Defining the Metabolic Pathways and Host-Derived Carbon Substrates Required for Francisella tularensis Intracellular Growth

The widespread onset of antibiotic resistance prioritizes the need for novel antimicrobial strategies to prevent the spread of disease. With its low infectious dose, broad host range, and high rate of mortality, F. tularensis poses a severe risk to public health and is considered a potential agent for bioterrorism. F. tularensis reaches extreme densities within the host cell cytosol, often replicating 1,000-fold in a single cell within 24 hours. This remarkable rate of growth demonstrates that F. tularensis is adept at harvesting and utilizing host cell nutrients. However, like most intracellular pathogens, the types of nutrients utilized by F. tularensis and how they are acquired is not fully understood. Identifying the essential pathways for F. tularensis replication may reveal new therapeutic strategies for targeting this highly infectious pathogen and may provide insight for improved targeting of intracellular pathogens in general.

I n order to establish a successful infection, intracellular bacterial pathogens must adapt their metabolism to utilize the nutrients available within the host cell, often in direct competition with the host's own metabolic processes and mechanisms for nutrient sequestration (1). Nevertheless, many of these microorganisms have evolved dedicated mechanisms to harvest and assimilate essential nutrients to proliferate within this specialized niche (2)(3)(4). Targeted strategies for carbon acquisition and assimilation fuel bacterial replication and often aid in the evasion of host cell defenses (5)(6)(7). Despite their importance, the metabolic pathways and host-derived carbon sources utilized by bacterial pathogens in vivo are generally not well understood (8,9).
Metabolites can be directly acquired from the host, salvaged from similar molecules, or synthesized de novo using host-derived sources of carbon, nitrogen, sulfur, etc. Bacteria that replicate within the host cell cytosol theoretically have access to the products and intermediates produced during major host metabolic processes that take place within this compartment, including glycolysis and amino acid biosynthesis. The actual concentrations of these products within an infected cell, however, are unclear. Rather, most nutrients are stored within complex structures, such as lipid droplets, glycogens, and proteins, and thus are not immediately available to intracellular pathogens (8).
Many bacteria employ active mechanisms to acquire host-derived carbon during intracellular growth. Mycobacterium tuberculosis and Chlamydia tracomatis, for instance, associate with host lipid droplets and utilize host-derived lipids for anabolic and catabolic purposes (10,11). Salmonella enterica serovar Typhimurium secretes effector proteins that stimulate the activation of Akt, a major metabolic regulator of host metabolism (12,13). This, in turn, stimulates host glycolytic flux and increases the concentration of glucose within the infected cell (13). Similar effector molecules actively alter host vesicular trafficking to direct nutrients to the Salmonella-containing vacuole (14). The observation that many pathogens employ active mechanisms to obtain carbon emphasizes that carbon acquisition within the host cell requires complex host-pathogen interactions, which are only beginning to be elucidated.
We previously demonstrated that Francisella tularensis induces host autophagy during infection, and that this pathway provides the pathogen with essential amino acid metabolites (15). Nevertheless, F. tularensis replicates to a considerable degree in the absence of autophagy, indicating that autophagy-derived nutrients are only a subset of the total required to support full F. tularensis intracellular proliferation (15). A transposon mutagenesis screen of Francisella tularensis subsp. holarctica LVS revealed that nearly half of the genes identified as essential for proliferation in macrophages encode proteins involved in metabolism or metabolite transport (16). These proteins include enzymes predicted to facilitate gluconeogenesis, glycerol catabolism, and amino acid transport, as well as purine, lipopolysaccharide (LPS), and fatty acid biosynthesis. Surprisingly, no glycolytic genes were identified during this screen. Glycolysis is a fundamental metabolic pathway that oxidizes carbohydrates to generate energy and provide precursor metabolites for other biosynthetic pathways. In contrast, the gluconeogenic pathway reverses the reactions of glycolysis during growth on nonglucose carbon substrates to replenish stores of glucose-6-phosphate (glucose-6P) and other essential metabolic intermediates when glucose concentrations are limited. One gene encoding a key gluconeogenic enzyme, glpX, was required for efficient intracellular replication (16). Indeed, glpX has repeatedly been identified as an important factor for virulence in genetic screens performed in F. tularensis Schu S4 and LVS (17)(18)(19). Furthermore, recent work by Brissac et al. demonstrated that gluconeogenesis is an essential metabolic pathway for Francisella novicida and F. tularensis LVS during growth in glucose-limiting conditions (20). These data suggest that F. tularensis intracellular proliferation may not be dependent on glycolysis but rather on gluconeogenesis to preferentially assimilate nonglucose carbon substrates within a host cell.
