Crystal structure of the S-adenosylmethionine-dependent mycolic acid synthase UmaA from Mycobacterium tuberculosis

The crystal structure of UmaA from M. tuberculosis was solved to 1.95 Å resolution in space group P3₂21.

Abstract

Mycobacterium tuberculosis is a Gram-positive bacillus that causes tuberculosis and is a leading cause of mortality worldwide. This disease is a growing health threat due to the occurrence of multidrug resistance. Mycolic acids are essential for generating cell walls and their modification is important to the virulence and persistence of M. tuberculosis. A family of S-adenosylmethionine-dependent mycolic acid synthases modify mycolic acids and represent promising drug targets. UmaA is currently the least-understood member of this family. This paper describes the crystal structure of UmaA. UmaA is a monomer composed of two domains: a structurally conserved SAM-binding domain and a variable substrate-binding auxiliary domain. Fortuitously, our structure contains a nitrate in the active site, a structural mimic of carbonate, which is a known general base in cyclopropane-adding synthases. Further investigation indicated that the structure of the N-terminus is highly flexible. Finally, we have identified S-adenosyl-N-decyl-aminoethyl as a promising potential inhibitor.

1. Introduction

Mycobacterium tuberculosis is a bacillus that primarily infects lung tissue and causes tuberculosis (TB). It is a leading cause of mortality worldwide, affecting approximately 23% of the world population in 2018 (Centers for Disease Control and Prevention, 2022 ▸). A hallmark of TB is its persistence: it often requires six months or more of multiple antibiotics to cure. In addition to the difficulty in carrying out long-term treatment due to patient noncompliance, misuse of anti-TB drugs leads to drug resistance. Two first-line antibiotics used to treat TB, isoniazid and rifampin, are ineffective in treating multidrug-resistant TB (MDR-TB), an emerging health threat with growing prevalence (Seung et al., 2015 ▸; World Health Organization, 2020 ▸; Centers for Disease Control and Prevention, 2022 ▸). In 2006, reports of extensively drug-resistant TB (XDR-TB) began to appear (World Health Organization, 2020 ▸). In addition to its resistance to first-line drugs, XDR-TB is also resistant to fluoroquinolones and second-line injectable drugs. The growing number of XDR-TB cases in several countries alarms us about the significant threat of drug resistance, and the discovery of new protein targets and antibiotics is necessary to combat drug-resistant TB.

Mycolic acid is an important lipid component of the M. tuberculosis cell wall because it is essential for the virulence and persistence of the bacterium (Barkan et al., 2009 ▸). Modifications are made to the meromycolate chains of mycolic acid to confer resistance to chemical injury and low antibiotic permeability (Barkan et al., 2009 ▸; Huang et al., 2002 ▸). A family of eight S-adenosylmethionine-dependent mycolic acid synthases, CmaA1, CmaA2, MmaA1–MmaA4, PcaA and UmaA, are responsible for these modifications. These enzymes use S-adenosylmethionine (SAM) to modify mycolic acids via methylation. Depending on the enzyme, modifications may include the addition of cyclopropyl, methyl, hydroxyl and methoxy groups (Marrakchi et al., 2014 ▸). UmaA is the least-understood member of this family, and studying this enzyme may shed light on the biosynthesis of mycolic acids, which is crucial for identifying targets and developing novel drugs.

Here, we present the crystal structure of M. tuberculosis UmaA (MtUmaA; PDB entry 7lxi), which consists of a SAM-binding domain and an auxiliary substrate-binding domain consisting of a Rossmann-like fold and a helical lid, respectively. The interactions of the active-site-bound compounds S-adenosyl­homocysteine, nitrate and hexaethylene glycol with UmaA were compared with the corresponding natural substrates SAM, bicarbonate and an olefinic fatty acid. We found that nitrate overlaps at the position containing bicarbonate, which acts as a general base in all known cyclopropane mycolic acid synthases. A structural alignment revealed that the SAM-binding region with the Rossmann-like fold is highly conserved among different species, while the auxiliary domain is highly variable (Holm et al., 2023 ▸). Further investigation using AlphaFold suggests that the N-terminus is very flexible. S-Adenosyl-N-decyl-aminoethyl (SADAE), an effective inhibitor of an ortholog in Escherichia coli, was identified as a promising potential inhibitor for UmaA based on its structure.

