X-ray Structure of 4,4′-Dihydroxybenzophenone Mimicking Sterol Substrate in the Active Site of Sterol 14α-Demethylase (CYP51)*♦

A universal step in the biosynthesis of membrane sterols and steroid hormones is the oxidative removal of the 14α-methyl group from sterol precursors by sterol 14α-demethylase (CYP51). This enzyme is a primary target in treatment of fungal infections in organisms ranging from humans to plants, and development of more potent and selective CYP51 inhibitors is an important biological objective. Our continuing interest in structural aspects of substrate and inhibitor recognition in CYP51 led us to determine (to a resolution of 1.95Å) the structure of CYP51 from Mycobacterium tuberculosis (CYP51Mt) co-crystallized with 4,4′-dihydroxybenzophenone (DHBP), a small organic molecule previously identified among top type I binding hits in a library screened against CYP51Mt. The newly determined CYP51Mt-DHBP structure is the most complete to date and is an improved template for three-dimensional modeling of CYP51 enzymes from fungal and prokaryotic pathogens. The structure demonstrates the induction of conformational fit of the flexible protein regions and the interactions of conserved Phe-89 essential for both fungal drug resistance and catalytic function, which were obscure in the previously characterized CYP51Mt-estriol complex. DHBP represents a benzophenone scaffold binding in the CYP51 active site via a type I mechanism, suggesting (i) a possible new class of CYP51 inhibitors targeting flexible regions, (ii) an alternative catalytic function for bacterial CYP51 enzymes, and (iii) a potential for hydroxybenzophenones, widely distributed in the environment, to interfere with sterol biosynthesis. Finally, we show the inhibition of M. tuberculosis growth by DHBP in a mouse macrophage model.

Sterol 14␣-demethylase (CYP51) 2 is a cytochrome P450 (P450, CYP) heme thiolate containing enzyme involved in biosynthesis of membrane sterols, including cholesterol in animals, ergosterol in fungi, and a variety of C24-modified sterols in plants and protozoa in most organisms in biological kingdoms from bacteria to animals (1). CYP51 has been a therapeutic target for several generations of azole antifungal agents including fluconazole, voriconazole, itraconazole, ravuconazole, and posaconazole (2). These drugs inhibit microbial growth by disrupting biosynthesis of ergosterol, a major component of fungal membrane. Protozoa share with fungi the requirement of ergosterol and ergosterol-related sterols for cell viability and proliferation (3). Inhibition of sterol biosynthesis has been proven to be effective in trypanosomatids (3)(4)(5) and Leishmania spp (6), which cause such tropical diseases as African sleeping sickness, Chagas disease, and leishmaniasis.
Although a paradigm for the CYP51 family of sterol 14␣demethylases, the biological and molecular functions of CYP51 Mt in M. tuberculosis remain poorly understood. The requirement of M. tuberculosis for host cholesterol to be taken up by macrophages and for subsequent intracellular survival (31) suggests that in vitro sterol-modifying CYP51 Mt may be involved in cholesterol-mediated cell entry (32). However, cyp51 has not been identified among cholesterol catabolic genes of Rhodococcus sp. RHA1 or M. tuberculosis H37Rv (33), leaving open the question of the biological function of CYP51 Mt .
Based on the distribution of temperature factor values (B-factors) in previously reported CYP51 Mt structures (12)(13)(14), an assumption was made that the enzyme undergoes conformational changes when the substrate binds and that these changes are dominated by the remodeling of the C helix and the BC-loop, facilitated by torsional flexibility of the remote GHand HI-loops. These highly dynamic regions are enclosed in mutational hot spots in azole-resistant clinical isolates of C. albicans (13,36,37), indicating their either direct or indirect (via protein dynamics) involvement in the substrate or inhibitor binding. A critical mutation hot spot, the BC-loop (36), could not be fully defined in previously reported structures due to insufficient electron density in this region.
Here, we report determination of the crystal structure of the CYP51 Mt -4,4Ј-dihydroxybenzophenone (DHBP) complex to a resolution of 1.95 Å. DHBP was identified previously as a top type I binding hit in a library of small synthetic organic molecules screened against CYP51 Mt (12). The newly determined structure confirms a potential for the benzophenone scaffold to mimic steroids and suggests its utility for the development of a novel class of inhibitors targeting flexible regions and dynamic states of CYP51 to overcome drug resistance acquired by fungi toward azole and triazole drugs. The structure reveals previously obscure interactions and highlights the role of the conserved aromatic residue Phe-89 (analogous to the mutational hot spot residue Tyr-132 in C. albicans), thus making a more accurate template for modeling the three-dimensional structures of fungal and prokaryotic orthologs.