To determine the specific host-derived carbon sources that facilitate rapid F. tularensis intracellular proliferation, we aimed to define the essential carbon metabolic pathways and metabolites required for F. tularensis intracellular and in vivo growth.

RESULTS
Gluconeogenesis, but not glycolysis, is essential for F. tularensis intracellular growth and virulence. Unlike most enzymatic reactions of the glycolytic pathway, the conversion between fructose 6-phosphate (F6P) and fructose 1,6-bisphosphate (FBP) is physiologically irreversible, and is catalyzed by enzymes specific to either glycolysis or gluconeogenesis. In F. tularensis, the glycolytic enzyme phosphofructokinase (PfkA) converts F6P to FBP, and the gluconeogenic enzyme fructose 1,6-bisphosphatase (GlpX) performs the reverse reaction ( Fig. 1). Deletion of pfkA should prevent F. tularensis from utilizing glucose or glucose 6-phosphate imported from the host, while deletion of glpX should prevent the bacterium from producing F6P during growth on gluconeogenic carbon sources. F6P is a precursor of the pentose phosphate pathway and is used for the de novo synthesis of lipopolysaccharide, peptidoglycan, pentose phosphates, and aromatic amino acids. We hypothesized that if glucose represents a major carbon source for F. tularensis within the host cell, then pfkA would be essential. Alternatively, if glucose is not a major source of carbon utilized by F. tularensis, then the gluconeogenic enzyme glpX would be required in order to synthesize sufficient F6P and glucose-6P from alternate carbon sources.
We first sought to confirm the predicted functions of pfkA and glpX for glycolysis and gluconeogenesis, respectively. Markerless, in-frame deletions were created for pfkA and glpX in Francisella tularensis subsp. tularensis Schu S4, and the deletion strains were grown in defined medium with either glycolytic or gluconeogenic carbon substrates. For all broth cultures, F. tularensis was grown in Chamberlain's defined medium (CDM) containing a low concentration (ϳ3mM) of 13 essential and nonessential amino acids and no other major carbon sources (21) (Text S1). In this medium, wild-type (WT) Schu S4 cells grew to low, but detectable levels, presumably by assimilating amino acids for protein synthesis or energy production ( Fig. 2A and Fig. S1A). Indeed, Brissac et al. recently demonstrated that supplementation of this medium with 30 mM select amino acids (threonine, proline, methionine, lysine, tyrosine, tryptophan, phenylalanine, asparagine, or serine) permits various degrees of F. novicida growth, suggesting that these amino acids may be utilized as carbon sources (20). Supplementation with either glucose or glutamate supported robust growth of WT Schu S4 cells ( Fig. 2A and Fig. S1A). A ΔpfkA mutant grew to WT levels during growth on glutamate, but did not grow on glucose, possibly due to glucose-mediated repression of alternative carbon catabolic pathways ( Fig. 2A). Importantly, though terminal optical density at 600 nm (OD 600 ) describes the overall capacity of each mutant to grow on different carbon substrates, closer attention to in vitro doubling times reveals subtle nuances in how each deletion affects growth. For instance, a ΔpfkA mutant reaches the same OD 600 as WT Schu S4, but grows at a much lower rate (Fig. S1B). As expected, the ΔglpX mutant grew on glucose but not on glutamate ( Fig. 2A and Fig. S1C). The growth defects of each mutant were restored to WT levels when the deleted genes were complemented in trans ( Fig. 2A and Fig. S1A to C).
To assess the importance of glycolysis and gluconeogenesis during F. tularensis intracellular growth, we utilized a luminescence reporter to monitor intracellular growth, as previously described (16), where F. tularensis Schu S4 strains harbor a plasmid expressing luciferase enzyme and substrate, as well as an addiction system to maintain the plasmid even in the absence of antibiotic selection. As demonstrated previously, an increase in the bacterial burden within the infected cell is directly proportional to an increase in reporter luminescence (16). Bone marrow-derived macrophages (BMDMs) were infected with WT Schu S4, the ΔpfkA mutant, or the ΔglpX mutant, each strain harboring the luminescence reporter. Twenty-four hours postinoculation, WT Schu S4 and ΔpfkA mutant cells grew to similar levels within the BMDMs, while the growth of the ΔglpX mutant was reduced approximately 10-fold relative to that of the WT and the ΔpfkA mutant (Fig. 2B). These data indicate that gluconeogenesis, but not glycolysis, is necessary for WT levels of F. tularensis intracellular growth and suggests that glucose does not represent a major carbon source within macrophage cells.
