Salmonella enterica serovar Typhimurium has three transketolase enzymes contributing to the pentose phosphate pathway

The genus Salmonella is responsible for many illnesses in humans and other vertebrate animals. We report here that Salmonella enterica serovar Typhimurium harbors three transketolases that support the non-oxidative branch of the pentose phosphate pathway. BLAST analysis identified two genes, STM14_2885 and STM14_2886, that together encode a putative transketolase (TktC) with 46–47% similarity to the known TktA and TktB isoforms. Assessing the mRNA and protein expression for each of the three transketolases, we determined that all are expressed in WT cells and regulated to varying extents by the alternative sigma factor RpoS. Enzyme assays with lysates from WT and transketolase-knockout strains established that TktA is responsible for >88% of the transketolase activity in WT cells. We purified recombinant forms of each isoenzyme to assess the kinetics for canonical transketolase reactions. TktA and TktB had comparable values for Vmax (539–1362 μm NADH consumed/s), Km (80–739 μm), and catalytic efficiency (1.02 × 108-1.06 × 109 m−1/s) for each substrate tested. The recombinant form of TktC had lower Km values (23–120 μm), whereas the Vmax (7.8–16 μm NADH consumed/s) and catalytic efficiency (5.58 × 106 to 6.07 × 108 m−1/s) were 10–100-fold lower. Using a murine model of Salmonella infection, we showed that a strain lacking all three transketolases is avirulent in C57BL/6 mice. These data provide evidence that S. Typhimurium possesses three transketolases that contribute to pathogenesis.

The genus Salmonella contains more than 2,500 different serovars that are responsible for a range of pathogenic illnesses not only in humans, but in nearly all vertebrate animals (1). In humans, Salmonella is one of the most prevalent food-borne pathogens worldwide, causing an estimated 93.8 million infections that result in ϳ155,000 deaths (2). Globally, the most prevalent serovar in humans is Salmonella enterica serovar Typhimurium (3). These facultative intracellular bacteria are capable of overcoming host innate immunity using an arsenal of defenses that alter or mitigate the immune response. Many of the defenses are dependent on the reducing power of NADPH (4 -8). Cellular NADPH is generated through the reduction of NADP ϩ primarily by enzymes within the pentose phosphate pathway (PPP), 3 a major route of carbohydrate metabolism in species ranging from bacteria and archaea to plants and animals (9 -12). The PPP can be divided into an oxidative (oxPPP) and a non-oxidative phase (non-oxPPP). The oxPPP is largely responsible for maintaining cellular redox homeostasis through the reduction of NADP ϩ to NADPH, which is critical for Salmonella virulence by protecting against innate host defenses (4,5). In the non-oxPPP of Salmonella, a series of enzyme-catalyzed reactions generate multiple metabolic intermediates that can be diverted to other pathways, including glycolysis, or serve as precursors for the generation of cellular constituents, such as nucleotides, aromatic amino acids, or lipopolysaccharide (Fig. S1) (11,13). Products from the non-oxPPP can also be channeled back into the oxPPP via gluconeogenesis, allowing carbon to cycle through the pathway. Ultimately, the non-oxPPP allows for dynamic carbon flux to multiple metabolic pathways (14).
Within the non-oxPPP, the dimeric, thiamine pyrophosphate (TPP)-dependent transketolase enzyme catalyzes the rate-limiting step through the reversible transfer of a ketol group to an aldose substrate, such as ribose-5-phosphate or erythrose-4-phosphate (15,16). During early studies analyzing variations in bacterial lipopolysaccharide, mutants of Escherichia coli or S. Typhimurium defective in transketolase activity were isolated (17,18). The E. coli transketolase mutants were unable to grow using D-ribose as a sole carbon source, but similar S. Typhimurium transketolase mutants would sometimes exhibit slow growth on the same medium. The phenotype in S. Typhimurium was not investigated further but suggests that other transketolases may be encoded by this bacterium. With the development of whole-genome sequencing, it is known that the S. Typhimurium strain 14028s genome harbors two genes, tktA and tktB, that encode two transketolase isoenzymes TktA and TktB, respectively (Fig. S2, A and B) (19). Herein, a third transketolase isoform in S. Typhimurium was identified based on sequence analysis. We present evidence that all three transketolase isoenzymes are functional and contribute to pathogenesis of this bacterium. Furthermore, we determined the regulation, biological contribution, and kinetic properties of each isoenzyme. Our findings offer deeper insights into the function and regulation of this central metabolic pathway.

