The cholesterol catabolic pathway occurs in most mycolic acid-containing actinobacteria, such as Rhodococcus jostii RHA1, and is critical for Mycobacterium tuberculosis (Mtb) during infection. FadD3 is one of four predicted acyl-CoA synthetases potentially involved in cholesterol catabolism. A ΔfadD3 mutant of RHA1 grew on cholesterol to half the yield of wild-type and accumulated 3aα-H-4α(3′-propanoate)-7aβ-methylhexahydro-1,5-indanedione (HIP), consistent with the catabolism of half the steroid molecule. This phenotype was rescued by fadD3 of Mtb. Moreover, RHA1 but not ΔfadD3 grew on HIP. Purified FadD3Mtb catalysed the ATP-dependent CoA thioesterification of HIP and its hydroxylated analogues, 5α-OH HIP and 1β-OH HIP. The apparent specificity constant (kcat/Km) of FadD3Mtb for HIP was 7.3 ± 0.3 × 105 M−1 s−1, 165 times higher than for 5α-OH HIP, while the apparent Km for CoASH was 110 ± 10 μM. In contrast to enzymes involved in the catabolism of rings A and B, FadD3Mtb did not detectably transform a metabolite with a partially degraded C17 side-chain. Overall, these results indicate that FadD3 is a HIP-CoA synthetase that initiates catabolism of steroid rings C and D after side-chain degradation is complete. These findings are consistent with the actinobacterial kstR2 regulon encoding ring C/D degradation enzymes.
Mycobacterium tuberculosis (Mtb), the aetiological agent of tuberculosis (TB), infects about one-third of the world's population, claiming over 1 million lives annually. With the emergence of extensively drug-resistant strains (XDR-TB) and its co-action with the HIV, Mtb remains a global threat for which new therapeutics are urgently needed (WHO, 2011). Part of the bacterium's success as a pathogen is its unusual ability to persist and replicate inside human macrophages (Russell, 2007). Mtb's adaptation to the host includes the pathogen's heavy reliance on lipids for growth (Russell et al., 2010). Studies suggest that genes necessary for lipid catabolism are preferentially expressed during intra-macrophage growth (Schnappinger et al., 2003; Homolka et al., 2010), and that growth on fatty acids is an important component of Mtb's virulence (McKinney et al., 2000). While less is known about Mtb's metabolism during the chronic stage of infection, granulomas are rich in cholesterol (Kim et al., 2010). The bacterium can use cholesterol as sole growth substrate in vitro (Pandey and Sassetti, 2008) and knockouts of several cholesterol catabolic genes attenuate Mtb's pathogenicity in various infection models (Pandey and Sassetti, 2008; Chang et al., 2009; Yam et al., 2009; Hu et al., 2010; Nesbitt et al., 2010). Based on these studies, it has been proposed that host-derived cholesterol is an essential nutrient, particularly during chronic infection.
The cholesterol catabolic pathway is encoded by over 80 genes in Mtb (van der Geize et al., 2007) (Fig. 1) and is highly conserved in mycolic acid-containing actinomycetes. Some of the latter, including Rhodococcus jostii RHA1, contain paralogous pathways responsible for the degradation of other steroids, such as cholate (Mohn et al., 2012). In Mtb and other actinomycetes, cholesterol is taken up by the ATP-dependent Mce4 transporter (Mohn et al., 2008). Subsequent catabolism appears to be organized according to three parts of the steroid molecule: the side-chain, rings A/B and rings C/D respectively. The side-chain is degraded by a process similar to the β-oxidation of fatty acids involving CoA thioester intermediates (Nesbitt et al., 2010). Degradation of the rings A/B requires the activities of two dehydrogenases, three oxygenases and a hydrolase (Yang et al., 2007; Capyk et al., 2009; Yam et al., 2009; Dresen et al., 2010; Lack et al., 2010) and yields 3aα-H-4α(3′-propanoate)-7aβ-methylhexahydro-1,5-indanedione (HIP), which contains steroid rings C and D (Fig. 1). Important aspects of this catabolism have yet to be elucidated. For example, while it has recently been demonstrated that a ring A/B-degrading enzyme acts on metabolites with partially degraded side-chains (Capyk et al., 2011), the extent to which side-chain and ring catabolism occur concurrently is unknown. Deletion of the igr locus, encoding some side-chain-degrading enzymes, yielded a mutant of Mtb that accumulated a HIP derivative harbouring a partially degraded side-chain (Thomas et al., 2011). This suggests that the 6C fragment derived from ring A/B is removed before the side-chain is fully degraded. Similarly, the mechanism by which HIP is degraded is also largely unknown. Interestingly, HIP and its analogues derived from other steroids, such as cholate, appear to be degraded by a single pathway in RHA1 encoded by the kstR2 regulon (Mohn et al., 2012). This regulon was originally identified in Mycobacterium smegmatis (Kendall et al., 2010) and comprises 15 genes in Mtb.
Among the cholesterol catabolic genes of unassigned function are several fadD genes, predicted to encode AMP-forming acyl-CoA synthetases. The latter use ATP and Coenzyme A (CoASH) to thioesterify substrates containing a free carboxylate via a two-step mechanism that proceeds through an acyl-adenylate intermediate (Groot et al., 1976) (Fig. 2). Acyl-CoA synthetases belong to the large family of acyl-adenylate/thioester-forming enzymes (Babbitt et al., 1992), and represent the primary route for fatty acid activation in microorganisms. Enzymes of this large family share three well-conserved motifs proposed to be implicated in substrate binding and/or catalysis (Gocht and Marahiel, 1994; Chang et al., 1997; Jung et al., 2000). In addition, several acyl-CoA synthetases contain a conserved lysyl residue, which is critical for acyl-AMP formation and is a target for reversible acetylation in at least some of these enzymes (Horswill and Escalante-Semerena, 2002; Starai et al., 2002; Xu et al., 2011).
