Discovery of a novel stereospecific β-hydroxyacyl-CoA lyase/thioesterase shared by three metabolic pathways in Mycobacterium tuberculosis

The vast number of poorly characterised enzymes in Mycobacterium tuberculosis (Mtb) is one of the key barriers precluding a better understanding of the biology that underpins pathogenesis. Here, we investigated the Mtb orphan enzyme Rv2498c to delineate its physiological role. Our results from in vitro enzymatic assays, phylogenetic analysis, X-ray crystallography and in vivo Mtb experiments, de-orphan Rv2498c as a multi-functional β-hydroxyacyl-CoA lyase/thioesterase (β-HAClyase/thioesterase) that participates in three different metabolic pathways: L-leucine catabolism, itaconate dissimilation, and glyoxylate shunt. Moreover, the deletion of the rv2498c gene from the Mtb genome resulted in attenuation in the mouse model compared to infection with the parent strain. To the best of our knowledge, this is the first report of an (R)-3-hydroxyl-3-methylglutaryl-CoA for leucine catabolism and an itaconate-specific resistance mechanism in Mtb.


Introduction
Mycobacterium tuberculosis (Mtb) is the aetiological agent of tuberculosis (TB). TB is one of the leading causes of death worldwide, accounting for approximately 1.3 million deaths in 2016, and is estimated to infect one-fourth of the global population. [1,2] The pathogenicity, physiological resiliency and plasticity of Mtb are notably complex, with humans serving as their only known reservoir, highlighting the effect of niche adaptation on pathogen evolution. [3,4] However, Mtb biology remains largely unexplored and over half of the enzymes in the proteome remain 'orphan enzymes', i.e. without a defined catalytic activity. [5,6] To date, delineating the in vitro activities and in vivo functional roles of hundreds of poorly characterised enzymes has been a significant challenge. [7] This difficulty is in part due to the unconventional nutrient assimilation of Mtb and its ability to persist in different metabolic states. [8][9][10][11][12] Without an adequate understanding of the fundamental biology underpinning infection and the associated metabolic networks, we continue to generate drug candidates that fail in eradicating the pathogen. Thus, a deeper understanding of the Mtb biology is pivotal to tackle the TB pandemic.
Here we report on the functional characterisation of the Mtb enzyme Rv2498c as a stereospecific multi-functional β-hydroxyacyl-CoA lyase/thioesterase (β-HAClyase/thioesterase) participating in three metabolic pathways: (1) L-leucine catabolism; (2) itaconate dissimilation; and (3) glyoxylate shunt. To the best of our knowledge, this is the first report that shows that Mtb can dissimilate itaconate, that Mtb leucine catabolism solely involves the (R)-HMG-CoA, and that Mtb has a second malate synthase. Additionally, we present Rv2498c ligand-bound crystal structures revealing the molecular foundations underlying the stereospecificity and of the β-hydroxyacyl-CoA thioester carbon-carbon cleavage. Overall, these findings offer new insights on the Mtb biology and physiological plasticity highlighting an example of a multi-functional enzyme.

Rv2498c has (R)-HMG-CoA lyase activity in vitro and ex vivo
Rv2498c is reported to encode the β-subunit CitE-like of citrate lyase complex. [7,13] Inconsistent with this annotation, the associated α-and γ-subunits needed to form the functional citrate lyase complex appear to be absent in Mtb. This disconnect prompted us to screen a diverse panel of commercially available CoA-thioesters using an UV-Vis HPLCbased assay (Supplementary Table 1).
In this screening, incubation of Rv2498c with HMG-CoA generated a product with a retention time similar to acetyl-CoA (Ac-CoA) and consumed exactly half of the HMG-CoA racemic mixture, suggesting absolute stereospecificity (Fig 1a); the stereoisomer (S)-HMG-CoA is a known metabolic intermediate in the leucine catabolic pathway, which is incompletely annotated in Mtb, while the role of the stereoisomer (R)-HMG-CoA is not well understood. [22] To unambiguously determine the stereospecificity of Rv2498c for HMG-CoA, we eliminated either the (R)-or (S)-isomer of HMG-CoA by taking advantage of the known stereospecific HMG-CoA lyases, P. aeruginosa PA0883 and PA2011. [18] Interestingly, we found that Rv2498c is a lyase specific for the (R)-HMG-CoA. Based on the observation that Ac-CoA is one product, the other likely product of C-C bond cleavage is acetoacetate (AAc), which is not visible in our UV-Vis HPLC assay. To directly detect the formation of AAc, we analysed the reaction products by 1 H-NMR spectroscopy and observed the formation of Ac-CoA and AAc from half of the (R/S)-HMG-CoA in the reaction mixture ( Fig. 1b). These results demonstrate the stereospecific (R)-HMG-CoA lyase activity of Rv2498c.
To validate the results obtained using purified recombinant Rv2498c, we investigated the stereo-chemical course of HMG-CoA degradation in cell-free protein extracts (CFPE) from Mtb H37Rv (parent) and from an rv2498c-knockout strain (rv2498cKO). Consistent with the results using recombinant Rv2498c, parent CFPE, but not rv2498cKO CFPE, degraded (R)-HMG-CoA (Fig. 1c). These results confirm the (R)-specific stereo-chemical course of HMG-CoA degradation in Mtb and demonstrate that Rv2498c is both necessary and sufficient for breakdown of the (R)-HMG-CoA in CFPE.

