|Home | About | Journals | Submit | Contact Us | Français|
Mycobacterium tuberculosis, the causative agent of tuberculosis, is an intracellular pathogen that shifts to a lipid-based metabolism in the host. Moreover, metabolism of the host lipid cholesterol plays an important role in M. tuberculosis infection. We used transcriptional profiling to identify genes transcriptionally regulated by cholesterol and KstR (Rv3574), a TetR-like repressor. The fadA5 (Rv3546) gene, annotated as a lipid-metabolizing thiolase, the expression of which is upregulated by cholesterol and repressed by KstR, was deleted in M. tuberculosis H37Rv. We demonstrated that fadA5 is required for utilization of cholesterol as a sole carbon source in vitro and for full virulence of M. tuberculosis in the chronic stage of mouse lung infection. Cholesterol is not toxic to the fadA5 mutant strain, and, therefore, toxicity does not account for its attenuation. We show that the wild-type strain, H37Rv, metabolizes cholesterol to androst-4-ene-3,17-dione (AD) and androsta-1,4-diene-3,17-dione (ADD) and exports these metabolites into the medium, whereas the fadA5 mutant strain is defective for this activity. We demonstrate that FadA5 catalyzes the thiolysis of acetoacetyl-coenzyme A (CoA). This catalytic activity is consistent with a β-ketoacyl-CoA thiolase function in cholesterol β-oxidation that is required for the production of androsterones. We conclude that the attenuated phenotype of the fadA5 mutant is a consequence of disrupted cholesterol metabolism that is essential only in the persistent stage of M. tuberculosis infection and may be caused by the inability to produce AD/ADD from cholesterol.
Circumstantial evidence implicates a role for cholesterol in M. tuberculosis infection. How this host lipid affects infection and the role of its metabolism by the bacteria is not clear. Foamy macrophages, laden with lipid droplets, are known to accumulate in lung granulomas in human infection (23) as well as in infected mouse lungs (4). Electron microscopy revealed that M. tuberculosis is in close apposition to lipid bodies in foamy macrophages and that the bacteria eventually merge with these lipid bodies and even accumulate intracellular lipids (22). In fact, since foamy macrophages contain high levels of cholesterol esters (19, 32), it is possible that the lipid bodies observed to merge with M. tuberculosis bacteria are composed of cholesterol esters. Macrophages infected with Mycobacterium leprae also accumulate cholesterol esters (17) that are thought to be responsible for the conversion of macrophages into foam cells. Significantly, bacteria from the sputum of human tuberculosis (TB) patients were observed to contain lipid bodies, and transcriptional profiling of bacteria from these samples demonstrated the induction of genes proposed to encode enzymes required for cholesterol utilization (13). These studies implicate host lipids and possibly cholesterol in the development of mycobacterial disease.
The first direct evidence suggesting that bacterial cholesterol metabolism is required for M. tuberculosis virulence came from studies on an mce4 mutant that is defective in cholesterol transport and attenuated for virulence in the mouse model (21). Cholesterol was shown to colocalize with bacteria in bone marrow macrophages, and metabolic labeling studies with M. tuberculosis demonstrated that the bacteria can convert cholesterol to both CO2 and phthiocerol dimycoserosates (PDIM) (21). Cholesterol degradation yields propionyl-coenzyme A (CoA) (35) which, in combination with CO2, is converted to methyl-malonyl-CoA, a metabolite used in methyl branched-lipid biosynthesis. Propionyl-CoA is also shunted to acetyl-CoA through the methylcitrate pathway. Thus, cholesterol is degraded by M. tuberculosis, and its metabolites contribute to flux through both catabolic and anabolic pathways.
A group of genes encoding enzymes annotated as involved in lipid metabolism was shown to be upregulated during incubation with cholesterol in the actinomycete Rhodococcus RHA1 and were proposed to be responsible for cholesterol catabolism in that organism (30). Some of these genes have orthologs in the M. tuberculosis genome, suggesting that a similar pathway functions in M. tuberculosis (Fig. (Fig.1),1), and the genes required for cholesterol A- and B-ring metabolism have been identified. The enzymatic activities of recombinant Hsd (34), KstD (16), HsaC (30), HsaD (30), and KshA/B (3) from M. tuberculosis are consistent with their predicted functions, as expected due to their high levels of similarity to other actinomycete orthologs, and their substrate specificities are consistent with cholesterol-derived metabolites being the physiological substrates. Recently hsaC, which encodes a cholesterol ring-cleaving dioxygenase, was shown to be required for optimal growth on cholesterol as a sole carbon source (33). Cholesterol was toxic to the mutant strain due to the accumulation of catechols, and it was somewhat attenuated in SCID mice and in guinea pigs, possibly due to cholesterol toxicity (33). In addition, the igr operon, which encodes putative β-oxidation enzymes, was found to be required for growth in the presence of cholesterol (6). Cholesterol was toxic to the mutant strain (6), which grew more slowly in mice from the time of initial infection (5). The precise function of igr-encoded enzymes has yet to be established.