EXPERIMENTAL PROCEDURES
Spectroscopic Binding Assays-CYP51 Mt double C37L/ C442A mutant with improved expression and purification qualities (14) was prepared as described elsewhere (12). Spectroscopic characterization of DHBP (ABCR GmbH & Co.KG) binding was conducted in a 1-ml quartz cuvette containing 2.5 M CYP51 Mt in 50 mM Tris-HCl, pH 7.5, and 10% glycerol. DHBP dissolved in Me 2 SO at 25 mM concentration was added in 0.5-l aliquots, resulting in a concentration increase from 12.5 to 125 M in 12.5-M increments. The same amounts of Me 2 SO alone were added to a reference cuvette followed by recording of difference spectra. Concentration of CYP51 Mt was determined from the difference spectra between the carbon monoxide-bound ferrous and water-bound ferric forms using the extinction coefficient of 91,000 M Ϫ1 cm Ϫ1 (38). To determine the dissociation constant (K D ), titration data were fitted using program GraphPad PRISM (Graphpad Software Inc.) to the rectangular hyperbola (A obs ϭ A max (S/(K D ϩ S)), where A obs is the absorption shift at 420 nM determined at any ligand concentration; A max is the maximal absorption shift obtained at saturation; K D is the dissociation constant for the protein-ligand complex; and S is the ligand concentration).
Crystallization, Data Collection, and Determination of Crystal Structure-Crystals of the CYP51 Mt -DHPB complex could not be obtained under conditions promoting crystal growth for the other previously characterized CYP51 Mt forms. Therefore, ab initio screening of crystallization conditions was performed using the nanoliter drop setter Mosquito (TTP LabTech) (hanging drop crystallization protocol) and high throughput screening kits purchased from both Hampton Research and Qiagen. Protein from the 1.2 mM frozen stock stored at Ϫ80°C in 20 mM Tris-HCl, pH 7.5, 200 mM NaCl, and 0.5 mM EDTA was mixed with DHBP dissolved in Me 2 SO at 100 mM stock concentration to final concentrations of 0.2 mM for protein and 2 mM for the ligand. The volume of the mixture was adjusted with 10 mM Tris-HCl, pH 7.5. The crystallization conditions generating small needle-shaped crystals were identified and optimized further in the 24-well crystallization plates. Larger needles of the diffraction quality were finally obtained by hanging drop vapor diffusion method from 1.2 M lithium sulfate, 0.1 M HEPES, pH 7.5, and 2% isopropyl alcohol. Prior to data collection, the crystals were briefly washed in the well solution supplemented with 20% glycerol and then flash-frozen in liquid nitrogen. Diffraction data were collected to a resolution of 1.95 Å at 100 -110 K at beamline 8.3.1, Advanced Light Source, Lawrence Berkeley National Laboratory. The images were integrated, and the intensities were merged by using the HKL2000 software suite (39). Diffraction statistics are shown in Table 1.
The structure was determined by molecular replacement using coordinates of BSPPA-bound CYP51 Mt (Protein Data Bank ID 2CIB) as a search model. The final atomic model (Table 1) was obtained after a few iterations of refinement using REFMAC5 (40) and CNS (41) and model building using the COOT (42) and O (43) programs. The quality of the structure (Table 1) was assessed by the program PROCHECK (44).
Macrophage Assay for DHBP Inhibitory Activity-Bone marrow-derived macrophages were extracted from C57/BL6 mice and seeded in culture dishes (ϳ5 ϫ 10 6 cells/dish). Macrophages were allowed to differentiate from 5 to 6 days in Dulbecco's modified essential medium containing 10% heat-inacti-vated fetal calf serum, 5% inactivated horse serum, 10% L-929-CSF cell supernatant, 10 mM HEPES, pH from 6.5 to 7.5, 1 mM sodium pyruvate, and 1% L-glutamine at 37°C and 5% CO 2 . Cells were then seeded in 96-well plates at a density of 1 ϫ 10 5 cells/well in Dulbecco's modified essential medium containing 10% fetal calf serum, 10 mM HEPES, pH from 6.5 to 7.5, 1 mM sodium pyruvate, and 1% L-glutamine and incubated overnight at 37°C and 5% CO 2 for adhesion. M. tuberculosis strain H37Rv from mid-exponential growth phase were collected by sedimentation and resuspended in phosphate-buffered saline (Invitrogen). Single cell suspension was made by three forced passages of suspension through a 0.40-m needle. Bacterial cell density was measured by microscopy using a Petroff-Hauser chamber, and infection of macrophages was performed in the 96-well plates, at three bacilli per a macrophage cell in triplicate wells. Macrophages were incubated at 37°C for 3 h and then washed twice with phosphate-buffered saline to remove the non-phagocytosed M. tuberculosis bacilli, and fresh medium containing DHBP or EPBA at working concentrations was added. Wells were periodically observed for cell viability or detachment under the microscope. For the assessment of colony-forming units, macrophages were lysed with 0.5% Triton X-100 in phosphate-buffered saline at days 1, 2, 3, 4, and 5 after infection, and serial dilutions in 0.08% Tween 80 in phosphatebuffered saline were plated on Middlebrook 7H11 agar plates.