Rv2498c (S)-citramalyl-CoA bound structure reveals the molecular basis for stereospecificity
To investigate the molecular basis of Rv2498c substrate stereospecificity, we determined the X-ray crystal structure of Rv2498c in variety of liganded states (Supplementary Table 2).
All structures showed the same trimeric arrangement of protomers previously described for the unliganded structure and for the oxaloacetate-and Mg 2+ -bound structure. [13] But in contrast to the previously reported structures, the C-terminus (50 residues) is well ordered, forming an -helix/β-hairpin/-helix motif that packs against the surface of the neighbouring protomer, capping its active site (Fig. 2a). The fact that the C-terminus is ordered in these ligand-bound structures is consistent with earlier observations that the position and organisation of the C-terminus might depend on the occupancy of the active site ( Fig.   2b). [20] Rv2498c has no sequence and modest structural similarity to the characterised family of TIM These three proteins were crystallised with propionyl-CoA in their active sites.
In the structures presented here of Rv2498c bound to (S)-citramalyl-CoA (PDB: 6AQ4) and to acetoacetate (PDB: 6AS5), the CoA moiety binds in a deep cleft at the base of which resides the active site Mg 2+ ion. The magnesium ion is coordinated by the 3-hydroxyl group and by an unidentate interaction with the terminal carboxylate of the ligand, as well as by Glu112, Asp138 and two water molecules (Fig. 2c, d). The carboxylate of the substrate is further positioned by polar interactions with the backbone NH atoms of Ala136, Glu137 and Asp138, while the 3-hydroxyl group lies close to Arg64. The 3-methyl group of the ligand contacts a hydrophobic surface formed by Gly135 and the side-chain atoms of Met133 and Val181 (Supplementary Fig. S1). Of note, the structure with (S)-citramalyl-CoA bound was obtained from crystals that were soaked with pyruvate and Ac-CoA, therefore indicating that the conformation of Rv2498c in the crystals is catalytic active.
Because the citramalyl group observed in the active site was in the S configuration, which corresponds to the R configuration at the 3-position of HMG-CoA (Supplementary Fig. S2), we could use this structure to directly model the mode of binding of (R)-HMG-CoA. In this case, the additional methanediyl unit (C4) found in HMG-CoA is readily accommodated in the pocket and can support a less distorted octahedral coordination of the Mg 2+ ion (the bite angle of the citramalyl group is ~70°; the modelled HMG moiety bite angle is ~85°).
Likewise, Molprobity calculations revealed shape and charge complementarity between the protein and modelled substrate, consistent with the active site supporting energetically favourable interactions with (R)-HMG-CoA.
[27] Attempts to model (S)-HMG-CoA into the active site generated electrostatically unfavourable interactions. Interestingly, with respect to the human (S)-HMG-CoA lyase (Fig. 2e, PDB: 3MP5), the differential positioning of the magnesium ion coordination with conserved Asp and Glu residues in the active site contributes to the stereoselectivity between the HMG-CoA stereoisomers ( Fig. 2f-h).

Phylogenetic analysis suggests Rv2498c might have alternative substrates
Aiming to conduct a comprehensive comparison of Rv2498c in the context of related enzymes, we constructed an amino acid sequence phylogenetic tree of the PFAM HpcH/HpaI aldolase/citrate lyase family (PF03328; Fig. 3 CoA. [19,21,29] Rv2498c also displays thioester hydrolase and malate synthase activities. Motivated by the results of the phylogenetic analysis (Fig. 3), we tested if (S)-citramalyl- Unexpectedly, Rv2498c was found to subsequently catalyse the hydrolysis of the thioester bond of (S)-malyl-CoA and β-methylmalyl-CoA both in vitro and in the parent vs rv2498cKO CFPE experimental set up (Fig. 4c, Supplementary Fig. S4), and is therefore able to synthesise malate and methylmalate by acting as a bifunctional (S)-malyl-CoA lyase/thioesterase as has been described for H. marismortui AceB.
[32] In strict agreement with our results using recombinant enzyme, we found that (S)-citryl-CoA is readily hydrolysed by both parent and rv2498cKO CFPEs, confirming that Rv2498c is not the enzyme responsible for (S)-citryl-CoA hydrolysis (Supplementary Fig. S5).

