1. Introduction
According to the World Health Organization, tuberculosis (TB) is one of the top 10 causes of death worldwide.
Mycobacterium tuberculosis (Mtb) is the leading cause of death from a single infectious agent [
1]. Recently, multiple studies have reported the spread of multidrug-resistant (MDR) and extensively drug-resistant strains, which cannot be cured with first-line and even second-line antitubercular (anti-TB) medications in the latter case [
2,
3]. The threat demands the development of novel drugs for anti-TB therapy, identification of their targets, and assessment of their metabolic stability.
One of the latest breakthroughs in this area was the discovery of SQ109 [
4], which is currently in Phase II/III clinical trials for the treatment of MDR pulmonary TB [
5]. SQ109 belongs to the 1,2-ethylenediamine class of anti-TB drugs [
6], consisting of adamantane head and geranyl tail, proposed to disrupt the synthesis of the complex Mtb cell wall [
7]. At least three mechanisms of action have been reported so far for the drug [
8]. First, it inhibits transport of trehalose monomycolates by Mycobacterial membrane protein Large 3 (MmpL3) from the cytoplasm. Second, it inhibits respiration by blocking menaquinone biosynthesis by MenA and MenG. Finally, it acts as an uncoupler, collapsing the pH gradient and membrane potential. SQ109 has demonstrated promising inhibition of cell growth and a very low spontaneous drug resistance rates [
6,
7]. SQ109 showed in vitro activity against the known resistant Mtb strains [
6] and
Mycobacterium bovis bacillus Calmette–Guérin (BCG) [
9]. It is also bactericidal against
Mycobacterium smegmatis (Msm) and
Mycobacterium aurum, although with reduced activity [
9]. Moreover, SQ109 is active against non-mycobacterium species [
10,
11,
12,
13]. In vivo studies demonstrated SQ109 effectiveness in murine TB model [
4]. SQ109 also interacts synergistically with other anti-TB drugs, such as rifampicin (RIF), isonizid, and bedaquiline [
9,
14], which is crucial for the combination therapy [
15].
SQ109 is effectively metabolized by human, dog, rat, and murine liver microsomes [
16]. Cytochrome P450s (CYPs), CYP2D6 and CYP2C19, were proposed to be primarily responsible for the metabolism in humans. However, to the best of our knowledge, no interactions with Mtb CYPs have been reported so far. Here, we show that three Mtb CYPs: CYP124, CYP125, and CYP142, can bind SQ109 with a ligand-binding constant, Kd
app = 3.4 ± 0.3 μM, 41 ± 3 μM and 52 ± 11 μM, respectively. CYP124 (Rv2266, CYP124A1) has previously assigned activity towards methyl-branched lipids of isoprenoid origin [
17,
18,
19,
20], while CYP125 (Rv3545c, CYP125A1) and CYP142 (Rv3518c, CYP142A1) are involved in the cholesterol catabolism [
19]. Among the tested enzymes, only CYP124 could oxidase SQ109 to a distinct product, detected by LC-MS with a turnover number of 0.60 ± 0.09 min
−1. A 1.25 Å crystal structure of the CYP124–SQ109 complex reveals two close conformations of SQ109, confirming the formation of a ω-terminal hydroxy product. Finally, given the molecular docking, we propose that the newly formed hydroxy group of SQ109 affects its binding with the prospective drug target, MmpL3, stabilizing SQ109-OH in the binding pocket [
21]. CYP124 represents the first example of a putative class of mycobacterial CYPs that might function similarly to drug-metabolizing human CYPs.
3. Discussion
The CYP family is well represented in Mtb, with 20 genes being identified while the function of the majority is still unknown. Their conservation during reductive genome evolution [
22] indicates the importance of Mtb CYPs for survival or/and pathogenicity. Micromolar binding affinity and detected catalytic activity of CYP124 with SQ109, obtained in the study, indicate the CYPs’ potential role in the metabolism of xenobiotics in Mtb. Given the crystal structure of the CYP124–SQ109 complex and the functional assay, we identified that CYP124 likely hydroxylates SQ109 at the ω-methyl group in the trans position. The ability to hydroxylate SQ109 was not detected for the other two Mtb steroid-metabolizing CYP enzymes—CYP125 and CYP142, although they were both able to bind the drug. We used the surrogate redox partners to detect the metabolism of SQ109 by CYP124. We cannot exclude that with cognate redox partners consecutive oxidation products might be also produced.
