Hydroxylation of Antitubercular Drug Candidate, SQ109, by Mycobacterial Cytochrome P450

Spreading of the multidrug-resistant (MDR) strains of the one of the most harmful pathogen Mycobacterium tuberculosis (Mtb) generates the need for new effective drugs. SQ109 showed activity against resistant Mtb and already advanced to Phase II/III clinical trials. Fast SQ109 degradation is attributed to the human liver Cytochrome P450s (CYPs). However, no information is available about interactions of the drug with Mtb CYPs. Here, we show that Mtb CYP124, previously assigned as a methyl-branched lipid monooxygenase, binds and hydroxylates SQ109 in vitro. A 1.25 Å-resolution crystal structure of the CYP124–SQ109 complex unambiguously shows two conformations of the drug, both positioned for hydroxylation of the ω-methyl group in the trans position. The hydroxylated SQ109 presumably forms stabilizing H-bonds with its target, Mycobacterial membrane protein Large 3 (MmpL3). We anticipate that Mtb CYPs could function as analogs of drug-metabolizing human CYPs affecting pharmacokinetics and pharmacodynamics of antitubercular (anti-TB) drugs.


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

CYP124, CYP125, and CYP142 Bind SQ109
Multiple substrates have been previously identified for CYP124, including cholesterol, vitamin D, their derivatives [18,20], and methyl-branched lipids [17]. Although its physiological function and, therefore, a natural substrate is not yet confirmed, there is a common pattern among all the substrates-a methyl-branched tail, tightly positioned for ω-hydroxylation in the distal site. Therefore, we tested an anti-TB drug candidate SQ109 to be a CYP124 substrate, as its geranyl tail is well mimicking the 2,5-dimethylhexane group.
We performed a spectrophotometric titration assay to estimate the binding affinity of the drug to the active site. The UV-visible difference spectrum of CYP124 shows type I response upon addition of SQ109, with a characteristic peak at 386 nm and a trough at 421 nm, consistent with the known methyl-branched substrates (Figure 1a) [17,18,20] The titration curve fits the tight-binding model with a Kd app = 3.4 ± 0.3 µM. The value is similar to that of for cholest-4-en-3-one (cholestenone), 3.4 ± 0.6 µM, and much stronger than that of for vitamin D3 (VD3) and 1α-hydroxy-vitamin D3 (1αOHVD3), 18 ± 2 µM and 34 ± 4 µM, respectively [20].
hydroxylate cholesterol and its immunoactive derivatives [19,20]. Similarly to CYP124, CYP125 showed type I response upon addition of SQ109 (Figure 1b). Surprisingly, SQ109 induced type II shift in the CYP142 spectrum, with a peak at 434 nm and trough at 414 nm (Figure 1c), indicating an apparent interaction of the diamine group with the heme iron. Compared to CYP124, both CYP125 and CYP142 bind SQ109 with significantly lower affinity (Kdapp = 41 ± 3 μM and 52 ± 11 μM, respectively).

CYP124 Catalyses Hydroxylation of SQ109
Next, we measured the catalytic activity in the reconstituted system, containing CYP124/125/142, SQ109, redox partners, and a NADPH-regenerating system. LC-MS detected the putative product after an hour of incubation. Among three proteins, only CYP124 showed activity against SQ109 with a turnover number of 0.60 ± 0.09 min −1 (Figure 1d). Although both CYP124 and CYP125 previously demonstrated close substrate specificity [20], different stereometry of the binding [19] might explain the absence of the product for the latter. MS-MS subsequently identified the CYP124 product. The parental molecular ion m/z was 331.4 for the substrate and 347.4 for the product, suggesting the addition of one oxygen atom to SQ109 (Figure 1f). The adamantane group can be easily identified by a characteristic peak at 135.2 ( Figure 1e). The peak at 69.2 in the substrate spectrum is of particular is colored blue, and the product of SQ109 after CYP124 hydroxylation (for parental molecular ion m/z = 347.4) is colored red. The inset shows fragment ion m/z = 85.2 found only in the product spectrum and the corresponding to monooxygenated SQ109 tail moiety (original fragment ion of untreated SQ109 m/z = 69.2).
Besides CYP124, we also tested Mtb CYP125 and CYP142 for SQ109 binding. These three enzymes were previously proposed to have close substrate specificity, as they were able to bind and hydroxylate cholesterol and its immunoactive derivatives [19,20]. Similarly to CYP124, CYP125 showed type I response upon addition of SQ109 ( Figure 1b). Surprisingly, SQ109 induced type II shift in the CYP142 spectrum, with a peak at 434 nm and trough at 414 nm (Figure 1c), indicating an apparent interaction of the diamine group with the heme iron. Compared to CYP124, both CYP125 and CYP142 bind SQ109 with significantly lower affinity (Kd app = 41 ± 3 µM and 52 ± 11 µM, respectively).

