Small-Molecule Scaffolds for CYP51 Inhibitors Identified by High-Throughput Screening and Defined by X-Ray Crystallography

Abstract

Sterol 14α-demethylase (CYP51), a major checkpoint in membrane sterol biosynthesis, is a key target for fungal antibiotic therapy. We sought small organic molecules for lead candidate CYP51 inhibitors. The changes in CYP51 spectral properties following ligand binding make CYP51 a convenient target for high-throughput screening technologies. These changes are characteristic of either substrate binding (type I) or inhibitor binding (type II) in the active site. We screened a library of 20,000 organic molecules against Mycobacterium tuberculosis CYP51 (CYP51_Mt), examined the top type I and type II binding hits for their inhibitory effects on M. tuberculosis in broth culture, and analyzed them spectrally for their ability to discriminate between CYP51_Mt and two reference M. tuberculosis CYP proteins, CYP130 and CYP125. We determined the binding mode for one of the top type II hits, α-ethyl-N-4-pyridinyl-benzeneacetamide (EPBA), by solving the X-ray structure of the CYP51_Mt-EPBA complex to a resolution of 1.53 Å. EPBA binds coordinately to the heme iron in the CYP51_Mt active site through a lone pair of nitrogen electrons and also through hydrogen bonds with residues H259 and Y76, which are invariable in the CYP51 family, and hydrophobic interactions in a phylum- and/or substrate-specific cavity of CYP51. We also identified a second compound with structural and binding properties similar to those of EPBA, 2-(benzo[d]-2,1,3-thiadiazole-4-sulfonyl)-2-amino-2-phenyl-N-(pyridinyl-4)-acetamide (BSPPA). The congruence between the geometries of EPBA and BSPPA and the CYP51 binding site singles out EPBA and BSPPA as lead candidate CYP51 inhibitors with optimization potential for efficient discrimination between host and pathogen enzymes.

In eukaryotic organisms, cytochrome P450 enzymes play important roles in many systems, including the biosynthesis of cholesterol, steroid hormones, and vitamins; the control of cardiovascular physiology and systemic blood pressure; drug metabolism; and chemical toxicology and carcinogenesis (23). With respect to their potential for pharmacological development, eukaryotic P450 enzymes may be divided into two groups, drug targets and drug-metabolizing enzymes. Well-established P450 drug targets include (i) the aromatase CYP19, required for the conversion of androgens into estrogens (30) and a key target in the treatment of breast cancer; (ii) sterol 14α-demethylases (CYP51), required for the biosynthesis of membrane sterols, including cholesterol in animals, ergosterol in fungi, and a variety of C-24-modified sterols in plants and protozoa (2), and a key target in the treatment of diseases caused by infectious microbes; and (iii) other biosynthetic sterol hydroxylases (20). The isoform-specific inhibition of P450 enzymes offers promise for the development of therapeutic, insecticidal, and herbicidal agents (7). High-throughput screening (HTS) libraries of small organic molecules and potential drugs against P450 enzymes may thus be key to selecting high-quality compounds in the lead identification stage (10) and later in evaluating the pharmacological properties of drug candidates and predicting drug-drug interactions and toxicity (18).

Cytochromes P450 are heme thiolate-containing enzymes whose spectral properties make them convenient targets for HTS technologies used to monitor changes in the visible region of their optical spectra—350 through 450 nm—which occur in response to the binding of a ligand, substrate, or inhibitor in the active site (23). The binding of a ligand to P450 elicits two major types of spectral changes: type I, characteristic of substrate binding, and type II, characteristic of azole inhibitor binding (see Fig. 1). The concentration dependence of the spectral changes allows an estimate to be made of a compound's binding affinity in the CYP active site. These properties are the basis of our experimental platform for HTS of small-molecule libraries against P450 drug targets.

FIG. 1.

Type I and type II spectral responses upon ligand binding in CYP51_Mt. (A) Type I spectral changes caused by binding of the nonsubstrate sterol estriol. (B) Type II spectral changes as a result of binding of an azole inhibitor, 4-phenylimidazole. The concentration dependence of the spectral changes allows the binding affinities of the ligands to be estimated.

The soluble CYP51 ortholog from Mycobacterium tuberculosis, CYP51_Mt (3), has two features that make it an attractive target for developing a screening assay and defining the binding of positive hits. First, it belongs to a family of sterol 14α-demethylases, whose biological function is the removal of the 14α-methyl group from sterol molecules generated during sterol biosynthesis in the majority of eukaryotic organisms, including pathogenic fungi and protozoa (2). Major antifungal drugs such as fluconazole, voriconazole, itraconazole, ravuconazole, and posaconazole are inhibitors of sterol 14α-demethylases. These drugs target the biosynthesis of ergosterol, a major component of fungal membranes, thereby preventing fungal growth.