The cholesterol catabolic cluster contains four fadD genes: fadD17, fadD18, fadD19 and fadD3, identified as Rv3506, Rv3513c, Rv3515c and Rv3561, respectively, in Mtb strain H37Rv (van der Geize et al., 2007). While fadD18 appears to be a truncated paralogue of fadD19, the three others encode full-length proteins sharing less than 30% amino acid sequence identity (Cole et al., 1998; Camus et al., 2002). FadD17 and FadD19 have been reported to catalyse the CoA thioesterification of long-chain fatty acids (Trivedi et al., 2004; Arora et al., 2009). Interestingly, deletion of fadD19 did not affect cholesterol catabolism in Rhodococcus rhodochrousDSM43269 but abolished the degradation of a steroid possessing a C24-branched side-chain (Wilbrink et al., 2011). Deletion of the fadD3 homologue in Comamonas testosteroni TA441, a Gram-negative steroid-degrading bacterium, led to the accumulation of HIP during growth on androsta-1,4-dien-9,17-dione (ADD), suggesting that HIP is the physiological substrate of FadD3 (Horinouchi et al., 2006). However, recent studies in Rhodococcus equi led to the proposal that a two-subunit CoA transferase encoded by ipdAB is involved in initiating HIP degradation (van der Geize et al., 2011). The fadD3 gene is part of the kstR2 regulon which, in RHA1, appears to be involved in the catabolism of steroids other than just cholesterol (Mohn et al., 2012). Regardless of their precise role, fadD19 and fadD3 were among those genes predicted to be essential for growth of Mtb on cholesterol based on deep-sequencing of transposon mutant libraries (Griffin et al., 2011).
Herein, we used a combination of molecular genetics and biochemical approaches to investigate the physiological role of FadD3 in Mtb and RHA1. Mutants of RHA1 were used to help define the role of FadD3 and to produce steroid metabolites for enzymatic studies. FadD3 of Mtb was purified and its substrate specificity was investigated. The results are discussed in terms of microbial steroid catabolism and the pathogenesis of Mtb.
Deletion of fadD3 in RHA1 disrupts growth on steroids
To investigate the physiological role of FadD3, the fadD3 gene of RHA1 (ro04595) was deleted and the mutant's phenotype characterized. The ΔfadD3 mutant did not grow as well as the wild-type strain on cholesterol, showing a significant lag phase (∼ 100 h), an increased doubling time (∼ 70 versus 45 h), and producing only ∼ 50% the final biomass (Fig. 3A). In contrast, the mutant's growth on pyruvate was not affected (data not shown). Analysis of solvent-extracted culture supernatants using gas chromatography-coupled mass spectrometry (GC-MS) revealed that the ΔfadD3 mutant completely degraded cholesterol and accumulated a major metabolite in near stoichiometric amounts (Fig. 3). As described below, this metabolite was HIP. In contrast, wild-type RHA1 did not accumulate any detectable metabolites during growth on cholesterol.
Since fadD3 belongs to the kstR2 regulon (Kendall et al., 2010) and this regulon is upregulated during growth of RHA1 on cholate (Mohn et al., 2012), we evaluated the ability of the ΔfadD3 mutant to grow on cholate as the sole organic substrate. As compared with wild-type RHA1, the phenotype of the mutant was very similar to that observed in the presence of cholesterol: the growth rate was lower (doubling time of 8.1 versus 3.7 h) and the growth yield less than half (Fig. 4A). As described elsewhere (Swain et al., 2012), RHA1 secreted several metabolites into the culture supernatant during growth on cholate, including 3,7(R),12(S)-trihydroxy-9-oxo-9,10-seco-23,24-bisnorchola-1,3,5(10)-trien-22-oate (THSBNC) and 1β(2′-propanoate)-3aα-H-4α(3″(R)-hydroxy-3″-propanoate)-7aβ-methylhexahydro-5-indanone (HHIDP), which were re-assimilated by the late stationary phase. In contrast, the ΔfadD3 mutant secreted a larger number of metabolites, several of which persisted in the supernatants of late stationary-phase cultures. Moreover, the major metabolite that accumulated in the supernatant of cholate-grown ΔfadD3 cells was not detected at any point during growth of the wild-type strain on cholate (Fig. 4B). This metabolite was extracted from the supernatant of a late stationary culture and identified by GC-MS to be 3aα-H-4α(3′(R)-hydroxy-3′-propanoate)-7aβ-methylhexahydro-1,5-indanedione (3′-OH HIP) [see Supporting information (SI)]. These results suggest that catabolism of rings C/D is required for full growth on cholate and that FadD3 initiates this catabolism in RHA1.
Characterization of accumulated metabolites
To identify the metabolites accumulated by the ΔfadD3 mutant during growth on cholesterol and to produce substrates for enzymological studies, a larger scale biotransformation was performed. A 200 ml culture of the ΔfadD3 mutant was grown on cholesterol. After 110 h, metabolites were extracted from the culture supernatant and purified by semi-preparative, normal-phase HPLC. One hundred and fifteen mg of cholesterol yielded 53 mg of a solid. GC-MS indicated this material comprised essentially a single compound (> 98%). This compound was identified as HIP based on the fragmentation pattern of the trimethylsilyl derivative, one-dimensional 1H nuclear magnetic resonance (NMR) and two-dimensional homonuclear correlated spectroscopy (COSY) (see SI). More particularly, the one-dimensional NMR assignments are consistent with those published for HIP (Horinouchi et al., 2005) with the exception of the two protons attached to carbon 7 which are slightly deshielded and one of the protons attached to carbon 3′ which is slightly more shielded in the present preparation. However, the latter assignment is consistent with that of the equivalent proton in methyl 1β(2′-propanoate)-3aα-H-4α(3″-propanoate)-7aβ-methylhexahydro-5-indanone (Thomas et al., 2011). Overall, HIP was obtained in ∼ 75% mole recovery from cholesterol. A second metabolite, obtained in lower amount (2.5 mg), was identified as 3aα-H-4α(3′-propanoate)-1β-hydroxy-7aβ-methylhexahydro-5-indanone (1β-OH HIP) by GC-MS and NMR (see SI). This metabolite, which co-purified with an unidentified contaminant, was proposed to accumulate during catabolism of HIP by a strain of Nocardia (Lee and Sih, 1967).
MtbfadD3 rescued the mutant phenotypes
To verify that the phenotype of the ΔfadD3 mutant did not result from a polar effect or a genetic rearrangement, the mutant was complemented with a multi-copy plasmid carrying the wild-type fadD3 gene from Mtb. The growth rate and yield of the resulting complemented strain, ΔfadD3c, on cholesterol were nearly identical to that of the wild-type strain (Fig. 5A). Moreover, GC-MS analysis demonstrated that, like the wild-type, the complemented strain did not accumulate HIP during growth on cholesterol (Fig. 5B).