2. Materials and methods

2.1. Macromolecule production

Cloning, expression and purification were conducted as part of the Seattle Structural Genomics Center for Infectious Disease (SSGCID; Myler et al., 2009 ▸; Stacy et al., 2011 ▸; Bryan et al., 2011 ▸) following standard protocols described previously (Choi et al., 2011 ▸; Bryan et al., 2011 ▸; Serbzhinskiy et al., 2015 ▸). The full-length protein (UniProt Q6MX39) encoding amino acids 1–286 was PCR-amplified from genomic DNA from M. tuberculosis strain ATCC 25618/H37Rv using the primers shown in Table 1 ▸. The gene was cloned into the ligation-independent cloning (LIC; Aslanidis & de Jong, 1990 ▸) expression vector pBG1861 (Choi et al., 2011 ▸) encoding a noncleavable 6×His fusion tag (MAHHHHHH-ORF). Plasmid DNA was transformed into chemically competent E. coli BL21(DE3)R3 Rosetta cells. The plasmid containing His-MtUmaA was tested for expression in 2 l of culture and was grown using auto-induction medium (Studier, 2005 ▸) in a LEX Bioreactor (Epiphyte Three Inc.) as described previously (Serbzhinskiy et al., 2015 ▸). The expression clone MytuD.00149.b.B1 and the purified protein MytuD.00149.b.B1.PW38903 are available at https://www.ssgcid.org/available-materials/ssgcid-proteins/.

Table 1. Macromolecule-production information.

Source organism	Mycobacterium tuberculosis (strain ATCC 25618/H37Rv)
DNA source	ATCC Representative, Sanger
Forward primer	5′-CTCACCACCACCACCACCATATGACTGAGTTAAGGCCCTTTTAC-3′
Reverse primer	5′-ATCCTATCTTACTCACTTACTTGGTCAGTGTGAACTGTCCC-3′
Expression vector	BG1861
Expression host	Escherichia coli BL21(DE3) Rosetta
Complete amino-acid sequence of the construct produced	MAHHHHHHMTELRPFYEESQSIYDVSDEFFSLFLDPTMAYTCAYFEREDMTLEEAQNAKFDLALDKLHLEPGMTLLDIGCGWGGGLQRAIENYDVNVIGITLSRNQFEYSKAKLAKIPTERSVQVRLQGWDEFTDKVDRIVSIGAFEAFKMERYAAFFERSYDILPDDGRMLLHTILTYTQKQMHEMGVKVTMSDVRFMKFIGEEIFPGGQLPAQEDIFKFAQAADFSVEKVQLLQQHYARTLNIWAANLEANKDRAIALQSEEIYNKYMHYLTGCEHFFRKGISNVGQFTLTK