SQ109 is a highly effective drug candidate against Mtb and, to a lesser extent, against Msm, with minimal inhibitory concentration (MIC) values being 0.3–0.6 μM and 9.4 μM, respectively [
8]. MmpL3 is suggested as one of the main targets of the compound in both organisms. Indeed, some mutations in the MmpL3 gene results in the emergence of the resistant strains [
7,
23]. The crystal structure of Msm MmpL3 in complex with SQ109 (PDB ID: 6AJG) showed that the drug disrupted core Asp-Tyr pairs (D256-Y646 and Y257-D645), apparently crucial for protein function [
24]. However, SQ109 did not block the Mmpl3 flippase activity in spheroplasts, suggesting other molecular targets [
25]. We consider a possibility that SQ109 is a prodrug, rather than a drug, which first needs to be activated by liver and/or Mtb CYPs. This assumption was first made by Chen et al. [
9] based on the rapid compound metabolism by microsomal P450s [
16]. The authors then hypothesized that the bactericidal activity of the drug might come from its metabolites, potentially produced by mycobacterial CYPs. They also noticed that the synergy with RIF [
9,
14] might be partly explained by the enhanced expression level of CYPs in RIF-treated mycobacteria [
26].
To test this idea, we performed molecular docking of the determined metabolite –SQ109-OH, to the crystal structure of the Msm MmpL3-SQ109 complex (PBD ID: 6AJG) [
24]. The structure shows that the newly formed OH group could fit within the SQ109 binding pocket and form favorable H-bonds (
Figure 4a). The top-ranked docking poses confirm the ability of SQ109-OH to H-bond with either S301 or the backbone oxygens of A637 and I297 (
Figure 4c). We also performed the molecular docking experiment using our homology-based model of Mtb MmpL3. Its binding pocket is somewhat similar to that of Msm, being different by only four residues, namely S301:A296, I319:T314, V638:L633, and L686:V681 (residue in Msm:residue in Mtb). The lost possibility of H-bonding with S301, in this case, might be compensated by bonding with S295 (
Figure 4e) or the Mtb-specific T314 (
Figure 4f), which was also confirmed by docking. The additional stabilization of SQ109-OH might facilitate the inhibition of the flippase activity observed for other Mmpl3-directed drugs [
25], such as AU1235 [
27] and BM212 [
28]. However, we cannot exclude that both SQ109 and SQ109-OH are active compounds, and further experiments are required to confirm the prodrug hypothesis. Taking together, our findings identify the first example of Mtb CYP capable of biotransformation of anti-TB drugs.
Activation of prodrugs by Mtb enzymes was previously demonstrated for isoniazid, pyrazinamide, and ethionamide [
29]. Different classes of enzymes catalyze these reactions: Catalase-peroxidase encoded by katG gene [
30], pyrazinamidase [
31], and mycobacterial Baeyer–Villiger monooxygenases [
32]. It also has been shown that Mtb acetyltransferases and phosphotransferases deactivate aminoglycosides (second-line anti-TB drugs). Xenobiotics (including anti-TB compounds) transformation in Mtb has also been shown through N–alkylation, amidation, ester hydrolysis, and the reduction of the nitro group [
29]. In this work, we extend the current knowledge and demonstrate the very likely involvement of the Mtb CYP enzyme in the hydroxylation of the anti-TB drug. Given the importance of this group of enzymes in the metabolism of xenobiotics in humans and the significant number of CYPs in Mtb, we suggest that Mtb CYPs may be involved in the metabolism of various classes of compounds. In this regard, the assessment of the anti-TB drug candidate’s metabolism using the whole cell-based system [
33,
34] or isolated Mtb enzymes could be a useful tool in anti-TB drug discovery.