CYP51_Mt may also be a potential M. tuberculosis therapeutic target on its own. It is one of the 20 different CYP enzymes encoded by the M. tuberculosis genome (6). The susceptibility of M. tuberculosis to the azole antifungal agents that target these enzymes suggests their important roles in M. tuberculosis physiology and, hence, their potential use as therapeutic targets (19). CYP51_Mt is the only M. tuberculosis CYP enzyme whose catalytic function has been demonstrated. Although due to the absence of the complete sterol biosynthetic pathway, M. tuberculosis cannot synthesize sterols de novo (6), CYP51_Mt can demethylate the sterols lanosterol, dihydrolanosterol, and obtusifoliol (4). Given that cholesterol is essential for the uptake of mycobacteria by macrophages and their subsequent intracellular survival (11), the activity of CYP51_Mt toward sterols may potentially be implemented in cholesterol-mediated entry into macrophages. Blocking this step may impair or abolish M. tuberculosis infectivity.

Here, we screened a commercial library of 20,000 synthetic organic molecules comprising a large selection of molecular scaffolds against CYP51_Mt. Compounds with micromolar affinities for CYP51_Mt were identified. Two of these compounds, 4,4′-dihydroxybenzophenone (DHBP; type I binding mode) and α-ethyl-N-4-pyridinyl-benzeneacetamide (EPBA; type II binding mode), were assessed for their ability to discriminate among different CYP families and inhibit M. tuberculosis growth in broth culture. We also determined the structural features of the EPBA binding mode by X-ray crystallography analysis of the CYP51_Mt-EPBA complex. Based on its structural similarity to EPBA, we identified and examined a second type II compound, 2-(benzo[d]-2,1,3-thiadiazole-4-sulfonyl)-2-amino-2-phenyl-N-(pyridinyl-4)-acetamide (BSPPA), which was not a part of the commercial screening library. BSPPA binding properties were similar to those of EPBA. The congruence between the geometries of EPBA and BSPPA and the CYP51_Mt binding site confers micromolar affinity on these compounds and makes them lead candidate structures for CYP51 inhibitors.

MATERIALS AND METHODS

Preparation of protein samples.

The CYP51_Mt gene with a sequence encoding a C-terminal His₄ tag was cloned into the NdeI-HindIII sites of the pET17b expression vector (Novagen) (4), and the vector was used to transform Escherichia coli strain HMS174(DE3). Transformants were grown for 5 h at 37°C in Luria-Bertani medium supplemented with 0.1 M potassium phosphate buffer (pH 7.5), 0.4% glycerol, and trace elements before the induction of CYP51_Mt expression by the addition of isopropyl-β-d-thiogalactopyranoside (IPTG; final concentration of 0.2 mM). Simultaneously, δ-aminolevulinic acid, a precursor of heme biosynthesis, was added to a final concentration of 1 mM. Following induction, the temperature was decreased to 20°C. After 20 h, cells were harvested and lysed by sonication. Insoluble material was removed from the crude extract by centrifugation (15,000 × g), and the supernatant was subjected to a series of chromatographic steps, including nickel-nitrilotriacetic acid agarose (QIAGEN) chromatography followed by S-Sepharose (Amersham Biosciences) and Q-Sepharose (Amersham Biosciences) chromatography. Protein was eluted from Q-Sepharose by a 0 to 0.5 M NaCl gradient. Fractions containing P450 were combined, concentrated using a Centriprep concentrating device (Millipore), and stored at −80°C. The double C37L-C442A and the triple C37L-C151T-C442A and C37L-F78L-C442A CYP51_Mt mutant forms (26) used for cocrystallization and binding assays and the recombinant M. tuberculosis CYP125 and CYP130 proteins used in spectroscopic assays were produced in the same way.

Automated library screening.

