Discovery of a glycerol 3-phosphate phosphatase reveals glycerophospholipid polar head recycling in Mycobacterium tuberculosis

Abstract

Functional assignment of enzymes encoded by the Mycobacterium tuberculosis genome is largely incomplete despite recent advances in genomics and bioinformatics. Here, we applied an activity-based metabolomic profiling method to assign function to a unique phosphatase, Rv1692. In contrast to its annotation as a nucleotide phosphatase, metabolomic profiling and kinetic characterization indicate that Rv1692 is a D,L-glycerol 3-phosphate phosphatase. Crystal structures of Rv1692 reveal a unique architecture, a fusion of a predicted haloacid dehalogenase fold with a previously unidentified GCN5-related N-acetyltransferase region. Although not directly involved in acetyl transfer, or regulation of enzymatic activity in vitro, this GCN5-related N-acetyltransferase region is critical for the solubility of the phosphatase. Structural and biochemical analysis shows that the active site features are adapted for recognition of small polyol phosphates, and not nucleotide substrates. Functional assignment and metabolomic studies of M. tuberculosis lacking rv1692 demonstrate that Rv1692 is the final enzyme involved in glycerophospholipid recycling/catabolism, a pathway not previously described in M. tuberculosis.

Each year, 1.4 million people succumb to tuberculosis, making Mycobacterium tuberculosis the deadliest bacterium affecting mankind (1). In addition, the dissemination of strains resistant to several antibiotics underscores the need for better understanding of this pathogen and for the development of novel vaccines and therapeutics (2, 3). Our approach to elucidating the unique pervasiveness of M. tuberculosis is through comprehensive discovery and characterization of metabolic pathways, as metabolism underlies survival of the bacteria both inside and outside the host and can contribute to phenotypic and genetic drug resistance.

The M. tuberculosis genome encodes for 4,043 genes, of which 3,933 encode proteins (4, 5). Many genes with essential functions, such as DNA replication, protein and RNA synthesis, and cell-division, have close homologs in other bacteria, and their functions are annotated largely on the basis of analysis of their counterparts. However, functions of at least one-third of the genes are unknown or putative (4). In particular, little is known about genes that are not conserved or conditionally important. Characterization of such genes presents a daunting task. M. tuberculosis is thought to be subjected to a myriad of conditions during its life cycle in the host, such as low pH, reactive oxygen and nitrogen species, and so on (6, 7). Understanding how M. tuberculosis adapts to and even thrives in such diverse environments is critical to our understanding of the pathology itself and for the discovery of novel therapeutics to treat tuberculosis.

We applied an innovative, activity-based metabolomic profiling (ABMP) approach to assign function to orphan (without a priori known substrates/products) mycobacterial enzymes and to discover unknown metabolic pathways. ABMP employs liquid chromatography coupled with accurate time-of-flight mass spectrometry to separate, identify, and quantify potential substrates and products in a mycobacterial extract of small molecules acting as a representative source of physiologically relevant substrates, cofactors, and allosteric regulators. One of the main advantages of ABMP is the lack of any inherent bias toward the number and type of substrates or toward the nature of a catalytic mechanism (8).

We applied this approach to an M. tuberculosis protein of an unknown function, Rv1692, annotated as a putative nucleotide phosphatase. Transposon saturation mutagenesis data indicate that rv1692 is not essential in vitro (9), but no information regarding is role in vivo is currently available. The enzyme belongs to the haloacid dehalogenase superfamily (HADSF), which contains a large number of enzymes that share a conserved core domain and yet catalyze diverse reactions (e.g., phosphatase, phosphonatase, phosphomutase, and dehalogenase) (10–12). The HADSF is further divided into three subfamilies, I, II (A and B), and III, according to the topology and the insertion of the cap domain that dictates substrate specificity (10–12). Subfamily I is characterized by a small α-helical bundle cap domain located between motifs I and II of the core domain. The subfamily II cap is located between motifs II and III of the core domain and consists of two different α/β folds, designated types IIA and IIB. Contrary to the other two subfamilies, subfamily III does not possess a cap domain and has only a core domain with a connecting loop serving in place of the cap domain (13).