Plates were incubated at 37°C for 3 weeks followed by the enumeration of colony-forming units. Data representing the averaged numbers of bacteria in the triplicate cultures plotted against compound concentration were fitted using program GraphPad PRISM.

RESULTS
Binding of DHBP in the Active Site of CYP51 Mt -DHBP binding was addressed by manual spectroscopic binding assays (Fig. 1), and a K D of 29.5 Ϯ 1.6 M was determined for the CYP51 Mt -DHBP complex, indicating 3-fold higher binding affinity for DHBP when compared with a non-substrate sterol estriol (K D ϭ 100 M) (14).
Determination of the Crystal Structure-The CYP51 Mt -DHPB complex crystallized under different conditions than those favoring crystallization of other CYP51 Mt forms (12)(13)(14). Crystals diffracted in the P2 1 2 1 2 1 space group having different unit cell dimensions, and hence, different packing interactions, apparently resulting from reshaping protein surface caused by the DHBP binding. Structure was determined to a resolution of 1.95 Å, and the coordinates were deposited with the Protein Data Bank (PDB ID code 2VKU).
Overall Structure of the CYP51 Mt -DHBP Complex-Unlike other structurally defined CYP51 Mt forms, the CYP51 Mt -DHBP structure is represented by a continuous electron density for all protein regions including the BC-loop and the C helix, which makes it the most complete CYP51 Mt structure determined to date. For comparison, 10 residues (85-94) within the BC-loop are disordered in the CYP51 Mt -estriol complex (PDB ID code 1X8V), 20 residues (85-104) in both the CYP51 Mt -EPBA and the CYP51 Mt -BSPPA complexes (PDB ID codes 2CI0 and 2CIB, respectively), and 10 residues (91-100) in the ligand-free CYP51 Mt (PDB ID code 1H5Z). Superimposition of the CYP51 Mt -DHBP complex with the other CYP51 Mt forms (Fig.  2) maps differences between the structures to the region comprising the BC-loop, the C helix, the H helix and its adjacent loops, and the C terminus of the G helix. No significant differences were observed in the FG-region, confirming that the substrate/inhibitor access to the active site in CYP51 Mt occurs µ µ ∆ FIGURE 1. DHBP binding to CYP51 Mt . The dissociation constant, K D , of DHBP binding was estimated from the absorbance shift at 420 nm during spectroscopic titration of CYP51 Mt with increasing concentrations of DHBP. A fragment of the series of difference spectra is shown in the inset. between the BC-loop and the C helix in the direction parallel to the heme plane and orthogonal to that in other P450 enzymes. The ϳ150°I helix kink persists in the CYP51 Mt -DHBP complex, although as a less destructive and smoother bend, largely preserving the H-bonding pattern of the ␣-helix. The short H helix, which connects G and I helices via the adjacent loops, is relocated up to 5.0 Å closer to the protein core (when compared with the 4-phenylimidazole-bound CYP51 Mt (PDB ID code 1E9X) having root mean square deviation with the CYP51 Mt -DHBP structures of 0.85 Å), thus transmitting the 2.3 Å motion of the I helix N terminus to the C terminus of the G helix, which is displaced 5.2 Å (Fig. 2). The BC-loop relocates up to 5.5 Å when DHBP binds in the active site.
The most distinct feature of the bound DHBP is that one of its two 4-hydroxylated phenyl rings (Fig. 4A, ring A) coincides with the 3-hydroxylated ring of estriol, suggesting for the polar hydroxyl group a role of a substrate recognition landmark, which apparently is neither sterol-specific nor CYP51-specific, because all CYP51 sterol substrates possess this same feature (Scheme 1A), and no contacts with any of the amino acid side chains of the protein are involved in the interactions. Thus, one of the 4-hydroxyl group of the DHBP is bound in the hydrophilic pocket formed by the backbone amide groups of Ile-322 (via a water molecule) and Ile-323 and the carbonyl group of Ile-323, whereas the side chains of the Tyr-76, Phe-78, Leu-321, Leu-324, and Val-434 stabilize ring A via hydrophobic contacts (Fig. 4). The oxygen atom of the carbonyl group connecting the rings A and B H-bonds (distance 2.9 Å) to the side chain of His-259. Ring B approaches the heme iron within 8 Å and is stabilized byinteractions with the side chains of Tyr-76, Phe-83, Met-79, and Phe-255, whereas its 4-hydroxyl group points toward the hydrophilic and/or charged groups of Gln-72, Arg-96, and the heme propionate moiety, the latter being bent toward the ligand.