We hypothesized that the severe intracellular growth defect observed for the ΔglpX mutant was due to the mutant's inability to synthesize sufficient levels of F6P and G6P from the catabolism of gluconeogenic carbon substrates. Therefore, supplementation with excess glucose should rescue ΔglpX mutant growth within cells. J774A.1 cells are a transformed macrophage cell line constitutively expressing c-Myc, and therefore import large quantities of glucose to increase glycolytic flux (22,23). To determine if excess glucose could restore the growth defect of the ΔglpX mutant, we infected J774A.1 cells with WT Schu S4, the ΔpfkA mutant, and the ΔglpX mutant, supplied the infected cells with either high-glucose (4.5 g/liter) or glucose-free Dulbecco minimal essential medium (DMEM), and then measured bacterial growth over a 36-hour period. WT Schu S4 and ΔpfkA mutant cells exhibited significant growth with or without glucose supplementation (Fig. S2A). As expected, the ΔglpX mutant strain did not replicate within J774A.1 cells cultured in glucose-free DMEM; however, intracellular replication was restored to WT levels with excess glucose (Fig. S2B).
The rescue of the ΔglpX mutant did not occur in primary BMDMs, as all BMDM infections were performed in high-glucose (4.5g/liter) DMEM (Fig. 2B). This observation suggests that the reduced level of glucose import and glycolytic flux exhibited by BMDMs, relative to those of J774A.1 cells, is insufficient to permit ΔglpX cells from acquiring adequate glucose from the host to restore WT growth properties, even when glucose is present at high concentrations in the medium. To test this, we attempted to rescue growth of ΔglpX in BMDMs by treating the host cells with 5-amnoimidazole-4carboxamide ribonucleotide (AICAR). AICAR is an analog of AMP (AMP) that stimulates activation of the major host metabolic regulator, AMP-dependent protein kinase (AMPK) (24). When activated, AMPK stimulates glucose uptake and energy production in part by increasing expression of major glucose transporters GLUT1 and GLUT4, and by increasing overall host glycolytic flux (25). We hypothesized that AICAR treatment of BMDMs cultured in high-glucose DMEM would restore ΔglpX mutant intracellular growth by stimulating glucose import. Indeed, while AICAR had little impact on the growth of WT Schu S4 within BMDMs (Fig. 2C), AICAR treatment significantly increased the intracellular growth of ΔglpX in BMDMs cultured in high-glucose DMEM (Fig. 2D). Altogether, our results support the conclusion that the inability of the ΔglpX mutant to fully assimilate gluconeogenic carbon sources results in attenuated growth during periods of glucose limitation.
We next tested whether F. tularensis similarly requires glpX and not pfkA for replication in a murine model of F. tularensis pulmonary infection. Groups of C57BL6/J female mice were infected intranasally with 100 CFU of WT, ΔpfkA, or ΔglpX Schu S4 strains. Three days postinfection, the lungs, livers, and spleens of the infected mice were harvested, homogenized, and plated for bacterial enumeration. Organ burdens for the ΔpfkA mutant strain were similar to those of WT Schu S4 (Fig. 2E). However, the number of CFU recovered from the lungs of mice infected with the ΔglpX mutant was similar to that of the original inoculum and below the limit of detection in the liver and spleen (Fig. 2E). These data align with our observations that glpX, and therefore gluconeogenesis, is necessary for F. tularensis replication in host cells, whereas pfkA and glycolysis are dispensable.
F. tularensis possesses multiple pathways that supply gluconeogenic substrates to support intracellular growth. Our data suggest that a ΔglpX mutant does not produce essential biosynthetic precursors from the nutrients available within the host cell. Since the deletion of glpX precludes the utilization of a large number of gluconeogenic carbon sources, such as glycerol, pentose sugars, amino acids, lactate, pyruvate, and tricarboxylic acid (TCA) cycle intermediates, we generated F. tularensis mutant strains unable to utilize some of these specific carbon sources. Like the conversion of F6P to FBP, the enzymatic conversion of pyruvate to phosphoenol pyruvate (PEP) during glycolysis is physiologically irreversible and must be bypassed during gluconeogenesis. The ATP-dependent decarboxylation of oxaloacetate to PEP is catalyzed by PEP carboxykinase (pckA) and is important for growth on carboxylic or amino acids. Alternatively, pyruvate-phosphate dikinase (ppdK) converts pyruvate to PEP and is required for growth on pyruvate, lactate, and some amino acids. The reactions catalyzed by each enzyme can independently fuel the gluconeogenic pathway to generate essential metabolic precursors necessary for growth. (Fig. 1).