As a first step to identify additional M. tuberculosis genes and enzymes involved in cholesterol metabolism, we describe the M. tuberculosis cholesterol regulon by transcriptional profiling. We show that a major portion of this regulon is under the control of the KstR transcriptional regulator, initially described in Mycobacterium smegmatis (15). fadA5 (Rv3546) is one of the genes regulated by cholesterol and by KstR. We demonstrate that a fadA5 mutant strain of M. tuberculosis does not grow on cholesterol as a sole carbon source in broth, thus establishing fadA5's involvement in cholesterol metabolism. We show that this gene is also required for persistence in the mouse model of M. tuberculosis infection. We demonstrate that cholesterol is not toxic in broth culture to the fadA5 mutant, in contrast to the hsaC and igr mutants, which showed cholesterol toxicity (6, 33). We provide the first evidence that cholesterol is converted to androst-4-ene-3,17-dione (AD) and 1,4-androstadiene-3,17-dione (ADD) by M. tuberculosis H37Rv and that these metabolites accumulate in the supernatant. We show that fadA5 is required for this activity and that FadA5 is a thiolase whose activity is presumed to be required for β-oxidation of the cholesterol side chain. These results suggest that one role of cholesterol metabolism in M. tuberculosis is to provide metabolites, possibly the prohormones AD and ADD, that allow persistence in the lung.
M. tuberculosis cultures were grown at 37°C in Middlebrook 7H9 liquid medium (Becton Dickinson), supplemented with 0.05% Tween-80, 10% albumin-dextrose-NaCl complex (ADN) (1), and 0.2% glycerol, or on Middlebrook 7H10 plates supplemented the same way. Kanamycin was added at 20 μg ml−1, and hygromycin was added at 100 μg ml−1. Growth on cholesterol as a sole carbon source was done by supplementing 7H9 medium with 1 mg ml−1 cholesterol made up in the nonionic surfactant tyloxapol (Sigma). The stock solution was 20 mg ml−1 cholesterol made in pure tyloxapol, autoclaved, and then boiled to dissolve the cholesterol completely. When this is diluted 20-fold into 7H9 medium, it must be boiled again to enable the formation of a uniform medium. Under these conditions, the cholesterol does not precipitate, and the growth of the culture can be followed by measuring optical density (OD) at 540 nm. When cholesterol was used as an inducer for transcriptional profiling, cultures were grown to mid-log phase in supplemented 7H9 medium, and then cholesterol (1 mg ml−1 final concentration) in Tween-80 (1%, wt/vol, final concentration) was added, and the cultures were incubated for an additional 24 h. Cholesterol stock solutions (20 mg ml−1) were prepared in aqueous Tween-80 (20%, wt/vol). Treatment with 1% Tween-80 only was used as the control.
DNA recombinant techniques were performed following standard procedures (24). All restriction enzymes and modifying enzymes were commercially obtained from Promega Laboratories or New England Biolabs and used according to the manufacturer's recommendations. Electroporation of M. tuberculosis and DNA Southern blot analysis from the M. tuberculosis chromosome were performed as previously described (18). The fadA5 mutation in M. tuberculosis was constructed using pSM270, as described previously (18), with two primers: 5′-GTGATGGAGTTGCGGGTGG-3′ (forward primer) and 5′-CTTGGTGGTATGGCTGGGC-3′ (reverse primer). The plasmid construct contained a deletion of the fadA5 gene, which was replaced with a kanamycin cassette. This plasmid was used to electroporate H37Rv, with selection for kanamycin and streptomycin, to obtain the first crossover recombinants. The second crossover recombinants were selected by resistance to sucrose and sensitivity to streptomycin. Southern blot analysis (data not shown) of presumptive recombinants indicated that strains had the predicted disruptions in fadA5 (strain ST76). Complementation of the fadA5 mutant was accomplished by cloning the entire open reading frame with 166 bp flanking at the N terminus into the integrative plasmid pMV306 (MedImmune, Inc.) that confers hygromycin resistance with two primers: 5′-TGGGCACGGTAGTTCTCCTAACTG-3′ (forward primer) and 5′-CTAGGTTAAATCCGCTCGATGATGG-3′ (reverse primer). This construct was electroporated into ST76, and selection for Hygr transformants resulted in strain ST93.
