Pyridoxal and α-Ketoglutarate Independently Improve Function of Saccharomyces cerevisiae Thi5 in the Metabolic Network of Salmonella enterica

ABSTRACT

Microbial metabolism is often considered modular, but metabolic engineering studies have shown that transferring pathways, or modules, between organisms is not always straightforward. The Thi5-dependent pathway(s) for synthesis of the pyrimidine moiety of thiamine from Saccharomyces cerevisiae and Legionella pneumophila functioned differently when incorporated into the metabolic network of Salmonella enterica. Function of Thi5 from Saccharomyces cerevisiae (ScThi5) required modification of the underlying metabolic network, while LpThi5 functioned with the native network. Here we probe the metabolic requirements for heterologous function of ScThi5 and report strong genetic and physiological evidence for a connection between alpha-ketoglutarate (αKG) levels and ScThi5 function. The connection was built with two classes of genetic suppressors linked to metabolic flux or metabolite pool changes. Further, direct modulation of nitrogen assimilation through nutritional or genetic modification implicated αKG levels in Thi5 function. Exogenous pyridoxal similarly improved ScThi5 function in S. enterica. Finally, directly increasing αKG and PLP with supplementation improved function of both ScThi5 and relevant variants of Thi5 from Legionella pneumophila (LpThi5). The data herein suggest structural differences between ScThi5 and LpThi5 impact their level of function in vivo and implicate αKG in supporting function of the Thi5 pathway when placed in the heterologous metabolic network of S. enterica.

IMPORTANCE Thiamine biosynthesis is a model metabolic node that has been used to extend our understanding of metabolic network structure and individual enzyme function. The requirements for in vivo function of the Thi5-dependent pathway found in Legionella and yeast are poorly characterized. Here we suggest that αKG modulates function of the Thi5 pathway in S. enterica and provide evidence that structural variation between ScThi5 and LpThi5 contributes to their functional differences in a Salmonella enterica host.

INTRODUCTION

The robust dynamic metabolism of microbes is characterized by metabolic modules and regulatory paradigms that are integrated to create a functioning complex system. Parameters that allow metabolic modules to function, or fail to function, across different organisms are generally poorly understood. Understanding these parameters is hampered by the fact that modules evolve in a specific metabolic network and are therefore optimized to function within the framework of the native organism. When moved by either horizontal gene transfer or metabolic engineering, incompatibilities can occur and prevent optional function (1, 2). These incompatibilities can be generated by depletion of metabolites, accumulation of toxic metabolites, inhibition of enzymes or pathways, lack of enzyme substrates, etc. It follows that some of these incongruities can be solved by altering either the metabolic network of the host or the module itself to restore a functional system (3). Identifying changes needed to generate a functioning system can result in insights about metabolic structure and integration and can be facilitated by focusing on a defined node of metabolism.

Thiamine biosynthesis is a well-defined metabolic network that has been used to investigate the plasticity of metabolism in Salmonella enterica (4, 5). Thiamine pyrophosphate (TPP), the biologically active form of Vitamin B1, is an essential cofactor for nearly all organisms (6). TPP is required throughout central carbon metabolism and specialized metabolite biosynthesis primarily to facilitate decarboxylation and ligase reactions (7). TPP is produced by bacteria, archaea, yeast and plants, where the two moieties 4-methyl-5-(2-hydroxyethyl)-thiazole-phosphate (THZ-P) and 4-amino-5-hydroxymethyl-2-methylpyrimidine-pyrophosphate (HMP-PP) are independently synthesized and combined to form thiamine monophosphate (TMP) (8). In a final step, TMP is phosphorylated to generate TPP (8). There have been two pathways described for the synthesis of the HMP-P moiety (Fig. 1). The majority of plants, archaea and bacteria use the SAM radical ThiC protein to convert a purine biosynthetic intermediate, 5-aminoimidazole ribotide, into HMP-P, formate and carbon monoxide (9–11). A second pathway was first described in yeast and subsequently found in the Legionellacae clade of bacteria (12–15). Labeling studies in Saccharomyces cerevisiae showed that the atoms in the pyrimidine moiety of thiamine originated from histidine and a B₆ vitamer, presumed to be pyridoxal phosphate (PLP) (Fig. 1) (12–14). Genetic analyses in S. cerevisiae determined that genes of the THI5 family (THI5/11/12/13) supported HMP synthesis, with a single isozyme being sufficient for growth (16).

FIG 1.

Thi5 and ThiC mechanisms of HMP synthesis.The two pathways to HMP-P are schematically represented. In the first, AIR, an intermediate in purine biosynthesis is converted to HMP-P by ThiC (exemplified by Salmonella enterica). In a Thi5 pathway (exemplified by Saccaromyces cerevisiae and Legionella pneumophila) HMP-P is generated from histidine and a B₆ Vitamer. In the B₆ vitamer, an “R” group represents either an aldehyde (pyridoxal-5′-phosphate), amine (pyridoxamine-5′-phosphate), or alcohol group (pyridoxine-5′-phosphate). Abbreviations: AIR, 5′-aminoimidazole ribotide; HMP-P, 4-amino-5-hydroxymethylpyrimidine phosphate; THZ-P, 4-methyl-5-β-hydroxyethylthiazole phosphate; TPP, thiamine pyrophosphate.

