Complexes of Trypanosoma cruzi Sterol 14α-Demethylase (CYP51) with Two Pyridine-based Drug Candidates for Chagas Disease

Background: Two pyridine derivatives were identified as promising drug candidates in animal models of Chagas disease. Results: They were tested as sterol 14α-demethylase (CYP51) inhibitors, and x-ray co-structures with T. cruzi CYP51 were determined. Conclusion: The structures explain the potency and selectivity of the compound. Significance: Structural information contributes to a better understanding of P450 inhibition and will facilitate rational design of pathogen-specific drugs. Chagas disease, caused by the eukaryotic (protozoan) parasite Trypanosoma cruzi, is an alarming emerging global health problem with no clinical drugs available to treat the chronic stage. Azole inhibitors of sterol 14α-demethylase (CYP51) were proven effective against Chagas, and antifungal drugs posaconazole and ravuconazole have entered clinical trials in Spain, Bolivia, and Argentina. Here we present the x-ray structures of T. cruzi CYP51 in complexes with two alternative drug candidates, pyridine derivatives (S)-(4-chlorophenyl)-1-(4-(4-(trifluoromethyl)phenyl)-piperazin-1-yl)-2-(pyridin-3-yl)ethanone (UDO; Protein Data Bank code 3ZG2) and N-[4-(trifluoromethyl)phenyl]-N-[1-[5-(trifluoromethyl)-2-pyridyl]-4-piperi-dyl]pyridin-3-amine (UDD; Protein Data Bank code 3ZG3). These compounds have been developed by the Drugs for Neglected Diseases initiative (DNDi) and are highly promising antichagasic agents in both cellular and in vivo experiments. The binding parameters and inhibitory effects on sterol 14α-demethylase activity in reconstituted enzyme reactions confirmed UDO and UDD as potent and selective T. cruzi CYP51 inhibitors. Comparative analysis of the pyridine- and azole-bound CYP51 structures uncovered the features that make UDO and UDD T. cruzi CYP51-specific. The structures suggest that although a precise fit between the shape of the inhibitor molecules and T. cruzi CYP51 active site topology underlies their high inhibitory potency, a longer coordination bond between the catalytic heme iron and the pyridine nitrogen implies a weaker influence of pyridines on the iron reduction potential, which may be the basis for the observed selectivity of these compounds toward the target enzyme versus other cytochrome P450s, including human drug-metabolizing P450s. These findings may pave the way for the development of novel CYP51-targeted drugs with optimized metabolic properties that are very much needed for the treatment of human infections caused by eukaryotic microbial pathogens.

in fungi, are required for the integrity of the eukaryotic cell membrane and as metabolic precursors for a variety of signaling molecules. For more than 40 years CYP51 has served successfully as the major target for antifungal azoles (6 -8), and by now it is quite generally recognized as a promising therapeutic target in Trypanosoma cruzi as well (9).
Endemic in South America, Chagas disease is a life-threatening malignancy with only two drugs available for treatment (benznidazole and nifurtimox), both having low efficiency in the chronic stage, during which the heart and gastrointestinal tract are affected predominantly. The disease is now becoming a severe global health problem, mainly because of human migration to other continents but also because of the broadening of the area of vector habitat; the triatomine insects that transmit the parasite ("kissing bugs") now populate two-thirds of the land in the United States (10) and have been found in Asia, Australia, and Africa. However, historically known as a disease of the poor, the infection does not attract the attention of pharmaceutical companies as being profitable, and for more than a century since its discovery the parasite has remained incurable.
Similar to fungi, in T. cruzi the endogenously synthesized sterol molecules are C24-alkylated (ergosterol-like) and therefore cannot be replaced in the parasite cells by host cholesterol (23). The shortage of endogenous sterols is especially harmful for the multiplying form of T. cruzi, intracellular amastigotes (24), the morphological form of the parasite that is prevalent at the chronic stage of infection. Accordingly, in amastigotes the expression level of the CYP51 gene is higher (25), and the effect of CYP51 inhibitors is about 3 orders of magnitude stronger than in the non-multiplying bloodstream trypomastigotes (22,25,26), the forms that are more relevant to the acute stage of infection and can be treated effectively with benznidazole. Depletion of endogenous sterols in T. cruzi due to CYP51 inhibition causes blebbing of the cellular membrane, mitochondrial swelling, alterations in the endoplasmic reticulum and Golgi, nuclear envelope detachment, and deterioration of the internal membranes, ultimately resulting in cell lysis and death (27,28).