Identification and in silico analysis of transketolases within Enterobacteriaceae
Aside from the two annotated transketolases in S. Typhimurium, two additional genes were found (STM14_2885 and STM14_2886) that, based on BLAST results and sequence analysis, form the components of a transketolase enzyme. S. Typhi-murium gene STM14_2886 is located on the complementary strand of the chromosome from 2,505,647 to 2,506,477, and the coding region of STM14_2885 begins within STM14_2886 at 2,504,701-2,505,654 (Fig. S2). Using NCBI conserved domains, we mapped out conserved functional regions within each S. Typhimurium transketolase and putative transketolase ( Fig.  1 and Fig. S3). Within both TktA and TktB, there were similar conserved dimerization domains in the N-terminal regions, approximately between residues 90 and 210 for each enzyme. The N termini of TktA and TktB also contained the TPP-binding domain for each enzyme. Analysis of STM14_2885 and STM14_2886 revealed that each protein contained both dimerization and TPP binding domains. STM14_2886 was classified as the N-terminal domain of transketolase containing the catalytic domain, whereas STM14_2885 was classified as the C-terminal domain with putative regulatory function. Predictive modeling of each putative transketolase was performed using Swiss-Model to identify conserved structures between the isoenzymes (Figs. S4 -S6) (20 -22). Based on structural similarities between proteins and the sequence analysis results that characterized each putative transketolase as two components of a transketolase enzyme. To maintain consistent nomenclature between other transketolases, we designated STM14_2886 as tktC 1 and STM14_2885 as tktC 2 , with

Three transketolase isoenzymes in S. Typhimurium
gene products TktC 1 and TktC 2 , respectively (Fig. S2, C and D). We further designated TktC 1 and TktC 2 as TktC when considered together. This was based on structural similarities between proteins and sequence analysis results that characterized each putative transketolase as two components of a transketolase.
The redundancy between the three transketolases in S. Typhimurium made us question the similarity between the isoenzymes in Salmonella and other Enterobacteriaceae. We made comparisons between amino acid sequences for each transketolase in S. Typhimurium, E. coli, and S. enterica serovar Typhi, the causative agent of typhoid fever (Tables S3-S5). The sequences for both TktA and TktB showed a high degree of similarity between species and serovars (Ͼ96%). The S. Typhimurium TktC sequence had the greatest difference, as it was only 47-48% similar to TktA and 41-46% similar to TktB in either E. coli or S. Typhimurium. After noting the differences between the number of transketolases in Salmonella and E. coli, we examined the genomes of other Enterobacteriaceae and determined the number of putative transketolase enzymes and transketolase genes (Table S6). Many bacteria in this group possessed only a single transketolase (e.g. Shigella dysenteriae), whereas some had a similar arrangement as S. Typhimurium (e.g. C. koseri). Of the organisms that were examined, C. rodentium possessed the greatest number of transketolase genes and putative isoenzymes. However, the reasons for these differences between organisms remain unclear.

Transketolase expression in S. Typhimurium shows partial growth phase and RpoS regulation
We assessed the expression and regulation of each transketolase gene and isoenzyme in S. Typhimurium during different growth phases. E. coli, another member of the family Entero-bacteriaceae, possesses genes for only two transketolases in its genome (tktA and tktB), and previous work demonstrated that they are both regulated by the alternative sigma factor RpoS (23). Based on the RpoS-dependent regulation of transketolases in E. coli, we compared the transcription and translation of the three S. Typhimurium transketolase isoforms in WT and rpoSdeficient strains (Fig. 2). Stationary-phase S. Typhimurium expressed lower levels of tktA mRNA compared with logarithmic phase. Although tktA was expressed at slightly higher levels in the absence of RpoS, the differences were not statistically significant. Whereas mRNA levels were down-regulated during stationary phase, TktA protein expression was unaffected. Expression of tktB mRNA and TktB protein in S. Typhimurium increased from logarithmic growth to stationary phase, and the up-regulation was at least partially dependent on RpoS. Unlike E. coli, which expresses tktB at levels comparable with tktA ( Fig.  S7), tktB expression in S. Typhimurium was nearly 100-fold lower than tktA during logarithmic phase, but only 4-fold lower during stationary phase. Expression of both tktC 1 and tktC 2 was up-regulated during stationary phase in an RpoS-dependent manner. Although mRNA levels of tktC were up-regulated, protein expression remained unchanged. As anticipated, katE and otsA followed the known pattern of RpoS regulation ( Fig.  S7) (24 -26). Taken together, these data demonstrate that the four genes encoding transketolases and the three protein isoenzymes in S. Typhimurium were all expressed under our tested conditions and exhibited varying degrees of growth phase-and RpoS-dependent regulation.