Roles of the Conserved Active Site Residues-Three of the eighteen active site residues, Tyr-76, Phe-83, and His-259, are invariant in the CYP51 protein family (supplemental Fig. S1). Moreover, Phe-83 is followed by glycine in all CYP51 proteins identified to date. The Phe-78 and Phe-255 are strictly specific for 14␣-demethylation of the sterol substrates carrying a single C-4 methyl group in ␣-configuration. Phe-78 was shown to impose steric constrains on binding sterol substrates having a methyl group in ␤-configuration (14), and therefore, is a key   discriminator between CYP51 sterol substrates single-and double-methylated at the C-4 atom (14,45). The "humanized" F78L CYP51 Mt mutant does not bind DHBP (12), suggesting an essential role for Phe-78 in the recognition of benzophenone scaffold.
The remaining 13 residues overlap among phyla. Among them, position 96 is the most phylum-specific, being occupied largely by arginine in bacteria and plants (supplemental Fig. S1, highlighted in yellow), methionine in protists (highlighted in pink), phenylalanine in animals (highlighted in green), and leucine (highlighted in orange) in most fungi, with an exception being phenylalanine in Schizosaccharomyces pombe, Cunninghamella elegans, and Ustilago maydis.
Nonspecific DHBP Binding-Three other DHBP molecules bound to the protein surface were identified. The first, having two alternative conformations, binds in the same crevice between the H and I helices as the second estriol molecule in the CYP51 Mt -estriol structure (14). The second one binds to a crevice formed by the cysteinyl loop on the proximal (in respect to heme) surface of the protein, and a third one binds between the AЈ and the FЈ helices. Both the second and the third DHBP molecules mediate protein-protein contacts, which might have facilitated crystallization of the CYP51 Mt -DHBP complex. Binding of multiple molecules may suggest the propensity for DHBP for nonspecific interactions with the proteins.
DHBP-BC-loop Interactions-A remarkable finding provided by the crystal structure of the CYP51 Mt -DHBP complex is elucidation of the function of residue Phe-89, which is important for substrate binding and conversion in CYP51 Mt (46). Well conserved Phe-89 (supplemental Fig. S1) was reported to be involved in fungal drug resistance, inhibitor binding, and the catalytic function of CYP51 in C. albicans (Tyr-132, according to C. albicans numbering) (36,(47)(48)(49)(50)(51), Histoplasma capsulatum (Tyr-136, according to H. capsulatum numbering) (52), and the causative agents of zygomycosis in humans, Rhizopus oryzae and Absidia corymbifera (53). Phe-89 interactions were obscure in the CYP51 Mt -estriol complex due to the missing electron density for this portion of the loop. In the CYP51 Mt -DHBP structure, the electron density for the BC-loop backbone and the majority of the side chains, including Phe-89, is defined, revealing no direct contacts of Phe-89 with the ligand. Instead, Phe-89 serves to maintain the loop architecture via the interactions with Phe-83, Arg-96, and Ser-252 (Figs. 3 and 4), supporting the function of Arg-96, which apparently recognizes the sterol side chain attached to the C-17 atom of the sterol nucleus (Scheme 1A). The position of Arg-96 suits for the interactions with the side chain of lanosterol docked in the active site based on the molecular dynamics simulations (37) (Fig. 4B).
Inhibitory Effect of DHBP and EPBA-Tuberculostatic activity of DHBP and EPBA was examined in the M. tuberculosis infection model using mouse primary bone marrow-derived macrophages. Both compounds inhibited M. tuberculosis growth, albeit to different degrees. A slight inhibitory effect of DHBP was observed at 50 M, which increased markedly as DHBP concentration increased to 100 M (Fig. 5A). Four days after infection, DHBP induced from 40 to 70% reduction in bacterial colony-forming units, when compared with the untreated control, whereas EPBA induced only 30% reduction (Fig. 5B). It should be noted that the highest EPBA concentra- tion tested was 30 M due to the compound toxicity at higher concentrations. On the contrary, no cell toxicity was detected for DHBP up to 200 M. The higher inhibitory effect of DHBP in macrophage model agrees with the results previously obtained in broth culture using the Alamar blue assay (12).