Rv2498c participates in L-leucine and itaconate catabolism
We interrogated the role of Rv2498c in Mtb metabolism by comparing growth and targeted LC-MS metabolic profiles of rv2498cKO, parent and a rv2498c-complemented strain where the rv2498c gene is present elsewhere in the chromosome (rv2498cKO::rv2498c). These four strains were cultured in chemically defined media of composition similar to Middlebrook 7H10 for solid medium or 7H9 for liquid medium but with a single carbon source present. In accordance to our substrate specificity results using purified recombinant Rv2498c and CFPEs, we hypothesised that Rv2498c could be involved in at least three metabolic pathways: (1) L-leucine catabolism, in which (R)-HMG-CoA is a catabolic intermediate; (2) itaconate dissimilation, a process that involves (S)-citramalyl-CoA formation and; (3) glyoxylate shunt, where Rv2498c could function alongside or substitute for the bona fide malate synthase GlcB (Rv1837c).
Consistent with a role of Rv2498c in L-leucine metabolism, no growth was observed on solid or in liquid medium for the rv2498cKO strain in the presence of L-leucine as the sole carbon source. We inferred that this growth deficiency was likely the outcome of the absence of HMG-CoA lyase activity, resulting in the accumulation of HMG, limited production of Ac-CoA, and/or altered branched-lipid metabolism. Accordingly, in the presence of L-leucine as sole carbon source, rv2498cKO, but not the parent or the complemented strains, accumulated HMG ( Fig. 4a), as well as other leucine catabolism intermediates such as methylcrotonoate, methylglutaconoate and hydroxymethylglutarate (Supplementary Fig. S6). We also observed a slight pH difference in the liquid medium of the Mtb rv2498cKO culture (pH 6.7) compared to the parent strain (pH 6.9) and attributed this acidification to an increased concentration HMG in the medium, as we could verify by LC-MS (Supplementary Fig. S6).
The phenotype of the cultures grown in medium with itaconate as sole carbon source was less clear-cut. A growth defect was observed for rv2498cKO compared to parent and complemented strains on solid medium, but none of the strains grew in liquid medium.
Itaconate is a strong inhibitor of isocitrate lyase, the first enzyme of the glyoxylate shunt, and it is therefore considered to be an antimicrobial compound. The observed dissimilar growth pattern could be due to increased exposure to itaconate in liquid compared to solid medium (e.g. biofilm in solid medium reduces itaconate exposure). A pathway for itaconate degradation has not been described for Mtb, but in mammals and in some bacteria it is understood to proceed via activation of itaconate to itaconyl-CoA, stereospecific hydration to form (S)-citramalyl-CoA, and C-C bond cleavage to form pyruvate and acetyl-CoA. [18,33] Consistent with the existence of this pathway in Mtb and with the involvement of Rv2498c in itaconate degradation, we observed accumulation of citramalate, the hydrolysis product of citramalyl-CoA, in the rv2498cKO strain when grown in itaconate as sole carbon source (Fig.   4b).
Similar to the cultures grown in itaconate liquid medium, no growth was observed for any of the strains in liquid medium containing glyoxylate, suggesting that Mtb is unable to use glyoxylate as a carbon source. However, rv2498cKO did show a decreased growth phenotype compared to the parent and the complemented strains on solid medium, perhaps because the cells are able to metabolise impurities in the agar. Nevertheless, when cells were grown with acetate or with glyoxylate as sole carbon source, we detected a discreet but reproducible accumulation of the TCA intermediate aconitate, a metabolite preceding the bifurcation of TCA and glyoxylate shunt, in rv2498cKO compared to parent and complemented strains. It is possible that in this instance, the rv2498cKO phenotype is obscured by the activity of the canonical malate synthase (Fig. 4c).
These in vivo experiments corroborate our in vitro and ex vivo biochemical data and offer further evidence that Rv2498c is a multi-functional enzyme participating in at least three metabolic pathways in Mtb: leucine catabolism, itaconate dissimilation, and glyoxylate shunt.

rv2498cKO strain is attenuated in a mouse infection model
To  selectivity. We suggest that Rv2498c carbon-carbon cleavage of (R)-HMG-CoA follows a mechanism similar to that proposed for (S)-HMG-CoA lyase described by Fu et al.
( Supplementary Fig. S9) and that the difference in the (S)-and (R)-stereospecificity for HMG-CoA can be partially attributed to the β-positioned hydroxyl group of the CoAthioester coordination with the divalent metal.
[46] We observed that the Rv2498c catalytic site has ordered water molecules, suggesting a catalytic mechanism for the carbon-carbon cleavage with water participation by shuttling protons and/or acting as the nucleophile in hydrolysis. [20,46,47] The lyase reaction likely proceeds via the reversal of standard retroaldol condensation, as previously suggested. [46,48] In conclusion, we have defined the enzymatic activities of Rv2498c as a β-HAClyase/thioesterase and present full-length crystal structures of the protein, which