The response of a metabolic network to a module exchange was probed in a study in which ScTHI5 was introduced into Salmonella enterica as a substitution for the native ThiC. The study showed that ScThi5 alone was not broadly able to substitute for ThiC in the synthesis of HMP-P. Rather, ScTHI5 expressed in trans was sufficient to complement the thiC strain of S. enterica only under a few limited growth conditions (17). Notably, ScThi5 was unable to support thiamine synthesis when glucose was the sole carbon source, a phenotype unaltered by the addition of cAMP, histidine or pyridoxal. These data indicated that an undefined metabolic change was required to integrate ScThi5 into the metabolic network of S. enterica. Supporting this notion, alleles of sgrR and ptsI were identified that restored ScThi5 function on glucose (17). Genetic analysis supported the conclusion that these suppressor alleles acted by decreasing flux through glycolysis, which increased ScThi5 function in vivo by an undetermined mechanism. Unexpectedly, Thi5 from Legionella pneumophila (LpThi5) restored growth to a S. enterica thiC mutant in multiple growth conditions, including when glucose was sole carbon source (15). Variants of LpThi5 were identified that generally phenocopied the behavior of ScThi5, leading to the hypothesis that differences between the orthologs may contribute to in vivo function in heterologous networks (15).

The broad understanding of thiamine metabolism and ease of genetic manipulation, make S. enterica an attractive model to probe the metabolic consequences of incorporating a heterologous module (Thi5) into a metabolic network. This study was initiated to further explore remodeling of the metabolic network that is required for ScThi5 to function in S. enterica. Data herein identified two metabolites, pyridoxal and α-ketoglutarate (αKG), that stimulate the function of ScThi5 in vivo in S. enterica. Results presented are consistent with a model in which glucose-specific suppressors act by decreasing flux through glycolysis resulting in increased alpha-ketoglutarate (αKG) via the TCA cycle and increased flux through the pentose-phosphate pathway leading to increased PLP synthesis. Additional growth conditions and mutations that increase αKG and/or PLP by other mechanisms also allow ScThi5 function, further supporting a significant role for these metabolites in generating a metabolic network permissive for ScThi5.

RESULTS

Demonstration that a Thi5 of bacterial origin (L. pneumophila) functioned to complement a S. enterica thiC mutant, while the metabolic network had to be modified for ScThi5 to function raised questions about i) differing metabolic networks in bacteria vs yeast, and ii) features of the Thi5 orthologs responsible for their differential function in vivo. Two chromosomal mutations that were previously isolated to allow ScThi5-dependent growth of a thiC mutant with glucose as a carbon source were characterized to address the former (17).

A variant of PtsI allows ScThi5 to function in S. enterica.

A mutant allele of ptsI encoding PtsI^R361H (ptsI611) supported ScThi5 dependent growth on glucose if expression of ScTHI5 was induced (Fig. 2). PtsI, or Enzyme I (EI) of the phosphotransferase system (E.C. 2.7.3.9), anchors a regulatory node in metabolism that coordinates flux between glycolysis and the TCA cycle in response to cellular levels of alpha-ketoglutarate (αKG), phosphoenolpyruvate (PEP), and pyruvate (18). The substitution in PtsI^R361H is located outside the active site and in a region of the enzyme that has a large ¹H_N/¹⁵N chemical shift when PEP or αKG binds (19). These data suggested the variant could have altered PEP turnover and/or regulatory behavior in response to αKG. Either of these effects have the potential to alter the metabolic network leading to an increase in ScThi5 function.

FIG 2.

ptsI611 allows ScThi5-dependent growth on glucose medium. An isogenic pair of thiC pTac85-ScTHI5 (pDM1625) strains that carry either wildtype ptsI (A) or the ptsI611 allele (B) in the chromosome were grown on NCE minimal medium with glucose as a sole carbon source (closed circles). Thiamine (100 nM; open circles) or IPTG (100 μM; open squares) was added to the medium. Growth was monitored by the increase in optical density at 650 nm with shaking at 37°C. Error bars indicate the standard deviation of three independent biological replicates.