RHA1 catabolizes steroid rings C and D
The accumulation of near stoichiometric amounts of HIP during the growth of the ΔfadD3 mutant on cholesterol suggests that RHA1 catabolizes HIP. To directly test whether steroid rings C and D are catabolized, RHA1 and ΔfadD3 were assessed for their ability to use HIP as sole source of carbon and energy. As shown in Fig. 6, RHA1 grew on HIP with a doubling time of 21 h while the ΔfadD3 mutant did not catabolize this compound. GC-MS analyses confirmed that HIP was depleted only in the RHA1 culture (data not shown). These results demonstrate that FadD3 is necessary for initiating catabolism of rings C and D, and are consistent with previous studies showing that several actinomycetes can metabolize HIP and analogues (Lee and Sih, 1967; Hashimoto and Hayakawa, 1977).
FadD3 substrate preference
FadD3 from Mtb was heterologously produced with an N-terminal affinity tag in Escherichia coli. The activity of the purified FadD3Mtb, with the tag removed, was assessed using a range of carboxylate-containing cholesterol metabolites and their analogues. These included a variety of two- and four-ringed structures carrying either a keto group or a side-chain of three, five or eight carbons in length, at C17 (Fig. 7). Reaction mixtures contained a steroid metabolite, FadD3Mtb, CoASH, ATP and Mg2+. Substrate depletion and product formation were followed by reversed phase HPLC. When the reaction was performed with HIP, all substrate peaks were depleted and two new peaks appeared in the HPLC chromatogram (Fig. 8). The first product peak had the same retention time (2.8 min) and electronic absorption spectrum as AMP. The second product peak eluted at 10.0 min and had a spectrum with features characteristic of a CoA thioester. Alkaline hydrolysis of this second product yielded HIP and CoASH as determined by GC-MS and HPLC analyses, respectively, indicating that the peak eluting at 10.0 min was HIP-CoA. Transformation of 3aα-H-4α(3′-propanoate)-5α-hydroxy-7aβ-methylhexahydro-1-indanone (5α-OH HIP) by FadD3Mtb also resulted in formation of two new peaks (data not shown), one of which was AMP. The second product eluted at 9.8 min and had a spectrum with features characteristic of a CoA thioester. The product was identified as 5α-OH HIP-CoA using electron-spray ionization mass spectrometry (ESI-MS) and GC-MS analysis of the hydrolysed CoA thioester (Fig. S1). FadD3 was similarly shown to catalyse the CoA thioesterification of 1β-OH HIP (data not shown). Controls performed without substrate, co-substrate or enzyme, respectively, indicated that CoA thioester formation was well coupled with ATP hydrolysis. FadD3Mtb did not transform the following compounds: 3aα-H-4α(3′-propanoate)-5α-hydroxy-7aβ-methylhexahydro-1-indanone-δ-lactone (HIL); 3-oxo-23,24-bisnorchol-4-en-22-oate (4-BNC); cholate; and 3β-hydroxy-5-cholesten-26-oate.
FadD3 did not transform HIP with a partially degraded side-chain
A common feature of the three HIPs transformed by FadD3 is that the cholesterol side-chain is completely degraded (C1 HIP numbering). To test whether FadD3 can transform a HIP derivative containing a partially degraded side-chain, 1β(2′-propanoate)-3aα-H-4α(3″-propanoate)-7aβ-methylhexahydro-5-indanone (HIDP; Fig. 7) was produced by incubating deoxycholate with the ΔcamB mutant of RHA1. Previous studies had established that this mutant transforms cholate mainly to HHIDP (Swain et al., 2012). Incubation of 80 mg deoxycholate yielded ∼ 2.5 mg HIDP after purification by normal-phase HPLC, representing ∼ 4% mole recovery. The reason for this low yield is unclear. HIDP's identity was determined by GC-MS (see SI), the characteristics of which were consistent with those of HHIDP (Swain et al., 2012) minus a trimethylsilaned-hydroxyl group. HIDP comprised ≥ 92% of the preparation as assessed by GC-MS. The major contaminant (≤ 2%) was tentatively identified as 1-ylidene(2′-propanoate)-3aα-H-4α(3″-propanoate)-7aβ-methylhexahydro-5-indanone (YHIDP), a desaturated analogue of HIDP (data not shown). Under reaction conditions similar to those described above, purified FadD3Mtb did not detectably catalyse the thioesterification of HIDP. This result indicates that FadD3 acts on a cholesterol metabolite whose side-chain at C17 is completely degraded.
Steady-state kinetic parameters of FadD3Mtb
To assess the substrate specificity of FadD3Mtb, steady-state kinetics were performed using a coupled spectrophotometric assay (Horswill and Escalante-Semerena, 2002). Briefly, the production of AMP was coupled to the oxidation of NADH through successive reactions catalysed by adenylate kinase, pyruvate kinase, and lactate dehydrogenase, respectively, under conditions in which AMP formation was rate-limiting. Control reactions confirmed full coupling between AMP and CoA thioester formation. Using this assay, the FadD3-catalysed thioesterification of both HIP (Fig. 9) and 5α-OH HIP (data not shown) displayed Michaelis–Menten kinetics at pH 7.3 and 22°C. As summarized in Table 1, the apparent specificity constant (kcat/Km) of FadD3Mtb for HIP was 165 times higher than for 5α-OH HIP. More particularly, the enzyme turned over 5α-OH HIP slightly faster than HIP (i.e. kcat ∼ 2× higher), but the apparent Km value for the former was ∼ 300-fold greater. Preparations of 1β-OH HIP contained at least one contaminant and were obtained in too low a yield to enable a complete steady-state kinetic study. However, the specific activity of FadD3Mtb for 1β-OH HIP was 2.4 μmol min−1 mg−1 under the standard assay conditions. While this value may be influenced by the contaminant, it is similar to that calculated for the same concentration of 5α-OH HIP (2.0 μmol min−1 mg−1), using the steady-state parameters. Finally, in the presence of 100 μM HIP, the apparent steady-state kinetic parameters of FadD3Mtb for CoASH were Km = 110 ± 10 μM, kcat = 7.5 ± 0.1 s−1 and kcat/Km = 6.6 ± 0.6 × 104 M−1 s−1.