Here, we report a detailed functional and structural characterization of Rv1692, revealing its substrate specificity and unique structural scaffold, as well as providing evidence that implicates it in an unexpected metabolic function in M. tuberculosis, the recycling/catabolism of glycerolphospholipid-derived polar heads.

Results and Discussion

Rv1692 Is a Unique HADSF Phosphatase.

HADSF enzymes contain a core Rossmann catalytic fold in which highly conserved residues coordinate a divalent metal. HADSF phosphatases act on chemically diverse substrates. Substrate specificity is determined by a structural insertion in the core, called a cap domain (10–14), which forms a lid over the active site cavity during catalysis. Rv1692 is a potential HADSF member conserved in mycobacteria. In addition to a clearly identifiable HADSF core, Rv1692 contains a weakly conserved cap domain that includes putative substrate specificity determinants (SI Appendix, Fig. S1), as indicated by sequence homology analysis. The closest characterized homolog of Rv1692 is NagD from Escherichia coli (28% identity of the HADSF core residues), a nucleotide phosphatase specific to UMP and GMP. However, unlike NagD, Rv1692 contains an extra 103-residue C-terminal region in addition to the core and cap domains (SI Appendix, Fig. S1). This region does not have an identifiable homolog. Phylogenetic analysis using Basic Local Alignment Search Tool (BLAST) and Molecular Evolutionary Genetic Analysis (MEGA5) (15, 16) indicates that the Rv1692 homologs that contain this extension of the HADSF core are confined to the Actinomycetales phylum (high GC content, Gram-positive bacteria). Interestingly, Rv1692 homologs from most pathogenic and opportunistic mycobacterial species such as M. tuberculosis, Mycobacterium ulcerans, and Mycobacterium marinum are found in group I (SI Appendix, Fig. S2 and Table S1), whereas most of the nonpathogenic homologs are found in group II, suggesting a possible role in disease. It is noteworthy that genes encoding additional NagD homologs lacking this extension are present in some species, suggesting a divergent function of Rv1692.

Search for Potential Substrates of Rv1692 by Activity-Based Metabolomic Profiling.

We cloned, expressed in E. coli, and purified Rv1692 (SI Appendix, Materials and Methods). Phosphatase activity assays indicate that despite its homology to NagD, UMP and GMP are not substrates of Rv1692 (SI Appendix, Table S2). This observation, together with the presence of an extra domain and a differentially evolved cap domain in Rv1692, suggest that Rv1692 has a distinct substrate specificity and function. To search for a substrate of Rv1692, we applied ABMP (8). In this method, we prepared an extract of polar metabolites from mycobacteria (17) and used it as a library of potential in vitro substrates. This small-molecule extract (SME) contains ∼1,700 unique ions (8). The identity and the abundance of ionizable polar small molecules in the SME are determined in both the absence and the presence of the purified Rv1692 and are analyzed in both positive and negative ion modes. Substrates and products are identified on the basis of their accurate mass (<5 ppm), retention time, and isotopic envelope. The ion with m/z = 325.0431 [M+H]⁺, which corresponds to UMP, did not show significant changes as a function of time and presence of Rv1692, which agrees with the lack of phosphatase activity with UMP (SI Appendix, Table S2). Unsupervised statistical analysis of the data was carried using Mass Profiler Professional, from Agilent Technologies. Of an initial 5,648 ions detected, only eight ions were identified as changing more than twofold (P < 0.01; SI Appendix, Fig. S3). Subsequent analysis indicated that all eight are false-positives.