His₆-PtsI^WT and His₆-PtsI^R361H were purified and characterized in vitro (Fig. S1). His₆-PtsI hydrolyzed PEP with a K_m of 0.26 ± 0.02 mM and a V_max of 881 ± 30.6 pmol Pi min⁻¹ mg⁻¹ (Table 1). His₆-PtsI^R361H had a K_m of 0.011 ± 0.001 mM for PEP, and a V_max of 252 ± 2.7 pmol Pi min⁻¹ mg⁻¹. These data showed the suppressing PtsI^R361H variant had a 23-fold higher affinity for PEP and a 3.5-fold decrease in maximal turnover. αKG modulates the activity of PtsI, and the two proteins were assayed in the presence of different concentrations of αKG to determine the inhibitory constant, K_i. The K_i for αKG was slightly increased in the PtsI^R361H variant over PtsI^WT (6.8 ± 0.3 mM αKG, 4.16 ± 0.34 mM, respectively) (Fig. 3; Table 1). Based on these data, αKG was modeled as an inhibitor of His₆-PtsI and His₆-PtsI^R361H. Non-competitive inhibition appropriately modeled the effect of αKG on both His₆-PtsI and His₆-PtsI^R361H (R²=0.9588 and 0.9626, respectively) consistent with previous data reported for the wild-type protein from Escherichia coli (20). Kinetic parameters for PEP hydrolysis by PtsI in isolation were similar to those determined when phosphotransferase activity was assayed in the complete PTS (21). The kinetic parameters for both His₆-PtsI and His₆-PtsI^R361H suggest that PtsI^R361H has an increased affinity for PEP with a decreased V_max. The effect of this substitution on catalysis of PEP under steady-state growth was estimated using the non-competitive model of enzyme inhibition (Equation 1). Steady state PEP and αKG concentrations were assumed to be 0.18 mM and 1.25 mM, respectively (22, 23). Under steady state, $V_{\mathit{app}}$ of PtsI = 277 pmol min⁻¹mg⁻¹ while $V_{\mathit{app}}$ of PtsI^R361H = 201 pmol min⁻¹mg⁻¹ (72% of wild type). In total these data minimized the possibility that αKG significantly altered the activity of the PtsI^R361H variant over the PtsI^WT. However, the data identified a decrease in turnover of the PtsI^R361H variant that could restrict flux through glycolysis, suggesting suppression occurred by a mechanism similar to that defined for the sgrR mutant (17). Disruption of glycolytic flux has been shown to increase flux through the TCA cycle and the pentose-phosphate pathway, suggesting these metabolic changes could be responsible for the increase in ScThi5 function in S. enterica (24–26).

TABLE 1.

Kinetic parameters^a of PtsI^WT and PtsI^R361H

	Km (PEP; mM)	Vmax (pmol Pi min−1 mg−1)	Ki (αKG; mM)	Kcat (min−1)
PtsI	0.26 ± 0.02	881 ± 30.6	4.16 ± 0.34	0.22
PtsIR361H	0.011 ± 0.001	252 ± 2.7	6.8 ± 0.3	0.063

^aHydrolysis of PEP was measured by monitoring the rate of Pi formation using a coupled assay with purine nucleoside phosphorylase and 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG) (19, 50, 51). Activity was determined with αKG titrated (0 – 3 mM) into the assay and the data used to determine the inhibitory constant (Fig. 3). The effect of αKG was modeled as a non-competitive inhibitor, and the K_m, V_max and K_i were calculated using Equation 1. K_cat was calculated using Equation 2.

FIG 3.

PtsI^R361H has an increased affinity for PEP and a lower V_max. The initial velocity of phosphoenolpyruvate (PEP) hydrolysis by (A) PtsI^WT and (B) PtsI^R361H was measured using a coupled assay to detect P_i formation with purine nucleoside phosphorylase and 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG). Reactions were performed in 50 mM HEPES pH 7.5, 1 mM MgSO₄, 200 μM MESG, 1 U/ml purine nucleoside phosphorylase, with 20 μM PtsI^WT or the PtsI^R361H variant at 37°C. αKG was titrated from 0 mM to 3 mM as indicated to determine the K_i for each PtsI protein. Error bars indicate the standard deviation of three technical replicates. The effect of αKG on the velocity of PEP hydrolysis was modeled as a non-competitive inhibitor, and the goodness of fit was evaluated. The non-competitive model fit the calculated K_m, V_max and K_i for both PtsI (R² = 0.9588) and PtsI^R361H (R² = 0.9626).

Loss of yggS allows ScThi5 function.

A previously identified, but uncharacterized (17), mutant that restored ScThi5-dependent growth of a S. enterica thiC mutant on glucose was also studied. The relevant strain did not have a mutation linked to ptsI or sgrR, loci characterized for suppression above and elsewhere (17), respectively. This suppressor was of particular interest because it allowed ScThi5-dependent growth of a S. enterica thiC mutant on multiple carbon sources (e.g., gluconate, galactose and glycerol) in addition to glucose. This feature distinguished it from the ptsI allele above and other characterized suppressors that had an impact only on glucose medium (17). Whole genome sequencing identified multiple SNPs in the strain, but our attention was drawn to an allele of yggS that was predicted to encode YggS^Q51Stop (yggS661). Standard transduction mediated genetic mapping analyses using an insertion near the yggSTU operon (in nupG) confirmed the causative mutation allowing ScThi5-dependent growth was linked to this locus. yggS is a gene of unknown function and a member of the COG0325 family that is conserved across all domains of life (27). The YggS protein purifies with PLP bound and mutations in the human homolog, PLPBP, are associated with Vitamin B6-dependent epilepsy (28, 29). An isogenic pair of strains (with or without the yggS lesion) were constructed in a ΔthiC1225 ΔaraCBAD background. Only the strain with the mutant allele of yggS achieved full growth on glucose, galactose and gluconate when ScThi5 provided the sole source of HMP (Fig. 4). Insertion-deletions of yggS, yggT and yggU were constructed and transduced into the relevant thiC/pScThi5 strain (DM16449). Of the three mutations, only the deletion of yggS allowed ScThi5-dependent growth, indicating the suppression phenotype was due to the loss of YggS and not to a polar effect on a downstream gene(s) (data not shown).

FIG 4.