DISCUSSION
A soluble ortholog of its eukaryotic membrane-associated counterparts, CYP51 Mt serves as a template for three-dimensional modeling of CYP51-inhibitor interactions in microbial pathogens. CYP51 Mt undergoes extensive remodeling of the BC-loop and C helix, the closed conformation for the BC-loop in the CYP51 Mt -DHBP complex having been observed for the first time here. A screen of synthetic organic molecules against CYP51 Mt (12) identified three type I hits (affinity range 25-50 M): DHBP, the estrogen metabolite 11-ketoestrone, and 2,7dihydroxy-9-fluorenone (Scheme 1B). Although structurally different, all three compounds have certain similarities of size, shape, and hydroxylation pattern.
Hydroxybenzophenones are industrial bulk products used as UV stabilizers in plastic surface coatings on food packaging. Hydroxybenzophenone derivatives are used as ingredients in sunscreen and other cosmetics to absorb UV light. These compounds are absorbed through human skin, and bioaccumulation may occur in wildlife and humans (64 -66). An inhibitory effect of benzophenones on CYP51 has not to our knowledge been reported previously. However, hydroxybenzophenones, including DHBP, have been found elsewhere to inhibit squalene cyclase, another enzyme in cholesterol biosynthesis (67), and have been identified as estrogen receptor ligands exhibiting estrogenic and antiandrogenic activities in a human breast cancer cell line (68,69). Further, a biochemical evaluation of nonsteroidal benzophenone-containing analogs of cholesterol suggests that the benzophenone scaffold can successfully replace the sterol tetracyclic nucleus in apolipoprotein-dependent cellular sterol efflux (70).
DHBP binds to CYP51 Mt with about the same affinity as to the estrogen receptor (68). The 4-hydroxyl group of DHBP, which is essential for hormonal activities of benzophenone derivatives (69), mimics the sterol C-3 hydroxyl group in the active site of CYP51 Mt . Two DHBP 4-hydroxyphenyl rings positioned in an angular arrangement contact the same amino acid residue as the tetracyclic estriol ( Fig. 4A) but, unlike estriol, they rearrange the BC-loop, revealing the interactions of the conserved aromatic functionality of Phe-89. Similar to the other mutation hot spot residues in CYP51, Phe-89 is not involved in direct interactions with the ligand bound in the active site, supporting our previous assumption that fungal resistance toward azole drugs related to CYP51 modifications evolved via alteration of CYP51 dynamics rather than through immediate protein-ligand contacts.
Lastly, tuberculosis is becoming increasingly difficult to treat as a result of the growing incidence of multidrug and extensively drug-resistant M. tuberculosis strains (71). Thus, the need for new therapeutic agents has become imperative. Although the biological function of CYP51 Mt remains unknown, we previously demonstrated the inhibitory effects of DHBP and EPBA on M. tuberculosis growth in broth culture (12). Antibacterial (72) and fungicidal (73,74) effects of modified benzophenones have been demonstrated elsewhere. Here, we confirm that the same range of DHBP concentrations that causes inhibition in liquid culture inhibits growth of M. tuberculosis in mouse macrophages without harming them. However, given the propensity of DHBP for nonspecific interactions with proteins, further analysis is required to establish a mechanism of DHBP action and to distinguish between the bacteriostatic or bactericidal nature of its antimycobacterial activity.
In summary, the binding mode of the aromatic compound DHBP to CYP51 was validated by determination of the x-ray structure of the CYP51 Mt -DHBP complex. DHBP binding triggers rearrangement of the protein BC-loop, allowing elucidation of the previously obscure role of the conserved Phe-89, reported elsewhere to be essential for substrate binding, catalytic conversion, and fungal drug resistance. Structurally different from sterols, DHBP binds CYP51 Mt via a type I mechanism, suggesting a possibility of exploiting a DHBP scaffold to design a novel class of CYP51 inhibitors targeting regions including the flexible BC-loop and the C helix. Such inhibitors may help to overcome the fungal drug resistance to antifungal azoles and triazoles, which has developed over the years. Another point highlighted by our study is the potential of hydroxybenzophenones to interact with sterol 14␣-demethylases and probably to affect sterol biosynthesis. Additionally, the interactions of benzophenone and fluorenone scaffolds with CYP51 Mt suggest the possibility of alternative catalytic functions for bacterial CYP51 enzymes, such as biodegradation of polycyclic aromatic hydrocarbons. In this regard, a precedent has recently been reported of CYP51 being recruited for production of antimicrobial plant defense compounds (75).