All mouse protocols were approved by the University of Medicine and Dentistry of New Jersey (UMDNJ) IACUC. Six- to eight week-old female C57BL/6 mice were infected by the respiratory route, as previously described (31), using a nose-only aerosolization apparatus (In-Tox Products) (29). Cultures of the various strains of M. tuberculosis were grown in Middlebrook 7H9 liquid medium (described above) to the logarithmic stage and diluted to 1 × 105 CFU ml−1 in phosphate-buffered saline (PBS)-Tween. This dilution (10 ml) was used in the nebulizer, resulting in the initial deposit of about 100 to 300 CFU/lung. Three mice were sacrificed at each time point, and the lungs, homogenized in PBS-Tween, were plated on 7H10 solid medium with serial dilutions. After 3 weeks of incubation at 37°C, the number of colonies was enumerated.
Preparation of RNA from M. tuberculosis grown in liquid cultures for microarray and quantitative reverse transcription-PCR (qRT-PCR) analysis using SYBR green was done as described previously (31). PCR primers, synthesized by Integrated DNA Technologies, are listed in Table S1 in the supplemental material. PCR conditions were as follows: a 10-min denaturation step at 95°C followed by 30 cycles of denaturation at 95°C for 0.2 min and annealing and extension at 67°C for 0.45 min. The extent of amplification was determined by generating a standard curve for each primer containing 0, 1,000, 10,000, and 100,000 theoretical copies of H37Rv cDNA. These reactions were run in tandem with the experimental samples. All values were normalized to the amount of 16S mRNA.
Microarray analysis was carried out at the Center for Applied Genomics (CAG) at UMDNJ. cDNA was labeled with Cy5 or Cy3 and hybridized to DNA chips containing oligomers of 70 bp representing all M. tuberculosis open reading frames. The hybridized chips were read using an Affymetrix Integrated GeneChip Instrument System equipped with a Fluidics Station and GeneChip Scanner 3000 TG. The data were filtered by removing all spots that were below the background noise or that were flagged as “bad.” Spots were considered to be below the background noise if the sum of the median intensities of the two channels was less than twice the highest average background of the chip. The chips were normalized by the print-tip Lowess method (11). Ratios were calculated for expression of genes in cultures with cholesterol treatment versus cultures without cholesterol and in the wild-type versus the kstR mutant; genes for which the ratios were greater than 1.5 were considered upregulated.
Three independent cultures of M. tuberculosis H37Rv, fadA5, and complemented fadA5 were grown as described for transcriptional profiling and were induced with cholesterol (1 mg ml−1) for 48 h. Both cell pellets and culture supernatants were autoclaved. Cell pellets were extracted by the method of Bligh and Dyer (2), concentrated to dryness, and resuspended in ethyl acetate (EtOAC). The culture supernatants were extracted with EtOAc twice. The aqueous layers were acidified to pH 5 and extracted with EtOAc twice. For all samples, the EtOAC extracts were washed with H2O three times and concentrated, and the concentrates were analyzed by liquid chromatography-mass spectrometry-UV detection (LC/MS/UV). A Waters Acquity Ultra Performance LC system (Milford, MA), equipped with a photodiode array (PDA) detector, and a single quadrupole (SQ) detector was used for identification of metabolites. Chromatography was performed with a C18 reverse-phase column (2.1-mm internal diameter by 100 mm; 1.7-μm particle size; Waters Acquity UPLC BEH) maintained at 55°C. The elution solvents were solvent A, consisting of 10% MeOH in H2O, and solvent B, consisting of MeOH, and the flow rate was 0.5 ml min−1. The eluting solvent was isocratic for 0.02 min with 100% solvent A, followed by a linear gradient to 44% solvent A in 0.3 min, a second linear gradient to 0% solvent A in 2 min, isocratic for 7 min with 0% solvent A, a third linear gradient to 100% solvent A in 0.5 min, and isocratic with 100% solvent A for 0.5 min. The mass spectrometer was operated with an atmospheric pressure chemical ionization source in positive ion mode with the source temperature set to 150°C; the desolvation temperature was set to 450°C with a corona voltage of 1.5 kV. The UV detection wavelength range was 200 to 400 nm. Radioactive samples were analyzed only by UV detection. Fractions were collected every 10 s and analyzed by liquid scintillation counting after the addition of 4 ml of Scintiverse II LSC cocktail.