Transketolase-deficient S. Typhimurium can grow with glucose, but not ribose, as the sole carbon source
A hallmark of transketolase-deficient E. coli is their inability to utilize pentoses as their only source of carbon (27,28). We Overnight cultures grown in LB were diluted 1:500 into fresh LB, and cells were harvested from logarithmic (A 600 ϭ 0.5) or stationary-phase (t ϭ 20 h) cultures. A, transcription of the four transketolase genes in S. Typhimurium was evaluated in WT and rpoS mutant bacteria using qRT-PCR. mRNA levels were normalized to the housekeeping sigma factor rpoD. Expression is represented as mean and S.D. of six or eight measurements from three or four independent biological replicates with histograms and error bars. Individual data points for each biological replicate are represented as circles and overlaid on each histogram. B, translational expression of each FLAG-tagged transketolase in S. Typhimurium was measured by Western blot analysis. Expression is represented as mean intensity and S.D. of each protein band with histograms and error bars. Data represent measurements from three independent experiments, and individual data points are overlaid on each histogram. Representative Western blots are depicted below the summary data, and each band corresponds to the bar directly above. Data in A and B were analyzed by one-way ANOVA followed by Tukey's post hoc test. Differences were considered significant (*) if p Ͻ 0.05. ns, not significant.

Three transketolase isoenzymes in S. Typhimurium
measured growth kinetics of our library of transketolase mutant S. Typhimurium strains in media supplemented with glucose or ribose to determine the role of each of the three putative transketolase isoenzymes in Salmonella pentose utilization. Each of the strains grew comparably with WT in LB broth (Fig. 3, A and D). All strains grew in M9 minimal medium supplemented with 0.2% D-glucose, although the ⌬tktA ⌬tktB ⌬tktC strain had a growth defect compared with WT ( Fig. 3, B and E). In contrast, when grown in M9 minimal medium supplemented with 0.2% D-ribose, the ⌬tktA and ⌬tktA ⌬tktC strains had growth defects (Fig. 3, C and F), and the ⌬tktA ⌬tktB ⌬tktC strain had no measurable growth (Fig. 3F). Importantly, when grown in M9 with only D-ribose, S. Typhimurium strains lacking two transketolases in any combination (i.e. ⌬tktA ⌬tktB, ⌬tktA ⌬tktC, or ⌬tktB ⌬tktC) exhibited measurable growth, indicating that the remaining enzyme in each case could catalyze the transketolase reaction.

Each of the three transketolase isoenzymes in S. Typhimurium demonstrates enzymatic activity in cell extracts
To further examine the individual contributions of each transketolase, we measured enzymatic activity in cell extracts derived from transketolase mutant strains ( Fig. 4 and Fig. S8). Cell extracts were obtained from each strain at both logarithmic and stationary phases and used in transketolase activity assays. The assays were performed with substrates for both canonical transketolase reactions: 1) xylulose 5-phosphate and ribose 5-phosphate (Fig. 4) and 2) xylulose 5-phosphate and erythrose 4-phosphate (Fig. S8). Loss of tktA corresponded with  Typhimurium. Transketolase enzymatic activity was measured from cell extracts of transketolase knockout strains harvested at either logarithmic (A 600 ϭ 0.5) or stationary (t ϭ 20 h) phase. The assay was performed using xylulose 5-phosphate and ribose 5-phosphate as substrates. Transketolase activity is expressed as units/mg of total protein input, where 1 unit is defined as the amount of enzyme that oxidized 1 mol of NADH/ min. Histograms and error bars, means and S.D. values of three or four independent experiments. Individual data points are overlaid on each histogram. Data were analyzed by one-way ANOVA followed by Tukey's post hoc test, comparing each knockout strain with WT at the same growth phase. Differences were considered significant (*) if p Ͻ 0.05. LOD for the assay was 0.002.