His-MtUmaA was purified using a two-step protocol consisting of Ni²⁺-affinity chromatography (IMAC) and size-exclusion chromatography (SEC). All chromatography runs were performed on an ÄKTApurifier 10 (GE) using automated IMAC and SEC programs according to previously described procedures (Bryan et al., 2011 ▸). Thawed bacterial pellets were lysed by sonication in 200 ml buffer consisting of 25 mM HEPES pH 7.0, 500 mM NaCl, 5% glycerol, 0.5% CHAPS, 30 mM imidazole, 10 mM MgCl₂, 1 mM TCEP, 250 µg ml⁻¹ AEBSF, 0.025% azide. After sonication, the crude lysate was clarified with 20 µl (25 units µl⁻¹) of Benzonase and incubated with mixing at room temperature for 45 min. The lysate was then clarified by centrifugation at 10 000 rev min⁻¹ for 1 h using a Sorvall centrifuge (Thermo Scientific). The clarified supernatant was passed over a 5 ml Ni²⁺–NTA His-Trap FF column (GE Healthcare) which had been pre-equilibrated with loading buffer composed of 25 mM HEPES pH 7.0, 500 mM NaCl, 5% glycerol, 30 mM imidazole, 1 mM TCEP, 0.025% sodium azide. The column was washed with 20 column volumes (CV) of loading buffer and was eluted with loading buffer plus 250 mM imidazole in a linear gradient over 7 CV. Peak fractions, as determined by UV absorbance at 280 nm, were pooled and concentrated to 5 ml. A SEC column (Superdex 75, GE) was equilibrated with running buffer consisting of 500 mM NaCl, 25 mM HEPES, 5% glycerol, 0.025% azide, 2 mM DTT pH 7.0. The peak fractions were collected and analyzed for the presence of the protein of interest using SDS–PAGE. The SEC peak fractions eluted as a single large peak, suggesting a monomeric enzyme. The peak fractions were pooled and concentrated to 26.15 mg ml⁻¹ using an Amicon purification system (Millipore). Aliquots of 110 µl were flash-frozen in liquid nitrogen and stored at −80°C until use for crystallization.

2.2. Crystallization

Purified His-MtUmaA was screened for crystallization in 96-well sitting-drop plates against the Microlytics MCSG1 (Molecular Dimensions) crystallization screen. Equal volumes of protein solution (0.4 µl) and precipitant solution (0.4 µl) were set up at 287 K against 80 µl reservoir in sitting-drop vapor-diffusion format. Suitable crystals (∼0.1 × 0.1 × 0.1 mm) were harvested directly from the screening tray condition C9 without optimization (see Table 2 ▸). A soaking solution was created from the reservoir solution by adding S-adenosylhomocysteine (SAH; Sigma–Aldrich) to a final concentration of 5 mM. 1 µl of this soaking buffer was placed in the adjacent well of the MRC2 plate. The crystals were transferred to the well containing soaking solution and the plate was resealed and allowed to incubate for 6 h. Crystals from the soaking solution were passed through a cryoprotectant composed of soaking buffer with 20% ethylene glycol. The crystals were then flash-cooled by harvesting them and plunging them directly into liquid nitrogen with cryoprotectant consisting of soaking buffer with 20% ethylene glycol.

Table 2. Crystallization.

Method	Vapor diffusion, sitting drop
Plate type	SWISSCI MRC2
Temperature (K)	287
Protein concentration (mg ml−1)	26.2
Buffer composition of protein solution	25 mM HEPES pH 7.0, 500 mM NaCl, 5% glycerol, 2 mM DTT, 0.025% sodium azide
Composition of reservoir solution	Microlytics MCSG1 screen condition C9: 32%(w/v) PEG 4000, 800 mM lithium chloride, 100 mM Tris–HCl pH 8.50. The crystal was soaked for 6 h in reservoir with 5 mM SAH and cryoprotectant (soaking buffer with 20% ethylene glycol).
Volume and ratio of drop	0.4 µl protein solution + 0.4 µl reservoir solution
Volume of reservoir (µl)	80

2.3. Data collection and processing

Data were collected at 100 K using a Rigaku FR-E+ SuperBright rotating-anode generator at a wavelength of 1.5418 Å with a Saturn 944+ CCD detector (see Table 3 ▸) and were reduced with XSCALE (Kabsch, 2010 ▸).

Table 3. Data collection and processing.

Values in parentheses are for the outer shell.

