Structural Diversities of Active Site in Clinical Azole-bound Forms between Sterol 14 (cid:1) -Demethylases (CYP51s) from Human and Mycobacterium tuberculosis *

To gain insights into the molecular basis of the design for the selective azole anti-fungals, we compared the binding properties of azole-based inhibitors for cytochrome P450 sterol 14 (cid:1) -demethylase (CYP51) from human (HuCYP51) and Mycobacterium tuberculosis (MtCYP51). Spectroscopic titration of azoles to the CYP51s revealed that HuCYP51 has higher affinity for ketoconazole (KET), an azole derivative that has long lipophilic groups, than MtCYP51, but the affinity for fluconazole (FLU), which is a member of the anti-fungal armamentarium, was lower in HuCYP51. The affinity for 4-phenylimidazole (4-PhIm) to MtCYP51 was quite low compared with that to HuCYP51. In the resonance Raman spectra for HuCYP51, the FLU binding induced only minor spectral changes, whereas the prominent high frequency shift of the bending

CYP51 1 is a crucial enzyme in the sterol biosynthesis pathway, which catalyzes the oxidative removal of the 14␣-methyl group of lanosterol and 24-methylene-24,25-dihydrolanosterol in yeast and fungi, obtusifoliol in plants, and 24,25-dihydrolanosterol in mammals to give the corresponding sterol intermediates (1,2). The reactions include three steps of successive conversion of the 14␣-methyl group to 14␣-hydroxymethyl, 14␣-cancoaldehyde, and 14␣-formyl intermediates followed by elimination of formic acid with concomitant introduction of the ⌬14,15 double bond into the sterol core (3,4).
Because CYP51s play key roles in sterol biosynthesis (5), they have been targets for anti-fungal drugs. Azoles are successful broad spectrum inhibitors for CYPs (6 -13). One of the well known targets for anti-fungal azoles is CYP51 in Candida albicans (CaCYP51), and the inhibition of ergosterol synthesis and accumulation of 14␣-methylsterol hinders cytoplasmic membrane syntheses and growth of fungus (6,12,13). CYP51 from Mycobacterium tuberculosis (MtCYP51) is also a potential candidate for the target molecules to treat tuberculosis (8 -11). Because of the increased incidence of fungal infection associated with AIDS, but also allied with organ transplantation, cancer therapy and intensive care, azole drugs have become central to anti-fungal therapy (14).
The well known anti-fungal azole inhibitors such as fluconazole (FLU) and ketoconazole (KET) shown in Fig. 1 have revolutionized treatment of some fungal infections (6 -8). However, these inhibitors are toxic to the liver of the host and have serious side effects because of their inhibition of other CYPs (12,14,15). Clinical anti-fungal or anti-bacterial inhibitors that inhibit the enzymatic activities of only fungus or bacterial CYPs without inhibition of other CYPs from the host are desired because for example human CYP51 is essential for human metabolism (12,14). Although some azole anti-fungals show selective inhibition of yeast and fungal CYP51 over their plant and human counterparts (16), cross-over inhibition of CYP51 in two different species still causes undesirable side effects and is one of the reasons for the continuing search for better agents (17,18).
To design new azole derivatives with high specificity for bacterial and fungal CYP51s, the molecular mechanism for specificity of the inhibitor binding to CYP51s needs to be elucidated. The problem of the specificity is being addressed empirically by exploring inhibitors of different structures (19,20) and by developing three-dimensional molecular models of CYP51-active sites based on primary sequence analysis and available structures for bacterial P450s (20,21). In addition, until recently known forms of mammalian and plant CYP51s were membrane-bound microsomal enzymes (22), which complicated structural studies of this protein by x-ray crystallography. Recently, a soluble CYP51 ortholog was found in Mt-CYP51 and the molecular structures of the 4-phenylimidazole (4-PhIm)-bound and FLU-bound forms of MtCYP51 are resolved (23), revealing that the bent I helix and the region between the BЈ and C helices interact with 4-PhIm and FLU. However, the structure of MtCYP51 has so far been the only available structure in CYP51s, and structural information of other CYP51s are still limited. For discovery of novel antifungal or anti-bacterial azoles that do not inhibit human enzymes, the detailed characterization of the structural diversities in the heme pockets between mammalian and bacterial CYP51s is essential.