Three transketolase isoenzymes in S. Typhimurium
a decrease of ϳ88% of detectable transketolase activity compared with WT, suggesting that under the tested conditions, TktA had the largest biological role. However, S. Typhimurium strains lacking two transketolases in any combination still yielded detectable enzymatic activity. The resulting activity was attributable to the remaining enzyme in each case, suggesting that whereas TktA provides the majority of transketolase activity in WT cells, all three isoenzymes provide a measurable contribution. The ⌬tktA ⌬tktB ⌬tktC strain yielded no detectable enzymatic activity using xylulose 5-phosphate and ribose 5-phosphate as substrates (Fig. 4), confirming that S. Typhimurium transketolase activity is encoded by a combination of three transketolases. When using xylulose 5-phosphate and erythrose 4-phosphate as substrates, there was very low NADH consumption detected in the ⌬tktA ⌬tktB ⌬tktC strain cell lysates (Fig. S8). This observation is probably due to erythrose-4-phosphate being used as a substrate in combination with residual cellular hexoses by transaldolase (Fig. S1). Ultimately, measurable transketolase activity in all except the ⌬tktA ⌬tktB ⌬tktC strain, combined with the inability of this strain to grow using ribose as the sole carbon source (Fig. 3), indicates that S. Typhimurium possesses three functional, complementary transketolases. Based on these results, we subsequently use Tkt 0 in reference to the ⌬tktA ⌬tktB ⌬tktC strain.

Ectopic expression of any S. Typhimurium transketolase isoenzyme complements Tkt 0
We constructed pBAD plasmids with WT alleles of each transketolase, relying on leaky expression from the pBAD promotor. The vectors were transformed into the Tkt 0 background individually, and the ability of each gene to complement the transketolase deficiency was tested. Each of the complemented strains grew in LB broth without any notable defects (Fig. 5A). When growth in M9 minimal medium supplemented with D-glucose was evaluated, the empty vector control plasmid caused a substantial fitness burden on the Tkt 0 strain, but all strains demonstrated measurable growth (Fig. 5B). The Tkt 0 strain did not grow in M9 minimal medium supplemented with D-ribose with only the empty vector control, but in trans com-plementation with tktA, tktB, or tktC allowed for growth under the same conditions (Fig. 5C). Notably, complementation of the Tkt 0 strain with individual tktC genes, tktC 1 or tktC 2 , did not restore growth in M9 minimal medium with D-ribose (Fig. S9), indicating that both subunits must be expressed for a functional transketolase.
We further evaluated the transketolase activity from the engineered expression vectors by measuring transketolase activity from cell lysates in the Tkt 0 background (Table S7). There was no detectable activity from the empty vector control or when tktC 1 or tktC 2 were expressed individually. There was, however, detectable activity from the Tkt 0 strain complemented with either tktA, tktB, or tktC when expressed in its WT arrangement.

Determining transketolase kinetics from purified recombinant enzyme
After verifying enzymatic activity from recombinant transketolase in the engineered expression vectors in vivo, we purified each enzyme. Constructs for TktA and TktB were readily purified to apparent homogeneity (Fig. S10), and kinetic parameters were determined using each of the substrates for both canonical transketolase reactions (Table 1 and Figs. 6 and 7). We were unable to purify the individual constructs for TktC 1 and TktC 2 , so we engineered a TktC 1 -TktC 2 fusion protein by eliminating the tktC 1 stop codon and inserting tktC 2 in-frame immediately downstream. This recombinant form of TktC, denoted here as TktC*, was successfully purified to apparent homogeneity (Fig. S10), and kinetic parameters are reported in Table 1 and Fig. 8. Protein structural modeling of TktC* revealed a structure comparable with that of TktC (Figs. S6 and S11). TktC* exhibited slightly lower K m values (23-120 M) compared with TktA and TktB (80 -730 M) for each of the tested substrates. Catalytic efficiencies (k cat /K m ) of TktA and TktB were comparable with each other (1.02 ϫ 10 8 to 1.06 ϫ 10 9 ) and were nearly 100 times greater than that of TktC* (5.58 ϫ 10 6 to 6.07 ϫ 10 7 ). Altogether, these results demonstrate that TktC 1 and TktC 2 together form a functional transketolase and provide foundational descriptions of three transketolase isoenzymes in S. Typhimurium.

Transketolase-deficient (Tkt 0 ) Salmonella are avirulent in mice
To determine whether transketolases are involved in S. Typhimurium pathogenesis, the virulence of the Tkt 0 strain in C57BL/6 mice was assessed (Fig. 9). Intraperitoneal inocula-tion of Tkt 0 S. Typhimurium was not lethal to the mice and animals did not exhibit any clinical signs of disease. These findings suggest that lack of transketolases and disruption of the non-oxPPP attenuates S. Typhimurium.