As peak recognition and retention time variations are likely to mask positive results obtained by automated methods, we carried out manual inspection of all mass spectral data obtained in the experiments in both positive and negative ionization modes. Indeed, manual inspection of the data indicated that one ion with m/z 171.0067 [M-H]⁻ disappeared completely on addition of Rv1692 (Fig. 1). The extracted ion chromatogram corresponding to this m/z did not display a single peak but, instead, displayed a set of peaks, which is characteristic of phosphorylated small molecules and their adducts with Na⁺ ion inter alia (Fig. 1A). Early elution time was consistent with acidic or phosphorylated compounds. A single empirical formula, C₃H₅O₆P, was obtained for the m/z of 171.0067 [M-H]⁻. The most likely metabolite from known bacterial metabolic pathways that matches this formula is glycerol phosphate (theoretical m/z 171.0064, Δppm = 1.75). LC-MS assays with synthetic D,L-glycerol 3-phosphate (G3P) are consistent with this conclusion (SI Appendix, Fig. S4). Therefore, our metabolomic data suggest that Rv1692 breaks down glycerol phosphate into glycerol and inorganic phosphate. It is important to stress that other phosphorylated metabolites detectable in our LC-MS–based assay such as hexose-phosphate, dihydroxyacetone phosphate/glyceraldehydes 3-phosphate, and glycerophospholipids polar heads did not show changes in absence or presence of Rv1692 in a time-dependant manner (SI Appendix, Fig. S5).

Fig. 1.

Activity-based metabolomic profiling reveals glycerol phosphate phosphatase activity of Rv1692. Metabolomic profiling of SME was carried out in the presence (red trace) or absence (black trace) of Rv1692 in negative ion mode. (A and B) Extract ion chromatograms in the absence and presence of Rv1692, respectively. (Insets C and D) Region of interested of the mass spectrum (from the retention time of 1–2.5 min). The green box was added to facilitate visualization of the correct ion, which is not present on Inset D. In the absence of Rv1692, we observe an ion with m/z 171.0067, which is within a 1.75-ppm error of the theoretical expected value for glycerol phosphate (m/z 171.0064). In the presence of Rv1692, the 171.0067 ion is completely consumed. The ion with m/z 171.0103 is within a 22.8-ppm error of 171.0064; therefore, it represents another compound. The data are representative of two independent experiments.

Rv1692 Is a G3P Phosphatase.

To directly investigate the substrate specificity of Rv1692, steady-state kinetic measurements were performed using a variety of substrates. In total, we tested 32 phosphorylated metabolites, mainly from established metabolic pathways, including chemical analogs of glycerol phosphate (e.g., linear compounds, cyclic sugars and nucleotides, phosphorylated coenzymes, and phosphorylated amino acids; Table 1 and SI Appendix, Table S2). Among all the substrates tested, G3P, glycerol 2-phosphate, ribulose 5-phosphate, L-glycerol 3-phosphate, and a nonphysiological substrate para-nitrophenyl phosphate (pNPP) exhibited measurable activity (Table 1). Rv1692 preferentially catalyzed the hydrolysis of G3P (V/K_m = 0.86 × 10³ M⁻¹⋅s⁻¹). This enzyme was also able to act on glycerol 2-phosphate (V/K_m = 0.14 × 10³ M⁻¹⋅s⁻¹) and D-ribulose 5-phosphate (V/K_m = 0.14 × 10³ M⁻¹⋅s⁻¹), albeit with significantly lower efficiencies. pNPP was also hydrolyzed with a drastically lower efficiency (V/K_m = 0.05 × 10³ M⁻¹⋅s⁻¹; Table 1). Interestingly, we did not observe any activity toward D-glyceraldehyde 3-phosphate or dihydroxyacetone-phosphate, which are metabolites closely related to glycerol 3-phosphate. This signifies either that both hydroxyl groups of glycerol 3-phosphate are required for binding or that sp² hybridized substrates are not accommodated at the active site. In addition, we observed that the stereochemistry was also critical for activity. Even though the values of K_m obtained for both L-glycerol 3-phosphate and G3P are the same within experimental error, at 0.89 ± 0.1 mM and 1.10 ± 0.20 mM, respectively, the turnover rate constant (k_cat) for G3P (0.77 ± 0.03 s⁻¹) is 40-fold higher than that for L-glycerol 3-phosphate (0.02 ± 0.01 s⁻¹). Therefore, Rv1692 is a glycerol-phosphate phosphatase with a preference for D-glycerol 3-phosphate over L-glycerol 3-phosphate.

