Many approaches have been used to determine how the intracellular environment influences the physiological state of a pathogen during infection. Most identify biochemical pathways that are expressed or are essential during infection, and use this information to infer both the metabolic state of the bacterium and the environment that imposes this state. However, the complexity of metabolic networks makes these inferences difficult, as a single enzyme can participate in multiple biochemical pathways, and the reliance on any pathway can be influenced by a number of different environmental factors. This work employed a unique combination of direct metabolomic profiling and genetic epistasis analysis to functionally connect a distinct feature of the intracellular environment with its metabolic consequence in the pathogen.
In order to understand the central metabolic adaptations required for the utilization of the host-derived carbon source, cholesterol, we globally surveyed the steady-state levels of primary metabolites during growth on this compound. This approach is fundamentally different from more commonly used profiling of mRNA or protein levels, which may not correlate with the metabolite levels (Jozefczuk et al.
). We found that the differential utilization of a pathway could be predicted by the abundance of both its specific intermediates and its more stable biosynthetic products. Labeling of proteinogenic amino acids has been used extensively to predict the flux through central carbon metabolism (Zamboni et al., 2009
), and our data suggest that pool sizes of individual amino acids are also likely stable enough to serve this purpose. Together, these data provided a holistic view of cellular metabolism and allowed us to identify pathways that were critical for cholesterol catabolism.
The relatively dramatic accumulation of the PrpD substrate, methylcitrate, suggested that the expression of this enzyme might be insufficient for the flux of propionyl-CoA derived from cholesterol. Indeed, it appears that several propionate units are liberated from each cholesterol molecule. In addition to the terminal sidechain carbons at least two other propionyl-CoA molecules may be produced (Van der Geize et al., 2007
), suggesting that as much as one third of the carbon derived from cholesterol could enter the MCC. While cholesterol appears to be a significant source of propionyl-CoA, the incomplete suppression of the ΔprpDC
mutant’s intracellular growth defect by Mce4 mutation indicates that either Mce4-independent cholesterol uptake pathways operate in this environment or that alternative carbon sources, such as odd chain fatty acids or branched chain amino acids, also contribute to the propionyl-CoA pool during infection.
The increased reliance on the MCC pathway during cholesterol catabolism was confirmed by demonstrating that transcriptional induction of the prpDC
genes via the Rv1129c protein was required for growth. The DNA binding function of the homologous RamB protein of Corynebacterium glutamicum
is posttranslationally regulated (Cramer et al., 2007
; Micklinghoff et al., 2009
), perhaps via binding of an intermediary metabolite. Thus, mycobacteria appear to use very similar regulatory proteins to control the pathways necessary for propionyl- and acetyl-CoA catabolism, and differential expression of these systems could be mediated via distinct allosteric regulation.
Consistent with previous observations (Munoz-Elias et al., 2006
; Savvi et al., 2008
), we found that the growth of mutants lacking sufficient MCC activity was inhibited by propionate, even in the presence of an additional carbon source. Several mechanisms have been proposed to mediate propionyl-CoA-related toxicity. For example, this metabolite has been shown to inhibit pyruvate dehydrogenase (PDH) and citrate synthase in other bacteria (Brock and Buckel, 2004
; Man et al., 1995
; Maruyama and Kitamura, 1985
). While the accumulation of glycolytic intermediates we observed in cholesterol-grown mycobacteria could be consistent with the these mechanisms, it is unclear whether the unusual PDH complex (Tian et al., 2005
), or the apparently redundant citrate synthase enzymes (CitA and GltA) of mycobacteria are inhibited by propionyl-CoA. Similarly, propionate-related toxicity in MCC-deficient Salmonella is due to the accumulation of a specific 2-methylcitrate isomer, which inhibits the gluconeogenic enzyme, fructose 1,6 bisphosphatase (Rocco and Escalante-Semerena
). However, the type II FBP expressed by Mtb bears little resemblance to the Salmonella enzyme. Thus, while abundant examples of this phenomenon have been reported, it is currently unclear whether these previously described mechanisms are relevant in mycobacteria.
These metabolic insights provided the information necessary to test whether the MCC requirement during intracellular growth was due to host cholesterol utilization. The ability of mce4
mutation to alleviate the intracellular growth defect produced by either prpDC
mutation or ICL/MCL inhibition implicates cholesterol as a major source of propionate in vivo
even under conditions in which cholesterol is not absolutely required for growth. This MCC mutant phenotype was reminiscent of strains lacking either the hsaC
gene (Yam et al., 2009
) or the “igr
” locus (Chang et al., 2007
; Chang et al., 2009
), which encode functions dedicated to earlier steps in cholesterol degradation. Like MCC mutants, these strains are intoxicated in the presence of cholesterol, and are unable to grow in resting macrophages or acutely infected animals. Together, these observations strongly argue that M. tuberculosis
obtains a significant amount of its carbon requirements from cholesterol.
In contrast to their essentiality during intracellular growth in macrophages, the prpDC
genes have been found to be dispensable for bacterial growth in the mouse model of TB (Munoz-Elias et al., 2006
). This observation was particularly surprising as cholesterol catabolism occurs throughout infection (Chang et al., 2009
; Pandey and Sassetti, 2008
; Yam et al., 2009
), and cell wall lipid alterations consistent with propionyl-CoA assimilation are evident in bacteria isolated from mouse tissue (Jain et al., 2007
). This apparent contradiction could be explained if the B12-dependent methylmalonyl pathway (MMP) were active under these conditions, either because the bacterium acquires B12 from the host, or some aspect of the host environment stimulates B12 production by the pathogen (Savvi et al., 2008
). Our observation that B12 supplementation reverses the growth defect of prpDC
mutants in macrophages, supports this hypothesis.
This work highlights propionyl-CoA assimilation as a critical feature of M. tuberculosis
metabolism in vivo
, and demonstrates that this requirement is a direct result of the nutritional environment in the mycobacterial phagosome. However, even within this relatively isolated compartment, the bacterium is likely to co-catabolize multiple carbon sources (de Carvalho et al.
), and inhibiting the bacterium’s ability to utilize individual nutrients is unlikely to represent a fruitful therapeutic strategy. In contrast, inhibition of at least three distinct steps in cholesterol catabolism (hsaC
, or MCC-encoding genes) results in metabolic toxicity that might be effectively exploited to treat TB. Thus, a fundamental understanding of mycobacterial physiology during infection is critical for identifying vulnerabilities that can be used for the rational design of more effective TB therapies.