A null allele of yggS allows ScThi5 dependent growth. An isogenic pair of thiC strains that contained pTac85-ScTHI5 (pDM1625) and either lack (DM17148; A to C) or contain (DM17147; D to F) the yggS661 allele were grown on NCE minimal medium with glucose, gluconate or galactose as a sole carbon source. Filled and empty circles show growth in the presence and absence of thiamine (100 nM), respectively. Filled squares depict growth in the presence of IPTG (100 μM) in the medium to induce expression of THI5. Growth was monitored by the increase in optical density at 650 nm with shaking at 37°C. Error bars indicate the standard deviation of three independent biological replicates.

The data showed that lack of a functional YggS was sufficient to allow ScThi5 to provide enough HMP for growth on all carbon sources tested. Subsequent selection for suppressors that allowed ScThi5-dependent growth on other carbon sources (i.e., pyruvate, galactose) yielded multiple alleles of yggS. Thus, the yggS locus appeared to be the main site of mutations that restored ScThi5-dependent growth of S. enterica on non-glucose carbon sources. Previous work in our lab and others has identified pleiotropic effects caused by the inactivation of yggS in S. enterica and E. coli. Significant for the work herein, a lesion in yggS disrupts homeostasis of B₆ vitamers and alters a variety of metabolite pools (27, 28, 30–32). Altered PLP homeostasis leads to an increase in total PLP and an altered PLP/PMP ratio, and cells lacking yggS excrete PLP into the growth medium (30). Additionally, strains lacking yggS have altered intracellular levels of branch-chain amino acids and ketoacids that are used in transamination reactions (28).

Increased αKG allows ScThi5 function in S. enterica.

Together the data above suggested that altering the metabolic network, via flux changes and/or metabolite pool sizes, influences the function of ScThi5 in S. enterica. Independent means to alter the metabolic network were explored. The standard minimal medium in the laboratory contains ammonia (16 mM) as the sole source of nitrogen. When glutamine (1 mM) was provided as the sole nitrogen source, ScThi5 function became sufficient for full growth of a thiC mutant on glucose, galactose or gluconate (Fig. 5).

FIG 5.

Glutamine as a nitrogen source allows ScThi5-dependent growth. A thiC strain with ScTHI5 on pTac85 was grown on minimal NCN medium with IPTG, glutamine (1 mM) as a nitrogen source and glucose (A), gluconate (B) or galactose (C) as a sole carbon source. Filled and empty circles signify the presence and absence of thiamine (100 nM), respectively. Growth was monitored by the increase in optical density at 650 nm with shaking at 37°C. Error bars indicate the standard deviation of three independent biological replicates.

Changing the nitrogen source impacts flux in multiple pathways and changes the level of numerous metabolites, including αKG. αKG levels in the cell are dynamic and increase when E. coli is grown with nitrogen sources other than ammonia, or when nitrogen assimilation by either glutamate dehydrogenase or glutamate synthase is perturbed (18, 23). In a series of experiments, αKG levels were modulated by nutritional shifts or genetic perturbations and the resulting impact on ScThi5 function assessed. Internal levels of αKG could not be increased by exogenous addition since S. enterica does not have a functional kgtP, the gene encoding the αKG transporter in E. coli (23, 34). However, dimethyl-ketoglutarate (dmKG) is a membrane-diffusible metabolite that is converted into αKG by esterases in the cell (20). A titration of dmKG on glucose medium (5 – 20 mM) showed that 10 mM allowed significant ScThi5-dependent growth of a thiC mutant on minimal glucose medium (Fig. 6, Table S1). When added in concentrations higher than 10 mM, dmKG was inhibitory to all strains (Table S1). Addition of dmKG (10 mM) similarly improved growth when galactose or gluconate were carbon sources (Table S2). The addition of ketobutryate, a ketoacid close in structure to αKG and reported to be elevated in a yggS strain (28), did not support ScThi5 function on glucose.

FIG 6.

Addition of dimethyl-ketoglutarate and/or pyridoxal impact ScThi5 function. A thiC strain with ScTHI5 on pTac85 was grown on minimal medium with IPTG and glucose as a sole carbon source. Other additions were as indicated; dimethyl-ketoglutarate (dmKG; 10 mM) (open squares), pyridoxal (PL; 100 nM) (closed triangles), both PL and dmKG (closed squares), or thiamine (Thi; 100 nM) (closed circles). Growth was monitored by the increase in optical density at 650 nm with shaking at 37°C. Error bars indicate the standard deviation of three independent biological replicates.

A mutation in gdhA (glutamate dehydrogenase) was introduced into the thiC/pScThi5 strain as another way to increase the cellular pool of αKG (23). Consistently, the ΔgdhA thiC/pScThi5 strain (DM17115) showed significant growth on minimal medium with ammonia as a nitrogen source when either glucose, galactose or gluconate were sole carbon sources (Fig. 7). In total, results of these three experiments were consistent with the hypothesis that increasing αKG in S. enterica resulted in increased function of ScThi5 in S. enterica, as judged by the level of growth supported by the protein.

FIG 7.

Altering route of nitrogen assimilation restores ScThi5 function. A thiC gdhA mutant strain expressing ScTHI5 from pTac85 was grown on minimal NCE medium with glucose (A), galactose (B) or gluconate (C) as a sole carbon source. Filled circles indicate presence of thiamine (100 nM) in the medium. Empty circles depict the presence of IPTG (100 μM) to induce expression of THI5 and squares indicate the presence of both 100 μM IPTG and 100 nM PL in the medium. Growth was monitored by the increase in optical density at 650 nm with shaking at 37°C. Error bars indicate the standard deviation of three independent biological replicates.