Screening was performed using a medium-throughput pipetting robot (Sciclone 3000; Caliper Lifesciences), a plate reader for absorbance scanning (Safire; Tecan), and various robots for washing and dispensing. The FMP-20 compound library containing 20,000 small organic molecules was purchased from ChemDiv (San Diego, CA). Compounds were solubilized at 10 mM stock concentrations in dimethyl sulfoxide (DMSO) and distributed in 0.4-μl aliquots into 384-well microtiter plates, which were stored sealed with aluminum foil at −20°C. Prior to use, 40 μl of 50 mM Tris-HCl, pH 7.5, containing 10% glycerol was added to each well to achieve a 100 μM concentration of the compound and reduce the DMSO concentration to 1%. Plates were subsequently incubated at 37°C for 15 min, followed by 5 min of sonication to allow suspensions of the most hydrophobic compounds to solubilize. However, wells were not inspected visually for precipitates or cloudiness upon dilution. Compound-specific changes in absorption spectra (310 to 450 nm) were recorded automatically after adding 10 μl of 4 μM wild-type CYP51_Mt in 50 mM Tris-HCl, pH 7.5, and 10% glycerol. Estriol and 4-phenylimidazole were used as references for type I and type II binding modes, respectively. Reference concentrations varied from 50 μM to 2.5 mM.

Manual spectroscopic binding assays.

Manual binding assays were conducted in a 1-ml quartz cuvette containing P450 proteins at concentrations ranging from 1 to 2.5 μM in a solution of 10 mM Tris-HCl, pH 7.5, and 10% glycerol, to which a compound dissolved in DMSO at a stock concentration (ranging from 10 to 100 mM) was added in 0.5-μl aliquots. The same amounts of DMSO alone were added to a reference cuvette, followed by the recording of difference spectra. Concentrations of CYP proteins at 450 nm were determined by using the extinction coefficient of 91,000 M⁻¹ cm⁻¹ for the carbon monoxide-bound ferrous species (22). Titration data were linearized by plotting S₀/ΔA against S₀, where S₀ is the total concentration of a ligand and ΔA is the change in absorption. The equilibrium dissociation constant (K_D) was extrapolated from the intercept of the linear plot with the S₀ axis.

Antimycobacterial activity assay.

The susceptibility of M. tuberculosis to DHBP and EPBA in broth culture was qualitatively evaluated by the Alamar blue method (34). Hygromycin (Roche) was used as a positive control. M. tuberculosis strain H37Rv-MPI was propagated in Middlebrook 7H9 broth (Becton Dickinson) supplemented with 10% albumin-dextrose-catalase (Becton Dickinson), 0.25% glycerol, and 0.1% Tween 80 at 37°C under shaking conditions to an optical density of 1.0 at 600 nm. Compounds EPBA and DHBP were dissolved in DMSO at stock concentrations of 100 and 40 mM, respectively. Both compounds were further diluted to various working concentrations ranging from 20 μM to 1 mM with the 7H9 medium lacking Tween 80 and distributed in 100-μl aliquots into the wells of a 96-well plate. For a negative control, the same amounts of DMSO alone were used. A 100-μl aliquot of a 1:25 dilution of an M. tuberculosis culture containing about 4 × 10⁵ bacteria was added per well. The plates were sealed and incubated at 37°C for 4 days, after which 50 μl of a freshly prepared 1:1 mixture of 10× Alamar blue reagent (Serotec) and 10% Tween 80 was added to each well. Plates were allowed to incubate for an additional 24 h at 37°C. Plates were read by visual inspection for a resazurin (a major component of Alamar blue) color change from blue to pink upon reduction to resorufin in wells containing live mycobacteria (21).

Crystallization, data collection, and determination of crystal structure.

For crystallization experiments, we used previously designed CYP51_Mt mutant forms (26) for improved expression and purification qualities. The C37L-C442A double mutant form of CYP51_Mt was mixed with a ligand dissolved in DMSO at a 100 mM concentration to obtain a final protein concentration of 0.2 mM and a final ligand concentration of 1 to 5 mM. Ligands were purchased from ChemDiv (San Diego, CA). Ligand-free crystals were obtained from the C37L-C151T-C442A triple mutant protein. The narrow crystallization screening grid (15 to 30% polyethylene glycol 4000, 2 to 12% isopropanol, 0.1 M HEPES, pH 7.5) previously developed to obtain CYP51_Mt crystals (25, 26) was used for crystallization by the vapor diffusion hanging-drop method. All diffraction data were collected at 100 to 110 K at the Southeast Regional Collaborative Access Team 22ID and Structural Biology Center 19ID beamlines, Advanced Photon Source, Argonne National Laboratory, Argonne, IL (Table 1). The images were integrated, and the intensities were merged by using the HKL2000 software suite (24). The structures were determined by molecular replacement using the CNS software suite (5) with coordinates of estriol-bound CYP51_Mt (Protein Data Bank identification no. [ID] 1X8V) as a search model.

TABLE 1.