T. cruzi CYP51, which has Ͻ25% identity to fungal orthologs, has been well characterized biochemically and structurally (25)(26)(27)29). First, the gene was cloned from the Tulahuen strain of the pathogen, the recombinant protein expressed in Escherichia coli and purified (29). We found that the T. cruzi CYP51 substrate preference toward the C4-double methylated eburicol (24-methylene-dihydrolanosterol) is largely defined by a single amino acid residue in the BЈ helix, Ile-105 (animal/fungilike), versus the corresponding Phe in the plant-like CYP51 orthologs from Trypanosoma brucei (30) and Leishmania infantum (31) that catalyze 14␣-demethylation of C4-mono-methylated obtusifoliol and norlanosterol. Studies on CYP51 inhibition revealed that susceptibilities of the protozoan enzymes to antifungal drugs often differ significantly from those of fungal CYP51 orthologs and led to the identification of several novel experimental inhibitory scaffolds (26,27,32,33). Determination of the x-ray crystal structures of T. cruzi (25,34), T. brucei (20,35), and L. infantum (31) CYP51s, ligand-free and complexed with several azole derivatives including 1,2,4triazole posaconazole (25) and 1,3-imidazole VNI (35) and with the mechanism-based suicide substrate 14␣-methylenecyclopropyl-dihydrolanosterol (32), established the molecular basis for CYP51 catalytic conservation (36). To be able to perform its ancient biological function across phylogeny at very low sequence identity, this P450 enzyme has a highly rigid substrate binding cavity, which in turn serves as the foundation of its druggability (32,37).
In the present work we tested T. cruzi CYP51 as a potential target for two pyridine-based compounds, UDO (EPL-BS1246) and UDD (EPL-BS0967), which have been identified as strong antiparasitic agents in cellular T. cruzi experiments and display high potency in vivo in a stringent mouse model of Chagas disease. We report UDO and UDD binding parameters and their inhibitory effects on sterol 14␣-demethylation activity and co-structures with T. cruzi CYP51, comparing them with the CYP51 complexes with posaconazole and VNI (Fig. 1). The structures provide valuable information regarding the features that contribute to the potency of these two new CYP51 inhibitors and their selectivity toward the T. cruzi enzyme ortholog. Particularly, they show that the peculiarities observed in the enzyme spectral responses to UDO/UDD binding reflect a longer coordination bond between the ferric heme iron and heterocyclic pyridine (versus azole) nitrogen, which is likely to underlie higher specificity because a weakened iron-nitrogen interaction would enhance the role of other, nonbonded, protein-ligand contacts. This is the first report of highly potent T. cruzi CYP51 inhibitors of a non-azole nature, and we believe that the information gained in this study will facilitate structure-based design of pathogen-specific drugs with optimized metabolic properties, which will have little or no effect on human drug-metabolizing P450s.

Testing UDO (EPL-BS1246) and UDD (EPL-BS0967) as Antiprotozoan Drug Candidates
The Compounds-UDO and UDD were synthesized as part of a Drugs for Neglected Diseases initiative (DNDi) drug discovery program to develop new treatments for Chagas disease and are the result of structure-activity relationship optimization of plant fungicide fenarimol (pyrimidine derivative), which was found to have moderate in vitro activity against T. cruzi (38). The synthesis was carried out by analogy to the methods reported elsewhere (38,39). Both compounds were prepared by a short synthetic sequence from readily available precursors. Racemic UDO was synthesized in five steps (three of them contiguous) from the key building blocks 3-fluoropyridine, 4-chlorophenylacetonitrile, and 1-[4-(trifluoromethyl)phenyl]piperazine, with a high overall yield (45%). Chiral resolution was achieved using supercritical fluid chromatography after which the S-enantiomer was identified as most active in in vitro assays. UDD was constructed in four steps, with a 12% overall yield, from 3-amino pyridine, piperidin-4-one, and 2-chloro-5-(trifluoromethyl)pyridine. As a trigonal nitrogen-centered ligand, UDD is achiral, and no further manipulation of the final product was required.
Cellular T. cruzi Assay-The assay was performed using the ␤-galactosidase-expressing Tulahuen T. cruzi as described previously (40). Briefly, L6 cells were plated into 96-well flat-bottom tissue culture plates and incubated at 37°C in 5% CO 2 for 24 h to allow cells to adhere. T. cruzi trypomastigotes were then added at a multiplicity of infection of 3, and plates were incubated for a further 48 h to allow infection to establish. Extracellular trypomastigotes were then removed, and compounds were added in seven-point serial dilutions performed in triplicate. After 96 h of incubation with the compounds, chlorophenol red-fl-D-galactopyranoside (CPRG, Roche Applied Science) was added. A color change following catabolization of the reagent by viable T. cruzi was observed, and absorbance was read at 530 nm using a Dynex microplate reader.
Cellular Leishmania donovani Assay-The assay was performed using the L. donovani MHOM/ET/67/L82 strain. Primary peritoneal mouse macrophages were infected at a multiplicity of infection of 1 to 10 with Leishmania parasites. Compounds were added after 2 h of infection. Parasite multiplication was compared with untreated-infected controls (100% growth) and uninfected controls (0% growth). After 5 days of incubation, parasite burdens (mean number of amastigotes/macrophage) were microscopically assessed after staining the cells with a 10% Giemsa solution. The results were expressed as percent reduction in parasite burden compared with untreated control wells.