Exogenous pyridoxal enhances ScThi5 function.

Data above showed that the metabolic changes caused by using glutamine as sole nitrogen source, or a lesion in yggS, restored full growth to the thiC/pScThi5 (Fig. 5 and 4, respectively) while attempts to specifically increase αKG pools with dmKG supplementation or a ΔgdhA mutation only partially restored growth to the thiC/pScThi5 strain in the absence of thiamine (Fig. 6 and 7, respectively). These data suggested that utilization of glutamine as a nitrogen source, or elimination of yggS, not only increased αKG pools but altered another metabolic feature necessary to allow ScThi5 activity sufficient for full growth. PLP is a substrate of Thi5 and past models suggested function of ScThi5 in S. enterica was limited by this metabolite or its delivery (35). When pyridoxal (100 nM; PL) alone was provided to the medium, limited ScThi5-dependent growth was allowed on glucose. However, when both dmKG and PL were provided, final cell density was restored to almost wild type, albeit with a slower growth rate (Fig. 6). Further, addition of PL dramatically enhanced the growth of the ΔgdhA thiC/pScThi5 strain on both glucose and galactose (Fig. 7). Together, these data suggest the metabolic network configuration of S. enterica requires cellular pools of both αKG and PLP be increased to support sufficient activity of ScThi5 to allow full growth. Further, the additive effect of the two metabolites suggests they have independent effects, each of which contributes to increased activity.

When growing on glucose, constricting flux through glycolysis could increase αKG and PLP levels by resulting in increased flux through the TCA cycle (αKG) and pentose-phosphate pathway (generating erythrose-4-phosphate, a precursor to PLP) (24–26). While a stimulatory role for PLP might be anticipated by its role as a substrate, it is unclear if levels of αKG influence ScThi5 function in a direct or indirect manner. An important consideration in this working model is the fact that S. enterica strains lacking yggS have increased PLP and increased levels of some ketoacids (28, 30), providing a potential explanation for the yggS lesion to restore full growth on all carbon sources used in this study.

Substitutions in LpThi5 change the metabolic network requirement.

The results above suggested changes that occurred in the metabolic network to allow ScThi5 to function in S. enterica. It was also of interest to understand how the difference in Thi5 homologs could accommodate or prevent function in a given metabolic context. Expression of the bacterial Thi5 (Legionella pneumophila; LpThi5) allowed full growth of a S. enterica thiC mutant without host modification (15). ScThi5 and LpThi5 share 48% amino acid identity, and limited site directed mutagenesis was used to probe conserved residues predicted to be in the active site (15). As expected, many variants of LpThi5 that altered the active site eliminated function (15). However, variants that altered the conserved GEFG motif (residues 118–121) involved in hydrogen bonding with the phosphate group of PLP in the active site (36) modulated but did not eliminate function. Unexpectedly, some variants affecting this motif retained the ability to complement an S. enterica thiC mutant on ribose, but not on glucose (15). This pattern of growth was notable because it was similar to the pattern allowed by the ScThi5 protein described herein. Analysis of the LpThi5^G118A and LpThi5^E119A variants showed a significant shift in Thi5-dependent growth compared to the LpThi5^WT, indicating a change in in vivo function. While LpThi5^WT allowed growth of a thiC mutant with glucose (15), the LpThi5^E119A variant did not, similar to the result obtained with ScThi5 (Fig. 8). More significantly, addition of both dmKG and PL to the growth medium dramatically improved the LpThi5^E119A-dependent growth (Fig. 8), similar to the results seen with ScThi5 (Fig. 6).

FIG 8.

LpThi5^E119A phenocopies ScThi5 dependent growth on glucose. A thiC strain with pTac85 containing either (A) ScTHI5 or (B) Lpthi5^E119A was grown on minimal medium with glucose as a sole carbon source in the presence of IPTG (100 μM) to induce expression of the respective thi5 gene. Further additions were as noted and include dimethyl-ketoglutarate (dmKG; 10 mM), pyridoxal (PL; 100 nM), or thiamine (Thi; 100 nM). Growth was monitored by the increase in optical density at 650 nm with shaking at 37°C. Error bars indicate the standard deviation of three independent biological replicates.

Further, when a lesion in gdhA was introduced into the thiC mutant to increase αKG during growth with ammonia as a nitrogen source (23), both LpThi5^G118A and LpThi5^E119A allowed growth (Fig. 9). Consistent with the additive effect of dmKG and PL shown for ScThi5 function, the addition of PL further improved the growth allowed by the gdhA deletion (Fig. 9). We hypothesize that ScThi5 has a lower affinity for PLP than the bacterial LpThi5 and the LpThi5^G118A, LpThi5^E119A variants reduce the normally high affinity of the bacterial enzyme for PLP. This difference may be compensated for by the PLP delivery system in yeast (SNZ3). The additive effect of αKG and PLP suggest αKG has a role of directly or indirectly activating Thi5 enzyme(s) or variants with compromised PLP binding.

FIG 9.