Table 1. Apparent steady-state kinetic parameters for AMP formation by FadD3Mtb.a
× 105 M−1 s−1
aSteady-state kinetics were performed at 22°C using 0.1 M HEPES, pH 7.3, 1 mM CoASH and 2 mM ATP. Values in parentheses represent standard errors.
Lactonization of 5α-OH HIP
The large difference in the kcat/Km values for HIP and 5α-OH HIP was somewhat unexpected given the metabolites' similar structures (see Fig. 7). Under certain solution conditions, 5α-OH HIP is converted non-enzymatically to HIL, which would not be transformed by FadD3. NMR spectroscopy was used to ensure that preparations of 5α-OH HIP did not contain significant amounts of the lactone. Briefly, the pD of 5α-OH HIP in deuterated buffer was lowered from 13 to 7. After 12 h, the NMR spectrum had not detectably changed, indicating that the 5α-OH HIP had not lactonized to any significant extent (data not shown). Thus, preparations of 5α-OH HIP used in enzymatic studies did not contain significant amounts of HIL.
As described above, neither HIDP nor HIL was detectably transformed by FadD3Mtb. In light of their structural similarity with HIP (Fig. 7), we tested whether either compound inhibited the enzyme. In presence of 10 μM HIP, a concentration equivalent to its apparent Km value, neither a 20-fold excess of HIDP nor a 40-fold excess of HIL affected the rate of HIP-CoA formation, as measured using the coupled assay (data not shown). This indicates that neither HIDP nor HIL significantly inhibits FadD3.
The current study establishes that FadD3 initiates the catabolism of steroid rings C/D in actinobacteria, catalysing the CoA thioesterification of the carboxylate that is generated by the HsaD-catalysed hydrolysis of di seco steroid metabolites (Fig. 1). Thus, a ΔfadD3 mutant of RHA1 grew to ∼ 50% the biomass of the wild-type strain on cholesterol and cholate and accumulated large amounts of HIP and 3′-OH HIP, respectively, in the culture supernatant. This growth yield is consistent with the accumulating metabolite containing about half of the carbon present in the steroid molecule. Moreover, the ability of RHA1, but not of ΔfadD3, to grow on HIP indicates that steroid rings C/D are catabolized. The ability of fadD3 from Mtb to rescue the mutant's phenotypes strongly suggests that the Mtb orthologue has the same physiological role. This was confirmed by the biochemical and kinetic studies, which demonstrate that FadD3Mtb efficiently catalyses the CoA thioesterification of HIP. Importantly, the enzyme did not transform HIDP, a HIP derivative harbouring an isopropionyl side-chain at position 1, or a variety of other carboxylate-bearing cholesterol catabolites. Taken together, these results establish that FadD3 is a HIP-CoA synthetase in Mtb and other mycolic acid-producing actinomycetes, and further indicate that the steroid side-chain must be completely degraded before rings C/D can be catabolized. The FadD3-catalysed thioesterification presumably initiates a round of β-oxidation to yield HIC-CoA, whose occurrence was suggested by van der Geize et al., (2011).
The role of FadD3 is consistent with a number of previous studies. First, studies in R. equi by Miclo and Germain showed that degradation of HIP begins by the ATP-dependent activation of HIP to HIP-CoA catalysed by an unidentified acyl-CoA synthetase (Miclo and Germain, 1990). Second, deletion of the fadD3 orthologue in C. testosteroni lead to accumulation of HIP, 3′-OH HIP and 3′,7-diOH HIP during growth on ADD, chenodeoxycholate and cholate, respectively (Horinouchi et al., 2006), suggesting that FadD3 also initiates steroid rings C and D catabolism in Gram-negative bacteria. Nevertheless, the role of FadD3 contradicts the proposed role of IpdAB, a predicted heterodimeric CoA transferase whose genes, like fadD3, occur in the kstR2 regulon of rhodococci and mycobacteria. IpdAB has been proposed to catalyse the production of HIP-CoA in the horse pathogen R. equi (van der Geize et al., 2011). However, this conclusion was based on the accumulation of HIP during biotransformation of 4-androsten-3,17-dione (4-AD) by an ipdAB mutant: the reaction catalysed by IpdAB has not been directly demonstrated. It is possible that the accumulation of HIP results from an indirect effect of inactivating ipdAB, and that IpdAB catalyses a reaction downstream of FadD3, as proposed for FadE30 (van der Geize et al., 2011).
Hydroxylated metabolites corresponding to 5α-OH HIP and 1β-OH HIP observed in the ΔfadD3 mutant have been reported in other cholesterol catabolic mutants. Thus, a ΔhsaA mutant of RHA1 accumulated significant amounts of the 9- and 17-hydroxylated analogues of 3-hydroxy-9,10-seconandrosta-1,3,5(10)-trien-9,17-dione (i.e. 3,9-DHSA and 3,17-DHSA where steroid C9 and C17 correspond to C5 and C1 of HIP respectively) during growth on cholesterol (Dresen et al., 2010). 17-OH metabolites, and by extension 1-OH HIP, may be formed during side-chain degradation. Thus, it has been proposed that Ltp2 (Rv3540c) catalyses retro-aldol cleavage of the side-chain to yield a ketone at C17 (Thomas et al., 2011). However, the cluster also includes hsd4A (Rv3502c) which may encode a 17-ketosteroid reductase (van der Geize et al., 2007). The physiological relevance of the 9-OH metabolites and 5-OH HIP is less clear. Although 5α-OH HIP-CoA has been proposed to be an intermediate of steroid catabolism (Miclo and Germain, 1990; van der Geize et al., 2011), the relatively poor specificity (kcat/Km) of HsaA for 3,9-DHSA (Dresen et al., 2010) and of FadD3 for 5α-OH HIP indicate that the oxo group is reduced after thioesterification of HIP. Thus, the occurrence of the hydroxylated analogues of upstream steroid metabolites may reflect the adventitious action of a reductase on those metabolites when they accumulate at high levels in a mutant.