Discussion
Data from the above experiments deliver multiple lines of evidence that S. Typhimurium possesses three functional transketolase isoenzymes, each with unique features and characteristics. Prior work has clearly demonstrated that regulation of E. coli transketolases depends on both growth phase and the alternative sigma factor RpoS (23, 29 -31). We observed a similar pattern of regulation in S. Typhimurium, most notably transcriptionally and translationally for TktB and transcriptionally for both genes of TktC. Interestingly, protein levels for all three transketolases in S. Typhimurium were at or near their highest expression during stationary phase. During stationary phase in Gram-negative bacteria, starvation conditions trigger RpoS, which subsequently promotes expression of genes that prepare the cell for survival (32)(33)(34). Additionally, RpoS induces cross-protection against diverse stresses, including osmotic, temperature, acid, oxidative, and nitrosative (4,(35)(36)(37)(38)(39). The contribution of RpoS to combatting stressful conditions is linked to its essential role in Salmonella virulence (40). Although a direct role for transketolases in virulence or resisting stress conditions has not been described previously, our results demonstrate that TktB and TktC are a part of the RpoS regulon, indicating that they may have a part in these cellular functions. The ability of each isoenzyme to catalyze the transketolase reaction highlights the overlapping and comple-  Table 1. X5P, xylulose 5-phosphate; R5P, ribose 5-phosphate; E4P, erythrose 4-phosphate.   Table 1. X5P, xylulose 5-phosphate; R5P, ribose 5-phosphate; E4P, erythrose 4-phosphate.  Table 1. X5P, xylulose 5-phosphate; R5P, ribose 5-phosphate; E4P, erythrose 4-phosphate.

Three transketolase isoenzymes in S. Typhimurium
mentary functions of each enzyme; however, the differences in expression and enzyme kinetics suggest that each S. Typhimurium transketolase may have a unique role in this bacterium. The reasons underlying why S. Typhimurium maintains three transketolases remain unclear. However, given that S. Typhimurium transits through varied and stressful environments, we predict that the individual transketolases in S. Typhimurium will each make unique contributions to cellular physiology. Even though TktA is responsible for the majority of transketolase activity in S. Typhimurium, both TktB and TktC are independently sufficient to allow growth under laboratory conditions. A similar phenomenon has been observed in E. coli, where up to 90% of transketolase activity was attributable to TktA (41). In the same study, a ⌬tktA mutant demonstrated slow growth on MacConkey agar containing ribose, whereas ⌬tktA ⌬tktB mutant showed no growth. Similarly, our study demonstrated that in the S. Typhimurium Tkt 0 strain complemented with tktC, even the low levels of transketolase activity were sufficient to allow growth in minimal medium with ribose as the sole carbon source (Table S7 and Fig. 5C). In other species for which transketolase kinetic constants have been determined, k cat values can be similar to 10-fold lower, as is the case with Saccharomyces cerevisiae, or more than 3 orders of magnitude lower, as in Bacillus methanolicus (42,43). These observations raise questions about the minimal level of transketolase activity required for growth and why WT S. Typhimurium utilizes three transketolase isoenzymes that exhibit high levels of transketolase activity. One possible answer to these questions is that transketolases are necessary for conditions other than routine growth, and each transketolase has unique properties. Our data indicate that transketolases are required for S. Typhimurium virulence, but polar effects on other virulence pathways resulting from transketolase dysregulation and the contributions of individual transketolases remain to be tested. Even though TktA and TktB are 75% identical and 86% similar, these two transketolases are differentially regulated by growth phase and RpoS. It is plausible that these two transketolases have their main effects during different growth phases or that TktB may be more beneficial during stressful conditions encountered during infection. An additional possibility is that TktA may be the most efficient enzyme and responsible for routine growth, but it may be more susceptible to oxidative and/or nitrosative stresses than the other isoenzymes, TktB and/or TktC. Such a situation is not without precedent among bacterial metabolic isoenzymes, as the two aconitase enzymes of E. coli, AcnA and AcnB, demonstrate opposing sensitivities to oxidative stress (44). It has also been shown that in Arabidopsis thaliana, transketolase and ribose-5-phosphate isomerase, both of which are enzymes of the non-oxPPP, are susceptible to nitration and nitrosylation (45)(46)(47). Additionally, the abundance of these enzymes increases when cellular levels of S-nitrosylation increase. S. cerevisiae mutants deficient in transaldolase, another enzyme of the non-oxPPP, were more sensitive to oxidative stress in a mechanism that appeared to be independent of NADPH, highlighting the importance of the non-oxPPP during stressful conditions (48).
S. Typhimurium transketolases are intriguing not only because the organism possesses three functional isoenzymes, but also because the third transketolase, TktC, is the product of two genes. This is, to our knowledge, the first time such a transketolase enzyme has been analyzed in prokaryotes. BLAST analysis did not identify any homologues of TktC in prokaryotes that were encoded by a single gene; however, the recombinant form of TktC engineered with the two genes fused together produced a functional and efficient transketolase. The observation that S. Typhimurium and a wide variety of other Enterobacteriaceae have maintained the two-gene arrangement suggests that this organization confers some advantage to the organisms. In S. Typhimurium, the two genes for TktC are highly divergent from both tktA or tktB, and even from STY2570-STY2571 in S. Typhi (Tables S3-S5), but the reasons underlying this genetic arrangement are unclear. By deducing the specific conditions under which TktC is most necessary and useful to the organism, we hope to reveal clues regarding why Salmonella and other organisms have this arrangement of transketolase isoenzymes. Ultimately, this study provides evidence for three transketolase isoenzymes that contribute to pathogenesis of S. Typhimurium, each with its own unique properties and characteristics. The results offer a foundation for future investigations into the functions of transketolases and other enzymes of the PPP.