M. tuberculosis fadA5 (Rv3546) was cloned into the BamHI site of the E. coli-M. smegmatis shuttle vector pSD31 (7) by PCR with Pfu Turbo (Stratagene) using primers NN78 (5′-GCC GCC GGA TCC ATG GGT TAC CCG GTC ATC-3′) and NN82 (5′-GGC GGC GGC GGA TCC TTA AAT CCG CTC GAT GAT G-3′) to introduce an N-terminal hexahistidine tag. The pSD31(fadA5) construct was electroporated into M. smegmatis mc2155, a single colony was selected on solid medium, and liquid cultures were grown in Middlebrook 7H9 medium supplemented with 0.2% glycerol, 0.05% Tween-80, 50 μg ml−1 hygromycin, 10 μg ml−1 cycloheximide, and 200 μg ml−1 ampicillin to an OD600 of ~1.7 at 35.5°C. Expression of fadA5 was induced by the addition of 0.2% acetamide at 35.5°C. After 24 h, cells were harvested and resuspended in 50 mM sodium phosphate, 300 mM NaCl, 10% glycerol, and 10 mM imidazole. Cells were disrupted by sonication, and cellular debris was removed by centrifuging at 15,000 rpm for 1 h. FadA5 was purified by immobilized metal affinity chromatography (IMAC) using Hisbind resin (Novagen, Madison, WI) following the manufacturer's protocol. Pure protein was exchanged into buffer lacking imidazole using an Amicon stirred cell (Millipore, Bedford, MA) with a YM-30 membrane (30,000-molecular-mass cutoff). Glycerol was added to FadA5 to a final concentration of 25%, and the enzyme was stored at −20°C. The identity of the purified protein was confirmed by tryptic digestion and matrix-assisted laser desorption ionization (MALDI) mass fingerprinting. FadA5 (~15 μg of pure protein) was denatured in 50 mM Tris-HCl-5 mM dithiothreitol (DTT), pH 8.0, containing 2 M urea. After a 1-h incubation at 37°C, 10 mM iodoacetamide was added to the reaction mixture to alkylate cysteines present on the protein. Alkylation was allowed to proceed for 1 h at room temperature in the dark. The product was digested with trypsin (300 ng) in 20 μl of 50 mM NH4HCO3 for 17 h. The tryptic fragments were extracted with 50% CH3CN/H2O-0.1% trifluoroacetic acid (TFA; 50 μl) three times. The combined extracts were dried and redissolved in 5 μl of 0.1% TFA-H2O. MALDI-time of flight (TOF) mass spectra were acquired in positive ion mode using a saturated solution of α-cyano-4-hydroxy cinnamic acid as the matrix. MALDI mass spectra were acquired on a Bruker Autoflex II TOF/TOF instrument. Data were analyzed using FlexAnalysis software and MS-Bridge (http://prospector.ucsf.edu/cgi-bin/msform.cgi?form=msbridgestandard).
Thiolase activity was measured in the degradative direction following the method of Thompson et al. (28) and Middleton (20). The acetoacetyl-CoA (AcAcCoA) contained less than 0.1% free CoA as determined by LC/MS/UV. Briefly, assay mixtures containing 100 mM Tris-HCl, pH 8.1, 25 mM MgCl2, 0 to 50 μM CoA, 0 to 200 μM acetoacetyl-CoA, and 5 mM Tris(2-carboxyethyl)phosphine were preincubated at 30°C for 5 min. A background rate of thiolysis was measured. Reactions were initiated by the addition of FadA5 (250 nM) to the assay mixture. Loss of the acetoacetyl-CoA-Mg2+-enolate was followed at 303 nm for 5 min in the scanning mode using a UV2550 UV/visible light spectrophotometer (Shimadzu Scientific Instruments) and its associated software, and the initial velocity was determined from the linear portion, at approximately 2 min. An extinction coefficient of 16.9 mM−1 cm−1 at 303 nm was used to determine the concentration of consumed acetoacetyl-CoA-Mg2+-enolate (20). The background rate of thiolysis in the absence of FadA5 was subtracted from the measured enzymatic rate. Steady-state kinetic parameters were obtained as apparent (app) rate constants at a fixed concentration of the second substrate (S) from hyperbolic fits of the apparent Michaelis-Menten equation (equation 1) to the data by using Grafit:
in which Vmapp is the apparent maximal velocity.