The requirement that the steroid side-chain be completely degraded prior to catabolism of rings C/D is supported by a recent study of a Δigr mutant of Mtb and contrasts to the emerging paradigm of the concurrent degradation of the side-chain and rings A/B. Thus, the Δigr mutant accumulated a methyl ester of HIDP during growth on cholesterol (Thomas et al., 2011). The presence of a partially degraded side-chain in this metabolite is consistent with the specificity of FadD3Mtb and indicates that the first two steroid rings can be degraded prior to complete side-chain degradation. The first indication that cholesterol side-chain and rings A/B catabolism occur concurrently in Mtb arose from the observation that 3-ketosteroid 9α-hydroxylase (KshA), which catalyses the opening of ring B with concomitant aromatization of ring A (Fig. 1), preferentially acts on CoA thioesters of partially degraded side-chain intermediates (Capyk et al., 2011). The results of Thomas et al. indicate that the Hsa enzymes that degrade ring A also act on metabolites with partially degraded side-chains.
The assigned role of FadD3 and the inability of the ΔfadD3 mutant to degrade HIP add to the burgeoning evidence that the kstR2 regulon encodes the degradation of steroid rings C and D. Originally identified in M. smegmatis (Kendall et al., 2010), the regulon is upregulated in RHA1 during growth on cholesterol (van der Geize et al., 2007) as well as on cholate (Mohn et al., 2012), suggesting its involvement in degradation of both steroids. Indeed, in bacteria containing more than one steroid catabolic cluster, such as RHA1 (McLeod et al., 2006), there is no obvious paralogue of the kstR2 regulon in these other clusters (Mohn et al., 2012). Although the roles of most of the kstR2 regulon genes remain to be elucidated (Kendall et al., 2010), other genes that appear to be involved in HIP degradation include fadE30, encoding an acyl-CoA dehydrogenase, and ipdAB, encoding a predicted CoA transferase. More specifically, based on studies in R. equi, it has been proposed that FadE30 catalyses the dehydrogenation of the thioesterified HIP side-chain (van der Geize et al., 2011). As noted above, it is possible that IpdAB catalyses a reaction downstream of FadE30.
Transposon mutagenesis studies indicate that fadD3 is required for normal growth of Mtb on cholesterol although the phenotype appears to be markedly less severe than predicted for other cholesterol catabolic genes (Griffin et al., 2011). The authors used a combination of high-density mutagenesis and deep-sequencing that has two advantages over previous microarray-based studies: (1) mutants can be mapped to a much higher resolution; and (2) the relative abundance of each mutant in different libraries can be compared over a much greater range (up to 1000-fold in this case). More particularly, the ratio of transposon mutants that were detected during growth on cholesterol versus glycerol suggest that disruption of fadD3 does not impair growth as severely as disruption of genes encoding enzymes that act immediately upstream of FadD3, such as kshA, hsaA, hsaC and hsaD, or downstream, such as fadE30 and ipdA. This partial impairment in Mtb is consistent with our finding that the ΔfadD3 mutant of RHA1 grows at a reduced rate on the cholesterol side-chain and rings A/B. Finally, the assigned role of FadD3 and the partial growth of fadD3 disruption mutants on cholesterol are also consistent with the transposon mutagenesis data that indicate that this gene is not essential for survival of Mtb in macrophages or mouse models of infection (Sassetti and Rubin, 2003; Rengarajan et al., 2005).
The relatively mild phenotype of fadD3 mutants as compared with those in genes encoding enzymes that act immediately upstream or downstream of FadD3 is intriguing inasmuch as disruption of these genes is predicted to lead to the accumulation of a CoA thioester except in the case of fadD3. As discussed above, disruption of kshA, hsaA, hsaC and hsaD are predicted to result in the accumulation of CoA thioesters of ring A/B metabolites while disruption of fadE30 and ipdA are predicted to lead to accumulation of CoA thioesters of ring C/D metabolites. A number of CoA thioesters have been shown to be toxic, including hydroxycinnamoyl-CoAs, which cause bacteriostatic effects in Acinetobacter (Parke and Ornston, 2004), and propionyl-CoA, which is toxic in Aspergillus nidulans (Brock and Buckel, 2004). Propionyl-CoA and/or propionyl-CoA-derived metabolites have also been proposed to be toxic in Mtb and some Gram-negative bacteria (Man et al., 1995; Rocco and Escalante-Semerena, 2010; Griffin et al., 2012). One explanation that has been proposed for the toxicity of these thioesters is that they sequester the coenzyme and thus deplete the CoASH pool (Man et al., 1995; Brock and Buckel, 2004; Parke and Ornston, 2004). Such an effect could help explain why the deletion of some cholesterol catabolic genes affects growth and pathogenesis more than the disruption of others. More specifically, the sequestration of CoA in the bacterial cell may contribute to the attenuated virulence that has been reported for the igr, kshA and hsaC mutants in Mtb (Chang et al., 2009; Yam et al., 2009; Hu et al., 2010) as well as for fadE30 and ipdAB mutants in R. equi (van der Geize et al., 2011). Clearly, more studies are required to determine (i) whether CoA thioesters represent a general class of toxic metabolites and (ii) their role in attenuating virulence.
Cholesterol catabolism has been identified as being critical for Mtb during infection (Pandey and Sassetti, 2008; Ouellet et al., 2011). Nevertheless, the precise role of this catabolism has yet to be elucidated. Undoubtedly, some of the apparently contradictory results are confounded by differences between Mtb strains, exemplified by the differential expression of cyp142 in H37Rv and CDC1551 (Johnston et al., 2010), and the limitations of various infection models. Determining the roles of cholesterol-degrading enzymes contributes to deciphering the mechanisms allowing the pathogen to utilize this growth substrate and, ultimately, to determining the role of this catabolism in virulence.