We generated markerless, in-frame deletions of ppdK and pckA. Growth characteristics of each mutant were analyzed in defined medium with specific glycolytic or gluconeogenic substrates to confirm the metabolic function of each enzyme. Both ΔppdK and ΔpckA mutants grew to levels similar to that of WT Schu S4 in CDM supplemented with glucose ( Fig. 3A and B). However, while the ΔpckA mutant grew to WT levels in CDM with or without excess glutamate, the ΔppdK mutant had a severe growth defect in CDM and in CDM supplemented with glutamate, similar to that of the ΔglpX mutant ( Fig. 3A and B). These data suggest that during growth in defined medium, F. tularensis preferentially synthesizes PEP from pyruvate (PpdK) and not from oxaloacetate (PckA).
We observed that both the ΔppdK and ΔpckA mutants grew to WT levels within the BMDMs (Fig. 3C). Furthermore, a ΔppdK ΔpckA double mutant replicated to significant levels within these cells, albeit at a lower rate, suggesting that these gluconeogenic pathways are not essential for F. tularensis growth within BMDMs (Fig. 3C). When we infected J774A.1 cells with the ΔppdK and ΔpckA mutants, we found that the ΔpckA mutant grew to WT levels within J774A.1 cells cultured with or without glucose supplementation (Fig. S2C). The ΔppdK mutant, however, exhibited significantly reduced growth within J774A.1 cells cultured without glucose (Fig. S2D). Intracellular proliferation of the ΔppdK mutant was restored to WT levels upon high glucose supplementation, indicating that ppdK may contribute to the assimilation of hostderived carbon in J774A.1 cells.
While we found that the organ burdens of the ΔppdK mutant were similar to those of WT Schu S4 in our murine model, we recovered significantly reduced numbers of the ΔpckA mutant from the lungs, livers, and spleens of infected mice (Fig. 3D). Furthermore, we recovered similar numbers of the ΔppdK ΔpckA double mutant relative to that of the ΔpckA single mutant. This indicates that pckA is required for optimal replication in a murine model of F. tularensis infection but that ppdK is dispensable (Fig. 3D).
Amino acids feed the gluconeogenic pathway through the TCA cycle. The attenuation of the ΔpckA mutant in mice suggests that F. tularensis relies on the metabolic pathway catalyzed by PckA during infection. Potential nutrients that can fuel the gluconeogenic pathway through PckA include TCA cycle intermediates or amino acids that feed into the TCA cycle. To discern between these possibilities, we evaluated the importance of glutamate dehydrogenase (gdhA) for F. tularensis intracellular growth. GdhA catalyzes the reversible oxidative deamination of glutamate to ␣-ketoglutarate, a TCA cycle intermediate (Fig. 1). F. tularensis is predicted to require GdhA to shuttle several amino acids into the TCA cycle, including glutamate, glutamine, proline, arginine, and potentially aspartate and asparagine. Therefore, if F. tularensis preferentially catabolizes amino acids and not TCA cycle intermediates, then a ΔgdhA mutant would likely be similarly attenuated relative to ΔpckA during in vivo growth.
To validate the predicted function of gdhA, we tested a ΔgdhA mutant for growth on glycolytic and gluconeogenic carbon sources in defined medium. As expected, we found that gdhA was required for growth in CDM or CDM supplemented with glutamate, but not in CDM supplemented with glucose (Fig. 4A). Because the ΔgdhA mutant grew to significant levels on glucose in defined media lacking glutamate, we reasoned that gdhA was dispensable for glutamate synthesis but was required for glutamate assimilation.
The ΔgdhA deletion mutant exhibited reduced growth in BMDMs that was restored upon expression of gdhA in trans (Fig. 4B), suggesting that GdhA-mediated carbon assimilation represents an important metabolic pathway during F. tularensis replication in BMDMs. When we infected J774A.1 macrophage cells with the ΔgdhA mutant, we found that the defect in bacterial intracellular replication during culture in glucose-free DMEM could be partially rescued with excess glucose, similar to that of the ΔglpX and ΔppdK mutants (Fig. S2E). Similarly, BMDMs cultured in high-glucose DMEM and treated with AICAR permitted significant growth of the ΔgdhA mutant relative to that in untreated BMDMs (Fig. 4C). These data suggest that the intracellular growth defect observed for the ΔgdhA mutant is at least in part due to the ability of this mutant to assimilate sufficient host-derived carbon.