Involvement of PL and αKG in LpThi5^G118A and LpThi5^E119A dependent growth. Strains containing Lpthi5^G118A (A, C) or Lpthi5^E119A (B, D) on pTac85 in a thiC (A, B) or thiC gdhA background (C, D) were grown on a minimal medium with glucose as a sole carbon source. Empty or closed circles indicate the absence or presence of thiamine (100 nM), respectively. Growth in the presence of IPTG (100 μM), or IPTG and pyridoxal (100 nM; PL) are indicated by empty or filled squares, respectively. Growth was monitored by the increase in optical density at 650 nm with shaking at 37°C. Error bars indicate the standard deviation of three independent biological replicates.

DISCUSSION

Incorporation of the Thi5 pathway for HMP-P biosynthesis from S. cerevisiae into S. enterica required modification of the metabolic network by reducing flux through glycolysis (17). The consequences of reducing flux through glycolysis, how this led to Thi5 function, and other mechanisms of suppression were unknown. The data herein suggests that to function, ScThi5 requires levels of PLP and αKG higher than those normally present in S. enterica. A stimulatory effect of PLP could be due to the need for increased synthesis and/or delivery of this cofactor, which is also substrate for Thi5. Significantly, the yeast metabolic network includes a PLP synthase dedicated to thiamine synthesis (35), which is not present in S. enterica.

A direct role for αKG in the reaction mechanism of Thi5 can be suggested. The proposed mechanism for Thi5 includes an oxidation of both the PLP and histidine substrates (37). Several oxygenases exist that facilitate oxidation of their substrates by using keto acids like αKG as co-substrates. These enzymes are found in diverse biochemical pathways, although thus far no examples that incorporate the co-substrate into the final product have been described (38). Importantly, several of the relevant enzymes are oxygen sensitive, and when purified under aerobic conditions they must be reconstituted anaerobically with the addition of strong reducing agents like dithionite. For example, when purified aerobically, enzymes like TauD spontaneously auto-oxidize and damage residues within the active site (39, 40). Such a characteristic would be consistent with a previous report describing an oxygen-sensitive reaction mechanism for Thi5 in vitro (37). Investigating these mechanistic hypotheses, and the possibility that Thi5 is a single-turnover enzyme (37) will require a dependable in vitro assay for activity, which is currently not available.

It was previously reported that unlike ScThi5, a bacterial enzyme (LpThi5) supported thiamine synthesis in S. enterica without alteration of the metabolic network (15). Herein we extended this work and showed that while Lpthi5 variants, presumed to be compromised in PLP binding lost the ability to function in S. enterica, Thi5-dependent growth was restored with the exogenous addition of both PL and αKG. These results supported the hypothesis that structural differences between the LpThi5 and ScThi5 proteins contribute to the functional differences displayed in S. enterica. The amino acid sequence of LpThi5 was modeled using the Phyre2 pipeline in intensive mode (41). The resultant model was aligned to a crystal structure of ScThi5 (PDB 4H67 chain D) to determine potential structural differences using Maestro (Schrodinger Inc.). ScThi5 has an extension of 20 amino acids at the c-terminus compared to LpThi5. This extension comprises alpha-helix 12 in the crystal structure and could influence the orientation of the two domains of the protein based on a proposed role of this helix in stabilizing Domain 2 (36). Additionally, based on the amino acid alignment and resulting structural alignment, LpThi5 has an eight amino acid truncation involving residues in alpha helix 6. Lack of these residues may constrict the active site, potentially increasing the affinity of LpThi5 for PLP which would be consistent with the working model herein, where αKG may be directly or indirectly activating the Thi5 pathway in vivo. Moving forward, this information will be valuable in efforts to define the biochemical mechanism of Thi5 action in vivo, and whether the effect of elevated αKG is direct or indirect.

Moving pathways between organisms can lead to incompatibilities that compromise function of either the pathway itself or fitness of the host (1, 2). Defining the solutions to such incompatibilities improves our metabolic understanding and builds knowledge to facilitate the construction of productive metabolic networks. For instance, the work herein suggests differences in network structure between L. pneumophila, S. enterica, and S. cerevisiae, that may help to determine why organisms use Thi5 over ThiC in the biosynthesis of thiamine-pyrophosphate. Furthermore, global metabolic knowledge can lead to a better understanding of enzyme mechanism (42), such as defining the role for αKG in Thi5 activity in vitro. If there is a difference between these two homologs in vivo or in vitro, this would begin to describe differences in structure-function relationships within the Thi5 pathway across organisms.

MATERIALS AND METHODS

Strains, media, and chemicals.

All strains of S. enterica are derived from strain LT2 and their relevant genotypes are described in Table 2. Rich media used in this study included Nutrient Broth (NB; 8 g/liter Difco Nutrient broth and 5 g/liter NaCl), or super broth (SB; 32 g/liter tryptone (Fisher Scientific), 20 g/liter yeast extract (Fisher Scientific), 5 g/liter NaCl with 0.05 N NaOH). Solid media contained 15 g/liter agar. Antibiotic concentrations in rich media are as follows: Ampicillin (Amp) - 100 mg/liter; Chloramphenicol −20 mg/liter; Kanamycin −50 mg/liter; Tetracycline - 20 mg/liter. Minimal medium was no-carbon E medium (NCE) (43) with 1 mM MgSO₄, 0.1 x trace minerals (44), with glucose, galactose or gluconate as a sole carbon source (11 mM). No carbon or nitrogen (NCN) (45) salts with glutamine (1 mM) as a nitrogen source and indicated carbon source at 11 mM was used as low nitrogen medium. Thiamine and pyridoxal were added as indicated to final concentration of 100 nM. IPTG was purchased from Gold Biotechnology, (St. Louis, MO), other chemicals were purchased from Sigma-Aldrich, (St. Louis, MO).