CYP124 Catalyses Hydroxylation of SQ109
Next, we measured the catalytic activity in the reconstituted system, containing CYP124/125/142, SQ109, redox partners, and a NADPH-regenerating system. LC-MS detected the putative product after an hour of incubation. Among three proteins, only CYP124 showed activity against SQ109 with a turnover number of 0.60 ± 0.09 min −1 (Figure 1d). Although both CYP124 and CYP125 previously demonstrated close substrate specificity [20], different stereometry of the binding [19] might explain the absence of the product for the latter. MS-MS subsequently identified the CYP124 product. The parental molecular ion m/z was 331.4 for the substrate and 347.4 for the product, suggesting the addition of one oxygen atom to SQ109 (Figure 1f). The adamantane group can be easily identified by a characteristic peak at 135.2 ( Figure 1e). The peak at 69.2 in the substrate spectrum is of particular interest, as it corresponds to amylene and disappears in the product spectrum. In contrast, a peak at 85.2 appears in the product, indicating that CYP124 catalyzed the hydroxylation of SQ109 in the amylene part.
The previous study of in vitro metabolism of SQ109 by the MS analysis showed that SQ109 (m/z 331.5) was effectively transformed by microsomes from various species to four different products with m/z 361 (M1), 347 (M2), 195 (M3), and 363 (M4) [16]. Recombinant human CYPs, CYP2D6 and CYP2C19, could extensively metabolize SQ109 to M1 and M4, while the latter could also produce M2 and M3. Notably, two different products have been assigned to M2. The first was proposed to be SQ109 with the oxidized adamantane ring. The second reaction was believed to ultimately result in the formation of the ketodiamine SQ109. In our study CYP124 produced a monooxygenated product with m/z ratio (346.4) close to those for the M2 metabolites. The first product was not detected on our MS-MS spectrum as the adamantane peak remained intact. However, we cannot exclude the possibility of the formation of the second product without the comparison with a reference MS. Overall, our finding suggests that Mtb CYP124 may also contribute to the metabolism of SQ109. Given our MS-MS data as well as previously reported ω-hydroxylation activity of methyl-branched chain fatty acids [17], we assume that CYP124 catalyzes chemically disfavored ω-hydroxylation of SQ109.