We expected the ΔgdhA mutant to be similarly attenuated relative to a ΔpckA mutant during growth in mice. Strikingly, when we assessed the requirement of gdhA for growth in our murine model of F. tularensis pulmonary infection, we found that the CFU recovered from the lungs, livers, and spleens of gdhA-infected mice were greatly reduced compared to those from the WT, and that the CFU recovered from the liver and spleen of the mice were reduced approximately 3-fold relative to those from the ΔpckA mutant (Fig. 4D). In addition to fueling the gluconeogenic pathway, ΔgdhA-mediated anaplerosis of the TCA cycle may be essential during infection to supply other essential metabolic precursors (e.g., oxaloacetate and/or acetyl coenzyme A [acetyl-CoA]) or to generate reducing power via the use of malic enzyme. This conclusion is consistent with our observation that excess glucose supplementation during growth in J774A.1 or AICAR-treated BMDMs only partially rescued bacterial proliferation of this mutant.
Glycerol catabolism is required for F. tularensis in vivo growth. Since the ΔglpX mutant was more severely attenuated in mice relative to the ΔppdK ΔpckA mutant, we reasoned that F. tularensis may assimilate additional carbon substrates besides those supplied through the gluconeogenic pathways catalyzed by Ppdk and PckA. In F. tularensis, glpA (G3P dehydrogenase) is predicted to be required for the catabolism of glycerol and G3P. We used the Targetron gene knockout system modified for use in Francisella to disrupt glpA in F. tularensis Schu S4 (26). Interestingly, we found that the generation of a ΔglpA mutant strain was only possible through the simultaneous introduction of a secondary mutation in glpK (Fig. S3A). In F. tularensis, glpK is located upstream of glpA and is predicted to encode a kinase responsible for the phosphorylation of glycerol forming G3P during glycerol catabolism (Fig. S3A). The disruption of G3P dehydrogenase in the presence of a fully functional glycerol kinase can lead to increased concentration of intracellular G3P. In Escherichia coli, excess G3P within the cell stimulates the synthesis of the toxic metabolite methylglyoxal (27). We suspect that a similar phenomenon may be responsible for the requirement of a secondary glpK mutation in an F. tularensis ΔglpA background.
We analyzed the growth properties of the glpKA disruption mutant in defined medium supplemented with glucose, glycerol, or G3P to confirm that glpA and glpK are required for growth on glycerol and G3P. As expected, the glpKA mutant grew to WT levels when cultured with glucose but not with glycerol or G3P (Fig. 5A). In fact, supplying the glpKA mutant with G3P led to significantly lower levels of bacterial replication relative to growth on glycerol or just CDM, possibly due to the toxic buildup of intracellular G3P. We found that growth of the glpKA mutant was restored on glycerol and G3P only when these two genes were expressed with the downstream  gene, glpF, despite the fact that sequencing of the surrounding genes in our glpKA mutant revealed no additional mutations in or around the coding sequence for glpF. glpF is predicted to encode a glycerol uptake facilitator and may be cotranscribed with glpA (Fig. S3A). As expected, we observed WT levels of growth on G3P but not on glycerol when glpA and glpF, but not glpK, were expressed in trans in the glpKA mutant (Fig. S3B), and growth on both glycerol and G3P was restored when the glpKA mutant was complemented with glpKAF in trans (Fig. 5A). These findings are summarized in Table S1.
The glpKA mutant replicated in J774A.1 cells to intermediate levels with or without supplemented glucose, indicating that this mutant can replicate within this cell line, but growth was not fully restored by the addition of excess glucose (Fig. S2F). We found that the glpKA disruption mutant did not replicate within BMDMs (Fig. 5B). Interestingly, growth within BMDMs was restored to WT levels upon in trans expression of glpA and glpF without glpK, suggesting that F. tularensis Schu S4 may assimilate G3P and not glycerol during intracellular growth (Fig. S3C). Growth of the mutant within BMDMs was similarly restored to WT levels upon complementation of glpKAF (Fig. 5B). Finally, replication of the glpKA mutant was significantly increased within BMDMs cultured with AICAR and excess glucose, demonstrating that, similar to the glpX mutant, the glpKA disruption mutant could be rescued by supplying an alternative carbon source (Fig. 5C).