To obtain insight into molecular mechanisms for specificity of azoles, we focused on two CYP51s from different species and characterized their azole binding properties by spectroscopies including absorption, resonance Raman, and EPR spectra. One of the CYP51s we focused on is MtCYP51, whose three-dimensional structure with and without azole inhibitors is available (23,24). Another CYP51 we examined here is derived from human (HuCYP51). Although HuCYP51 plays key roles in metabolizing steroids (25) and is one of the human CYPs that is not desired to be inhibited (25), the enzyme is a microsomal membrane-bound protein, and the difficulty in crystallization prevents us from elucidating the secondary and tertiary structures. The structural and functional comparison between the two CYP51s would provide some new insights into design of novel azoles with high specificity to bacterial and fungal CYPs.
To investigate the inhibitory effects of the azoles to the two CYP51s, we measured dissociation constants (K d ) of three heme iron-ligated azole inhibitors, FLU, KET, and 4-PhIm, from the absorption titration. In addition, we also estimated K d of 2-phenylimidazole (2-PhIm), an inhibitor supposed to bind to CYP51 without the ligation to the heme iron (26). The heme environmental structures of azole-bound and unbound CYP51s were characterized by the EPR and resonance Raman spectroscopies, which provide information on the heme environmental changes induced by the azole binding. Based on the functional and structural comparison of two CYP51s, the structural diversities in the heme pockets between mammalian and bacterial CYP51 are discussed.
To avoid the oligomerization, the membrane-spanning leader sequence was deleted from the N terminus in HuCYP51. Furthermore, we introduced the L247Q (Leu-247 3 Gln) mutation into the N terminusdeleted HuCYP51, which enabled us to purify the enzyme without detergents (31). In this study, HuCYP51 represents the L247Q mutant of the truncated form, not the wild-type enzyme. The activity of this truncated mutant (HuCYP51) is the same as full-length wild type human CYP51 2 as reported in other microsomal P4505 (32,33), and the mutation would not induce significant structural change on the heme environmental structure of the human enzyme.
Supernatant of lysate from the cells expressing CYP51s was diluted with 50 mM Tris-HCl (pH 7.4) containing 20% glycerol, 0.1 mM phenylmethylsulfonyl fluoride, and 20 mM mercaptoethanol and the mixture was applied onto the Ni 2ϩ -nitriloacetic acid-agarose (Qiagen) column equilibrated with the same buffer. The protein was eluted with 50 mM Tris-HCl (pH 7.4) containing 80 mM imidazole and applied onto a hydroxyapatite Bio-Gel (Bio-Rad) column equilibrated with the same buffer. After being washed with equilibrating buffer, HuCYP51 was eluted with a linear gradient of 10 -250 mM phosphate buffer (pH 7.4) containing 20% glycerol. UV-visible spectra indicated that the histidine tag is not coordinated to the heme iron in MtCYP51 and HuCYP51 (23,(27)(28)(29)(30).
Spectroscopic Measurements-Electric absorption spectra of the enzymes were recorded on a Lambda 19 spectrometer (PerkinElmer Life Sciences). To determine the K d values for the azole binding, we subtracted the spectrum of the inhibitor-bound form from that of the inhibitor-free form (27,28). Dilution effects of solvents of azoles were accounted for by multiplication of the spectra by the appropriate factor. The concentration of azoles and the peak absorbance differences were then fitted to the following Equation 1, where [I] is the concentration of azole and A and A max are the observed difference in absorption at given concentrations of azole ligand and the difference in absorption at saturation, respectively. For the measurements of the resonance Raman spectra, the azolebound and -unbound ferric forms were excited at 413.1 nm with a krypton ion laser (BeamLok 2060, Spectra Physics) at room temperature (34). The Raman scattering was detected with a single polychromator (DG-1000, Ritsu) equipped with a liquid nitrogen-cooled chargecoupled device (CCD3200, Astormed, or CCD-100PB, Princeton Instruments). The spectra were analyzed by the same method as reported previously (34). The enzyme concentration for the resonance Raman measurements was 0.05 mM. The concentrations of the inhibitors we used for the measurements were 5, 2.5, 2.5, and 5 mM for 4-PhIm, FLU, KET, and 2-PhIm, respectively. EPR spectra were measured on a Varian E-12 spectrometer equipped with an Oxford ESR-900 liquid helium cryostat. The microwave frequency was X-band (9.22 GHz), and the measurements were carried out at 15 K. The microwave power and modulation were 10 mW and 1 mT, respectively, at 15 K. The concentration of the enzymes was 0.1 mM. The concentrations of inhibitors for the EPR measurements were 3, 3, 3, and 10 mM for 4-PhIm, FLU, KET, and 2-PhIm, respectively.