Data collection and refinement statistics

Parameter	Value(s)a or determination for:
	EPBA (PDB ID 2CI0)	BSPPA (PDB ID 2CIB)	Ligand-free structure (PDB ID 2BZ9)
Data collection parameters
Wavelength (Å)	1.00931	0.97933	0.99199
Resolution (Å)	1.53	1.50	2.20
No. of unique reflections	65111	69465	41999
Redundancy	4.4 (3.9)	5.5 (5.1)	3.4 (2.5)
% Completeness	97.9 (89.6)	98.2 (98.4)	97.3 (88.7)
Space group	P212121	P212121	P21
Cell dimensions (a, b, c) (Å)	45.0, 85.6, 110.5	44.9, 85.9, 111.0	38.9, 110.3, 99.6 (β = 90.67°)
No. of molecules in asymmetric unit	1	1	2
% Solvent content	40	40	40
% Rsymb	4.2 (13.5)	6.5 (55.5)	11.2 (35.6)
I/σ(I)c	47.7 (9.0)	27.5 (2.8)	13.7 (3.2)
Refinement parameters
No. of reflections used in refinement	62,309	65,316	39,163
Rcrys(Rfree)d (%)	20.3 (22.7)	20.4 (22.6)	21.5 (27.5)
No. of atoms in:
Protein	3,337	3,373	7,020
Heme	43	43	86
Substrate	18	29	None
Water	364	358	216
Wilson plot B value (Å2)	22.1	21.2	10.3
Mean B-factor (Å2) for:
Protein	24.9	23.4	Chain A, 22.6; chain B, 23.3
Heme	23.8	19.7	Chain A, 18.5; chain B, 21.4
Substrate	29.5	22.4	NA
Water	35.1	33.4	25.5
All	25.9	24.3	23.0
Root mean square deviation for:
Bond length (Å)	0.005	0.006	0.007
Bond angle (degrees)	1.2	1.2	1.3
% Protein residues in most favored/additional allowed/generously allowed regionse	Chain A, 91.5/8.2/0.3	Chain A, 90.9/8.8/0.3	Chain A, 90.6/9.1/0.0; chain B, 90.4/9.3/0.3

^aNumbers in parentheses correspond to the highest-resolution shell. PDB, Protein Data Bank; NA, not applicable.

^bR_sym = ΣI_i − <I>/ΣI_i, where I_i is the intensity of the ith observation and <I> is the mean intensity of reflection.

^cI/σ(I) is a measure of a signal-to-noise ratio, where I is the intensity of the reflection and σ(I) is the estimated standard deviation equal to [Σ(I_i−<I>)²/(N−1)]^1/2, where I_i is the intensity of the i^th observation, <I> is the mean intensity of reflection, and N is the number of measurements.

^dR_cryst = Σ‖Fo − Fc‖/ΣFo, where Fo and Fc are the observed and calculated structure factors, respectively, and R_cryst is calculated with the working reflection set. R_free is the same as R_cryst but calculated with the reserved reflection set.

^eRamachandran statistics calculated by the program PROCHECK (15).

The final atomic models (Table 1) were obtained after a few iterations of refinement using CNS (5) and manual model building with the program O (14). Ramachandran statistics (15) (Table 1) indicate no outliers, with the exception of one residue, A46, in chain A of the ligand-free structure (Protein Data Bank ID 2BZ9). From one to four N-terminal residues, depending on the crystal form, were missing in all protein chains due to insufficient electron density in the N-terminal region. Twenty BC-loop residues, all C-helix residues, and five GH-loop residues were also missing in the ligand-bound forms. In the ligand-free form, both molecules in an asymmetric unit lacked from 10 to 14 BC-loop residues and 4 or 5 GH-loop residues for the same reason.

Protein Data Bank accession numbers.

The atomic coordinates and structure factors determined in this study (Protein Data Bank IDs 2CI0, 2CIB, and 2BZ9) have been deposited in the Protein Data Bank, Research Collaboratory for Structural Bioinformatics, Rutgers University, New Brunswick, NJ (http://www.rcsb.org/).

RESULTS

HTS assay.

The HTS assay was developed based on the optical spectral properties of P450 enzymes to elicit both type I and type II spectral changes (Fig. 1). Type I changes shift the ferric heme Soret γ band from 418 nm to 388 to 394 nm (showing a peak at ∼390 nm and a trough at ∼420 nm in the difference spectra) (Fig. 1A), indicating the expulsion of an axial water molecule from the heme iron coordination sphere and the transition of the heme iron from the low-spin hexacoordinated to the high-spin pentacoordinated state. Type II changes shift the ferric heme Soret γ band to 425 to 436 nm (showing a trough at ∼416 nm and a peak at ∼436 nm in the difference spectra) (Fig. 1B), indicating the replacement of a water molecule, a weak axial ligand, with a stronger one, usually one with a nitrogen-containing aliphatic or aromatic group. The spin and coordination states of the heme iron remain unchanged in type II binding.