We then assessed the importance of glycerol catabolism for F. tularensis during growth in mice and found that the number of CFU recovered from the lungs of mice infected with the glpKA disruption mutant was similar to the original inoculum and below the limit of detection in the livers and spleens of mice (Fig. 5D). These data suggest that a glpKA mutant colonized, but did not proliferate or disseminate, in a murine model of F. tularensis infection. F. tularensis Intracellular Carbon Metabolism ® Data from our mutational analysis suggest that glycerol represents an essential host-derived source of carbon during F. tularensis intracellular growth. However, we could not exclude the alternative possibility that F. tularensis attenuation may be due to a toxic buildup of metabolites or disruption of proper metabolic regulatory mechanisms in our mutant strains. To delineate these possibilities, we sought to reduce the concentration of available glycerol within BMDMs and to examine the impact on WT F. tularensis intracellular proliferation. A significant bulk of host glycerol stores are sequestered as triglycerides in host lipid droplets (28). During lipolysis, a series of enzymatic reactions free glycerol from cellular triglyceride stores and release it into the cytosol of the host (28,29). We hypothesized that F. tularensis may exploit this process to establish a source of glycerol during intracellular growth. Atglistatin is a selective inhibitor of adipose triglyceride lipase (ATGL), an enzyme responsible for the first catalytic step of lipolysis (30). When we infected Atglistatin-treated BMDMs with WT F. tularensis, we observed that Atglistatin treatment significantly reduced F. tularensis intracellular burden in a dose-dependent manner (Fig. 6A). Importantly, these concentrations were not cytotoxic to BMDMs (Fig. S4A). To verify these findings, we used Cre-Lox recombination to generate ATGL deficient BMDMs. BMDMs derived from C57Bl6/J or ATGL-flox mice were treated with Cre recombinase gesicles during differentiation. Cre-treated BMDMs isolated from ATGL-flox mice demonstrated approximately 60% knockdown of ATGL expression based on reverse transcription-quantitative PCR (qRT-PCR) (Fig. S4B). This was associated with a significant reduction in F. tularensis replication within ATGL knockdown BMDMs (Fig. 6B). From these data, we conclude that host lipolysis is important for sustaining F. tularensis growth, and that host-derived glycerol represents a primary source of carbon necessary for fueling F. tularensis in vivo replication.

DISCUSSION
Previous work by our group and others highlight the importance of amino acid metabolism for F. tularensis replication and virulence (15,(31)(32)(33). Furthermore, Brissac et al. recently demonstrated that gluconeogenesis is vital for F. tularensis subsp. holarctica LVS and F. novicida growth during periods of glucose limitation (20). Here, we have similarly demonstrated that gluconeogenesis is essential for intracellular and in vivo growth for the highly virulent F. tularensis subsp. tularensis Schu S4, while pfkA, and thus glycolysis, is dispensable. Additionally, through systematic mutational analysis, we identified specific metabolic pathways essential for F. tularensis virulence. We found that ΔglpX, ΔpckA, ΔgdhA, and ΔglpA mutant strains were attenuated during growth in a mouse model of F. tularensis pulmonary infection, suggesting that these pathways may by critical for the efficient assimilation of host-derived carbon. These findings are summarized in Table S2. The metabolic pathways required for F. tularensis growth varied based on the infection model. We found that pckA was important for growth in mice, while ppdK was essential for WT levels of growth within a J774A.1 transformed macrophage cell line. The differential requirements of these genes suggest that F. tularensis may utilize alternate gluconeogenic pathways for growth in different environments, as the bacterium may preferentially assimilate different host-derived carbon sources, perhaps based on availability. As transformed macrophages undergo altered metabolism relative to primary cells, it is likely that the carbon sources available to F. tularensis are distinct within these models. For instance, J774A.1 metabolism is subject to the "Warburg effect," in which these cells significantly increase glucose uptake and aerobic glycolysis, leading to high intracellular concentrations of lactate (34). F. tularensis may exploit this metabolic aberrance and primarily assimilate lactate during replication within these cells. As ppdK is required for F. tularensis assimilation of lactate (Fig. 1), this may explain the requirement of ppdK specifically in J774A.1 cells.
We found that ppdK, and not pckA, is essential for growth on glutamate in defined medium. F. tularensis possesses an additional gluconeogenic enzyme (malate dehydrogenase [MaeA]) responsible for the synthesis of pyruvate from malate, which can then be converted to PEP through PpdK (Fig. 1). Previous work has suggested little or no utilization of the oxidative branch of the pentose phosphate pathway during F. tularensis growth (20). Bypassing the oxidative branch of the pentose phosphate pathway means that F. tularensis must use an alternative mechanism for the generation of the essential cofactor, NADPH. It is possible that the bacterium relies on an NADP ϩdependent malic enzyme for the production of NADPH during growth on glutamate defined medium. As the conversion of TCA intermediates to PEP through malic enzyme bypasses PckA but requires PpdK, this would provide a possible explanation for why ppdK and not pckA is the preferred gluconeogenic pathway during growth on glutamate.