Diffraction source	Rigaku FR-E+ SuperBright rotating-anode generator
Wavelength (Å)	1.5418
Temperature (K)	100
Detector	Rigaku Saturn 944+ CCD
Crystal-to-detector distance (mm)	50
Rotation range per image (°)	1.00
Total rotation range (°)	150
Exposure time per image (s)	20
Space group	P3221
a, b, c (Å)	103.22, 103.22, 49.15
α, β, γ (°)	90, 90, 120
Resolution range (Å)	50–1.950 (2.000–1.950)
Total No. of reflections	208503 (8889)
No. of unique reflections	22257 (1641)
Completeness (%)	99.900 (100.000)
Multiplicity	9.4 (5.4)
〈I/σ(I)〉	26.410 (2.23)
R meas	0.056 (0.699)
CC1/2	99.8 (74.9)
Overall B factor from Wilson plot (Å2)	36.7

2.4. Structure solution and refinement

The structure was solved using direct refinement as the unit-cell parameters and space group were identical to a previously solved structure (PDB entry 7l9u). All solvent molecules were removed from the initial model and placed directly in the unit cell, followed by rigid-body refinement and subsequent full refinement using phenix.refine (Liebschner et al., 2019 ▸) and finally by manual rebuilding using Coot (Emsley et al., 2010 ▸). The test set for cross-validation was generated using the XDS suite, specifically XDSCONV, with 5% of the reflections randomly assigned to the test set for R_free calculation. The quality of the electron density was very good, allowing us to build the entire main chain except for the first three N-terminal residues and a region between residues 199 and 202. During the model-building process, we discovered several blobs of electron density that were inconsistent with water molecules. We were able to successfully identify these unknown blobs using the unmodeled blob-building features in Coot. The final model included 186 waters, one hexaethylene glycol (PEG-6), two nitrates, one phosphate, one chloride and one ethanediol. The quality of the structure was checked using MolProbity (Chen et al., 2010 ▸). All data-reduction and refinement statistics are shown in Table 4 ▸. All images were prepared using PyMOL (version 2.2; Schrodinger). Coordinates and structure factors have been deposited with the Protein Data Bank (https://www.rcsb.org; Berman et al., 2000 ▸) with accession code 7lxi.

Table 4. Structure solution and refinement.

Values in parentheses are for the outer shell.

Resolution range (Å)	43.07–1.95 (2.00–1.95)
Completeness (%)	99.9
σ Cutoff	F > 1.34σ(F)
No. of reflections, working set	20178 (1353)
No. of reflections, test set	2062 (109)
Final Rcryst	0.1654 (0.2283)
Final Rfree	0.1961 (0.2567)
No. of non-H atoms
Protein	2241
Ion	14
Ligand	23
Water	186
Total	2464
R.m.s. deviations
Bond lengths (Å)	0.007
Angles (°)	0.733
Average B factors (Å2)
Protein	35.9
Waters	39.3
Nitrate	53.25
Nitrate (active site)	23.25
Phosphate	62.25
Hexaethylene glycol (PEG-6)	41.50
1,2-Ethanediol (EDO)	35.75
S-Adenosyl-L-homocysteine (SAH)	28.25
Chloride	38.00
Ramachandran plot
Most favored (%)	93.3
Allowed (%)	6.7
Rotamer outliers	3

2.5. Docking

Structure-based blind docking was performed by submitting the PDB files for UmaA and (E)-hexacos-17-enoic acid to the CB-Dock2 web server (Liu et al., 2022 ▸). We inspected the potential position of the olefinic fatty acid and selected a conformation based on low-energy solutions (low Vina scores).