Identification of CYP51 inhibitors by HTS.

A commercial library of 20,000 synthetic organic compounds comprising a large variety of small molecule scaffolds was screened against wild-type CYP51_Mt. Three type I candidates with binding affinities ranging from 25 to 50 μM were identified, one being the estrogen metabolite 11-ketoestrone (Fig. 2A, 3529-0059). In this group of reagents, DHBP (Fig. 2A, 0627-0373), an industrial bulk product used for UV protection, had the highest binding affinity (K_D of 28 μM, as confirmed by a manual spectroscopic assay). Type II candidates were more abundant and had overall higher binding affinities for CYP51_Mt than type I candidates. Six type II hits having binding affinities in the 5 to 10 μM range (Fig. 2 B), four hits with affinities in the 10 to 20 μM range, four hits with affinities in the 20 to 40 μM range, and eight hits with affinities in the 40 to 50 μM range were identified. Among the top type II hits (Fig. 2B), two compounds were distinguished by notable structural similarities, including an N-(4-pyridyl)-formamide moiety (Fig. 2) and a phenyl ring, both attached to the same carbon atom, a structural motif not seen in clinically approved antifungal azole drugs. Based on these structural features, a number of other compounds that were not part of the primary screening library were subsequently screened. Compound BSPPA (Fig. 2C) was singled out from this pool as having the same high binding affinity for CYP51_Mt as EPBA.

FIG. 2.

Top HTS hits. (A) Type I HTS hits (K_D, ∼25 to 50 μM). (B) Type II HTS hits (K_D, ∼5 to 10 μM). Dissociation constants confirmed for selected compounds in manual binding assays are provided below their chemical structures. The N-(4-pyridyl)-formamide moiety is highlighted in gray. Chiral carbon centers are marked by an asterisk. (C) BSPPA, selected based on structural similarity to EPBA. Compounds are identified by the numbers in the ChemDiv, Inc., product library catalog. (D) Fluconazole, a clinical antifungal drug.

Inhibition of M. tuberculosis by EPBA and DHBP.

The inhibitory effects of two HTS hits, one for each binding mode (DHBP for type I and EPBA for type II), on M. tuberculosis in liquid medium were examined (Fig. 3). The inhibitory effect of DHBP was evident at 100 μM, while 200 μM DHBP prevented M. tuberculosis growth completely. At the same time, the inhibitory effect of EPBA was evident at 250 μM, while 500 μM EPBA completely prevented M. tuberculosis growth. DMSO alone did not inhibit M. tuberculosis growth in liquid medium at the range of concentrations used.

FIG. 3.

Inhibition of M. tuberculosis growth by DHBP and EPBA. The inhibition of M. tuberculosis (Mtb) growth in broth culture was assessed by the Alamar blue assay utilizing a resazurin color change from blue to pink upon reduction to resorufin in wells containing live mycobacteria. The experiment was performed in duplicate.

EPBA and DHBP binding specificities.

EPBA and DHBP binding specificities for the CYP51_Mt mutant forms and the M. tuberculosis CYP enzymes from different protein families, CYP125 and CYP130, were examined. EPBA bound both CYP51_Mt, with a K_D of 5 μM, and the F78L active-site CYP51_Mt mutant form, with a K_D of 20 μM (Fig. 4A). By comparison, the K_Ds for the structurally different compound fluconazole (Fig. 2D), a clinically used antifungal drug, were 30 and 70 μM, respectively. Phenylalanine at position 78 was selected for mutagenesis because it is highly specific for protozoa and plant species metabolizing 4α-methylated sterols (Fig. 5), except in Trypanosoma cruzi CYP51, which has isoleucine at this position and strongly favors 4α,β-dimethylated sterol substrates (17). In humans and other animals metabolizing 4α,β-dimethylated sterols, position 78 is invariably occupied by leucine. EBPA showed 50- and 400-fold drops, respectively, in binding affinity for the CYP125 and CYP130 reference proteins compared to that for CYP51_Mt (Fig. 4C). DHPB binding was strictly specific to CYP51_Mt (K_D of 28 μM), and no binding to the F78L mutant form or to CYP125 or CYP130 was observed.

FIG. 4.

EPBA and BSPPA binding specificity. Linearization in the form of the S₀/ΔA-versus-S₀ plot of the UV-visible spectral titration data obtained by adding EPBA (A) or BSPPA (B) to CYP51_Mt or its F78L mutant form in 5 μM increments or by adding EPBA (C) to CYP125 or CYP130 in 100 μM increments.