Anti-T. cruzi Activity in Vivo-Female Swiss mice ϳ8 weeks old were obtained from the Animal Resources Centre (Perth, Western Australia). All animal experiments were carried out with approval of the Animal Ethics Committee of Murdoch University as described (38). Briefly, compounds were suspended in a hydroxypropyl methylcellulose suspension vehicle (0.5% w/v hydroxypropyl methylcellulose, 0.4% v/v Tween 80, and 0.5% v/v benzyl alcohol in deionized water) and given by oral gavage. Induction of acute T. cruzi infection was achieved by infecting the mice with 50,000 Tulahuen trypomastigotes via intraperitoneal injection (five mice/group). Compound administration was commenced on day 8 post-infection and continued daily for 20 days. Parasitemia was determined by collecting blood by tail prick and counting live trypomastigotes under the microscope using a Neubauer hemocytometer. Efficacy was monitored by comparing parasitemia levels in compoundtreated versus vehicle only-treated groups. If the parasitemia levels were no longer detectable in the blood following compound treatment, animals were immunosuppressed with cyclophosphamide 10 days after the cessation of treatment. Cyclophosphamide was administered intraperitoneally at 50 mg/kg/ day in three cycles; each cycle comprised 4 days of treatment followed by 3 days of rest. Parasitemia was determined by collecting blood as detailed above, and parasitological cure was confirmed by PCR on tissue samples (38). A more extended description of in vivo data will be published elsewhere.
Inhibition of Human Drug-metabolizing CYPs-Inhibition of hepatic cytochromes P450 was assessed in human liver microsomes using a substrate-specific approach of monitoring metabolites formed by each specific CYP enzyme. Metabolic reactions included phenacetin-O-deethylation (1A2), tolbutamide methylhydroxylation (2C9), S-mephenytoin 4Ј-hydroxylation (2C19), dextrome-thorphan-O-deethylation (2D6), and testosterone-6-␤-hydroxylation (3A4). Each substrate was added at a concentration less than or equal to the K m for the metabolic pathway, and microsomal protein concentrations and assay incubation times were optimized for each reaction to ensure linear metabolite formation based on initial rates (Table  1). Microsomes were suspended in phosphate buffer and incubated at 37°C in the presence of probe substrates. Reactions were initiated by the addition of an NADPH-regenerating buffer system and were quenched at the appropriate times using acetonitrile. Samples were centrifuged and concentrations of metabolites assessed by LC/MS. The IC 50 was taken as the concentration at which there was a 50% reduction in the rate of metabolite formation relative to the metabolite formation rate in the absence of inhibitor. Additionally, the effects of UDO and UDD on CYP3A4 were compared with the effects of posaconazole and VNI (Fig. 1) using a commercial kit (BD Biosciences screening kit) as described (22); the drug concentration range was 16,000 to 2.4 nM. The IC 50 for the manufacturer's positive control inhibitor ketoconazole at these conditions was ϳ8 nM).

Testing UDO (EPL-BS1246) and UDD (EPL-BS0967) as CYP51 Inhibitors
Protein Engineering, Expression, and Purification-For functional studies, including ligand binding and enzymatic activity assays, the full-length CYP51 expression constructs, designed as described previously (29, 31) (GenBank TM accession numbers AY856083 and EF192938 for T. cruzi and L. infantum CYP51, respectively), were used. For crystallization, we utilized the T. cruzi CYP51 N-terminally truncated construct, where the membrane anchor sequence (first 31 residues upstream of Pro-32) was replaced with the 5-amino acid sequence fragment MAKKT (described in Ref. 25) for co-crystallization with posaconazole. In all cases the CYP51 genes were His tag-engineered at the C terminus to facilitate purification, subcloned into the pCW expression vector at NdeI (5Ј) and HindIII (3Ј) sites, and expressed in E. coli strain HMS174(DE3) (Novagen). For T. cruzi CYP51, the purification procedure included sequential chromatography on nickel nitrotriloacetate-agarose (Qiagen) and carboxymethyl-Sepharose (Sigma) (29). L. infantum CYP51, human CYP51, and cytochrome P450 reductase were purified as reported previously (30,31,41). For crystallization purposes complexes with the inhibitors were obtained by saturating T. cruzi CYP51 with the compound during the last purification step by adding a 10 mM dimethyl sulfoxide stock solution of the inhibitor to the washing and elution buffers, with a final concentration of 20 M. Spectral Ligand Binding Assay-All spectra were recorded at 22°C on a dual beam Shimadzu UV-240IPC spectrophotometer. The P450 concentration was determined from the Soret band intensity of the reduced CO complexes using visible absorption difference spectroscopy (⌬⑀ 450 -490 ϭ 91 mM Ϫ1 cm Ϫ1 ) (42). Spectral titrations were carried out at a P450 concentration of 0.4 M in a 50 mM potassium phosphate buffer, pH 7.4, containing 200 mM NaCl and 0.1 mM EDTA. The optical path length was 5 cm. Ligand binding was monitored as a red shift in the Soret band maximum, reflecting coordination of the heterocyclic nitrogen to the P450 heme iron (26). Difference spectra were generated by recording the P450 absorbance in a sample cuvette versus the absorbance in a reference cuvette, both containing the same amount of the protein. Compounds were added to the sample cuvette in the concentration range of 0.1 to 2.0 M from 0.2 mM stock solutions in dimethyl sulfoxide. The titration step was 0.1 M. At each step, the corresponding volume of dimethyl sulfoxide was added to the reference cuvette. The apparent dissociation constants of the enzyme-inhibitor complexes (K d ) and the maximum spectral change (⌬A max ) were calculated using a quadratic function for tight binding ligands by fitting the data for the ligand-induced spectral change (peak to trough absorbance changes in the difference spectra (⌬A)) versus total ligand concentration to the following equation (43) in Prism (GraphPad Software), where [L] and [E] are the concentrations of the ligand and the enzyme used for the titration, respectively.