3. Results and discussion

3.1. Overview of structure

The crystal structure of SAM-dependent mycolic acid synthase (UmaA) from M. tuberculosis was determined to 1.96 Å resolution. The quality of the experimental electron density was excellent, allowing us to build the entire protein backbone except for residues 1–3 and 199–202. Along with the SAH, the final model included hexaethylene glycol (PEG-6) and nitrate in the active site. The PEG-6 that we modeled was likely to be derived from the PEG 4000 used in the crystallization conditions. PEG 4000 is inherently a polydisperse mixture of PEG molecules with an average molecular weight of 4000 Da. Our analysis is consistent with an actual PEG-6 molecule rather than partial density of a longer molecule as there were discrete ends to the molecule, which were capped at the ends by protein. The source of the nitrate is unknown. It could have come through in the purification from the cells or have been present in the growth medium. Fortuitously, the PEG-6 and nitrate may bind in the active site similarly to the olefinic substate and carbonate general acid/base. The protein comprises 16.5% β-strands, 48% helices and 35.5% loop regions (Laskowski et al., 2018 ▸). UmaA is a monomer composed of two discontinuous domains: a SAM-binding domain and a substrate-binding auxiliary domain (Fig. 1 ▸). The SAM-binding domain, composed of residues 43–171, 206–226 and 276–286, is the core region of the protein and consists of a Rossmann-like fold. The Rossmann-like fold classifies UmaA as a class I methyltransferase, the most common of the five classes of SAM-dependent MTases, which are grouped based on their catalytic domain folds (Schubert et al., 2003 ▸). This domain resembles a sandwich fold of alternating β–α–β motifs with the seven β-strands flanked by α-helices on each side (Blum et al., 2021 ▸). The sequence re-enters the helical lid domain between β-strand 5 and α-helix 12 (residues 172–205) and between β-strands 6 and 7 (residues 227–275). Like other class I methyltransferases, UmaA contains a highly conserved glycine-rich region (the GXGXG motif) between β-strand 1 and α-helix 4, which is important for binding to SAM (Martin & McMillan, 2002 ▸). The substrate-binding auxiliary domain (residues 4–42, 172–205 and 227–275) may have resulted from an insertion mutation at the middle of the Rossmann fold and is composed of α-helices and loops that hover over the SAM-binding domain like a helical lid; it displays a higher level of variability in its fold, which corresponds to the diverse range of substrates that can be methylated.

Figure 1.

Structure of the SAM-dependent mycolic acid synthase UmaA from M. tuberculosis. UmaA is a monomer consisting of a SAM-binding domain and a substrate-binding auxiliary domain. The SAM-binding region, shown in blue, contains residues 43–171, 206–226 and 276–286. The substrate-binding region, shown in yellow, contains residues 4–42, 172–205 and 227–275. PEG-6, SAH and nitrate are shown as sticks. The GXGXG motif is shown in magenta. PEG-6 and SAH are shown in green with standard atom coloring.

3.2. Ligand–protein interactions

SAH interacts with UmaA at the binding cleft within the Rossmann-like fold. Fig. 2 ▸(a) shows polar interactions between SAH and the residues in the binding site of the protein. The two O atoms of the carboxylate group in SAH form hydrogen bonds to Tyr32 and Thr33, while the amino group in SAH forms hydrogen bonds to the carbonyl O atom on the backbone of Gly71 and Ile135. The hydroxyl groups on the ribofuranose moiety in SAH form hydrogen bonds to Gly73, Thr93, Gln98 and Leu94. Gly71 and Gly73 are glycine residues that are part of the highly conserved GXGXG motif, which supports the finding that the glycine-rich loop is essential for SAM-MTases that bind to SAM (Martin & McMillan, 2002 ▸). The adenine ring interacts through main-chain and side-chain hydrogen bonds with amino-acid residues including Gly121, Trp122 and Phe141.

Figure 2.

SAH interactions. (a) Interactions between SAH and UmaA. Polar interactions are shown as yellow dashed lines. The coloring is the same as in Fig. 1 ▸. (b) Interactions between nitrate and UmaA with the superposition of PcaA from M. tuberculosis (PDB entry 1l1e) and the cyclopropane fatty-acyl phospholipid synthase from L. acidophilus (PDB entry 5z9o). PDB entry 1l1e is shown in dark gray and PDB entry 5z9o is shown in light gray. (c) Superposition between PEG-6 and the docked fatty acid (E)-hexacos-17-enoic acid. The blue and yellow colors are projected onto the molecular surface and represent the substrate-binding and SAM-binding domains, respectively. Hydrophobic residues are shown in orange. (d) A theoretical model of the reaction mechanism. A methyl group was added to the SAH molecule using the Builder function in PyMOL. The nitrate was replaced with carbonate. The fatty acid was the (E)-hexacos-17-enoic acid docked in (c). SAM, carbonate and fatty acid are shown in green with standard atom coloring.