FIG. 5.

CYP51 active-site residues across the different phyla. Fragments of the multiple sequence alignment of 69 sequences of CYP51 isoforms from different organisms show the residues of the substrate binding site in CYP51 as deduced from the estriol-bound form (26). Accession numbers for CYP51 in the Swiss-Prot/TrEMBL (http://us.expasy.org/sprot) or NCBI (http://www.ncbi.nlm.nih.gov) database are given next to the names of the host organisms. Alignment was performed using the maximum a posteriori algorithm as implemented in the BCM Search Launcher (http://searchlauncher.bcm.tmc.edu/multi-align/multi-align.html) (31). The figure was generated with ESPript (12). M. bovis, Mycobacterium bovis; M. avium, Mycobacterium avium; M. ulcerans, Mycobacterium ulcerans; M. smegmatis, Mycobacterium smegmatis; N. farcinica, Nocardia farcinica; M. capsulatus, Methylococcus capsulatus; T. brucei, Trypanosoma brucei; T.b. gambiense, Trypanosoma brucei gambiense; L. amazonensis, Leishmania amazonensis; L. major, Leishmania major; A. fumigatus, Aspergillus fumigatus; V. inaequalis, Venturia inaequalis, V. nashicola, Venturia nashicola; L. maculans, Leptosphaeria maculans; T. acuformis, Tapesia acuformis; O. yallundae, Oculimacula yallundae; B. graminis, Blumeria graminis; B. fuckeliana, Botryotinia fuckeliana; M. fructicola, Monilinia fructicola; U. necator, Uncinula necator; M. grisea, Magnaporthe grisea; N. crassa, Neurospora crassa; M. graminicola, Meloidogyne graminicola; P. digitatum, Penicillium digitatum; P. italicum, Penicillium italicum; E. nidulans, Emericella nidulans; S. cerevisiae, Saccharomyces cerevisiae; C. glabrata, Candida glabrata; A. gossypii, Ashbya gossypii; C. albicans, Candida albicans; C. tropicalis, Candida tropicalis; P. carinii, Pneumocystis carinii; S. pombe, Schizosaccharomyces pombe; C. neoformans, Cryptococcus neoformans; F. neoformans, Filobasidiella neoformans; U. maydis, Ustilago maydis; C. elegans, Cunninghamella elegans; and C. krusei, Candida krusei.

Crystallization of the CYP51_Mt-EPBA complex.

In addition to having a micromolar affinity, EPBA was found to profoundly affect the crystallization properties of CYP51_Mt. First, crystals of the CYP51_Mt-EPBA complex appeared within an hour under virtually all conditions of the narrow crystallization screening grid containing combinations of polyethylene glycol 4000, isopropanol, and HEPES, pH 7.5, which were previously found to promote the growth of CYP51_Mt crystals over weeks of incubation (25, 26). Second, broader screening of the CYP51_Mt-EPBA complex under the 96 sets of crystallization conditions of the index kit (Hampton Research) unexpectedly resulted in crystal growth in >20% of the various index formulations (22 of 96). In contrast, we were unable to obtain crystals of the CYP51_Mt-DHBP complex.

The crystal structure of the CYP51_Mt-EPBA complex, determined to a resolution of 1.53 Å (Table 1), revealed that EPBA binds in the active site via (i) coordination with the heme iron through a lone pair of nitrogen electrons; (ii) two water molecule-mediated hydrogen bonds with residues H259 and Y76, which are invariable in the CYP51 family; and (iii) hydrophobic contacts between the phenyl moiety and amino acid side chains in the phylum- and/or substrate-specific cavity of CYP51 (Fig. 6). Meanwhile, the ethyl moiety protrudes into the active-site opening created by a bend of the I helix and an open conformation of the BC loop (25) (Fig. 7A).

FIG. 6.

Stereo view of EPBA bound in the active site of CYP51_Mt. The electron densities of EPBA (green) and the active-site residues located within 4 Å of the ligand are represented by a fragment of a 2Fo-Fc map (cyan), where Fo is the observed structure factor and Fc is the calculated structure factor, contoured at 2.0 σ. Two water molecules mediating ligand-hydrogen bonding contacts between the amide nitrogen of EPBA and H259 and the carbonyl oxygen of EPBA and Y76 are shown as red spheres. Images were generated using the program SETOR (9).

FIG. 7.