Transcriptional profiling data acquired as described above were deposited in the Gene Expression Omnibus (GEO) database under accession number GSE13978.
To determine the effect of cholesterol on the expression of M. tuberculosis genes, mid-log-phase cultures of H37Rv growing in standard medium were incubated with 0.1% (wt/vol) cholesterol made up in 20% Tween-80 or with only Tween-80. RNA was isolated at 3 h and 24 h after addition of cholesterol, and gene expression was analyzed using microarrays. Fifty-two of the M. tuberculosis genes which were upregulated are found in an 83-gene region of the chromosome that corresponds to a 223-gene region in Rhodococcus RHA1 (30) proposed to encode proteins that catabolize cholesterol, and we will refer to this 83-gene region as the “Cho-region” (Fig. (Fig.2).2). For full results, see the GEO database (GSE13978). We validated the microarray results by analyzing the expression levels of 16 genes by qRT-PCR. In all cases, the values were normalized to levels of 16S RNA. The genes chosen included 14 that were upregulated by cholesterol according to microarray analysis and 2 that were not upregulated but were of interest for functional considerations. The qRT-PCR results (see Table S2 in the supplemental material) were all consistent with the microarray results.
KstR is a transcriptional repressor of many genes annotated as involved in lipid metabolism in M. smegmatis (15), orthologs of which are found in the 223-gene region of Rhodococcus corresponding to the Cho-region. To determine the genes under the control of KstR in M. tuberculosis, a transposon- disrupted mutant in Rv3574, encoding kstR, as well as its parent strain, CDC1551, were grown to mid-log phase, and RNA was isolated 24 h later for gene expression analysis by microarray. We obtained this transposon mutant strain from the Tuberculosis Animal Research and Gene Evaluation Taskforce (TARGET; NIH/NIAID contract NO1-AI30036). Many of the genes in the Cho-region were upregulated in the mutant strain, as predicted (15) (Fig. (Fig.2).2). For full results, see the GEO database (GSE13978). We did not perform qRT-PCR on the RNA used to analyze the KstR regulon since the expression of 16S RNA, used for normalization, was affected in the kstR mutant (data not shown).
We reasoned that some of the β-oxidation genes induced by cholesterol might encode enzymes required for cholesterol catabolism. fadA5 is adjacent to, but transcribed in the reverse direction of, the igr operon (Rv3545c-Rv3540c). FadA5 has been proposed to function in β-oxidation of the propionyl side chain of metabolite 5 shown in Fig. Fig.11 (30). We prepared a strain of H37Rv carrying a mutation in fadA5 (Rv3546), a putative thiolase and one of the genes upregulated by cholesterol and repressed by KstR, and tested its ability to grow on cholesterol as a sole carbon source. In these experiments, we dissolved cholesterol in pure tyloxapol. Unlike Tween-80, tyloxapol cannot be catabolized by M. tuberculosis, and no growth was observed in control 7H9-tyloxapol medium (Fig. (Fig.3).3). Inclusion of tyloxapol prevented bacterial clumping and enabled us to monitor cell growth by light scattering at 540 nm. Wild-type H37Rv grew on cholesterol as a sole carbon source, but the fadA5 mutant was unable to grow (Fig. (Fig.3).3). The ability to grow on cholesterol was restored in the complemented strain. The fadA5 mutant grew normally in standard 7H9 medium supplemented with glycerol and ADC (10% albumin-dextrose-catalase).
Wild-type H37Rv and the fadA5 mutant as well as the complemented strain were used to infect mice by the aerosol route. The fadA5 mutant initially grew as rapidly as the wild-type strain, and then at about 8 weeks, the number of CFU in the mouse lungs decreased about 10-fold compared with the level in the wild-type strain (Fig. (Fig.4).4). The virulence of the fadA5 mutant was restored when the strain was complemented with the wild-type gene integrated at the att site. The fact that the fadA5 mutant strain grew normally for the first 4 to 5 weeks in mouse lungs strongly implies that cholesterol utilization as a carbon source is not required, at least during the stage of infection prior to the onset of the immune response.