Previous studies report that cyclopropane synthases utilize bicarbonate (Iwig et al., 2005 ▸) as a general base during catalysis because it interacts with conserved amino-acid residues that can participate in general acid/base chemistry. In our structure, we have nitrate in the active site. The trigonal planar shape of nitrate is likely to mimic the binding position of bicarbonate. Indeed, PcaA from M. tuberculosis (PDB entry 1l1e; r.m.s.d. 1.2 Å; 58% sequence identity) and cyclopropane fatty-acyl phospholipid synthase from L. acidophilus (PDB entry 5z9o; r.m.s.d. 3.8 Å; 35% sequence identity) superimposed onto UmaA (Huang et al., 2002 ▸; Ma et al., 2019 ▸). Fig. 2 ▸(b) shows the nitrate from UmaA and the bicarbonate from the homologs and that the interacting residues (Cys, Glu, His and Tyr) are conserved.

Also, a study by Huang et al. (2002 ▸) analyzed the structural and catalytic differences between SAM-dependent methyltransferases and synthases that modify mycolic acids. CmaA1, CmaA2 and PcaA are cyclopropane synthases, introducing cyclopropane rings to enhance cell-wall rigidity and resistance. CmaA1 acts at the distal position, while CmaA2 and PcaA modify the proximal site. In contrast, MmaA4 catalyzes oxygen incorporation, forming hydroxylated mycolates, keto-mycolates and methoxy-mycolates. A key distinction is their general base mechanism: CmaA1, CmaA2 and PcaA use a bicarbonate ion in the active site, while MmaA4 has a glutamate residue in the same position, potentially facilitating proton abstraction via a water-mediated mechanism. To delve deeper into the structural features of these enzymes, we aligned CmaA1 (PDB entry 1kpg; r.m.s.d. 1.3 Å; Huang et al., 2002 ▸), CmaA2 (PDB entry 3hem; r.m.s.d. 1.3 Å; D. Barkan, Z. Liu, J. C. Sacchettini & M. S. Glickman, unpublished work), PcaA (PDB entry 1l1e; r.m.s.d. 1.2 Å; Huang et al., 2002 ▸), MmaA2 (PDB entry 1tpy; r.m.s.d. 1.1 Å; C. V. Smith & J. C. Sacchettini, unpublished work) and MmaA4 (PDB entry 7q2h; r.m.s.d. 2.1 Å; Galy et al., 2021 ▸) with UmaA. In our structural and sequence alignments (Figs. 3 ▸a and 3 ▸b), we have Gly136, which is small enough to accommodate a carbonate molecule. The carboxylate functional group of Glu146 is in the same position as the carbonate (Fig. 3 ▸a). Taken together, UmaA is likely to be a SAM-dependent ‘cyclopropane-adding’ mycolic acid synthase.

Figure 3.

UmaA may add cyclopropane rings to mycolic acids. Structural (a) and sequence alignment (b) with UmaA paralogs (PDB entries 1kpg, 1l1e, 1tpy, 3hem and 7q2h; the coloring is the same as in Fig. 4 ▸a). In (a) the nitrate is depicted as sticks shown in white with 2F_o − F_c omit electron density. Glu146 (PDB entry 7q2h primary-sequence numbering) is depicted as blue sticks. In (b) the predicted general base (carbonate or Glu) is indicated with a green triangle. Residues that make polar interactions with carbonate are indicated with blue circles. The secondary-structure elements shown are α-helices (α), 3₁₀-helices (η), β-strands (β) and β-turns (TT). Identical residues are shown in white on a red background and conserved residues are shown in red. This figure was created using ESPript (Robert & Gouet, 2014 ▸)