CYP51 Inhibition Assay-The enzymatic activity of T. cruzi, L. infantum, and human CYP51s was reconstituted in vitro as described previously (29,31,41) using eburicol, obtusifoliol, and lanosterol, respectively, as the substrates. Briefly, the reaction mixture contained 1 M CYP51, 2 M cytochrome P450 reductase, 100 M dilauroyl-␣-phosphatidylcholine, 0.4 mg/ml isoctrate dehydrogenase, and 25 mM sodium isocitrate in 20 mM MOPS (pH 7.4), 50 mM KCl, 5 mM MgCl 2 , and 10% glycerol. After the addition of the 3 H-radiolabeled sterol substrate (ϳ2000 cpm/nmol, final concentration 50 M) the mixture was preincubated for 5 min at 37°C in a shaking water bath (GCA Precision Scientific). The reaction was initiated by the addition of 100 M NADPH and stopped by extraction of the sterols with ethyl acetate. The reaction products were dried, dissolved in methanol, and analyzed by a reverse-phase HPLC system (Waters) equipped with a ␤-RAM detector (INUS Systems, Inc.) using a Nova-Pak C18 column (particle size 4 M, 3.9 ϫ 150 mm) and a linear gradient consisting of water:acetonitrile: methanol (1.0:4.5:4.5) (solvent A) and methanol (solvent B) at a flow rate of 1 ml/min. Under these conditions the retention  50 , drug concentration that gives half-maximal response in cellular growth reduction; ED 50 , drug regimen that is required to reach parasitological cure in 50% animals; IC 50 , drug concentration required to inhibit the activity of the enzyme by 50%. *, data from Villalta et al. (22); **, data from Keenan et al. (38). time for eburicol, obtusifoliol, and lanosterol was 26, 23, and 25 min, respectively. The retention time for the reaction intermediates, 14␣-alcohol and 14␣-aldehyde derivatives of obtusifoliol (30) (as seen in Fig. 3, L. infantum CYP51), was 10 and 12 min, respectively. The potencies of UDO and UDD to inhibit T. cruzi, L. infantum, and human CYP51 were compared as the inhibition of enzymatic activity in a 1-h reaction (25,26) with posaconazole serving as positive control.
Crystallization, Data Collection, and Structure Determination-The co-crystals were obtained by hanging drop vapor diffusion. The 250 M protein solution in 20 mM potassium phosphate buffer, pH 7.2, containing 200 mM NaCl, 0.1 mM EDTA, 1 mM dithiothreitol, 10% glycerol, and 8 mM Cymal-4 saturated with the inhibitor as described above, was mixed with an equal volume of a precipitant solution containing 50 mM potassium phosphate buffer, pH 7.6, 0.2 M magnesium chloride hexahydrate, and 25% (w/v) PEG 4000 and was equilibrated against the precipitant solution at 22°C. Crystals were prescreened using Bruker Microstar microfocus rotating anode x-ray generator/ Proteum Pt135 CCD area detectors, and those that diffracted to ϳ3.0 Å resolution were subsequently used for data collection at a synchrotron (Advanced Photon Source, Argonne National Laboratory). Data were collected at LS-CAT, beamline 21ID-F. The diffraction images were integrated using Mosflm (44) and scaled with Aimless (CCP4 Program Suite 6.3.0 (45)) in the P3(1)21 space group to resolutions of 2.8 Å (UDO) and 2.9 Å (UDD). The solvent content was estimated with the Matthews probability calculator. In both cases a single solution with one molecule in the asymmetric unit was found with PhaserMR using the posaconazole-bound T. cruzi CYP51 structure (Protein Data Bank ID code 3K1O) as the search model. Iterative models of the protein-inhibitor complexes were built and refined using COOT (46) and Refmac (47) in CCP4 (45). The statistics for data collection, refinement, and validation are shown in Table 2. The protein chain is seen from residues 28 ("A" in the N-terminal MAKKT sequence) to 478 (of 482) for T. cruzi CYP51-UDO and from residues 29 to 478 for T. cruzi CYP51-UDD. The electron density for both UDO and UDD was well defined (Fig. 2), showing a single orientation of the inhibitor molecule within the enzyme binding cavity and full occupancy. The coordinates and structure factors have been deposited in the Protein Data Bank under ID codes 3ZG2 (CYP51-UDO) and 3ZG3 (CYP51-UDD).