Classification of transketolase domains
The NCBI-ProteinID for each S. Typhimurium transketolase (TktA, ACY90124; TktB, ACY89466; STM14_2886/TktC 1 , ACY89325; STM14_2885/TktC 2 , ACY89324) was used as the query sequence to search the NCBI protein database. Each sequence was then subjected to analysis to identify conserved domains (49 -52). Specific residues predicted to be involved in dimerization and TPP binding were also classified this way.

Bacterial strains and plasmids
S. enterica serovar Typhimurium strain ATCC 14028s was used as the WT strain and as the background for mutant con-

Three transketolase isoenzymes in S. Typhimurium
struction for this study. Mutant strains (Table S1) were created following the method described by Datsenko and Wanner (53). Primers with homologous extensions that corresponded to the gene to be knocked out were used to amplify a kanamycin resistance gene flanked by flippase recognition targets (FRTs) from the plasmid pKD13 using TaKaRa LA Taq high-fidelity polymerase (Thermo Fisher Scientific) ( Table S2). The resulting product was purified and concentrated using the Thermo Fisher Scientific GeneJET PCR purification system (Thermo Fisher Scientific) and was transformed into competent S. Typhimurium strain TT22236 containing the plasmid pTP223 that expresses an isopropyl-␤-thiogalactopyranosideinducible -Red recombinase enzyme. Mutations were transduced into WT S. Typhimurium using P22 phage. Pseudolysogens were eliminated via streaking of colonies on Evans blue uranine agar to screen for phage-free cells. Excision of the kanamycin resistance cassette was achieved by electroporating strains with the temperature-sensitive plasmid pCP20, which encodes a FLP recombinase. The FLP recombinase catalyzes recombination of the FRT sites flanking the kanamycin resistance cassette, thereby removing the cassette and leaving a single FRT scar. PCR analysis using primers outside of the gene of interest was used to confirm deletion of the gene.
Complementation and expression vectors were engineered in pBAD plasmids using the specified primers (Table S2). WT alleles of genes were inserted into the pBAD multiple cloning site using the indicated restriction enzymes and subsequent ligation. Resulting constructs were under the control of an arabinose-inducible pBAD promoter with an N-terminal His 6 tag. A tktC 1 -tktC 2 fusion without overlapping genes was created using seamless cloning (Invitrogen). The start and stop codons were removed from both tktC 1 and tktC 2 , and tktC 2 was moved to be in-frame immediately downstream of tktC 1 (Table  S2). The resulting plasmid construct, designated pBAD::tktC*, consisted of the fused genes under an arabinose-inducible promoter with an N-terminal His 6 tag.

Bacterial growth
Strains were inoculated from frozen bacterial stocks and grown to stationary phase overnight (20 h) in LB broth at 37°C with constant agitation. Overnight cultures of strains were diluted 1:500 or 1:250 (as indicated) into fresh LB or M9 minimal medium supplemented with 0.2% D-glucose or 0.2% D-ribose, and optical density was determined by measuring absorbance at 600 nm (A 600 ) every 15 min while incubated at 37°C with constant agitation in a Cytation 5 microplate reader (Biotek, Winooski, VT). Blank values for each condition were obtained by subtracting the absorbance at A 600 of media alone from the absorbance at A 600 of the bacterial cultures. For gene and protein expression analysis and for transketolase activity assays, overnight cultures were subcultured 1:500 into fresh LB and grown at 37°C with constant agitation to mid-logarithmic phase (A 600 ϭ 0.5) or stationary phase (20 h).