We considered the possibility that attenuation of the fadA5 mutant strain was due to cholesterol toxicity. We grew the wild-type and mutant strains in complete medium in the presence and absence of cholesterol, using the same medium as used for the transcriptional profiling assays. Strains were grown to the exponential phase and diluted into medium with and without cholesterol (2.6 mM in 20% Tween), and growth curves were determined using assays for CFU. The concentration of cholesterol used is typical of concentrations employed for induction of actinomycete cultures and higher than used in studies of hsaC (0.5 mM) (33) and igr (0.1 mM) (6) toxicity. The results show that cholesterol does not inhibit the growth of H37Rv and its fadA5 mutant (Fig. (Fig.5),5), and, in fact, the strains grew slightly better in the presence of cholesterol. We conclude that cholesterol, or the accumulation of a cholesterol-derived metabolite, is not toxic for fadA5.
The FadA5 sequence is predicted with 100% probability to fold into a thiolase structure (27). In addition, the two sequence motifs (PF00108 and PF02803 ) that are signature motifs for thiolases are conserved in FadA5 (14). These motifs contain the active-site cysteine nucleophile and a general acid and a general base, histidine and cysteine, that catalyze the thiolase reaction.
FadA5 was heterologously expressed as an N-terminally His6-tagged protein in M. smegmatis and purified to homogeneity using immobilized-metal ion affinity chromatography (Fig. (Fig.6).6). The protein identity was confirmed by MALDI-TOF mass analysis of a peptide mixture generated by trypsin digestion. Eighty-six percent of the protein sequence was covered, confirming that the protein was expressed correctly in frame. The purified enzyme was assayed in the degradative direction with acetoacetyl-CoA (AcAcCoA) and coenzyme A as the substrates to yield acetyl-CoA at pH 8.1. The apparent rate constants for AcAcCoA were measured at a fixed concentration of CoA (50 μM). The apparent rate constants for CoA were measured at a fixed concentration of AcAcCoA (150 μM). Fully saturating concentrations were not used due to substrate inhibition, as has been previously observed with many thiolases. The steady-state kinetic parameters were the following: KMapp (CoA), 15 ± 1 μM; kcatapp (CoA), 0.018 ± 0.001 s−1; KMapp (AcAcCoA), 464 ± 207 μM; and kcatapp (AcAcCoA), 0.076 ± 0.002 s−1 (Fig. (Fig.7).7). Thus, FadA5 is a thiolase. However, the activity with these substrates is slow relative to known thiolases (28), and, thus, the physiological substrate is most likely not acetoacetyl CoA.
We grew cultures of H37Rv, the fadA5 mutant, and the complemented fadA5 mutant in complete medium to the logarithmic stage of growth. Cholesterol was added to the cultures, and an aliquot was removed from each to which [4-14C]cholesterol was added. Samples were withdrawn from the radioactive cultures, and at 48 h, more than 90% of the starting cholesterol was consumed in both wild-type and mutant cultures. The cultures with no [4-14C]cholesterol were collected 48 h after the addition of cholesterol, lipids of both cell pellets and culture supernatants were extracted, and the extracts were analyzed by LC with MS and UV detection. The extracts of the cell pellets of both wild-type and mutant contained the same metabolites. In contrast, the wild-type strain secreted two metabolites into the culture supernatants that were absent in the fadA5 mutant cultures (Fig. (Fig.8;8; see also Fig. S1 in the supplemental material). The two metabolites retained the 14C label from cholesterol and accounted for 2% of the 14C label in the cultures. They had maximum wavelengths (λmaxs) of 245 and 243 nm and had MH+ parent ions at 285.2 and 287.2, respectively (see Fig. S2 in the supplemental material). These spectroscopic data as well as the retention times were consistent with assignment to predicted metabolites ADD and AD (Fig. (Fig.11 and and9).9). Coinjection of the samples with authentic AD confirmed the identity of the second metabolite. Complementation of the fadA5 mutant strain restored production of ADD and AD (Fig. (Fig.8).8). Several metabolites unique to fadA5 were observed in the culture supernatants. These metabolites had lost the 4-14C label, suggesting that the A ring of cholesterol is still degraded in fadA5. However, the structures of these metabolites were not further elucidated due to the small quantities isolated. This observation is consistent with metabolism of cholesterol in other actinomycetes for which it has been found that side chain and ring degradation occur concomitantly (8).