RESULTS
As shown in Fig. 1, UDO and UDD are highly efficient against T. cruzi, in both cellular and in vivo experiments. The compounds were identified upon testing of a number of pharmacologically optimized derivatives of pyrimidine-based fungicide fenarimol, and each contains a heterocyclic (pyridine) ring with a basic nitrogen atom, thus suggesting the possibility of its coordination to the P450 heme iron. On the other hand, their inhibitory effects on all five major human drug-metabolizing P450s (48) are very weak (Table 1), and the whole organism primary in vitro screen that was used to drive the discovery process is mechanistically nonspecific. Therefore, it was of special interest to verify whether they would inhibit the parasite CYP51.

CYP51 Spectral Responses to Ligand Binding
Recombinant CYP51s are usually purified in the hexacoordinated low spin form, with the Soret (␥) band absorbance maximum at ϳ417 nm (3, 29 -31, 49) and the oxygen atom from a water molecule serving as the sixth (distal axial) ligand for the ferric heme iron (50). If a stronger ligand replaces the water molecule in the iron coordination sphere, the Soret maximum shifts to the longer wavelength (red shift or type II spectral response). Consequently, in the difference spectra a trough and a peak appear on the left and the right side of an isosbestic point, respectively. Accordingly, when the iron in T. cruzi CYP51 becomes coordinated to an azole ring nitrogen, the trough/ isosbestic point/peak in the difference spectra emerge at 407/ 419/428 nm in the case of imidazole N3 (VNI) (Fig. 3D) and at 392/415/425 nm in the case of triazole N4 (posaconazole) (Fig.  3C). The difference spectra elicited in T. cruzi CYP51 by UDO and UDD, however, revealed the trough at 389 nm, the isosbestic point at 411 nm, and a peak at 423 and 422 nm, respectively (Fig. 3, A and B). The Soret band absolute absorbance maximum was red-shifted only to 421 nm (versus 424 nm in VNI-bound CYP51 and 423 nm in posaconazole-bound CYP51 (Table 3), hence suggesting a weaker interaction of the sixth ligand with the heme iron. On the other hand, the apparent K d values calculated from T. cruzi CYP51 spectral responses to UDO (19 nM) and UDD (32 nM) were found comparable to those obtained for VNI and posaconazole (Table 3), with the titration curves all showing saturation at about an equimolar ratio of inhibitor/enzyme (Fig. 3A). Thus, although the pyridine ring nitrogen can be characterized as a weaker iron-coordinating atom, based on the apparent K d values, both compounds seemed to form strong complexes with T. cruzi CYP51. Interestingly, for L. infantum CYP51, in which the spectral responses to the binding of UDO and UDD are quite similar in their shape (Fig. 3, E and F), the calculated K d values were about 1 order of magnitude higher (156 and 419 nM, respectively (Table 3)), whereas the responses of the human enzyme ortholog suggested tight binding with the position of the peak in the difference spectra red-shifted to 424 nm (Fig. 3, G and H) with apparent K d values of 69 nM (UDO) and 26 nM (UDD) ( Table 3).

Inhibition of CYP51 Activity in Vitro
As discussed previously (26,37), the sensitivity of the CYP51 reaction reconstituted in vitro is limited by (a) the affinity of the CYP51-cytochrome P450 reductase complex formation (K d Ϸ 1 M, so that further lowering the P450 concentration sharply decreases its turnover rate) and (b) by very poor solubility of the sterol substrates, which in turn does not allow the use of higher substrate excess over the enzyme. Therefore, to identify the most potent CYP51 inhibitors under our standard conditions, we used a 1-h CYP51 reaction, with the molar ratio inhibitor/ enzyme/substrate being 2/1/50 (20,25,34). In these conditions, in the absence of inhibitors all of the substrate is converted into the product, and most compounds (even among those displaying complete inhibition of the initial rate of the reaction) show little inhibitory effect (9) because over time they are replaced in the enzyme active center by substrate.
T. cruzi CYP51-UDO and UDD, however, completely inhibit T. cruzi CYP51 activity under these conditions (Fig. 4, A and B). Moreover, lowering the concentration of T. cruzi CYP51 in the reaction mixture to 0.2 M (which leads to about a 40-fold decrease in the enzyme turnover) did not reveal any product formation at inhibitor/enzyme molar ratios of Ն1 in all three cases (not shown). These data indicate that although based on the position of the Soret band the maximum iron-pyridine interaction can be characterized as weaker, the potencies of these compounds to inhibit T. cruzi enzyme activity are similar to that of posaconazole and agree with the low apparent K d values.
L. infantum CYP51-The inhibitory effects of both compounds on the activity of L. infantum CYP51 are slightly weaker, as 32 and 60% of the substrate conversion is observed at a 2-fold molar excess of UDO and UDD, respectively (Fig. 4A); the molar ratios inhibitor/enzyme required to slow down the L. infantum CYP51 reaction by 2-fold are 1.5 and 2.8 (Fig. 4C). Again, lower inhibitory potencies correlate with the lower apparent binding affinities and are in agreement with the weaker antiparasitic effect of UDD against leishmanial cells (Fig. 1).