4. Materials and Methods
4.1. Cloning, Expression, and Purification of Recombinant CYP124
cDNAs encoding CYP125 (gene
Rv3545c), CYP142 (gene
Rv3518c), and CYP124 (gene
Rv2266) were amplified by PCR genomic DNA of Mtb H37Rv (obtained from The Vyshelessky Institute of Experimental Veterinary Medicine, NASB, Minsk, Belarus). Expression plasmids for each protein were generated using the vector pTrc99a. The proteins were expressed and purified as described previously [
18]. The cDNA encoding spinach Ferredoxin-1 (Fdx1) was amplified from the total RNA isolated from
Spinacia oleracea seedlings. Adrenodoxin reductase-like flavoprotein (Arh1, A18G mutant) expression construct was provided by Prof. Rita Bernhardt (Saarland University, Saarbrucken, Germany). Fdx1 and Arh1 were expressed in
Escherichia coli and purified using metal-affinity and anion-exchange chromatography.
4.2. Substrate Binding Studies
To determine Kd
app values of the CYPs, optical titration was performed using a Cary 5000 UV-VIS NIR dual-beam spectrophotometer (Agilent Technologies, Santa Clara, CA, USA) in 1-cm pathlength quartz cuvettes. Stock solutions of the SQ109 (Sigma-Aldrich, St. Louis, MO, USA; #SML1309, 98+% by HPLC) were prepared at a concentration of 10 mM in 45% hydroxypropyl-beta-cyclodextrin (Sigma-Aldrich, St. Louis, MO, USA). The equivalent volume of 45% hydroxypropyl-beta-cyclodextrin solution was added in the reference cell at each SQ109 сoncentration point. Titration was repeated at least three times, and Kd
app was calculated as described previously [
18].
4.3. Catalytic Activity Assay
We tested the catalytic activity of Mtb CYPs similar to [
20]. Mtb CYPs were reconstituted in 50 mM potassium phosphate (pH 7.4) containing 0.5 µM CYP, 2 µM spinach Fdx1, 0.5 µM Arh1, 100 µM SQ109, 1 mM glucose-6-phosphate (Sigma-Aldrich, St. Louis, MO, USA), 1 U/mL glucose-6-phosphate dehydrogenase (Sigma-Aldrich, St. Louis, MO, USA), and 0.4 mM β-NADPH (Sigma-Aldrich, St. Louis, MO, USA). The proteins (CYPs, Fdx1, and Arh1) were pre-incubated with SQ109 in the buffer solution for 10 min at 30 °C. The reaction was started by adding an NADPH-regenerating system containing glucose-6-phosphate, glucose-6-phosphate dehydrogenase, and β-NADPH. After 1 h of incubation at 30 °C, the reaction was stopped by boiling, and the reaction mixture was extracted using dichloromethane and subjected to the LC-MS analysis. The activity of CYP124 was estimated graphically from the LC-MS results, allowing us to estimate a Turnover Number—nmoles of metabolized product/nmole of CYP/min.
4.4. Identification of SQ109 Product
An Agilent 1200 series HPLC instrument equipped with an Agilent Triple Quad 6410 mass-spectrometer (Agilent Technologies, Santa Clara, CA, USA) was used. The samples were analyzed by gradient elution on a Zorbax Eclipse Plus C18 column (Agilent Technologies, Santa Clara, CA, USA; 2.1 × 50 mm; 1.8 µm). TFA (0.1 % v/v in water) was used as mobile phase A and acetonitrile as mobile phase B. The gradient was 20%–50% B in 0–7 min. The flow rate was 400 µl/min. The column temperature was maintained at 35 ± 1 °C. The mass-spectrometry experiments were performed with an electrospray ionization source (ESI) in positive-ion mode. The nebulizing gas flow rate was set at 9.5 l/min, the drying gas temperature at 350 °C, the capillary voltage at 4000 V, and the nebulizer at 35 psi. The ESI-MS/MS analysis was done in product-ion mode with different values of the fragmentor (135, 150, and 200 V) and the collision energies (10 and 20 V).