Binding of EBPA and BSPPA in the CYP51_Mt active site. EPBA (A) and BSPPA (B) bound in the active site of CYP51_Mt, which is represented by a space-filled model. Both compounds (cyan) and the heme edge (green) are clearly seen through the active-site opening created by the bent I helix, an open conformation of the BC loop, and missing electron density for the C helix. The ethyl (A) and benzothiadiazole sulfonamide (B) groups protrude into the bulk solvent through an open space of the active-site entrance. In panel A, the heme propionate side chain is shown in two alternative conformations. In panel B, the benzothiadiazole ring of BSPPA flips over to make stacking contacts with the pyridine ring of the same molecule and the Y76 side chain. Some contacts are also made with the heme propionate chain apparently stabilizing it in a single conformation. Images were generated using VMD software (13).

Active-site structure of the CYP51-BSPPA complex.

The crystal structure of the CYP51_Mt-BSPPA complex, determined to a resolution of 1.5 Å (Table 1), confirmed that both BSPPA and EPBA bind in the active site through the same set of interactions, with the qualification that the sulfonamide group of BSPPA is within interaction distances from the M79, F83, and F255 residues while the benzothiadiazole ring flips over to make stacking contacts with both the pyridine ring of BSPPA and Y76 and interaction contacts with the heme propionate chain (Fig. 7B). No other interactions are involved in benzothiadiazole ring binding since the CYP51_Mt active site is open to the bulk solvent. It should be noted, however, that this scheme may not be valid for other CYP51 enzymes whose three-dimensional structures are yet to be determined.

Despite the fact that racemic mixtures were used for cocrystallization with both EPBA and BSPPA, only specific stereoisomers were found bound in the active site of CYP51_Mt, namely, the R-stereoisomer for EPBA and the S-stereoisomer for BSPPA. EPBA was subsequently tested in the binding assay as the racemate and as the S-enantiomer, both obtained by chemical synthesis. No significant differences in the binding affinities for CYP51_Mt were observed.

DISCUSSION

New molecular scaffolds for CYP51 inhibitors were identified in an HTS assay designed to detect the spectral shift of the P450 ferric heme Soret band in response to ligand binding (Fig. 1). Two HTS hits, DHBP (type I) and EPBA (type II) (Fig. 2), one for each binding mode, were examined for the inhibition of M. tuberculosis growth in liquid culture. Although neither DHBP nor EPBA is a potent M. tuberculosis inhibitor, our data provide evidence that M. tuberculosis growth may be affected through the inhibition of CYP51_Mt (Fig. 3). The gene encoding CYP51_Mt has previously been determined to be nonessential both by Himar1-based transposon mutagenesis (27, 28) and by the genome-wide transcriptome analysis of M. tuberculosis gene expression during macrophage infection (29). Important future work is to further address the mechanism(s) employed by these compounds to achieve the inhibitory effect.

Hydroxybenzophenones have been reported earlier to be antibacterial agents (8). Benzophenone-related compounds have also been found to inhibit squalene cyclase, another enzyme in cholesterol biosynthesis (16). Various benzophenones, including DHBP, have also been identified as the ligands of steroid receptors, thus suggesting the potential of these compounds to mimic steroids (32). Photo-cross-linking is a typical mode of action for benzophenones that may contribute to their activity in optical assays (1, 33). On the other hand, the antimicrobial or sterol biosynthesis inhibitory effects of N-pyridinyl-benzeneacetamides have not, to the best of our knowledge, been described previously. DHBP and EPBA both efficiently discriminate between CYP51_Mt and two other M. tuberculosis CYP enzymes, CYP130 and CYP125. The binding affinity of EPBA for CYP130 and CYP125 is reduced 50- and 400-fold, respectively (Fig. 4), compared to that for CYP51_Mt, while DHPB binding is strictly specific to CYP51_Mt and no binding to either CYP125 or CYP130 or the F78L CYP51_Mt mutant form was observed.

The crystal structures of CYP51_Mt in complex with EPBA and the structurally related compound BSPPA were determined to 1.53- and 1.5-Å resolutions, respectively, revealing the binding mode of these compounds in the active site. A rigid planar structure of the N-(4-pyridyl)-formamide moiety originating from the chiral carbon center enables both EPBA and BSPPA to bind heme as an axial ligand via coordination with the heme iron through a lone pair of nitrogen electrons and to span the protein active site by utilizing water molecule-mediated hydrogen bonding between the amide nitrogen of EPBA and H259 and the carbonyl oxygen of EPBA and Y76, bringing a second aromatic function, the phenyl ring, into interaction with the active-site residues Y76, F78, F255, H259, L321, M433, and V434 (Fig. 6). A third substituent, the ethyl group in EPBA or the benzothiadiazole sulfonamide group in BSPPA, largely protrudes into the bulk solvent through an open space of the active-site entrance (Fig. 7). In principle, a third substituent may serve to provide specific protein-ligand interactions with other CYP51 forms, if their active sites are protected from the bulk solvent by the BC loop. The observation that the EPBA S-enantiomer displays the same binding affinity as the racemate, whereas exclusively the R-enantiomer is found in the crystal, may suggest a rapid isomerization of the asymmetric carbon atom under assay conditions via the formation of a stabilized benzyl cation as an intermediate of racemization.