We were surprised to find that a ΔgdhA mutant demonstrated significantly reduced growth within a mouse compared to a ΔppdK ΔpckA double mutant. If gdhA is required solely for gluconeogenic purposes, we would expect that these two mutants would be similarly attenuated, as a ⌬ppdK ΔpckA double mutant theoretically halts the gluconeogenic conversion of TCA cycle intermediates to glucose. However, during replication within a mouse, gdhA may be additionally required for anaplerosis of the TCA cycle or glutamate biosynthesis. Furthermore, it was recently demonstrated that glutamate import plays a critical role in oxidative stress defense and phagosomal escape during F. tularensis infection (32). Thus, the attenuation of this mutant may be in part due to its inability to withstand oxidative stress within the phagosome to reach the host cell cytosol. However, because growth of ΔgdhA can be partially rescued by supplying J774A.1 cells (Fig. S2E) or AICAR-treated BMDMs (Fig. 4C) with excess glucose, we conclude that this pathway is primarily involved in carbon acquisition during F. tularensis intracellular growth.
Unlike ppdK and pckA, we found that a glpKA mutant was attenuated for growth in all models tested, highlighting the importance of glycerol catabolism for F. tularensis pathogenesis. Based on the annotated genomic sequence of F. tularensis subsp. tularensis Schu S4, a ΔglpA mutant strain cannot assimilate glycerol or G3P (35). During our investigation, we found that disrupting glpA in F. tularensis Schu S4 resulted in an independent polar mutation in glpK that prevented growth on glycerol. As expected, genetic complementation of our glpA mutant strain with glpA, but not with glpK, rescued growth on G3P but not on glycerol (Fig. S3C). However, our partially complemented strain replicated to WT levels within BMDMs, suggesting that within this cell type, G3P and not glycerol is available for F. tularensis metabolism. This conclusion is consistent with the fact that glycerol is actively phosphorylated by the host to prevent its efflux from the cell. Of note, unlike F. tularensis subsp. tularensis and F. novicida, F. tularensis subsp. holarctica can only metabolize G3P and not glycerol. As F. tularensis possesses a small, decaying genome adapted to an intracellular lifestyle, this may reflect an interesting evolutionary example that supports our prediction that F. tularensis specifically metabolizes G3P within the cell (36).
Despite occupying similar niches, intracellular bacterial pathogens have evolved distinct methods to meet their respective nutritional requirements. Many pathogens, such as Salmonella enterica, Legionella pneumophila, and enteroinvasive Escherichia coli species, preferentially assimilate glucose during intracellular growth (5,37,38). In contrast, Shigella flexneri downregulates genes involved in glucose catabolism and favors the assimilation of C 3 substrates during growth within the cytosol (39). Listeria monocytogenes relies on two major carbon substrates (glycerol and glucose 6phosphate) to fuel distinct catabolic and anabolic pathways during cytosolic replication (40). Our data suggest that the primary carbon substrate utilized by F. tularensis during intracellular growth varies depending on the model of infection. This is not surprising, considering that the host range of F. tularensis subsp. tularensis Schu S4 includes over 250 species, and that within these hosts, F. tularensis infects numerous cell types, including macrophages and dendritic, endothelial and epithelial cells (41). In order to replicate within such a diverse range of hosts, F. tularensis must adapt its metabolism to the carbon sources available from the environment, which can vary significantly from host to host and between cell types. Thus, we suspect that the extraordinary ability of F. tularensis to proliferate within such a wide range of hosts is in part due to the pathogen's capability to sense and adapt to the fluctuating availability of nutrients over the course of its infectious lifestyle.
When available, F. tularensis will consume glucose. The intracellular growth defect of the ΔglpX mutant in J774A.1 cells was rescued by supplying excess glucose (Fig. 3B). Furthermore, F. tularensis subsp. holarctica LVS replication within J774A.1 and THP-1 macrophage cells leads to a significant reduction in host intracellular glucose (20). However, the nutrient concentrations within these established cell lines do not reflect the physiological conditions encountered by F. tularensis during infection, and it is likely that in physiological conditions, glucose limitation forces F. tularensis to utilize nonglucose carbon substrates. Indeed, a transcriptomic analysis of the F. tularensis metabolic network during extracellular and intracellular growth suggests that significant changes in carbohydrate metabolism occur when the pathogen transitions to an intracellular lifestyle (42). Our data support the proposed model that in the absence of glucose, F. tularensis will primarily utilize alternate carbon sources, such as amino acids or C 3 substrates derived from the host.