Table 1.

Binding and steady-state kinetic parameters for Rv1692

Substrates	Kd, μM*	kcat, s−1	Km, mM	Hill no.	kcat/Km, s−1⋅M−1 (×103)
G3P	165.3 ± 22.3	0.77 ± 0.03	0.89 ± 0.01	2	0.86 ± 0.08
D-ribulose 5-phosphate	201.8 ± 35.1	0.18 ± 0.05	1.30 ± 0.40	2	0.14 ± 0.06
Glycerol 2-phosphate	328.7 ± 23.6	0.18 ± 0.01	2.01 ± 0.18	2	0.09 ± 0.01
Para-nitrophenyl phosphate	60.8 ± 2.7	1.00 ± 0.20	19.30 ± 5.30	1	0.05 ± 0.02
L-glycerol 3-phosphate	158.2 ± 22.2	0.02 ± 0.01	1.10 ± 0.20	1	0.02 ± 0.01

Assays were performed in Tris⋅HCl 50 mM 1mM MgCl2 at pH 7.5, using the EnzCheck phosphate detection kit.

^*Kd measurements were performed by determining the quenching of intrinsic tryptophan fluorescence, with 1 μM Rv1692 in buffer Tris HCl 50 mM at pH 7.5.

Altogether, these data indicate that the substrate-binding pocket of Rv1692 can efficiently accommodate small linear carbon backbones, containing two free hydroxyl groups, but not cyclic sugars or nucleotides. The V/K_m value for G3P (∼10³ M⁻¹⋅s⁻¹) is below the threshold of 10⁴ M⁻¹⋅s⁻¹ expected for intermediate metabolism enzymes, suggesting a function outside core metabolism. Until now, only two D-glycerol 3-phosphate phosphatase activities from bacteria have been reported: one from Bacillus licheniformis (18) and the other from Corynebacterium glutamicum (19). Regarding the latter, Linder et al. found that the G3P phosphatase deletion mutant strain showed a significantly reduced growth rate but had a final biomass comparable to the wild-type strain when cultured in a medium containing 40 g/L glucose and 1 mM G3P as carbon sources. These data are in accordance with the inability of the mutant strain to rapidly metabolize G3P. The authors of that study concluded that this G3P phosphatase activity may function to prevent the intracellular accumulation of G3P under metabolically unbalanced conditions.

Divalent Metal Ion Specificity of Rv1692.

HADSF phosphatases require divalent metals, usually Mg²⁺ or Mn²⁺, for activity (20). To probe the specificity of Rv1692 toward divalent metals, we performed steady-state kinetic measurements of hydrolysis of pNPP as a function of concentration of the divalent metal. The kinetic parameters are presented in SI Appendix, Table S3. Among the divalent metals tested, Mg²⁺, Co²⁺, Mn²⁺, and Ni²⁺ supported catalysis. Co²⁺ and Mn²⁺ support eight- and fourfold higher catalytic efficiency than Mg²⁺ (V/K_act = 27.20 × 10³ M⁻¹⋅ s⁻¹, 13.27 × 10³ M⁻¹⋅s⁻¹, and 3.27 × 10³ M⁻¹⋅s⁻¹ for Co²⁺, Mn²⁺ and Mg²⁺, respectively). For Ni²⁺, the catalytic efficiency (V/K_act = 1.25 × 10³ M⁻¹⋅s⁻¹) is approximately half of that for Mg²⁺. Higher concentrations of divalent metals lead to enzyme inhibition (SI Appendix, Table S3). Although the specificity for Mg²⁺ is eight- and fourfold fold lower than those for Co²⁺ and Mn²⁺, respectively, Mg²⁺ is likely the physiologically relevant catalytic divalent metal ion. The free intracellular concentration of Mg²⁺ in E. coli is ∼1–2 mM (21), which is similar to the K_act value for Mg²⁺, whereas intracellular concentrations for Mn²⁺ (0.01 μM; ref. 22) and Co²⁺ (0.001 μM; ref. 23) are considerably smaller than the corresponding K_act values.