Human CYP51-The activity of the human enzyme, however, is not affected under these conditions; a 180-fold molar excess of UDO and a 35-fold molar excess of UDD were necessary to achieve a 2-fold inhibition (Fig. 4D). Thus, regardless of the low apparent K d values calculated from the titration curves, both compounds (especially UDO) inhibit human CYP51 much more weakly than their parasitic counterparts.
Interestingly, the relative potencies of UDO and UDD to inhibit human and Leishmania CYP51s are reversed; whereas UDO is roughly 2-fold stronger than UDD as the inhibitor of L. infantum CYP51, for the human enzyme UDD has a 5-fold higher potency than UDO. To better understand the molecular basis for the potency/selectivity of these compounds, we determined their x-ray structures in complex with T. cruzi CYP51.

Structural Characterization of the CYP51 Inhibitor Complexes
Inhibitor Binding Mode-As expected for type II P450 ligands, each of the inhibitors interacts with the heme, its pyridine ring nitrogen forming a coordination bond to the heme iron. The pyridine rings are in favorable conformation, perpendicular to the heme plane. Yet, in both cases the Fe 3ϩ -N coordination bond is rather long, 2.31 Å for UDO and 2.34 Å for UDD (Fig. 2). This is about 0.3 Å longer than the coordination bond in the CYP51 complexes with imidazoles (2.0 -2.04 Å in VNI (PDB code 3GW9) (35), the tipifarnib derivative JFK (PDB code 3TIK) (20), or NEE (PDB code 4h6o) (34),) and more than 0.2Å longer than the coordination bond in the CYP51 complexes with triazoles (2.07-2.15 Å) (PDB codes 3K1O, 3KHM, (25), and 3L4D (31)) (see examples in Table 4). A weaker coor- dination bond explains the smaller red shift in the Soret band absorbance maximum and apparently reflects the decrease in the elecronegativity of the iron-coordinating atom (imidazole Ͼ triazole Ͼ pyridine). Overall, the orientation of UDO and UDD within the CYP51 binding cavity is similar to VNI and almost identical to that of posaconazole, with the longer (tworing) arm of each compound directed toward the substrate access channel entrance (Fig. 5, A-C). Overall CYP51 Structure-Similar to what was observed previously for complexes with imidazole and triazole inhibitors (20,25,34,35) and the substrate analog MCP (32), the binding of UDO or UDD does not induce any significant conformational changes at the overall structure level (Fig. 5D). The r.m.s. deviations for the C␣ atoms of UDO-CYP51 and UDD-CYP51 relative to posaconazole-CYP51 (PDB code 3K1O) are only 0.535 and 0.567 Å, respectively, and between the two structures reported here the r.m.s. deviation is 0.453 Å. As appears common for CYP51s, the highest deviations are displayed by a few residues at the N and C termini, the GH loop region as well as by the structural elements surrounding the entrance into the substrate access channel, especially the FG loop area. Although the functional role of the GH loop flexibility in CYP51s remains unclear, the movements around the FЉ helix, which forms the upper portion (the lid) of the substrate channel entrance, may reflect the opening and closing of the channel that takes place in vivo upon binding of the bulky sterol molecule and release of the 14␣-demethylated reaction product back into the microsomal membrane. The other secondary structural elements, including the side chains of the residues contacting the inhibitor, remain in very similar positions in both structures (a stereo view of their superimposition can be seen in Fig. 6C).
UDO-CYP51 Complex-UDO (M r 460, 16.5 Å in length) has a molecular volume of 549 Å 3 (Accelrys), which is considerably smaller than that of posaconazole (673 Å 3 , 25.5 Å in length (25). Nevertheless, it forms van der Waals contacts with 18 amino acid residues (distance Ͻ 4.5 Å) of the enzyme (Fig. 6A). Six of these residues (Val-102, Tyr-103, Met-106, Phe-110, Ala-115, and Tyr-116) are from the BЈ helix and BЈ/C loop, the major CYP51 substrate recognition site. In the sterol-bound form of the enzyme, these secondary structural elements encircle the whole skeleton of the substrate molecule from its ␤-side (32). Leu-127 is from helix C (CYP51-specific area of the binding cavity). In the substrate-bound CYP51 it interacts with the distal portion of the sterol arm (C26 -C28). Altogether, these seven residues flank the inhibitor from the upper portion of the enzyme active site. Helix I forms the distal wall of the cavity and provides four residues, Ala-287, Phe-290, Ala-291, and Thr-295. Leu-356, Met-358, and Met-360 are from the K/␤1-4 loop, and Met-460 is from the tip of the ␤4 hairpin, which also forms the bottom of the entrance into the substrate access channel. Again, the majority of these residues are involved in the interaction with the substrate (32) and reveal a high level of conservation within phyla (36), thus providing a potential molecular basis for the phylum specificity of CYP51 inhibitors. Finally, Pro-210, Val-213, and Phe-214 (FЉ helix) are from the lid of the substrate channel entrance. The involvement of these three residues is specific for UDO binding (Fig. 6A, upper left  corner).