TABLE 2.

Strains, plasmids, and primers

Strain/plasmid/primer	Relevant genotype	Source
Strain
DM13337	ΔthiC1225 ΔaraCBAD	Lab strain collection
DM16449	ΔthiC1225 ΔaraCBAD/pDM1625	(15)
DM17114	ΔthiC1225 ΔaraCBAD ΔyggS638::Km/pDM1625	This study
DM17115	ΔthiC1225 ΔaraCBAD ΔgdhA631::Cm/pDM1625	This study
DM17147	ΔthiC1225 ΔaraCBAD nupG::Tn10d(Tc) yggS661/pDM1625	This study
DM17148	ΔthiC1225 ΔaraCBAD nupG::Tn10d(Tc)/pDM1625	This study
DM17305	ΔthiC1225 ΔaraCBAD yfeA85::Tn10d(Tc)/ pDM1625	This study
DM17306	ΔthiC1225 ΔaraCBAD yfeA85::Tn10d(Tc) ptsI611/pDM1625	This study
DM16126	ΔthiC1225 ΔaraCBAD/pDM1674	(15)
DM16127	ΔthiC1225 ΔaraCBAD/pDM1675	(15)
DM17329	ΔthiC1225 ΔaraCBAD ΔgdhA631::Cm/pDM1674	This study
DM17330	ΔthiC1225 ΔaraCBAD ΔgdhA631::Cm/pDM1675	This study
Plasmid
pDM1625	pTac85-ScTHI5 (codon-optimized)	(15)
pDM1674	pTac85-Lpthi5G118A	(15)
pDM1675	pTac85-Lpthi5E119A	(15)
pDM1677	pTEV5-ptsI	This study
pDM1678	pTEV5-ptsI611	This study
Primer
NheI ptsI F	TAGGGCTAGCATGATTTCAGGCATTTTAGCATCC	This study
XhoI ptsI R	TAGGCTCGAGTTAGCAGATTGTTTTTTCTTCAATGA	This study
yggS F	GTCACAATTTCTTCCATATTCATATAAACATCCTCGGAAAGTGTAGGCTGGAGCTGCTTC	This study
yggS R	ACAGCTCAATTACCGTTGAGAGCAGGAAGGTCAACGTATTCATATGAATATCCTCCTTAG	This study
yggT 5′	TCGTGATTACACAAAAAATtaaGGAAAACTGAGGAACGCCGTGTAGGCTGGAGCTGCTTC	This study
yggT 3′	GCCGTAACACCAGACCGTCTTCGCAGCGAGTCACAGCACTCATATGAATATCCTCCTTAG	This study
yggU 5′	CGACAGGCAACATGTTGCTGCCGGGGCTGTGGATGGCGTTGTGTAGGCTGGAGCTGCTTC	This study
yggU 3′	GCACTTTACCGGCGTTGCCGGTTGCGAGAACAACTTTTTGCATATGAATATCCTCCTTAG	This study

Genetic methods.

Suppressor mutants (ptsI611, yggS661) were isolated in a thiC/pScThi5 strain of S. enterica on minimal glucose medium after diethyl sulfate mutagenesis as reported (17). The causative mutation in each strain was identified by Illumina whole-genome sequencing, linked to transposon Tn10d(Tc) using standard genetic techniques and reconstructed into the DM13337 (ΔthiC1225 ΔaraCBAD) background using the high-frequency generalized transducing mutant of bacteriophage P22 (HT105/1, int-201) (46), as previously described (17, 47). Insertion-deletion mutations of yggS, yggT, and yggU were constructed by Lambda Red recombineering using the primers in Table 2, where the full coding region of each gene was replaced with a kanamycin resistance cassette (48). The presence of each insertion-deletion was confirmed by colony PCR before moving the insertion-deletion into a DM13337 background by P22-mediated transduction. Other mutations were moved by transduction into the DM13337 background using standard genetic approaches as needed.

Growth analysis.

Growth of S. enterica strains were monitored at OD₆₅₀ in 96-well plates with a BioTek ELx808 plate reader. Strains were grown for 16–20 h in NB Amp (2 ml) at 37°C before pelleting and resuspension in an equal volume of 0.85% NaCl. The cell suspension was used to inoculate (1%) the appropriate media (200 μl). Plates were incubated at 37°C with medium shaking and data were plotted using Prism 8 (GraphPad).

Molecular techniques.

Plasmids were constructed using standard molecular techniques. Plasmid DNA was isolated using the PureYield Plasmid MiniPrep System (Promega, Madison, WI). Q5 DNA polymerase (New England Biolabs, Ipswich, MA) was used to amplify ptsI and ptsI611 from genomic DNA of the appropriate strain with primers synthesized by Integrated DNA Technologies (Coralville, IA) listed in Table 2. The resulting PCR products were purified using a PCR purification kit (Qiagen, Venlo, Limburg, The Netherlands). NheI and XhoI restriction endonucleases (New England Biolabs, Ipswich, MA) were used to digest the PCR product. The digested product was then purified using the PCR purification kit and ligated into the pTEV5 plasmid, generating pDM1671and pDM1672. The identity of each plasmid insert was confirmed by sequencing by Eton Bioscience, Inc, Research Triangle Park, NC.