A GNAT-Like Fused Region Is Required for Rv1692 Solubility.

To clarify the role of the C-terminal region in the activity of Rv1692, we crystallized Rv1692 with Mg²⁺ and Ca²⁺ and determined these two structures (SI Appendix, Table S4). The crystal structures show that Rv1692 contains a HADSF fold and forms a dimer similarly to other HADSF phosphatases (10). The C-terminal extension of Rv1692, which is absent in other characterized HADSF members, resembles a small GCN5-related N-acetyltransferase (GNAT) fold (24) (Fig. 2). This region is fused to the HADSF catalytic domain and does not appear to be an independent domain. To check for an allosteric effect of this domain on the phosphatase activity, we measured the phosphatase activity of Rv1692 with G3P as substrate in the presence or absence of typical GNAT cosubstrates acetyl-CoA (AcCoA) and succinyl-CoA (SucCoA) or their product, CoA. (SI Appendix, Table S5). We also tested whether or not this fused GNAT domain possessed acetyltransferase or hydrolase activities. No effect on the phosphatase activity and no acetyl- or succinyltransferase or hydrolase activities were detected (SI Appendix, Table S6). This was further confirmed by ¹H-NMR, in which no formation of acetyl-glycerol or generation of CoA was observed (SI Appendix, Fig. S6). Altogether, these results indicate that this GNAT region is not likely to be involved in acetyl group transfer using AcCoA and SucCoA, it could nonetheless be a regulatory domain with a hitherto unknown mechanism of action.

Fig. 2.

Crystal structure of Rv1692. (A) Dimerization of Rv1692 is enabled through the cap domains of the monomers. The HADSF core, cap, and fused GNAT regions are highlighted in orange, gray, and green, respectively. (B) A 90° rotated view of the dimer of Rv1692 (C) A cartoon representation of a monomer showing the C-terminal GNAT region fused to the HADSF core.

Finally, we generated two Rv1692 truncation mutants lacking the GNAT region to assess its role in activity and/or structural stability. The truncated variants (Rv1692_1–265 or Rv1692_1–267) were well expressed in E. coli but were insoluble (SI Appendix, Fig. S7), indicating that the GNAT region is required for the solubility of the HADSF fold of Rv1692 and is potentially needed for the structural integrity of this enzyme. Indeed, deletion of the GNAT region exposes several hydrophobic residues of Rv1692 to solvent (SI Appendix, Fig. S1), which likely causes aggregation of the deletion mutants and may even cause their misfolding. Rv1692 is the first member of HADSF that contains an integral GNAT-like region, thus defining a new subfamily, subfamily C, belonging to the HADSF II group.

The Active Site and Substrate Binding.