RESULTS
Azole Titration by Absorption Spectra-To elucidate specificity of the azole inhibitors, we determined the dissociation constant of the azoles by using spectral titration in the Soret region. The Soret peaks of the azole-unbound ferric forms were detected at 417 nm in both of MtCYP51 and HuCYP51. The peak at 417 nm is characteristic of the 6-coordinate low spin state (6cLS) (27,28), indicating that a water molecule is coordinated to the heme iron. By the addition of 4-PhIm, KET, and FLU, the Soret peaks were red shifted to 422, 425, and 420 nm, respectively (data not shown). Such a red shift in the Soret peak was also found for the binding of imidazole to P450cam (35), and these red shifts (3-8 nm) are derived from the ligation of the aromatic nitrogen of the azoles to the heme iron (27)(28)(29).
Calculated K d values were summarized in Table I. The dissociation constants of MtCYP51 were virtually identical to those in previous study (5 M for KET and 10 M for FLU) (27,28). Both of the enzymes showed the highest and lowest affinities for KET (19 M for MtCYP51 and 8.0 M for HuCYP51) and 2-PhIm (2100 M for MtCYP51 and 9100 M for HuCYP51), respectively. The extremely low affinity for 2-PhIm can be ascribed to no ligation to the heme iron (26). The high affinity for KET and FLU has been supposed to be caused by the interactions between side chains of amino acid residues in the inhibitor binding site and the lipophilic groups of FLU and KET (27)(28)(29)(30). Although the azole compounds having lipophilic groups have high affinity, the specificity of the azole compounds is significantly different between the two enzymes. Whereas 4-PhIm and KET exhibit higher affinity for HuCYP51 than for MtCYP51, FLU can bind more tightly to MtCYP51. This indicates that the inhibitor binding sites would be significantly different in these two CYP51 enzymes.
Such different environments of the inhibitor binding sites between two enzymes are also evident by the spectral changes in the titration of 2-PhIm. Based on Equation 1, the plots of absorbance changes at 396 and 417 nm against the azole concentrations showed the 1:1 binding of 2-PhIm to CYP51s (data not shown), but 2-PhIm cannot coordinate to the heme iron because of the structural constraint (26). In MtCYP51, the addition of 2-PhIm increased the absorbance at 396 nm in compensation for the decrease at 417 nm as illustrated in Fig.  2A. This indicates the structural transition from the 6cLS state to the 5-coordinate high spin state (5cHS). The intensity of the Soret peaks at 396 and 417 nm was almost the same in the saturated condition, showing that the fraction of the 5cHS state was ϳ50%. Such a transition of the spin state was also detected in the binding of isoniazid and benzimidazole to CYP51s (9), and these azole compounds are supposed to kick out the water molecule coordinating to the heme iron without the ligation of these azoles. Thus, addition of 2-PhIm displaces the coordinated water molecule from the heme iron in MtCYP51, resulting in the formation of the 5cHS state. On the other hand, addition of 2-PhIm to HuCYP51 induced only minor spectral changes in the Soret region (Fig. 2B). In 2-PhIm-bound HuCYP51, the water molecule is still ligated to the heme iron to keep the 6cLS state. The difference of the amino acid residues in the inhibitor binding sites between the two enzymes alters the binding position of 2-PhIm, leading to the different spin states in the presence of 2-PhIm.