Expression and Purification of PtsI and PtsI^R361H proteins.

E. coli BL21-AI strains containing pDM1677 or pDM1678 were inoculated into each of two flasks containing NB Amp (50 ml) and grown overnight at 30°C with shaking. Each of two Fernbach flasks containing SB Amp (1.5 L) were inoculated from an overnight culture (1%), and the flasks were incubated at 37°C with shaking (180 rpm) until OD₆₅₀ reached 0.15. At that point, the temperature was lowered to 30°C and when the OD₆₅₀ reached 0.6, expression was induced with IPTG (1 mM) and arabinose (0.02% wt/vol). Incubation at 30°C continued for 20 h prior to harvesting by centrifugation. The cell pellet was stored at −80°C until use.

Twenty-five grams of the cell pellet was resuspended to a total volume of 50 ml in Buffer A [50 mM HEPES (Fisher Scientific), 150 mM NaCl, 20 mM Imidazole (Fisher Scientific) pH 7.5 at 4°C] with DNase (0.025 mg/ml), lysozyme (1 mg/ml) and phenylmethylsulfonyl fluoride (0.1 mg/ml) and incubated on ice for 1 h. The cell suspension was lysed at 1.45 kbar using a Constant Systems Limited One Shot (United Kingdom), and the resultant cell lysate was cleared at 48,000 × g (50 min, 4°C). Particulates from the cell extract were removed using a 0.45 μM PVDF filter (Millipore) and the filtered extract was injected onto two pre-equilibrated 5 ml HisTrap HP Ni-Sepharose columns connected in sequence. The protein was washed with 5 column volumes of Buffer A, 5 column volumes of 4% Buffer B (Buffer A + 480 mM Imidazole (Fisher Scientific), pH 7.5 at 4°C) and eluted from the column with a gradient of Buffer B from 4% to 100% over 10 column volumes. Fractions containing His₆-PtsI (or His6-PtsI^R361H) as determined by SDS-PAGE were combined and concentrated by centrifugation using a 50 kDa filter (Millipore), exchanged into 50 mM HEPES buffer with 10% glycerol, pH 7.5 at 4°C, using a PD10 column following the manufacturer’s instructions (GE Healthcare), flash-frozen in liquid nitrogen and stored at −80°C until use. Protein concentration was determined by extinction coefficient using the theoretical molecular weight and A₂₈₀ extinction coefficient of His₆-PtsI as determined by the ExPASy Protparam database (ε₂₈₀ = 30620 M⁻¹cm⁻¹) (49). Protein purity was evaluated by SDS-PAGE coupled with densitometry, and both enzymes were purified to > 95% purity (Fig. S1).

PEP hydrolysis by PtsI.

The enzymatic activity of His₆-PtsI and His₆-PtsI^R361H was measured by coupling the formation of inorganic phosphate from the hydrolysis of PEP with the conversion of P_i and 2-amino-6-mercapto-7-methylpurine ribonucleoside (MESG) to ribose-1-phosphate and 2-amino-6-mercapto-7-methylpurine by purine nucleoside phosphorylase (PNP) (19, 50, 51). Formation of 2-amino-6-mercapto-7-methylpurine was monitored at 360 nm using the EnzChek Phosphate assay kit (ThermoFisher Scientific). Briefly, 20 μM His₆-PtsI or His₆-PtsI^R361H were incubated with 1 mM MgSO₄, 200 μM MESG and 1 U/ml PNP, in 50 mM HEPES, pH 7.5 at 37°C for 10 min. The reaction was initiated with the addition of PEP and the change in absorbance at 360 nm was monitored continuously for an hour. To monitor inhibition, alpha-ketoglutarate was added to a final concentration between 1 and 3 mM during the 10-minute incubation at 37°C. The V_i measurements were obtained during the first 10 min where the rate of change in absorbance was linear and < 10% of available PEP was consumed. Rates of P_i formation were calculated from a standard curve of A₃₆₀ from 0 – 10000 pmol KH₂PO₄ in triplicate (ΔAbs = 2.54 × 10⁻⁵pmol P_i, R² = 0.9994). The effect of αKG on PtsI and PtsI^R361H activity was modeled using the non-competitive inhibition model. The following equation was used in Prism 8 and the fit was evaluated.

$$v_{\mathit{app}}=\frac{V_{\mathit{max}}[\mathit{PEP}]}{\left(1+\frac{\left[\alpha \mathit{KG}\right]}{K_i}\right)\left(K_m+[\mathit{PEP}]\right)}$$ (Equation 1)

The k_cat of each enzyme was calculated using the following equation, where E_t is the quantity of enzyme in the reaction.

$$V_{\mathit{max}}=k_{\mathit{cat}}\left[E_t\right]$$ (Equation 2)

Data availability.

The data that supports the findings of this study are available in the supplementary material of this article, and additional information is available upon request from the authors.