UDD-CYP51 Complex-UDD (M r 466, 15.6 Å in length) is even slightly smaller than UDO (molecular volume 535 Å 3 ), but it also forms van der Waals contacts with 18 amino acid residues, and again, only three of those residues, Ile-105 (BЈ helix), Leu-130 (C helix), and Met-284 (I helix), are UDD-specific (Fig.  6B). Leu-130 and Met-284 are seen in the lower right corner of Fig. 6B, and 105 is the residue that defines the T. cruzi CYP51 substrate preference toward the C4-double methylated sterol substrates (29). In the T. cruzi CYP51 complex with UDD, Ile-105 is involved in the interaction with the inhibitor because, as seen in Fig. 5A, the UDD molecule, with a structural skeleton even more rigid than that of UDO, is slightly more crescentshaped. Another feature specific to UDD T. cruzi CYP51 costructure is that the trifluoromethylbenzene ring of the inhibitor wedges between the heme plane and BЈ/C loop, pushing away Tyr-116 (Fig. 6D, right). Because of its offset position (4.02 Å) from the hydroxyl group of the porphyrin ring D propionate, the OH of Tyr-116 cannot donate a hydrogen bond to the heme as it does in the enzyme complex with UDO (Fig. 6D, left) as well as with posaconazole (25). Displacement of the Tyr-116 side chain was observed in the co-structures of T. cruzi CYP51 with VNF (25) and NEE (34). As discussed previously (35,36), disruption of this heme support from the protein moiety may also add to the UDD inhibitory potency. Assuming that the inhibitor molecule in human CYP51 adopts an orientation similar to that in the complex with T. cruzi CYP51, relatively stronger inhibition of the enzyme by UDD might also be due to disruption of the heme support from the corresponding tyrosine (Tyr-145 in human CYP51).
To summarize, the binding of both UDO and UDD to T. cruzi CYP51 produces the enzyme-inhibitor complexes that display long iron-nitrogen coordination bonds and extensive van der Waals interactions with the amino acid side chains of the active site and do not require structural rearrangements in the protein.

DISCUSSION
It has been known for years that cytochromes P450 are inhibited by heterocyclic compounds, with two components being important for the strength of the inhibition: 1) coordination of a basic heterocyclic atom with the heme iron (which lowers the P450 potential to be reduced) and 2) weaker but multiple interactions of the non-coordinated portion of the inhibitor molecule with the protein moiety (51). To date, within this type of P450 inhibitor, azoles have certainly remained the most potent, with 1,3-imidazole and 1,2,4-triazole derivatives serving as clinical antifungal drugs and agricultural fungicides (CYP51 inhibition), and yet, especially within the last decade, becoming quite notorious for their off-target activity (particularly, their inhibition of other human CYPs (48,(52)(53)(54)(55)(56)(57) and relatively high propensity to induce resistance (58,59)). Although several heterocyclic compounds other than azoles have been identified as T. cruzi CYP51 inhibi-tors, their inhibitory effects on the enzyme activity in vitro are much weaker, e.g. ChemDiv C155-0123 or indomethacin-pyridines are quite easily replaced by substrate during the time course of the CYP51 reaction (9). To our knowledge, this is the first report of non-azole type II ligands in which the potency to inhibit CYP51 activity is comparable with that of the most effective azoles (Fig. 4A). This structural study suggests that, because of weaker interaction between the heme iron and pyridine nitrogen of the inhibitors, the strength and selectivity of UDO and UDD must derive mainly from the fine topological fit between the rigid small molecules and the T. cruzi CYP51 binding cavity. Thus,   although UDO and DDD are much smaller than posaconazole, they form contacts with as many as 18 amino acids within the T. cruzi CYP51 active site. Among them, the residues from the C-terminal portion of the BЈ/C loop, helix C, and the N-terminal helix I (Fig. 6, A and B) might be of special value to ensure CYP51 selectivity because this part of the cavity appears to be CYP51-specific (31,32). As a comparison, posaconazole forms contacts with 24 amino acids of T. cruzi CYP51. Yet only 13 of these are from the CYP51 active site cavity, whereas the 11 others are located on the protein surface, encircling the entrance into the substrate access channel (25). Mutations in some of these residues have been shown to be responsible for the development of CYP51-based posaconazole resistance in fungi (25). Besides, it is the synthesis of the posaconazole long arm (which in the enzyme-inhibitor complex protrudes through the channel entrance above the protein surface) that makes the drug highly expensive (60). VNI (35), which has a molecular volume comparable with UDO/UDD, forms contacts with only 15 protein residues; its potency/selectivity toward trypanosomal CYP51s is strengthened by the hydrogen bond network that connects, via the inhibitor carboxamide fragment, the BЈ and I helices in the protozoan enzymes, but it is probably not formed in the complex of VNI with the human CYP51 (9).