The Electronic Structure of Heme in the Azole-bound Forms, Resonance Raman Spectra-To follow the heme environmental changes associated with the azole binding, we measured resonance Raman spectra of the ferric enzymes with and without the azoles. Fig. 3 shows the high frequency region of the inhibitor-unbound and -bound ferric state with 413.1 nm excitation in which the heme oxidation state marker ( 4 ), spin state marker ( 3 ), and the lines susceptible to the spin state and coordination structure ( 2 and 10 ) were detected (36 -42). In the inhibitor-unbound ferric state of MtCYP51 (Fig. 3A, a), 2 , 3 , and 10 appeared at 1583, 1504, and 1638 cm Ϫ1 , respectively, corresponding to the 6cLS state (37,41). The same spectral pattern was observed for HuCYP51. These observed frequencies of the marker lines in the two CYP51s were almost identical to those in CYP101, CYP102, and CYP121, which indicates that the heme of CYP51s are planar as found for other CYP enzymes (37)(38)(39).
By binding of 4-PhIm, FLU, or KET, the 3 lines for both of the enzymes were slightly but definitely shifted to the lower wave numbers (from 1504 -1501 cm Ϫ1 for MtCYP51 and from 1502-1500 cm Ϫ1 for HuCYP51), whereas other marker lines such as 4 and 10 were not perturbed. Previous studies revealed that the 3 line is quite sensitive to the donor atom of the axial ligand (40,41). Although the 2 line is also supposed to reflect the donor atom of the heme iron (39), the 2 line in the CYP51s was less sensitive compared with the 3 line. In thromboxane synthase (CYP5), the binding of imidazole shifted the 3 line from 1503-1502 cm Ϫ1 , and a similar shift in the 3 line was also observed for imidazole-ligated P450cam (from 1503-1500 cm Ϫ1 ) (40,41). The azole-induced low frequency shift of the 3 line in the CYP51s therefore indicates the binding of the nitrogen atom of these azole compounds to the heme iron. In contrast, addition of 2-PhIm induced no significant shift in the 3 line for HuCYP51, implying that 2-PhIm cannot be ligated to the heme iron. In 2-PhIm-bound MtCYP51 however, an additional 3 line appeared at 1488 cm Ϫ1 , which is characteristic of the 5cHS state (37,38). Appearance of the 5cHS state was also indicated in the absorption spectra ( Fig. 2A).
In addition to these spectra for the high frequency region, the resonance Raman spectra in the low frequency region demonstrate structural perturbations on heme peripheral groups by the azoles binding (43,44) (Fig. 4). Several lines including stretching modes of the porphyrin skeleton and bending modes of heme peripheral groups are detected in the low frequency region (36,43,44). The most prominent change by the addition of the azoles was found in the bending mode of the vinyl groups (␦ vinyl ). In MtCYP51 (Fig. 4A), the ␦ vinyl lines appeared at 413 cm Ϫ1 for the azole unbound form, whereas the 4-PhIm-, FLU-, or KET-bound forms set the ␦ vinyl line at 426 cm Ϫ1 . Because the Raman shift of the ␦ vinyl line is sensitive to orientation of the heme vinyl groups, the wave number shift in the ␦ vinyl line suggests orientational perturbation of the vinyl groups upon the binding of the azoles. The binding of 4-PhIm, FLU, or KET perturbs the environmental structure of the heme vinyl group in MtCYP51. Although a shoulder peak appeared around 426 cm Ϫ1 by addition of 2-PhIm, the major ␦ vinyl line was still detected at 413 cm Ϫ1 , indicating that the environmental changes around the vinyl groups by binding of 2-PhIm are rather small compared with those by other azoles. The two 3 lines appeared at 1488 and 1503 cm Ϫ1 in 2-PhIm-bound MtCYP51, although this did not reflect the behavior of ␦ vinyl line.