Cholesterol is clearly important in many aspects of M. tuberculosis infection (13, 17, 19, 21, 32). However, the role of M. tuberculosis cholesterol metabolism is not yet clear. Cholesterol may serve as a nutritional source or as a starting material for the production of secondary metabolites that regulate the host-bacteria relationship. Elucidation of the essential in vivo function(s) of cholesterol metabolism genes has been difficult to deconvolute because of accumulation of toxic metabolites in cholesterol metabolism mutants studied to date (6, 33). In this work, deletion of the fadA5 gene blocked cholesterol metabolism without toxicity effects and allowed us to delineate when cholesterol metabolism is essential to bacterial survival in the mammalian host.
We present evidence that an important role of cholesterol metabolism by M. tuberculosis is to provide steroid metabolites during the chronic phase of infection. Moreover, metabolism of cholesterol as a primary carbon source is not required during the growth phase of infection. This conclusion is based on the observation that the fadA5 mutant grows normally in mouse lungs at first, but the number of bacteria in the lungs declines after induction of the cellular immune response, demonstrating a classical “persistence” phenotype. If cholesterol were the sole carbon source available in the mouse lung, the fadA5 mutant, which cannot utilize cholesterol as a sole carbon source, should not be able to grow. The proposed role for fadA5 in the persistence phenotype is consistent with the observation of Pandey and Sassetti (21) that disruption of mce4, encoding a cholesterol transporter, has an effect on the chronic stage of mouse infection and not the initial stages. The fact that fadA5, which is upregulated by incubation with cholesterol, is also upregulated during growth of M. tuberculosis in both human macrophages (10) and in mouse lungs (9) is consistent with the idea that the bacteria are in contact with cholesterol during in vivo growth.
Furthermore, we show that M. tuberculosis H37Rv metabolizes cholesterol to androst-4-ene-3,17-dione and androsta-1,4-diene-3,17-dione, which are found in the culture supernatant, and that the attenuated fadA5 mutant lacks this activity, which is restored, together with virulence, in the complemented strain (Fig. (Fig.4).4). Growth of the fadA5 mutant is not inhibited by incubation with cholesterol (Fig. (Fig.5),5), and therefore toxicity does not account for its attenuation as has been proposed for hsaC (33) and the igr operon (6). We conclude that the function of FadA5 is required for the production of androsterone metabolites. These metabolites may play a role in the host immune response that results in persistence of M. tuberculosis infection. Alternatively, the attenuated phenotype of the fadA5 mutant may be a consequence of disrupted cholesterol catabolism that is only essential in the persistent stage of M. tuberculosis infection.
Our data confirm that fadA5 encodes a β-ketoacyl-CoA thiolase as annotated. Based on the identities of isolated metabolites in other actinomycetes (25, 26), two β-oxidation cycles have been proposed in the pathway for cholesterol side chain degradation that ultimately yields the 17-keto androsteroids, e.g., AD and ADD. However, the specific gene products responsible for catalyzing these reactions have not been identified. In M. tuberculosis, we propose FadA5 as a candidate enzyme for thiolytic cleavage of the β-ketoacyl intermediate(s) in cholesterol side chain metabolism (Fig. (Fig.9).9). Thiolytic cleavage of these intermediates is necessary for the production of the androsteroids. This function is more consistent with the data than a catalytic role in β-oxidation of the propionyl side chain of metabolite 5. Disruption of enzymes required for the metabolism of intermediate 5 would be expected to result in an increased accumulation of AD and ADD because the intermediate is downstream of AD and ADD in the pathway. Confirmation of the precise step or steps catalyzed by FadA5 requires further mechanistic investigation of the enzyme and the cholesterol metabolism pathway, which we are pursuing.
We acknowledge financial support from the National Institutes of Health (grants AI065251 and HL53306 to N.S.S., A1044856 and AI065987 to I.S., DK007521, to N.M.N., and NIH/NIAID NO1-AI30036 [TARGET contract]) and from the New York State Technology and Research Program (FDP C040076 to N.S.S.).
Editor: S. M. Payne
Published ahead of print on 12 October 2009.
†Supplemental material for this article may be found at http://iai.asm.org/.