 × g) for 10 min at 4°C. The cell extract was moved to a new tube, brought to a volume of 2.5 ml with 0.85% NaCl, and loaded onto a Sephadex G-25M PD-10 desalting column (Pharmacia). Proteins were eluted with 3.5 ml of 0.85% saline, and the protein concentration was determined with the Bradford assay (39). Approximately 30 μl of extract (corresponding to ∼13 μg protein) was assayed at 30°C in a total reaction volume of 200 μl containing 130 mM Tris (pH 7.5), 0.02 mM PLP, and 1 mM 5,5′-dithiobis-(2-nitrobenzoic acid) (Sigma). A volume of 8 μl of 50 mM cystathionine, to yield a final concentration of 2 mM, was added to start the reaction, and absorbance at 412 nm was monitored over time. Experiments were performed in triplicate, and results are reported as averages and standard deviations.

Transaminase B (IlvE) assays.

IlvE activity was measured in permeabilized cell extracts as described previously (22). Strains were grown to late log phase in 5 ml of minimal glucose medium, pelleted, washed with 50 mM KPO₄ (pH 8), and frozen overnight at −20°C. Frozen cell pellets were thawed on ice, resuspended in 50 mM KPO₄ (pH 8), and lysed using PopCulture (Novagen). Aliquots of PLP (50 μM) and 2-ketoglutarate (10 mM) were mixed with ∼30 μl of the permeabilized cell suspension, and isoleucine was added (20 mM) to initiate the reaction. Reaction mixtures were incubated for 20 min at 37°C, and IlvE activity was determined based on the amount of 2KMV formed. 2KMV was derivatized with 2,4-dinitrophenylhydrazine (DNPH) prior to hydrazone formation, followed by organic extraction. The organic layer was treated with 0.5 N HCl and then removed and mixed with 1.5 N NaOH to allow chromophore formation. Absorbance of the resulting aqueous layer (containing the chromophore) was measured at 540 nm using a SpectraMax M2 microplate reader (Molecular Devices). The protein concentration of each lysate was determined using the bicinchoninic acid (BCA) assay (Pierce). Activity is reported as nanomoles of 2KMV per minute per milligram of protein per cell lysate. The resulting data were analyzed by one-way analysis of variance (ANOVA), using GraphPad Prism 7.0, and Tukey's test was used to assess significant changes in IlvE activity (P = 0.05).

MetC purification.

metC was amplified from S. enterica LT2 using primers MetC_For_NdeI, containing an engineered 5′ NdeI restriction site, and MetC_Rev_XhoI, containing an engineered XhoI site in place of the stop codon. Amplification was performed by PCR using cloned Pfu DNA polymerase. PCR conditions were as follows: denaturation at 95°C for 1 min, annealing at 52°C for 1 min, and extension at 72°C for 1 min. The resulting 1.2-kb fragment was digested with NdeI and XhoI and ligated into the pET20b vector (Novagen), which had been digested with NdeI and XhoI. The plasmid (pMC12) complemented metC mutant TT14 (data not shown). MetC was overexpressed from pMC12 in E. coli BL21(AI), according to the manufacturer's protocol (Novagen). Cells from the resulting cultures were broken at 15,000 lb/in² in a French pressure cell, at 4°C. Cell debris was removed by centrifugation at 42,000 × g for 30 min at 4°C. Proteins were purified using a column containing Superflow Ni²⁺ resin (Qiagen), according to the standard protocol (Novagen). Fractions containing MetC were concentrated at 30 lb/in² under argon gas, using a 10,000-molecular-weight-cutoff membrane (Amicon YM10). The protein was dialyzed in binding buffer and stored at −80°C.

Serine dehydratase (IlvA) assays.

IlvA was provided by K. Hodge-Hanson and was purified as described previously (7). The RidA preparation was the same one as used in previously described assays (12). Reaction mixtures (300 μl) consisted of 50 mM Tris-HCl (pH 9) and 0.6 μM IlvA. Reactions were initiated by the addition of l-serine (60 mM) and were monitored continuously at 230 nm for 120 s in a 96-well quartz plate, using a SpectraMax M2 microplate reader. Initial rates were determined based on the increase in A₂₃₀, corresponding to pyruvate formation. The impact of RidA (0.3 μM) or MetC (0.45 μM) on the rate of pyruvate formation was assessed by adding the proteins to IlvA reaction mixtures at the indicated concentrations and monitoring changes in A₂₃₀.