Materials
All biological and chemical reagents were purchased from Sigma Aldrich or Fisher Scientific, unless stated otherwise. The following reagent was obtained through BEI Resources, NIAID,

Purification of recombinant proteins
Cell pellets were suspended in buffer (50 mM HEPES (pH 7.5), 5 mM MgCl2 containing lysozyme (hen egg white), DNase (TURBO, Thermo Fisher Scientific Inc.), and Complete ® protease inhibitor cocktail EDTA-free (Pierce)) that is twice the volume of cell pellet. The suspensions were stirred on ice for 2 hr and then sonicated 3x 30 sec on ice.
The cell lysates were centrifuged at 20,000 rpm, 4 °C, for 1 hr. The clear supernatant was incubated at 4 °C for 3 hr with Ni-NTA resin that has been equilibrated with buffer containing 10 mM imidazole. The resin slurry was then transferred to a glass Econo-column (Bio-Rad).
The column was washed with the same buffer, and then performed a step-wise discontinuous imidazole gradient up to 500 mM imidazole. The fractions were analysed by SDS-PAGE. The protein fractions were pooled and concentrated using Vivaspin, and then purified by gelfiltration. The protein fractions were analysed by SDS-PAGE, pooled and concentrated using Vivaspin. The protein concentration was determined by BSA assay. The proteins were stored at -20 °C as 50% glycerol stock or -80 °C as 25% glycerol stock.

Phylogenetic analysis
A seed multiple sequence alignment (MSA) of selected members of the HpcH_HpaI aldolase/citrate lyase family (PF03328) was built based on an alignment of the structures available in the PDB. [2] Structures were visualised and edited using Chimera and the multiple structural alignment was computed by Mustang. [3,4] The PF03328 sequences in the fullalignment were downloaded from PFAM and aligned to the seed MSA using MAFFT E-INSi --add option. [5] Partial and highly divergent sequences were removed and redundancy was reduced to 90% ID or less using Jalview. [6] The model of protein evolution used was LG+I+G+F with  of 1.64 and p-inv of 0.01, as selected by Prottest 3.4. [7] Bootstap repeats and consensus trees were generated with PHYLIP. [8] Pairwise distances were calculated with TREEPUZZLE via the puzzleboot script. [9] Distance trees were calculated with BIONJ. [10] For the acyl-CoA lyase tree, sequences were aligned with MUSCLE and a maximumlikelihood tree was calculated with PhyML. [11,12]

Cell free protein extract preparation
Mtb cultures were grown at 37 °C in 7H9 medium supplemented with Tyloxapol and ADC to

Other techniques
DNA sequencing was performed by GATC Biotech (Konstanz, Germany). Protein concentration was determined by bicinchoninic acid (BCA) assay (Pierce BCA Protein Assay Kit), using bovine serum albumin as the standard.

LC/MS metabolomics
Methods are as described by Larrouy-Maumus et al. [16] HPLC-based activity assays Each CoA-thioester was detected as a single peak after being separated by HPLC with a Poroshell 120 EC-C18, 2.7 m, 4.6 x 50 mm column (Agilent) using the following elution condition: 1-min isocratic elution at 2% acetonitrile in buffer, followed by a 10-min linear gradient of 2-20% acetonitrile, with 2-min isocratic elution at 20% acetonitrile in buffer, and then back to 1-min isocratic elution at 2% acetonitrile in buffer at a flow rate of 0.5 ml/min.

H Proton nuclear magnetic resonance spectroscopy
Samples were prepared in D2O or D2O phosphate buffer (0.1 M, pD = 7.2). Proton nuclear magnetic resonance (δH) spectra were recorded on Bruker Avance III HD 400 (400 MHz), Bruker Avance III 600 (600 MHz), or Bruker Avance III HD 800 (800 MHz). All chemical shifts were quoted on δ-scale in ppm, with residual solvent as internal standard.

Construction of Mtb rv2498cKO mutant and complements
The construction of Rv2498c knockout followed methods described by Parish et al. [17] Briefly, an unmarked in frame deletion of the Rv2498c gene was made by amplifying 1.

Metabolite preparation
Mtb strains were grown at 37 °C in 7H9 medium supplemented with Tyloxapol and ADC to       Metabolic profiles show quantitative measurements of metabolite from filter culture growth on chemically defined media with added substrate (10 mM). The data are shown as mean values ± SD from three biological replicates.