There is electron density in the substrate-binding domain that is consistent with a partial PEG from the PEG 4000 mixture and we modeled it as PEG-6. We studied how the substrate-binding domain interacts with PEG-6 and a long-chain fatty acid, which PEG-6 is likely to mimic. The substrate-binding domain contains a binding pocket lined with nonpolar residues, which allows the association of long, nonpolar ligands such as meromycolic acids. The crystal structure of UmaA shows a long, polar molecule (PEG-6) extended in the pocket of the domain. In contrast, crystal structures of related enzymes contain long, nonpolar ligands that adopt a U-shaped conformation in the lipid-binding site (Barkan et al., 2009 ▸; Cronan & Luk, 2022 ▸; Hari et al., 2018 ▸; Huang et al., 2002 ▸; Ma et al., 2019 ▸). The difference in ligand conformations may be due to the polarity of PEG-6 compared with the other ligands found in the crystal structures of these related enzymes. To understand how a nonpolar molecule would interact with UmaA, we docked (E)-hexacos-17-enoic acid, an olefinic fatty acid similar to a meromycolic acid, using CB-Dock2 and selected the conformation with the lowest Vina score of −7.2 kcal mol⁻¹ among the five given binding poses (Liu et al., 2022 ▸). This fatty acid contains a cis double bond between C17 and C18, where the proximal double bond in the meromycolic chain of mycolic acids is typically located (Marrakchi et al., 2014 ▸). The unsaturated fatty acid traces along and overlaps with PEG-6 until it makes a U-shaped turn at its cis double bond between C17 and C18 (Fig. 2 ▸c). The region around 4.0 Å of the fatty acid is shown in orange. Most of the residues in this region, Tyr32, Glu139, Ile168, Met176, Val183, Phe190, Met191, Leu204, Tyr231, Tyr264, Cys268, Thr170, Phe271, Phe272, Ile276 and Ser277, are nonpolar and interact with the alkyl chain through van der Waals forces.

3.3. Catalytic mechanism

The observed interactions of UmaA with these ligands give insights into a potential mechanism. Fig. 2 ▸(d) presents a theoretical model of SAM, bicarbonate and (E)-hexacos-17-enoic acid positioned on the molecular surface of UmaA and meeting at the interface between the two domains. Using the PyMOL Builder function, we attached a methyl group to the S atom of SAH and replaced the N atom of nitrate with carbon. The olefinic fatty acid makes the U-shaped turn at its double bond between C17 and C18, which positions the double bond near the methyl group attached to the S atom in SAM and under Tyr32. Consequently, the double bond can perform an S_N2 attack on the electrophilic methyl group of SAM, which results in a methylated carbon and a secondary carbocation adjacent to it. Tyr32, located approximately 6.0 Å above the double bond, can stabilize the carbocation intermediate by π–cation interactions. Bicarbonate plays the role of a general base and deprotonates the added methyl group, which can produce a cyclopropyl group (Iwig et al., 2005 ▸). The actual product that UmaA generates remains unclear. Previous experiments demonstrated that UmaA catalyzes the conversion of the double bond of phospholipid-linked oleic acid to a methylene stearic acid in vitro (Meena et al., 2013 ▸). Other research has shown that MSMEG0913, which was initially thought to be the UmaA ortholog in M. smegmatis, adds a methyl group adjacent to either a trans double bond or a trans cyclopropyl group at the proximal position of mycolic acids (Laval et al., 2008 ▸). However, they concluded that MSMEG0913 was not orthologous because its absence could not be compensated after introducing UmaA into M. smegmatis. They also did not observe changes to the lipid composition in M. tuberculosis after deleting its UmaA gene (Laval et al., 2008 ▸).