Crystal Structure of CYP124-SQ109 Complex
To determine the binding mode and the hydroxylation position, we co-crystallized SQ109 with CYP124 as described previously [20]. The structure of the CYP124-SQ109 complex was determined with 1.25 Å resolution (Protein Data Bank accession number, PDB ID: 6T0J). The crystallographic data collection and refinement statistics are given in Table 1. The overall structure resembles the CYP124-VD3 complex (PDB ID: 6T0G; Cα-Cα RMSD being 0.2 Å; Figure 2a) [20]. The 2mF o −DF c composite omit map unambiguously shows SQ109 density in two conformations, which differ only in the ethylenediamine-adamantane part. At the same time, geranyl tails in both conformations are positioned above Fe in the typical geometry for ωhydroxylation ( Figure 3a). Distances between the heme iron and the carbon atoms of the branched methyl groups are 4.0 Å and 6.1 Å, which favorites the hydroxylation of the ω-methyl group in the trans position. Thus, CYP124 catalyzes the regio-and stereo-selective hydroxylation of SQ109.  The geranyl tail is tightly enclosed in the narrow tunnel with a broadening to accommodate ω-5 methyl branching position (methyl at C3 of SQ109, Figure 3b). In contrast, adamantane and ethylenediamine are positioned more freely in a wide cavity in the upper part of the binding pocket. The interactions of SQ109 are mostly hydrophobic (L60, F63, T96, and F212 interact with the adamantane; F107, I111, L198, L263, V266, A267, F416, and I417 fix the aliphatic chain), while in the predominant conformation (A), the drug could potentially form a weak hydrogen bond with HOH879 (Figure 3a,b). The structure of the complex again emphasizes remarkable flexibility, which allows CYP124 to accommodate a variety of chemically dissimilar substrates. Indeed, the  [20] and phytanic acid (PA; PDB ID: 2WM4; magenta) [17]; (c-e) CYP124 ligand-binding cavities for ligand-free form, SQ109 and VD3 complexes, respectively. deteriorate the CYP cycle, it may affect the reaction speed, as we have previously observed for VD3 and 1αOHVD3 [20].

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 ωmethyl group in the trans position. The ability to hydroxylate SQ109 was not detected for the other Binding of SQ109 induces similar structural rearrangements, previously reported for phytanic acid (PA) complex and also observed for other substrates (Figure 2a) [17,20]. Briefly, paired movement of αF (K184-G199), BC-loop (S89-D115), and straightening of αI at N-terminal (S253-T271) results in the formation of a narrow tunnel, suitable for long methyl-branched hydrocarbons. The N-terminal of αD (K132-A139) is bent towards the proximal part of heme, contributing to the formation of the putative binding site for the redox-partner. Finally, FG-loop (G199-D208) and ordered "lid"-loop (W55-H69) enclose the active site from the solvent. The result of these transitions is shown for ligand-free, SQ109, and VD3 complexes (Figure 2c-e, respectively).
The geranyl tail is tightly enclosed in the narrow tunnel with a broadening to accommodate ω-5 methyl branching position (methyl at C3 of SQ109, Figure 3b). In contrast, adamantane and ethylenediamine are positioned more freely in a wide cavity in the upper part of the binding pocket. The interactions of SQ109 are mostly hydrophobic (L60, F63, T96, and F212 interact with the adamantane; F107, I111, L198, L263, V266, A267, F416, and I417 fix the aliphatic chain), while in the predominant conformation (A), the drug could potentially form a weak hydrogen bond with HOH879 (Figure 3a,b). The structure of the complex again emphasizes remarkable flexibility, which allows CYP124 to accommodate a variety of chemically dissimilar substrates. Indeed, the superposition of all known substrates shows that the upper part of the active site can adjust to substantially different head radicals, while methyl-branched tail remains pinned for ω-hydroxylation (Figure 2b; RMS distance between eight terminal carbons of the substrates not exceeding 0.6 Å).
Two common features between CYP124 substrates have been previously identified: A methyl-branched tail and a polar head [17]. The importance of the polar group remains unclear, as, in the majority of published complexes, it does not form H-bonds with the protein [17,20]. Moreover, in the CYP124-VD3 crystal structure (PDB ID: 6T0G), the polar A-ring is disordered in the upper part of the pocket (Figure 2e) [20]. The absence of stabilizing H-bonds might result in the emergence of two conformations of the adamantane group. Although less polarity of the head group is not likely to deteriorate the CYP cycle, it may affect the reaction speed, as we have previously observed for VD3 and 1αOHVD3 [20].

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]  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.

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.

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].

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).

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(HCO 2 ) 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 1 1 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).

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].