Several crystal structures of NagD-like HADSF enzymes are available, but only a few of them contain bound substrates; for example, a predicted phosphatase from Saccharomyces cerevisiae with glycerol 3-phosphate bound (PDB code 3RF6), human pyridoxal 5′-phosphate phosphatase with pyridoxal 5′-phosphate bound (PDB code 2CFT), and human phospholysine/phosphohistidine phosphatase in complex with 4-(2-hydroxyethyl)-1-piperazine ethanesulfonic acid (PDB code 2X4D). In these structures, the cap domain is in a closed conformation. From comparing our structures of Rv1692 with these and other HADSF structures, it can be observed that the substrate binding pockets of both Rv1692 structures are in open conformation (SI Appendix, Fig. S8 A–C). We modeled the cap domain in the closed conformation to analyze the shape of the pocket and the substrate-interacting surface (SI Appendix, Fig. S8 D–F). The closed conformation of the cap of Rv1692 was modeled on the basis of the structures of the human pyridoxal 5′ phosphate phosphatase and the predicted phosphatase from S. cerevisiae. Because the conserved residues of the catalytic domain involved in Mg²⁺-scissile phosphate coordination are positioned nearly identically, they serve as reference, making such modeling highly predictive. In the model, the substrate binding pocket closely resembles that of the S. cerevisae phosphatase, where the cap domain provides a very similar surface for binding to glycerol 3-phosphate (Fig. 3A) and cannot accommodate larger substrates such as nucleotides or sugars (Fig. 3 B–D). In particular, residues I156 and P157 must contact the bound substrate similarly to residues W209 and A210 of the S. cerevisae protein. In agreement with substrate specificity studies, larger sugar phosphates and nucleosides cannot be accommodated in the small active site of Rv1692.

Fig. 3.

Substrate binding pocket of Rv1692. (A) The modeled pocket in closed conformation with modeled glycerol 3-phosphate (G3P; ligand from PDB code 3RF6). (B–D) The pocket with modeled adenosine 5′phosphate (AMP; from PDB code 3OCV), uridine 5′-phosphate (U5P; from PDB code 3OPX), and 1,7-di-O-phosphono-L-glycero-β-D-manno-heptopyranose (GMB; from PDB code 3L8G), respectively, showing clash of the ligands with Rv1692 active site residues.

Deletion of rv1692 in M. tuberculosis Results in an Accumulation of G3P and G3P-Containing Lipid Polar Heads.

To examine the G3P phosphatase activity of Rv1692 in vivo and to define its biological role, an in-frame, unmarked deletion mutant of rv1692 was constructed in M. tuberculosis H37Rv (Δrv1692) (SI Appendix, Fig. S9). LC-MS analysis of the metabolome of wild-type, Δrv1692, and complemented strains was performed as previously described (8, 25). Targeted metabolomic analysis shows an increase of the pool size of glycerol phosphate in the Δrv1692 compared with the wild-type (Fig. 4). The pool size of reference metabolites such as aspartic acid, glutamine, histidine, and serine remain unchanged (Fig. 4). This result is in accordance with the glycerol phosphate phosphatase activity assigned to Rv1692. Genetic complementation restored the glycerol phosphate pool to wild-type levels (Fig. 4), confirming that the effect on glycerol phosphate pool size is not a result of a polar effect on a downstream gene or other experimental artifacts. Notably, the level of specific lipid polar heads containing glycerol phosphate was also perturbed in the deletion mutant. Glycerophosphoinositol and phosphatidylglycerol levels increased twofold or more, but glycerophosphoethanolamine level remained unchanged (Fig. 4). All three of these polar heads are found in M. tuberculosis’s glycerophospholipids (26). It is noteworthy that these lipid polar heads are not biosynthetic intermediates in glycerophospholipogenesis, and therefore might be indicative of a glycerophospholipid catabolic pathway. Importantly, these glycerol-phosphate-containing polar heads are not substrates of Rv1692, as no changes were observed during ABMP. Also worth noting is that the amount of hexose phosphate remained unchanged in the Δrv1692 compared with the wild-type strain, suggesting that glycerol derived from Rv1692 is unlikely to significantly affect the level of gluconeogenesis.

Fig. 4.

Targeted metabolomic analysis of M. tuberculosis Δrv1692 reveals glycerophospho polar head accumulation. Accumulation of glycerol-phosphate and glycerol-phosphate-containing polar heads in M. tuberculosis Δrv1692 is consistent with the functional assignment. The abundances of representative metabolites from the central-carbon metabolism and polar head catabolism are presented as ratios of Δrv1692/wild-type and complemented/wild-type. The data are representative of two independent experiments.