While Y76 and H259 are invariable in the CYP51 family, positions L321, M433, and V434 are specific for certain groups of CYP51 enzymes within a phylum, and positions F78 and F255 are phylum specific (Fig. 5). Thus, F78 serves as a key discriminator between 4α-methylated and 4α,β-dimethylated sterol substrates due to its localization within van der Waals distance of the C-4 atom of a sterol (26), which sterically hinders the binding of sterols that have the methyl group in the β-configuration. This role for F78 has subsequently been demonstrated by catalytic activity assays with trypanosome forms of CYP51 (in which F78 corresponds to position 105 according to T. cruzi numbering) (17). A single mammalian-like F78L substitution in CYP51_Mt results in fourfold drops in EPBA and BSPPA binding affinities (Fig. 4A and B) and completely eliminates the DHBP binding.

The amino acid residue at position 255 is unlikely to play a direct role in the discrimination of sterol substrates based on the methylation status of C-4. Nevertheless, F255 is strictly specific to plant and protozoan CYP51 isoforms (Fig. 5). Therefore, the phenyl ring of EPBA or BSPPA binds in a strictly phylum- and/or substrate-specific cavity of CYP51 and potentially could be modified to maximize binding and tune an inhibitor toward a specific form of CYP51. Thus, inhibitors utilizing stacking interactions with aromatic residues in the 78 and 255 positions may be optimized to efficiently discriminate between a protozoan pathogen and a mammalian host, since in vertebrates both of these positions are always occupied by a leucine (Fig. 5). It is apparent that functional groups other than phenyl rings attached to the chiral carbon center may also be explored, while the N-(4-pyridyl)-formamide bulk may serve as a universal building block in the design of species-specific CYP51 inhibitors.

In CYP51_Mt-ligand structures determined both in the present work and previously, (25, 26), the BC loop is either in an open conformation or in an undefined electron density state, which creates the quite unusual condition for P450 of open access to the active site and the exposure of heme to the bulk solvent. Moreover, in the presence of EPBA or BSPPA having a micromolar affinity and notably promoting the crystallization of CYP51_Mt, heme is exposed to the bulk solvent even more than in previously determined structures, so that one of the heme propionate side chains adopts an alternative conformation in the CYP51_Mt-EPBA complex (Fig. 7A). Since all these crystals were obtained in the same P2₁2₁2₁ space group with virtually identical unit cell dimensions and with only one molecule in an asymmetric unit, some possibility remained that the open conformation might be a result of crystal-packing interactions. Here we present the only structure of CYP51_Mt in a P2₁ space group (Protein Data Bank ID 2BZ9), determined to a resolution of 2.2 Å, having two molecules in an asymmetric unit, which eliminates this possibility. Both molecules in the structure have the BC loop open despite its exposure to the different crystal-packing contacts.

In summary, an HTS assay utilizing general spectral properties of P450 was developed and applied to the CYP51 isoform from M. tuberculosis. A number of compounds in type I and type II binding modes were identified which bind this protein with micromolar affinity. The inhibitory effect on M. tuberculosis growth in broth culture was demonstrated for two HTS hits, one for each binding mode. The binding modes of one of the top HTS hits, EPBA, and of the rationally selected, structurally related BSPPA were validated by the determination of X-ray crystal structures. These structures revealed that only one particular enantiomer of each compound binds in the active site of CYP51_Mt. The N-(4-pyridyl)-formamide moiety of both compounds interacts with the residues H259 and Y76, which are invariable in the CYP51 family, and with the heme prosthetic group and thus may serve as a universal block for building CYP51 inhibitors. As a result, EPBA efficiently discriminates between CYP51_Mt and two other M. tuberculosis CYP proteins, CYP130 and CYP125. In addition, the phenyl moiety attached to the same carbon center makes contacts in a strictly phylum- and/or substrate-specific CYP51 cavity and thus can be modified to maximize binding to species-specific CYP51 enzyme isoforms. These data allow us to predict that highly selective CYP51 inhibitors could be designed based on the EPBA template. These inhibitors may be further structurally optimized for specific forms of CYP51 to efficiently discriminate between host and pathogen enzymes, thereby reducing potential side effects of therapeutic drugs.