Bacterial metabolic pathways must be coordinated to reduce unnecessary energy expenditure and maximize fitness. In E. coli, key branch points in the glycolytic pathway are controlled by feed-forward/feedback inhibition. For instance, the conversion of F6P to FBP by PfkA is stimulated by ADP and inhibited by the downstream metabolite PEP (43). Conversely, the reverse reaction (catalyzed by fructose 1,6-bisphosphatase) is inhibited by AMP and glucose-6P (44). Carbon catabolite repression is poorly understood in F. tularensis; however, instances of catabolite repression have been described in other Gammaproteobacteria, including Pseudomonas aeruginosa and S. Typhimurium (45). Therefore, it is likely that F. tularensis also employs regulatory mechanisms to inhibit the utilization of alternative carbon substrates in the presence of a preferred carbon source such as glucose. We observed significant growth attenuation for a pfkA mutant in CDM supplemented with glucose relative to that in CDM alone or in CDM with glutamate. Similarly to that in E. coli, the buildup of glucose-6P may allosterically inhibit the activity of GlpX and prevent growth on gluconeogenic carbon sources, such as glutamate or other amino acids that are present at low concentrations in the medium.
Central carbon metabolism represents arguably the single most important cellular process in the context of bacterial viability and virulence. Energy generation, precursor biosynthesis, virulence factor expression, cell division, etc. are all contingent on a bacterium's ability to acquire and utilize sufficient carbon to fuel these processes. Targeting bacterial catabolic and anabolic pathways is a promising strategy for combating pathogenic organisms such as F. tularensis. Indeed, it is well established that F. tularensis purine auxotrophs are attenuated during infection, and these mutants have been suggested as potential candidates for use as a live vaccine (46,47). Similarly, targeting other essential metabolic pathways, such as gluconeogenesis, glycerol catabolism, or amino acid catabolism, either through drug or vaccine development, may constitute a means for limiting the spread of this deadly pathogen. Overall, by identifying the specific metabolic pathways and nutrients utilized by F. tularensis during intracellular growth, our findings begin to unravel the complex host-pathogen relationship exploited by F. tularensis during infection and further our understanding of F. tularensis pathogenicity.

MATERIALS AND METHODS
Bacterial and cell culture. Francisella tularensis subsp. tularensis Schu S4 was obtained from BEI Resources and maintained in a biosafety level 3 (BSL-3) facility. Detailed protocols for bacterial and cell culture maintenance and manipulation are described in Text S1.
Plasmid vectors and bacterial genetics. Markerless, in-frame deletions were generated through allelic exchange, as previously described, for all F. tularensis deletion strains except for glpA (48). The glpA gene was disrupted using the Targetron system modified for use in Francisella species, as previously described (26). Detailed methods for mutant generation and complementation are provided in Text S1.
Growth curves. Overnight cultures of F. tularensis SchuS4 grown in CDM were diluted to an OD 600 of 0.05 in 200 l of CDM or modified CDM in a 96-well plate (Corning). Each major carbon source was added to a final concentration of 0.4%. Cultures were incubated in an Infinite 200M Pro series plate reader (Tecan) at 37°C with orbital shaking. Bacterial intracellular growth within J774A.1 or BMDM cells was determined by measuring the luminescence of Schu S4 harboring the luminescence reporter plasmid pJB2 or pJB3, described in Text S1. During infection, J774A.1 and BMDM cells were cultured in high-glucose (4.5 g/liter) or glucose-free, pyruvate-free DMEM (Gibco) supplemented with 10% dialyzed fetal bovine serum (FBS). When stated, BMDMs were pretreated 2 h prior to infection with 150 M AICAR (Cayman Chemical) or Atglistatin (Cayman Chemical). Atglistatin cytotoxicity was measured using a Vybrant MTT cell proliferation assay kit (Thermo Fisher), following the manufacturer's protocol. Detailed methods are provided in Text S1.
Mouse infections. Groups of 6-to 8-week-old female C57BL6/J mice (Jackson Labs) were inoculated intranasally with 100 CFU of F. tularensis Schu S4 WT or mutant strains. Infected and control mice were housed in a recirculating air system (Techniplast) within a BSL-3 facility. At 3 days postinfection, mice were sacrificed and the lungs, livers, and spleens were harvested and homogenized using a Biojector (Bioject). The homogenates were serially diluted and plated onto chocolate or MMH (modified Mueller Hinton) agar to quantify organ burdens.