3.4. Conformational heterogeneity of the N-terminus

A close look at our structural alignment shows that the helical lid domain, particularly the N-terminus, was the most variable region among the enzymes (Fig. 4 ▸a). The crystal structures of MmaA2, CmaA1 and CmaA2 have an N-terminus that forms a helix close to the meromycolic acid-binding site. This binding site is shaped like a long tunnel that can accommodate the olefinic fatty acid, SAM and carbonate. Substrates can access the binding site in the open state, while the catalytic reaction occurs in the closed state. When closed, the lid covers one end of the tunnel and sequesters the reactants upon binding to the active site. The other end of the tunnel is open, possibly allowing mycolic acids of different lengths to bind. Crystal structures of UmaA and MmaA4 have an N-terminus with poor electron density, making it difficult to accurately place the amino-acid residues. The difference in conformations of the N-terminus between paralogs suggests that it is highly dynamic (Fig. 4 ▸a). We cannot exclude the possible influence of crystal packing and the noncleavable N-terminal His₆-tag on the observed conformation and disorder.

Figure 4.

Conformational differences in the N-terminus. (a) Structural alignment of the N-terminus of UmaA in comparison with paralogs. UmaA and MmaA4 (PDB entry 7q2h) have N-terminal ends that protrude into the solvent, while the other structures (PDB entries 1l1e, 1tpy, 1kpg and 3hem) have N-terminal ends that close onto the binding site. (b) Structural alignment of the N-terminus of the crystal structure of UmaA and the predicted structure of UmaA from AlphaFold. The crystal structure shows poor or missing 2F_o − F_c omit electron density (1σ) at the N-terminus. AlphaFold predicts the structure of the N-terminal region with low confidence, which is depicted as a gradient from blue to red.

To further study the N-terminus, we compared the crystal structure of UmaA with the predicted structure of UmaA from AlphaFold (Jumper et al., 2021 ▸; Varadi et al., 2022 ▸). Fig. 4 ▸(b) displays a significant deviation in the N-terminal region between the experimental and predicted structures. AlphaFold predicted that the N-terminal region would form a helix that closes onto the active site, but this prediction had low confidence, progressively becoming lower as it approached the N-terminus. In contrast, the crystal structure of UmaA shows that its N-terminal region projects outwards but does not have a well defined structure. The electron density in this region is weak, with several missing residues (the side chains of Phe7, Glu9, Glu10 and Gln12; Fig. 4 ▸b). The ambiguity from the experimental structure, complemented by the low confidence in the theoretical structure, suggests that the N-terminal region is very flexible. Taken together, these findings highlight the dynamic role of the N-terminal region as a gate controlling enzymatic function and suggest a potential target for inhibitors that could disrupt this mechanism.

3.5. A potential inhibitor for UmaA

Given the role of UmaA in the metabolic pathway of mycolic acid synthesis, identifying potential inhibitors for UmaA and related enzymes is of great importance. The hydrophobic pocket offers a potential site for inhibition with a nonpolar molecule. SADAE, an analog of SAH, was studied via superimposition into the active site. While SAH contains the amino acid homocysteine, SADAE contains an alkyl chain in its place. SADAE was identified as a very effective inhibitor of an ortholog in E. coli (IC₅₀ of 11 µM) and an even more potent inhibitor of mycolic acid methyltransferases in vitro (Guianvarc’h et al., 2006 ▸). A crystal structure of MmaA4 complexed with SADAE (PDB entry 3ha7) superimposed onto UmaA with an r.m.s.d. of 1.6 Å (Vaubourgeix et al., 2009 ▸). The structural similarities of the SAH moiety may allow it to fit and interact with the same residues as SAH (Fig. 5 ▸). The alkyl chain spans into the hydrophobic pocket, where it interacts with the nonpolar residues by hydrophobic inter­actions. These properties suggest that SADAE can be used as a potential compound for further drug development.

Figure 5.

S-Adenosyl-N-decyl-aminoethyl (SADAE), a potential inhibitor of SAM-dependent mycolic acid synthases, superimposed onto UmaA. SADAE was taken from the structure of MmaA4 (PDB entry 3ha7). SAH is shown in green, while SADAE is shown in purple.

Supplementary Material

PDB reference: UmaA from Mycobacterium tuberculosis, 7lxi

Funding Statement

This project has been funded in whole or in part with Federal funds from the National Institute of Allergy and Infectious Diseases, National Institutes of Health, Department of Health and Human Services under Contract No. 75N93022C00036.