The notion that good fit between the non-ligated portions of UDO/UDD and the topology of the CYP51 binding cavity should play a greater role in these complexes is also supported by our results on the inhibition of L. infantum and human CYP51s. In the L. infantum ortholog, which has a single amino acid difference from T. cruzi CYP51 within the active site, Phe-104 (corresponding to Ile-105 in T. cruzi (31)) protrudes deeper into the substrate binding cavity. In the L. infantum ortholog (one amino acid difference from T. cruzi CYP51 within the active site), the phylum-specific Phe-104 (corresponding to T. cruzi CYP51 Ile-105) protrudes deeper into the substrate binding cavity. When L. infantum and T. cruzi CYP51 structures are superimposed, it is seen that this residue comes into a close contact with UDO and especially with UDD (see Fig. 7A as an example). The distance Phe-103-UDO (2.1 Å) is too short and would induce a strongly repulsive van der Waals potential between the contacting atoms. Hence, in the complex with L. infantum CYP51, the inhibitor molecule must adopt a slightly different position. As a result, the geometric fit is impaired and the inhibitory potency decreases (Fig. 4, A and C), which is also reflected in the lower binding affinity. On the contrary, the inhibitory effects of UDD, and especially UDO on human CYP51 (where 12 of the 25 amino acids in which side chains line the surface of the active site cavity differ from those FIGURE 6. Views of the T. cruzi CYP51 active site illustrating the interactions with UDO (salmon-colored carbon atoms) and UDD (gray-colored carbon atoms). A, binding of UDO. The inhibitor molecule is shown as a spherical model, and the carbon atoms of the side chains of the inhibitor-contacting amino acid residues are depicted as light blue sticks; the corresponding secondary structural elements are shown as semitransparent gray ribbons and numbered; the carbon atoms of the heme are in yellow. UDO-specific T. cruzi CYP51 residues are underlined. B, binding of UDD (spherical model, gray-colored carbon atoms). The carbon atoms of the side chains of the inhibitor-contacting residues and the semitransparent ribbon of the corresponding secondary structural residues are colored pink and numbered. UDD-specific T. cruzi CYP51 residues are underlined. C, comparative location of the inhibitor-contacting residues in the superimposed T. cruzi CYP51 co-structures with UDO and UDD. The P450 orientation is the same as in A and B. This is a stereo view. The side chains of the inhibitorcontacting residues remain in very similar positions, with only the long arms of Met-106 and Met-460 being shifted to the right in the UDO structure to better accommodate the piperazinyl ring of the inhibitor. D, while both of the conserved in the CYP51 family heme supporting hydrogen bonds (between the hydroxyl group of Tyr-103 and the propionate of porphyrin ring A as well as between the hydroxyl group of Tyr-116 and the propionate of porphyrin ring D are preserved in the T. cruzi CYP51 complex with UDO, binding of UDD requires wedging of its aromatic ring between Tyr-116 and the heme plane, which results in disruption of the Tyr-116 OH-porphyrin ring D hydrogen bond. in T. cruzi CYP51 (Fig. 7B)), are negligible (Fig. 4D), although their binding affinities appear to be very high. This apparent discrepancy (low apparent K d values, yet weak inhibition) was observed before for other inhibitory scaffolds (e.g. posaconazole, VNF (25), and VNI (35)) and can result from the differences in the central portion of the human CYP51 I helix (25,37) that has a low energy loop-like region preceded by a leucine residue (Leu-310) instead of bulkier Phe-290 in the parasitic CYP51s (Fig. 7B). The larger space combined with the elevated flexibility of the P450 core helix in close proximity to the heme iron area might diminish the steric hindrances for the heterocyclic ring of an inhibitor to approach the heme but will not be helpful in holding it within the P450 active center tightly enough to prevent its replacement with the substrate in the enzymatic reaction.
Finally, a weaker coordination bond between the catalytic heme iron and the less electronegative nitrogen of the pyridine ring implies less of a negative impact of UDO and UDD on the P450 reduction potential. This provides a reasonable explanation for the observed weak influence of UDO and UDD on the activity of the five major human drug-metabolizing P450s ( Fig.  1and Table 1), a highly desirable feature for any new chemotherapeutic agents, as it decreases the risk of potential of in vivo toxicity, particularly when due to adverse drug-drug interactions (48).
In conclusion, UDO and UDD are promising new drug candidates for Chagas disease chemotherapy. Although their testing against other T. cruzi strains (recently shown to display high variability in their CYP51 genes (28)) is required prior to proceeding to clinical trials, the compounds represent a valuable alternative for a new CYP51 inhibitory chemotype. The structural information gained in this study will provide guidance on options for further improvement of the scaffolds and analysis of their structure-activity relations. More generally, the data suggest that balancing the impact of heme-heterocycle and apoprotein-ligand interactions may be highly helpful in minimizing the off-target activity of CYP51 inhibitors and subsequently directing the design of pathogen-selective drugs.