Chemicals and reagents
ATP, PEP, CoASH, NADH, cholesterol, 5α-cholestane, cholic acid, deoxycholic acid, 2-hydroxypropyl-β-cyclodextrin, l-lactic dehydrogenase, adenylate kinase and pyruvate kinase were purchased from Sigma-Aldrich (St. Louis, MO). Sodium pyruvate and sodium benzoate were purchased from MG Scientific (Pleasant Prairie, WI) and Fisher Scientific (Ottawa, ON) respectively. 3-Oxo-23,24-bisnorchol-4-en-22-oic acid (4-BNC) was purchased from Steraloids, (Newport, RI). 3aα-H-4aα(3′-Propanoic acid)-5α-hydroxy-7aβ-methylhexahydro-1-indanone-δ-lactone (HIL) and 3β-hydroxy-5-cholesten-26-oic acid were kindly provided by Dr Robert van der Geize (University of Groningen, the Netherlands). Restriction enzymes and T4 DNA ligase were purchased from New England Biolabs (Ipswich, MA). Expand High Fidelity PCR system and GoTaq DNA polymerase were purchased from Roche (Laval, QC) and Promega (Madison, WI) respectively. Oligonucleotides were purchased from Integrated DNA Technologies (San Diego, CA) through the Nucleic Acid Protein Service Unit (NAPS) at the University of British Columbia (UBC). PolyHis-tagged tobacco etch virus proteinase (TEVpro) was obtained as described previously (Kapust et al., 2001). All other reagents were of HPLC or analytical grade. Water for buffers was purified using a Barnstead Nanopure DiamondTM system (Dubuque, IA) to a resistivity of at least 18 MΩ.
Strains and plasmids used in this study are provided in Supporting information (Table S1). RHA1 (Seto et al., 1995) and derivative strains were cultivated in baffled flasks at 30°C under aerobic conditions (200 r.p.m.) in 100 ml of M9 mineral medium (Sambrook and Russell, 2001) supplemented with trace elements, vitamin B1 (Bauchop and Elsden, 1960) and 1.0 mM cholesterol or 1.0 mM cholate. For growth on HIP, cells were cultivated at 30°C and 200 r.p.m. in 16 × 150 mm glass tubes containing 6 ml of M9 mineral medium supplemented with trace elements, vitamin B1 and 1.0 mM HIP. Cholesterol was added in water, then autoclaved. Cholic acid was added aseptically to autoclaved water from a solution in DMSO. HIP was added aseptically from a filter-sterilized solution in water. M9 salts and supplements were then added aseptically. For the complementation assays, the media were supplemented with thiostrepton (2.5 μg ml−1) and chloramphenicol (34 μg ml−1), added aseptically from solutions in DMSO. Growth was monitored spectrophotometrically at 600 nm. E. coli strains were grown at 37°C (200 r.p.m.) in Luria–Bertani (LB) broth. Bacto agar, 1.5% w/v, (Difco) was used for solid media. As appropriate, kanamycin (25 μg ml−1), chloramphenicol (34 μg ml−1) or carbenicillin (50 μg ml−1) was added to LB media.
Monitoring of metabolite accumulation
Samples of 0.9 ml were collected from cultures at regular time intervals and were amended with 5α-cholestane to 0.06 mM as an internal standard. The samples were acidified with glacial acetic acid (90:1 v:v ratio) and metabolites were extracted at room temperature with an equal volume of ethyl acetate, dried under nitrogen and dissolved in 50 μl of pyridine. The extracts were derivatized at room temperature with 50 μl of bis(trimethylsilyl)-trifluoroacetamide/trimethylchlorosilane and analysed by gas chromatography-coupled mass spectrometry (GC-MS) on an Agilent 6890 series GC equipped with an HP-5ms 30 m × 250 μm capillary column (Hewlett-Packard, Palo Alto, CA) and an HP 5973 mass-selective detector. The GC conditions were as follows: the injector temperature was set at 280°C, the transfer line temperature was 290°C and the flow rate was 1 ml min−1 with helium. The temperature programme of the oven was 104°C for 2 min, increased to 290°C at a rate of 15°C per min, then held at 290°C for 15 min. The MS conditions were set to electron emission scanning at 40–800 m/z, at 1.97 scans per second. Relative abundance and absolute concentrations were calculated on the basis of total ion current signal, with samples normalized on the basis of the internal standard.
Cloning of fadD3
DNA was propagated, digested, ligated and transformed using standard protocols (Sambrook and Russell, 2001). Plasmid DNA was purified and transformed into E. coli by electroporation using a MicroPulser from Bio-Rad (Hercules, CA) with Bio-Rad 0.1 cm GenePulser cuvettes. The fadD3 gene was amplified from Mtb H37Rv genomic DNA with primers Rv3561-F and FadD3-PET-R (Table S2) using a polymerase chain reaction containing 100 ng template DNA, 10 units of the GoTaq DNA polymerase, 50 μM of each dNTP, 0.4 M betaine, and 60 pmol of each oligonucleotide in a volume of 200 μl. The reaction was subject to 25 temperature cycles using a Veriti Thermal cycler (Life Technologies, Burlington, ON) as follows: 95°C for 30 s, 62°C for 30 s and 72°C for 180 s. The fadD3 amplicon was digested with NdeI and HindIII, and ligated into pTipQc2 (Nakashima and Tamura, 2004) to yield pTipFD3. To generate the affinity-tagged FadD3 protein, the Rv3561-HEV-F and FadD3-PET-R primers (Table S2) were used to amplify the gene using pTipFD3 as the template DNA and the same conditions as above. The amplicon was digested with NdeI and HindIII, and ligated into pET-41b(+) (EMD Millipore) to yield pETVFD3. FadD3 produced from this construct carries an N-terminal hexahistidine tag that can be proteolytically removed using TEVpro. The proteolytically treated FadD3 is identical in amino acid sequence to wild-type protein except that the N-terminal methionine is replaced by a serine. DNA was sequenced at NAPS at UBC.
RHA1ΔfadD3 (ΔfadD3 hereafter) was generated by the previously described sacB counterselection technique (van der Geize et al., 2001). Briefly, the fadD3 flanking regions were amplified from RHA1 genomic DNA using the fadD3-up-F and fadD3-up-Rev primer pair and the fadD3-down-F and fadD3-down-Rev primer pair respectively (Table S2). Triple-ligation using T4 DNA ligase, with the cut sites for KpnI, HindIII and XbaI, was performed to clone the amplicons into pK18mobsacB (Schäfer et al., 1994) to produce pK18fadD3. The mutagenic construct pK18fadD3 was introduced in wild-type RHA1 by conjugation as described elsewhere (Swain et al., 2012). To confirm the deletion, kanamycin sensitive, sucrose-resistant colonies were screened by colony PCR using the fadD3up-1063F and fadD3down-100R primers (Table S2). For the complementation assays, the pTipFD3 plasmid (see above) was electroporated into ΔfadD3 to generate the complementation strain ΔfadD3c, whereas pTipQc2 was electroporated in RHA1 and ΔfadD3 to generate the control strains RHA1[pTipQc2] and ΔfadD3[pTipQc2].