Another Raman line sensitive to the azole binding is the bending mode of the propionate groups (␦ prop ). The Raman shift of the ␦ prop line has been reported to reflect formation of a hydrogen bond between the heme propionate group and an amino acid residue in the heme pocket (36,40,45). The azoleunbound form of MtCYP51 exhibited the ␦ prop line at 386 cm Ϫ1 . Slight shifts to 388 and 389 cm Ϫ1 were observed for the FLUand KET-bound MtCYP51, whereas the ␦ prop line for the 4-PhIm-and 2-PhIm-bound forms appeared at 384 and 379 cm Ϫ1 , respectively. Although such shifts in the ␦ prop line might be caused by the cleavage of the hydrogen bond of the heme propionate group with surrounding amino acid residues, the typical shift associated with the cleavage of the hydrogen bond is the low frequency shift of ϳ10 cm Ϫ1 as found for thromboxane synthase and myoglobin (40,45). Therefore for MtCYP51 it is more likely that the hydrogen bond with the heme propionate group is not disrupted by the binding of the azoles. Presumably the binding of the azoles to MtCYP51 induced some structural changes near the propionate groups of the porphyrin ring, but the structural perturbation is rather small and depends on the molecular structure of the azoles. In HuCYP51, the two bending modes, ␦ vinyl and ␦ prop , are also susceptible to be influenced by the azole binding. However, the spectral perturbations by the azole binding in HuCYP51 are substantially different from those in MtCYP51 as displayed in Fig. 4B. Although the binding of 4-PhIm or KET shifted the ␦ vinyl line from 420 to 432 cm Ϫ1 , the binding of FLU did not affect the position of the ␦ vinyl line. The shifts of the ␦ prop line by the azole bindings in HuCYP51 are also quite different from those in MtCYP51. 4-PhIm, FLU, and KET showed the high frequency shifted ␦ prop line in HuCYP51. These spectral changes in the two enzymes suggest that the interactions of the azoles with the vinyl and propionate groups are different in the two CYP51s.
It should be noted here that the spectral patterns of the FLUand KET-bound forms of MtCYP51 are quite similar, whereas the close spectral similarity was found for the 4-PhIm-and KET-bound forms of HuCYP51. Considering that 4-PhIm is a poor inhibitor for MtCYP51 and that the binding of FLU to HuCYP51 was also rather weak compared with that of another lipophilic azole, KET, the spectral pattern in the low frequency region would be related with the binding affinity of the azole inhibitors. The positions of the lines in MtCYP51 and HuCYP51 were summarized in Tables II and III, respectively.
The Electronic States of Heme Iron in Azole-bound Forms, EPR Spectra-To further characterize the azole binding in the two CYP51s and to elucidate the molecular mechanism for the specific azole binding, the EPR spectra for the azole-unbound and -bound forms of the CYP51s were examined. Fig. 5 shows the EPR spectra of the azole unbound, 4-PhIm-, FLU-and KET-bound forms of MtCYP51 (Fig. 5A) and HuCYP51 (Fig.  5B) at 15 K. The sets of the g values (g z , g y , g x ) of azole-unbound MtCYP51 (A, a) and HuCYP51 (B, a) (13). The slight but definite difference in the g values of these two CYP51s reflect the different environment of the vinyl groups because the ␦ vinyl line in the resonance Raman spectra shows the significant frequency shift between the azole-unbound MtCYP51 and HuCYP51. As reported previously (13), the heme environmental structure of CYP51s is almost retained in broad range of the species corresponding to the 6cLS state, but the electronic state of the heme iron depends on the interactions of the heme peripheral groups and the surrounding amino acid residues.
By binding of the azole inhibitors to the heme iron, g x , and g z values were shifted to high and low field, respectively, for both of the enzymes. These shifts can be interpreted as decreased heme rhombicity because of the coordination of the nitrogen atom of the azole ring (35, 46 -48). The decreased rhombicity was also reported in the 4-PhIm-bound form of P450cam (2.45/ 2.5, 2.25, 1.89) (46). On the other hand, the addition of 2-PhIm did not induce significant changes in the low spin region of the EPR spectra (data not shown), supporting no ligation of 2-PhIm to the heme iron. The only spectral change by the addition of 2-PhIm was the appearance of a small signal at 7.96 in MtCYP51, which is derived from the S ϭ 5/2 species (35, 46 -48) and is consistent with the results of the absorption and resonance Raman spectra.