4.5. Crystallization, Data Collection, and Crystal Structure Determination
CYP124–SQ109 was crystallized by a sitting drop approach in 96-well crystallization plates with commercially available kits (Qiagen, Hilden, Germany) at 20 °C with 1:1 protein/mother liquor ratio with the ligand concentration of 100 μM. Crystals were obtained from 0.3 M Mg(HCO2)2 and 0.1 M Tris pH 8.5. Glycerol (20%) as cryoprotectant was added before flash-freezing in liquid nitrogen.
The diffraction data were collected at the European Synchrotron Radiation Facility (ESRF) beamline ID23-1. The data collection strategy was optimized in BEST [
35]. The data were processed in the XDS software package [
36]. The crystallographic data collection statistics are given in
Table 1.
The phase problem was solved by molecular replacement in Phaser [
37] from PHENIX [
38], where the generated poly-ala model of ligand-free CYP124 (PDB ID: 2WM5) [
17] was utilized as a starting model. The space group was P12
11 and contained one molecule per asymmetric unit. The model was subsequently rebuilt in PHENIX.AutoBuild [
39]. PHENIX.Refine [
40], and Coot [
41] were used for model refinement. In the last refinement steps, the mF
o−DF
c map unambiguously showed SQ109 in the active site (
Figure 4c). The final resolution cut-off was determined by the application of paired refinement [
42]. The quality of the resulting model was analyzed by PHENIX.MolProbity [
43] and Quality Control Check web server (
https://smb.slac.stanford.edu/jcsg/QC/).
The Structure for CYP124 in complex with SQ109 was deposited in Protein Data Bank with accession number 6T0J. The diffraction images and processing data were deposited to Integrated Resource for Reproducibility in Macromolecular Crystallography [
44] (
http://proteindiffraction.org/) with accession number 6T0J.
The figures containing electron density and molecular structures were generated using PyMOL (Schrödinger LLC, New York, NY, USA).
4.6. Molecular Docking
A 3D conformer of SQ109-OH was created in PyMOL by adding a hydroxy group to SQ109 from its complex with Msm MmpL3 (PDB ID: 6AJG) [
24]. Ligand torsion trees were created in AutoDock Tools [
45]. The structure of the Msm MmpL3 was taken from its complex with SQ109 (PDB ID: 6AJG) [
24]. The structure of the Mtb MmpL3 was modeled using the Phyre
2 web server [
46]. The model of the first 757 residues was based on the homology with Msm MmpL3-ICA38 complex (PDB ID: 6AJJ) [
24], scored with 100% confidence by Phyre
2. Notably, the C-terminal domain of Mtb MmpL3 (P758-L944) was not modeled due to the absence of the alignment coverage. However, it was not used for the docking. Polar hydrogens of the protein molecules and the ligands were assigned in PyMOL. Molecular docking was performed, following two protocols. For the first one, we used AutoDock Vina [
47] with default settings except for exhaustiveness, which was set to 10. For the second protocol, we used our in-house modification of AutoDock Vina and the Convex-PL [
48] scoring function for rescoring, augmented with additional descriptors that account for conformational flexibility and solvation. We have recently applied this protocol for pose prediction in [
49], where it is discussed in more detail. We ran several AutoDock Vina simulations to obtain more diverse docking poses and clustered all resulting conformations with a 1 Å threshold using RDKit (Open-source cheminformatics,
http://www.rdkit.org). To confirm the docking protocols, we compared the docked SQ109 poses with crystallographic ones (PDB ID: 6AJG;
Figure 4a,b,d) [
24]. The only significant difference was in the amylene part of the molecule, which may be attributed to the structure’s low resolution. We visually inspected the ten top-ranked docking poses produced by both protocols.
The figures illustrating interactions between SQ109-OH and MmpL3 proteins were generated using PyMOL and
PLIP [
50].