According to these results and the kinetic parameters, Rv1692 does not appear to be a core metabolic enzyme but, rather, is involved in the recycling/catabolism of glycerophospholipid polar heads. To provide independent evidence for lipid polar head recycling/catabolism in M. tuberculosis, we performed experiments using U-¹³C glycerol as sole carbon source and followed the incorporation of the label in the wild-type, Δrv1692, and complemented strains. An increase in labeling could be expected if Rv1692 was involved in the catabolism of lipid polar heads, and a decrease in labeling could be expected if Rv1692 was involved in intermediate metabolism, decreasing the flux of glycerol phosphate to glycolysis and gluconeogenesis. A significant increase in the labeling of glycerophospho-ethanolamine and glycerolphospho-inositol is observed in Δrv1692 compared with wild-type and complemented strains (SI Appendix, Fig. S10). This result is fully consistent with a decrease in the rate of degradation/catabolism of these lipid polar heads in the absence of Rv1692.

Altogether, the results presented here strongly suggest that Rv1692 is involved in glycerophospholipid polar heads recycling/catabolism at the last step (Fig. 5). Such pathways have been partially characterized in E. coli lysates by Albright et al. (27). In this study, all enzymatic activities necessary for glycerophospholipid polar head recycling were identified in protein lysates. It was also found that polar heads were obtained from phospholipids after action of phospholipases A1 and A2, which can be localized in the wall, inner membrane, and cytosol. In this last compartment, these authors found a phosphodiesterase activity that converted phosphoethanolamine into ethanolamine and glycerol 3-phosphate. Such phospholipase activities have been identified in M. tuberculosis. Schué et al. (28) isolated two secreted cutinases-like proteins: Rv1984c, which preferentially hydrolyzed medium-chain carboxylic esters and monoacylglycerols, and Rv3452, which behaved like a phospholipase A2. The other lipases/cutinases are most likely cell-wall-associated and surface-exposed (29). Concerning the subsequent steps in recycling of phospholipids, in the M. tuberculosis genome, Rv0317c (glpQ2) and Rv3842c (glpQ1) are annotated as glycerophosphoryl diester phosphodiesterases that may degrade polar heads generating glycerol 3-phosphate and the corresponding alcohols. It is interesting to point out that G3P itself might induce accumulation of these lipid polar heads by direct inhibition of the phosphodiesterases. G3P levels in M. tuberculosis are known to vary dramatically, depending on the carbon source used (25), and therefore the nutritional status of the cell might directly regulate glycerophospholipid catabolism via inhibition of GlpQ phosphodiesterases. These activities and their potential inhibition by G3P remain to be confirmed experimentally.

Fig. 5.

Schematic representation of a potential polar head recycling pathway used in M. tuberculosis. On enzymatic action of secreted or cell wall-associated cutinases/lipases, the phospholipids polar heads are released. These polar heads are subsequently cleaved by glycerophosphoryl diester phosphodiesterases (Rv0317c and Rv3842c) generating G3P. G3P can then be acylated by Rv1551, followed by Rv2182c and Rv2482c, to form 1,2-diacyl-sn-glycerol 3-phosphate, committing it to phospholipids biosynthesis. Alternatively, G3P can be cleaved by Rv1692, generating glycerol and inorganic phosphate. Glycerol could function as an osmolyte, used in secondary metabolite production, or leave the cell by passive or facilitative diffusion.