Although the g values of the KET-bound form were almost identical in the two enzymes, the spectral pattern was substantially different in the 4-PhIm-and FLU-bound forms. The binding of 4-PhIm induced two sets of EPR signals, 2.55, 2.26, 1.86 and 2.47, 2.26, 1.90 in MtCYP51 (Fig. 5A, b). Such discrete sets of the g values were also observed for the 4-PhIm bound form of P450cam (46) and the imidazole bound form of endothelial nitric-oxide synthase (47,48), which are ascribed to two orientations of the imidazole ring relative to the heme plane. The two sets of the g values therefore correspond to multiple binding orientations of the azole inhibitor and suggest that the binding specificity of 4-PhIm is lower than that of other azoles in MtCYP51. In the FLU-bound form of MtCYP51, the g values were not shifted from those of the azole unbound form, but the significant signal broadening was observed. The binding of FLU perturbs the electronic state of the heme iron in MtCYP51.
It is quite interesting that the multiple sets of the g values were also found for the azole-bound HuCYP51, but the azole inhibitor showing the multiple sets of the g values is FLU, not   , some other unresolved EPR signals around g ϭ 2.4 and 1.9 were also detected in the FLUbound HuCYP51. These multiple sets of the g values in the FLU-bound form of HuCYP51 imply that the binding specificity of FLU in HuCYP51 is lower than that in MtCYP51.

Correlation of the Inhibitor Binding Modes and Spectroscopic Properties of Two CYP51s-As clearly shown by resonance
Raman spectroscopy, the binding of 4-PhIm to MtCYP51 is different from that of FLU and KET in that the low frequency shifted ␦ prop line was observed, although these three azoles can coordinate to the heme iron. The low frequency shifted ␦ prop line was also encountered in the 2-PhIm-bound form in which 2-PhIm is not coordinated to the heme iron. The ligation of 4-PhIm to the heme iron in MtCYP51 would therefore be rather weak, corresponding to the large K d value in the titration experiment. The low affinity of 4-PhIm to the heme iron of MtCYP51 is also supported by the two sets of the EPR signals in the 4-PhIm-bound form. As reported previously (47,48), multiple sets of the g values in the EPR spectrum for the inhibitor-bound forms imply heterogeneous binding of the inhibitor, clearly showing the weak and less specific binding of 4-PhIm to MtCYP51.
On the other hand, it is FLU, not 4-PhIm, that shows the unique spectroscopic property in HuCYP51. The resonance Ra-man spectrum of FLU-bound HuCYP51 was quite similar to that of the azole unbound form. The bending mode of the heme vinyl group, ␦ vinyl , was not shifted by binding of FLU, whereas clear high frequency shifts were detected in the 4-PhIm-and KET-bound forms. The EPR spectrum for FLU-bound HuCYP51 gave multiple sets of the g values as found for 4-PhIm-bound MtCYP51, indicative of the weak and less specific inhibitor binding of FLU to HuCYP51. In fact, the K d value of FLU in HuCYP51 is extremely large (120 M) compared with another lipophilic azole, KET (8.0 M) (Table I). However, the affinity of 4-PhIm (190 M) was lower than that of FLU in HuCYP51, whereas 4-PhIm-bound HuCYP51 showed a high frequency shift in the ␦ vinyl line and a single set of the g values. The lipophilic group of FLU would increase the affinity to HuCYP51 as found for MtCYP51, but some specific interactions between FLU and surrounding amino acid residues would be lost in HuCYP51, leading to less specific binding. Thus, the spectroscopic property and binding mode of the azoles are well correlated, suggesting that the structural factors affecting the resonance Raman and EPR spectra also regulate the binding mode of the azoles for the CYP51 enzymes.