Gene expression
Gene expression was measured in three or four independent samples from WT and ⌬rpoS::FRT S. Typhimurium at logarithmic and stationary phases. At the indicated time point, 1.5 ml of bacterial culture was harvested by centrifugation for 1 min at 12,000 ϫ g. Culture supernatant was removed, and the cell pellet was resuspended in 0.5 ml of RNazol (Molecular Research Center, Inc., Cincinnati, OH) and processed following the manufacturer's protocol for total RNA isolation. DNA was depleted from samples using Ambion TurboFree DNase (Thermo Fisher Scientific), and RNA was purified using a GeneJet RNA purification and concentration minikit (Thermo Fisher Scientific). RNA was quantified spectrophotometrically using the Take3 application and Cytation 5 microplate reader (BioTek), and 1.5 mg of each sample was reverse transcribed using the High Capacity cDNA Synthesis kit (Applied Biosystems, Foster City, CA) following the manufacturer's protocol. An equal amount of sample was diluted in buffer for no reverse transcriptase controls. Two technical replicates of each sample were analyzed by qRT-PCR (25 ng/reaction, total volume 20 l) with Bullseye EvaGreen Master Mix (MidSci, St. Louis, MO) using the CFX Connect system (Bio-Rad) with the following cycling parameters: 1) 95°C for 10 min; 2) 95°C for 3 s and 60°C for 30 s; repeated 40 times. Dilutions of S. Typhimurium genomic DNA were used to measure primer efficiency for each gene that was assessed, and all sets had greater than 97% efficiency (Table S2). RNA quality was assured by comparing Ct values for primers at two locations in the 5Ј region of rpoD. Amplification specificity was determined by melt curve analysis following qRT-PCR. Relative expression of each gene was calculated as relative to the housekeeping sigma factor rpoD using the 2 Ϫ⌬⌬Ct method (54).

Western blotting
Salmonella strains grown to logarithmic or stationary phase in LB broth were harvested by centrifuging 1 ml of cells at 12,000 ϫ g for 1 min and resuspending the pellet in 500 l of lysis buffer (0.1 M Tris, 4% SDS, pH 7.8). Samples for TktA::FLAG expression analysis were diluted 1:10 due to high expression, but representative Western blotting images show undiluted samples. All samples were immediately boiled in sample loading buffer (Bio-Rad), and denatured proteins were separated on an Any kDa Mini-PROTEAN TGX Precast protein gel (Bio-Rad). Equal loading of each cell lysate was determined using the BCA assay kit with BSA as a standard (Thermo Fisher Scientific) and stain-free gel assessment before blotting proteins to a polyvinylidene difluoride membrane. Membranes were blocked with 5% milk in TBST for 2 h, washed, incubated 2 h with mouse ␣-FLAG antibody (Invitrogen, Catalogue no. MA121315, lot RK241385) diluted 1:1000 in TBST, washed, incubated 2 h with goat ␣-mouse-HRP (Invitrogen, catalogue no. 31460, lot SA245916) diluted 1:2000 in TBST, washed, and imaged using Western Sure Premium ECL (LI-COR, Lincoln, NE) with a ChemiDoc MP gel imaging system (Bio-Rad). Band intensity for each lane was determined using ImageLab software and local subtraction algorithms (Bio-Rad).