Protein production and purification
FadD3Mtb was produced using E. coli Rosetta(DE3)pLysS freshly transformed with pETVFD3. Cells were grown at 16°C, 200 r.p.m. in LB broth supplemented with 25 μg ml−1 kanamycin and 34 μg ml−1 chloramphenicol. When the culture reached mid-log phase (OD600 of 0.6), isopropyl β-d-1-thiogalactopyranoside was added to 0.5 mM to induce fadD3 expression. Cells were incubated for a further 20 h, harvested by centrifugation at 4°C (4000 g × 10 min) and frozen until further use. The cell pellet from 5 l of culture was suspended in 20 ml of 9 mM sodium phosphate, pH 8.0 containing 10% glycerol, 50 μg ml−1 DNase I and the Complete Mini Protease Inhibitor cocktail (Roche, Laval, QC). Cells were subjected to five rounds of 25 s bead beating using a MP Biomedicals FastPrep-24 bead beater (Solon, OH) set to 5.0 with 5 min incubation on ice between rounds. Cell debris was removed by centrifugation (17 000 g × 20 min, 4°C) and the supernatant was further clarified by ultracentrifugation (255 000 g × 2 h, 4°C). The cytoplasmic extract was recovered, supplemented with 10 mM imidazole and loaded onto 5 ml Ni-Sepharose 6 Fast Flow resin (GE Healthcare). The tagged protein was eluted using ice-cold 20 mM sodium phosphate, pH 8.0 containing 50 mM NaCl and 500 mM imidazole. The recovered protein was dialysed overnight at 4°C against 2 l of 25 mM HEPES, pH 7.5, 50 mM KCl, then digested with TEVpro. Briefly, 50 mg of polyHis-tagged FadD3 was incubated at 22°C for 22 h in presence of 1 mg TEVpro in a total volume of 10 ml containing 50 mM Tris-HCl, pH 8.0, 0.5 mM EDTA and 1 mM DTT. Cleaved FadD3 was separated from cleaved tag and the TEVpro by passing the mixture on a 0.4 ml Ni-Sepharose column equilibrated with ice-cold 20 mM sodium phosphate, pH 8.0, 300 mM NaCl, 10 mM imidazole. The FadD3Mtb was recovered in the flow-through and dialysed overnight at 4°C against 2 l of 25 mM HEPES, pH 7.5, 50 mM KCl. FadD3Mtb was concentrated at 4°C by centrifugation in an Ultracel 10 K (EMD Millipore) to a final concentration of 85 mg ml−1, flash-frozen as beads in liquid nitrogen and stored at −80°C.
Preparation of steroid metabolites
Metabolites were produced by incubating steroids with specific RHA1 mutants and purifying them by high-performance liquid chromatography (HPLC). To produce HIP and 1β-OH HIP, the ΔfadD3 mutant was grown in baffled flask at 30°C (200 r.p.m.) in 1 l of M9 mineral media supplemented with trace elements, vitamin B1, 20 mM benzoate and 0.2 mM cholesterol. At mid-log (OD600 ∼ 3.0), cells were harvested by centrifugation at room temperature, and resuspended in 200 ml of M9 mineral media supplemented with traces elements, vitamin B1 and 1.5 mM cholesterol (115 mg). The cell suspension was incubated for 110 h at 30°C with shaking (200 r.p.m.). Cells were again harvested by centrifugation, the supernatant was acidified to pH ∼ 5 and the metabolites were recovered using three successive extractions with one equivalent of ethyl acetate. The organic fractions were combined and evaporated to dryness. The residue was suspended in 1.5 ml of ethyl acetate, passed through a 0.2 μm filter and loaded onto a 250 × 10 mm Luna 5 μm silica(2) column (Phenomenex, Torrance, CA) connected to a Waters binary pump 1525 HPLC module (Milford, MA) equipped with a Waters 2996 photodiode array detector and operated at 11.1 ml min−1. The metabolites were eluted using a 25 min gradient from 0 to 59.7% ethyl acetate in hexanes and 0.5% acetic acid. The eluate was monitored at 254 and 285 nm. Under these conditions, HIP and 1β-OH HIP had retention times of 18.0 and 24.0 min respectively. Fractions of HIP and 1β-OH HIP were recovered, evaporated to dryness under nitrogen and stored at −80°C. To obtain HIDP, the ΔcamB mutant (Swain et al., 2012) was grown and incubated essentially as described above except that 0.2 mM cholate was used to generate the biomass and 2.0 mM deoxycholate (80 mg) was used instead of 1.5 mM cholesterol for the biotransformation. Metabolites were recovered from the supernatant as described above, suspended in 1.5 ml of ethyl acetate, filtered and loaded onto a 150 × 3 mm Luna 5 μm silica(2) column (Phenomenex) connected to a Series 1100 HPLC module (Agilent Technologies, Santa Clara, CA) operated at 0.5 ml min−1. HIDP was purified using a linear gradient of 0 to 7.4% isopropanol in hexanes and 0.5% acetic acid over 18.5 min. The eluate was monitored at 230 nm. Under these conditions, HIDP had a retention time of 17.5 min. HIDP was recovered, evaporated to dryness and stored at −80°C.
HIP, 1β-OH HIP and HIDP were dissolved to concentrations of 50 mM in methanol. 4-BNC was dissolved to 50 mM in 94% ethanol containing 60 mM sodium hydroxide. HIL solution (50 mM) was prepared in water. 5α-OH HIP solution (100 mM) was prepared by dissolving HIL in 0.5 M sodium hydroxide. Cholic acid solution (100 mM) was prepared in 100% DMSO. 3β-hydroxy-5-cholesten-26-oic acid solution (1 mM) was prepared in 10% 2-hydroxypropyl-β-cyclodextrin.