Structural Factors Affecting the Bending Modes of Vinyl and Propionate Groups in Resonance Raman Spectra-As discussed above, the structural perturbations around heme vinyl and propionate groups affect the affinity of the azoles for the CYP51 enzymes. Based on the crystallographic structure of MtCYP51 (23), possible amino acid residues interacting with 2-vinyl and 4-vinyl groups are Phe-387 and Phe-399, respectively (Fig. 6a). By binding of the azoles to MtCYP51, the positional shift was observed for Phe-399, perturbing the interaction with the 4-vinyl group (23). Although Phe-399 is located on the L helix and far from the inhibitor binding site, a hydrogen bond between Gln-403 located near Phe-399 on the L helix and Ser-261 on the I helix constituting the inhibitor binding site is formed (Fig.  6a). It is therefore plausible that the azole binding induces some structural changes of the I helix and the structural perturbation on the I helix is propagated to the L helix through the hydrogen bond between the two helices, leading to the positional changes of Phe-399 and the shift of the ␦ vinyl line in the resonance Raman spectra of the azole bound CYP51s.
Another sensitive marker for the azole binding is the ␦ prop line. As shown in the x-ray structure of MtCYP51, a highly conserved hydrogen bond network including Gln-72 (near the BЈ helix), Tyr-76 (near the BЈ helix), Arg-326 (1-4 sheet), His-392 (near the L helix), and heme propionate groups is formed in the ABЈ loop and the L helix (Fig. 6b) (23). These hydrogen bonds are supposed to be essential to maintain the integrity of the inhibitor binding site (23). The C helix is also located near the heme propionate group and sensitive to the inhibitor binding (23) (Fig. 6c). Because of the large thermal fluctuation, side chains of amino acid residues on the C helix would not interact with the heme peripheral group in the azole unbound form, whereas the temperature factors for the side chain of some amino acid residues on the C helix were decreased to form hydrogen bonds of the heme 6-propionate group with Lys-97 and Arg-95 on the C helix in the 4-PhIm-and FLU-bound forms of MtCYP51, respectively (23). It is therefore plausible that the binding of the inhibitor induces structural perturbation in the region of the ABЈ loop and reduces the structural fluctuation of the C helix, which alters the hydrogen bond network including the heme propionate groups and results in the frequency shifts of the bending mode of the heme propionate group.
Structural Origin for the Specificity of Azoles-As discussed in the previous section, the azoles interact with the side chains of the amino acid residues on the I helix and in the region between the BЈ and C helices. It is likely that the different inhibitor specificity between two CYP51s would be based on the amino acid sequence of these regions. One clue to identify the key amino acid residues for inhibitor specificity is the binding property of FLU. The amino acid sequence of the I helix is rather conserved and the sequence homology between MtCYP51 and HuCYP51 is high, but Phe-255 in MtCYP51 is replaced with Leu-310 in HuCYP51 (Fig. 7a). The replacement of a bulky phenylalanine residue with a flexible leucine residue would enlarge the inhibitor binding site, which relieves the steric repulsion to the inhibitors and results in the enhanced affinity of HuCYP51 for 4-PhIm and KET. However, affinity of FLU was drastically decreased in HuCYP51, which cannot be explained by the reduced steric hindrance in HuCYP51. Although we cannot exclude the possibility that some favorable interactions between Phe-255 and FLU such asstacking of the aromatic rings are formed only in MtCYP51, the crystal structure of FLU-bound MtCYP51 shows that the phenyl ring of Phe-255 is not parallel to 2,4-difluorophenyl or the azole ring of FLU (23), and it is unlikely that specific interactions are disrupted by the substitution of the amino acid residue in HuCYP51.