Transketolase activity assays
Cells were grown to logarithmic or stationary phase as described above with 100 g/ml ampicillin and 100 M isopropyl-␤-thiogalactopyranoside (Thermo Fisher Scientific) when appropriate. For logarithmic-and stationary-phase cells, 1 ml and 100 l of culture was harvested, respectively. Cells were Three transketolase isoenzymes in S. Typhimurium washed in 1 ml of PBS, pelleted by centrifugation, and resuspended in 500 l of 50 mM Tris, pH 7.5. Samples were freezethawed at Ϫ75°C and then sonicated on dry ice (5 s on, 10 s off, repeated four times). Samples were centrifuged at 12,000 ϫ g at 4°C for 10 min to remove cell debris. Supernatants were used for the assay as described previously (55) (Fig. S12). Briefly, protein concentration was determined using the BCA assay kit with BSA as a standard (Thermo Fisher Scientific). For the enzymatic reactions, each well contained a final volume of 100 l consisting of 50 mM Tris (pH 7.5), 0.24 mM MgCl 2 , 10 M thiamine pyrophosphate, 0.5 mM NADH, 3 units of glycerol 3-phosphate dehydrogenase, 10 units of triosephosphate isomerase, 0.5 mM D-xylulose 5-phosphate, and 0.5 mM D-ribose 5-phosphate or D-erythrose 4-phosphate (all reagents obtained from Sigma-Aldrich). The reaction proceeded at 37°C, and the decrease in A 340 nm was monitored with a Cytation 5 microplate reader (BioTek). One unit of enzyme was defined as the amount of enzyme that oxidized 1 mol of NADH per minute. Transketolase enzyme activity for cell lysates was expressed as units per milligram of total protein. Limits of detection (LOD) and quantification (LOQ) were determined from all replicate blank readings for each assay (56,57). LOD was calculated as mean plus three S.D. values; LOQ was calculated as mean plus 10 S.D. values.
Purified transketolase enzyme kinetics were measured using the same reaction as above, with the exception that one sugar substrate was present in excess while the other sugar co-substrate was present at the concentration indicated (Figs. 6 -8).
Reactions containing excess D-xylulose 5-phosphate had a concentration of 2 mM, reactions containing excess D-ribose 5-phosphate had a concentration of 10 mM, and reactions containing excess D-erythrose 4-phosphate had a concentration of 1 mM (all materials were obtained as described above, with the exception of D-xylulose 5-phosphate, which was purchased from Carbosynth (Compton, Berkshire, UK)). Enzymatic activity for each reaction was calculated as M NADH consumed/s.

Enzyme purification
Under physiological conditions, transketolases dimerize to form functional enzymes (16), and because of the high similarity between S. Typhimurium transketolase isoenzymes, we purified recombinant protein from the Tkt 0 (⌬tktA ⌬tktB ⌬tktC) background to eliminate the possibility of heterodimer formation and co-purification. Overnight cultures were subcultured 1:500 into fresh LB with ampicillin and allowed to grow to A 600 ϭ 0.5, at which time arabinose was added to a final concentration of 0.2%. Strains were grown for an additional 4 h, at which time 50 ml of culture was harvested via centrifugation. Supernatants were removed, and cell pellets were resuspended in 10 ml of equilibration/wash buffer (50 mM sodium phosphate, 300 mM sodium chloride, 10 mM imidazole, pH 7.4). For TktC* purification, equilibration/wash buffer also contained Protease Inhibitor Mixture (Sigma-Aldrich). Cell suspensions underwent five freeze-thaw cycles from Ϫ80 to 37°C. Lysozyme (Thermo Fisher Scientific) was added to each sample during thaw of cycle 4. After the last thaw, samples were incubated on dry ice while sonication was performed with the following conditions: 30 s on, 30 s off, repeated six times beginning at 50% and increasing by 10% each cycle. Cell debris was removed by centrifugation at 3,000 ϫ g for 30 min at 4°C. Supernatant was then added to 3 ml of HisPur cobalt spin columns (Thermo Fisher Scientific), and the manufacturer's protocol was followed for purification of His-tagged proteins using native buffer conditions, with the exception that purified protein was eluted using an increasing gradient of imidazole (10, 25, 50, 75, 100, 150, and 200 mM). The protein from the final elution was used for all subsequent assays. Protein concentration was determined by BCA assay using BSA as a standard (Thermo Fisher Scientific). The purity of each transketolase was determined by loading 150 -250 g of each sample into an Any kDa Mini-PROTEAN TGX stain-free gel (Bio-Rad) and analyzing the lane using ImageLab software (Bio-Rad).

Mouse infections
Eight-week-old C57BL/6 mice bred in the University of Colorado School of Medicine animal facility according to institutional animal care and use committee guidelines were used to assess the role of transketolases in Salmonella pathogenesis. Individual animals were inoculated intraperitoneally with ϳ200 cfu of Salmonella grown overnight to stationary phase in LB broth. Mouse survival was monitored for 3 weeks. Mice manifesting signs of distress (e.g. low spontaneous activity and ruffled coat) were humanely euthanized by CO 2 inhalation followed by cervical dislocation.

Statistical analysis
Data from gene and protein expression, as well as transketolase activity from cell lysates, were analyzed using GraphPad Prism (version 6.01) with one-way analysis of variance (ANOVA) followed by Tukey's post hoc test. Results were considered statistically significant (*) when p Ͻ 0.05. Enzyme kinetics were determined using GraphPad Prism (version 6.01) with the Michaelis-Menten least-squares (ordinary) fit method.