HPLC-based activity assay
Reactions were performed in 100 μl of 100 mM HEPES, pH 7.3 containing 5 mM MgCl2, 2.5 mM ATP, 1 mM CoASH, 2 μM FadD3Mtb and 1 mM of the tested substrate (Fig. 7), except for 3β-hydroxy-5-cholesten-26-oate that was used at a final concentration of 0.2 mM. Reaction mixtures were incubated for 4 h at 22°C, then stopped by adding an equal volume of methanol and incubated on ice for 10 min. Methanol was evaporated at room temperature under low pressure and the mixture was centrifuged (16 000 g × 10 min, 4°C). The supernatant was recovered, passed through a 0.2 μm filter and stored on ice. Reaction products were analysed using a Waters 2695 Separations HPLC module (Milford, MA) equipped with a Waters 2996 photodiode array detector and a Luna 3 μm PFP(2) 50 × 4.6 mm column (Phenomenex), using a linear gradient of 0 to 90% methanol in 0.1 M ammonium acetate, pH 4.5 over 20 min at 1 ml min−1. The eluate was monitored at 258 nm.
Purification and analysis of CoA thioesters
CoA thioesters were purified by HPLC on a Luna 3 μm PFP(2) 50 × 4.6 mm column (Phenomenex) equilibrated with 0.1 M ammonium acetate, pH 4.5. One millilitres of samples were injected onto the column. CoA thioesters were eluted using a 20 ml gradient of 0 to 90% methanol in 0.1 M ammonium acetate, pH 4.5 and the eluate was monitored at 258 nm. Under these conditions, the retention times for HIP-CoA, 5α-OH HIP-CoA and 1β-OH HIP-CoA were 10.0, 9.8 and 9.7 min respectively. Identities of the CoA thioesters were established by subjecting them to alkaline hydrolysis for 1 h at pH 13, 50°C. The solutions were then acidified to pH ∼ 5 with glacial acetic acid and examined by HPLC to verify the generation of CoASH. The hydrolysed products were extracted with ethyl acetate, dried, dissolved in pyridine, derivatized with bis(trimethylsilyl)-trifluoroacetamide/trimethyl-chlorosilane and subjected to GC-MS analysis to confirm the generation of HIP, 5α-OH HIP or 1β-OH HIP. Additionally, purified sample of 5α-OH HIP-CoA was subjected to electron-spray ionization mass spectrometry (ESI-MS) using a Bruker Esquire-LC ion trap mass spectrometer equipped with an ESI ion source, which was operated in the positive ion mode. Sample was dissolved 1:9 in methanol to a final concentration of 30 pmol μl−1. Five microlitres was injected at a flow rate of 200 μl min−1 in 1:9 water:methanol. The scanned mass range was m/z 50–2000 Da.
Steady-state kinetic studies of FadD3Mtb were performed using a spectrophotometric assay that couples AMP formation to NADH oxidation (Horswill and Escalante-Semerena, 2002). Reactions were typically performed in a total volume of 0.25 ml of 0.1 M HEPES, pH 7.3 containing 1 mM CoASH, 2 mM ATP, 2 mM PEP, 5 units pyruvate kinase, 5 units adenylate kinase, 20 units lactate dehydrogenase, 100 μM NADH, 5 mM MgCl2 and different amounts of HIP (0–100 μM) or 5α-OH HIP (0–600 μM). The specific activity for 1β-OH HIP was determined at a single concentration (500 μM). The mixture was incubated 5 min at 22°C and the reaction was initiated by adding 5 nM or 100 nM FadD3Mtb for HIP or hydroxy-HIPs kinetics respectively. Initial velocity was recorded at 340 nm over 1 min. Apparent steady-state kinetic parameters for CoASH (0–1500 μM) were determined in the presence of 100 μM HIP. Kinetic parameters were evaluated by fitting the Michaelis–Menten equation to the data using the least-squares fitting and dynamic weighting options of LEONORA (Cornish-Bowden, 1994). For the inhibition studies, reactions containing 10 μM HIP and 5 nM FadD3 were performed in the presence of HIDP (100–200 μM) or HIL (200–400 μM).
NMR measurements were carried out at the UBC Chemistry High-Resolution NMR Analytical Services and Research Support Facility, with approximately 1 mg of HPLC-purified HIP dissolved in deuterated methanol (CD3OD) for one-dimensional NMR or deuterated chloroform (CDCl3) for homonuclear two-dimensional correlated spectroscopy (COSY). All NMR spectra were recorded at 298.1 K on a Bruker Avance 400 MHz spectrometer equipped with a 5 mm BBI-Z inverse broadband probe with a Z-gradient coil.
NMR spectra of HIL, 5α-OH HIP and 1β-OH HIP were recorded at 300 K using a 500 MHz (for HIL) or 600 MHz spectrometer (for hydroxy-HIPs). HIL was prepared to 10 mM in deuterated water (D2O) and pD was measured to be 7.0. 5α-OH HIP was generated by dissolving HIL in D2O to 20 mM then adjusting with deuterated sodium hydroxide (NaOD) to pD 13 to allow for hydrolysis of HIL. HPLC-purified 1β-OH HIP was prepared to ∼ 15 mM in D2O. One-dimensional proton spectra of HIL, 5α-OH HIP (diluted to 10 mM with D2O) and 1β-OH HIP were obtained at the aforementioned pD. Additionally, homonuclear COSY spectra were recorded for the hydroxy-HIPs. To monitor the lactonization of 5α-OH HIP, 20 mM 5α-OH HIP was diluted to 10 mM using deuterated sodium phosphate buffer (NaD2PO4), and pD was adjusted to 7.0. One-dimensional proton spectra were recorded every 10 min for 12 h.
This research was supported by an operating grant from the Canadian Institutes for Health Research (CIHR) to L.D.E. Dr Robert van der Geize is thanked for providing HIL and 3β-hydroxy-5-cholesten-26-oic acid. Dr Lawrence MacIntosh, Dr Mark Okron and Antonio Ruzzini are thanked for providing assistance with the NMR data acquisition. Jason Rogalski is thanked for his help with acquiring MS data. Israël Casabon is a recipient of a postdoctoral fellowship from the Fonds de Recherche en Santé du Québec and the Michael Smith Foundation for Health Research.