G3P can then follow three paths. In the first path, G3P can be acylated by Rv1551 and then by Rv2182c, Rv2482c to generate 1,2-diacyl-sn-glycerol 3-phosphate leading to phospholipids after addition of the polar heads. In the second path, G3P can be directed for glycolysis and gluconeogenesis. In the third path, described here, G3P is broken down by Rv1692 to generate glycerol and inorganic phosphate. Glycerol can leave the cell by passive or facilitated diffusion, serve as an osmolyte, or be used as a central building block in secondary metabolite or lipid antigen biosynthesis. For example, Layre and colleagues described the discovery of a glycerol-based lipid antigen, glycerol monomycolate, in M. tuberculosis (30). In addition, diarabinosyl glycerol dimycolate has also been described in M. tuberculosis (31). These observations strongly suggest that Rv1692 may function in recycling/catabolism of glycerol and inorganic phosphate, thus acting in the last step of catabolism of glycerophospholipid polar heads (Fig. 5). It is important to stress that this catabolic pathway has neither been described in M. tuberculosis before nor predicted based on genomic information, despite the fact that some of the enzymes from this pathway in other organisms have defined homologs in M. tuberculosis (Fig. 5). In addition to the evidence presented in this study and the identification of homologs of all enzymes required for lipid polar head catabolism, a few studies support its relevance in vivo. Fatty acid catabolism is well documented in M. tuberculosis. Glycerophospholipid catabolism provides lipid polar heads and fatty acids, which could be degraded through β-oxidation, generating energy and biosynthetic precursors via Krebs and glyoxylate cycles and gluconeogenesis (32–35). Glycerophospholipid turnover has also been verified in related species, such as Mycobacterium smegmatis and Mycobacterium phlei (36). Finally, Rv3842 is essential in vivo during macrophage infection (37). Rv3842 encodes the phosphodiesterase, which breaks down lipid polar heads, releasing glycerol phosphate (37).

In conclusion, by using a combination of metabolomic profiling, enzymology, genetics, and structural studies, we established Rv1692 as the first G3P phosphatase from M. tuberculosis. The enzyme contains a GNAT region fused to the HADSF fold, which is essential for structural integrity. This phosphatase is likely involved in the recycling of glycerol phosphate-containing lipid polar heads, catalyzing the last predicted reaction of this pathway in M. tuberculosis. This additional feature is essential for structural integrity. Homologs of Rv1692 are found in other Actinomycetales, suggesting that this function is not restricted to mycobacteria. Further studies will determine the role of this recycling pathway in M. tuberculosis during growth in vitro, in macrophages, and in animal models of infection.

Materials and Methods

Materials.

All chemicals were purchased from Sigma. Chromatographic columns were purchased from GE Healthcare. EDTA-free protease inhibitor was purchased from Roche. BL21(DE3) pLysS cells and Ni-NTA resin were purchased from EMD.

Bioinformatic Analysis.

The primary sequence alignment was performed with ClustalW (15). Tree topography and the evolutionary distance are mapped by the neighbor-joining method and constructed using MEGA5 (16).

Preparation of SME for ABMP.

SME was prepared as described earlier (8). Briefly, the Mycobacterium bovis bacillus Calmette–Guérin pellet was suspended in 10 mL acidic ACN solution (acetonitrile, 0.2% acetic acid), and cells were disrupted by sonication. Soluble extract was obtained by centrifugation at 20,000 × g for 10 min at 4 °C and then flash-frozen and lyophilized. Lyophilized SME was suspended in 20 mM Tris HCl at pH 7.4, and insoluble material was removed by centrifugation. Aliquots were stored at −80 °C.

ABMP.

Samples (typically 20 µL) of the SME were incubated for various lengths of time in the presence or absence of Rv1692 (5 µM) and MgCl₂ (4 mM) in a final volume of 250 µL 20 mM Tris⋅HCl at pH 7.5. Cold ACN containing 0.2% acetic acid was used for quenching, yielding 70% (vol/vol) ACN mixtures. After centrifugation at 20,000 × g for 10 min at 4 °C, samples were stored at −80 °C until analyzed by LC-MS, as described in SI Appendix, Materials and Methods.

Bacterial Strains and Cultivation.

Construction of rv1692-deleted and complemented M. tuberculosis H37Rv strains was performed using a suicide plasmid method (38) (SI Appendix, Materials and Methods). Analysis of the metabolites obtained from these strains was performed as described by de Carvalho et al. (25) (SI Appendix, Materials and Methods).