An alternative interaction site with azoles is the region between the BЈ and C helices. The crystal structure of the MtCYP51-FLU complex indicates that the 2,4-difluorophenyl FIG. 7. a, sequence alignment of MtCYP51, CaCYP51, and HuCYP51 in the regions of the BЈ-C helices and the I helix. Residues identical or homologous are shaded in dark and light, respectively. Cylinders on the sequences represent the BЈ, C, and I helices (BЈ helix, the residues 78 -83 in MtCYP51; C helix, the residues 92Ϫ99 in MtCYP51; I helix, the residues 245Ϫ267 in MtCYP51). b, the amino acid residues interacting with FLU in the region of the BЈ and C helices of MtCYP51 (Protein Data Bank code 1EA1 and Ref. 23). The BЈ and C helices are displayed in yellow ribbon models. The random coils are shown in white lines. Carbon, nitrogen, oxygen, sulfur, fluorine, and iron atoms are indicated in green, blue, red, yellow, purple, and purple, respectively. ring of FLU is located near Arg-96 and Leu-100 in the inhibitor binding site (Fig. 7b) (23). The structure of the 4-PhIm-bound form also revealed that Tyr-76, Phe-78, Met-79, and Phe-89 form a hydrophobic binding site for the phenyl group of 4-PhIm. Although the amino acid residues of the hydrogen bond network constituting the inhibitor binding site, which includes the propionate group, Tyr-76, Arg-326, and His-392, in MtCYP51 are completely conserved in CYP51s, the sequence homology for the C helix is relatively low, and the hydrophilic patch including Arg-95, Arg-96, and Lys-97 in MtCYP51 is not conserved (Fig. 7a). In HuCYP51, these hydrophilic amino acid residues are replaced with hydrophobic amino acid residues (Val-151, Phe-152, and Leu-153).
The amino acid substitution of the C helix in HuCYP51 therefore enhances hydrophobicity for the inhibitor binding site, leading to the high affinity of the hydrophobic group in 4-PhIm and KET for HuCYP51. However, FLU has the hydrophilic triazole rings and hydroxyl group, which would reduce the affinity for the hydrophobic inhibitor binding site. In addition, hydrophilic interactions are also suggested between the fluorine atom of difluorophenyl group in FLU and Arg-96 because the carbon-fluorine bond is significantly polarized (49). The negatively charged character of the fluorine atom might be stabilized by the positive charge of Arg-96 in MtCYP51. On the other hand, the hydrophobic inhibitor binding site in HuCYP51 would reduce the affinity for the 2,4-difluorophenyl group in FLU and decrease the affinity of FLU. The hydrophobicity in the C helix is crucial for the specific binding affinity of the azoles for CYP51s.
Furthermore, amino acid substitution of the BЈ helix can also be a factor to regulate the inhibitor specificity of CYP51s. Although most of the amino acid residues are conserved in the BЈ helix, amino acid residues interacting with the azoles in MtCYP51, Phe-78 and Met-79, are changed to Leu and Thr in HuCYP51, respectively (Fig. 7a). These amino acid residues form a pocket for the phenyl group of 4-PhIm, and the amino acid replacement of Phe-78 and Met-79 with Leu and Thr, respectively, would enlarge the pocket and stabilize the binding of 4-PhIm in HuCYP51.
Thus, we can conclude that the region between the BЈ and C helices is the key region for the azole recognition in MtCYP51 and HuCYP51. This region is one of the hot spots for mutation in CYP51s, and some of the mutants are azole resistant (6,18), and the BЈC loop is supposed to be a substrate entry pathway into the heme pocket in MtCYP51 (21,24). The primary roles of the region between the BЈ and C helices in azole recognition have also been suggested in another CYPs from human liver (50). The crystallographic structure of the warfarin bound form in CYP2C9 indicated that the BЈC loop constitutes the substrate binding site and that the interactions with the substrate are crucial for the substrate specificity (51). The detailed structural characterization of the region between the BЈ and C helices could be a clue to discovering novel anti-fungal and anti-bacterial drugs.
Summary-To investigate the structural origin of the inhibitor specificity in CYP51s, we characterized structural changes associated with the azole binding in bacterial and human CYP51s by electronic absorption and resonance Raman and EPR spectroscopies. We correlated the Raman shift of the bending modes for the heme vinyl and propionate groups by azole binding with the binding mode of azoles. The EPR spectra of the azole bound forms also reflect the binding specificity of the azoles, and a clear relationship was found between the EPR spectra and the binding mode. Together with the sequence alignment between the two CYP51s and the crystal structure of the azole bound bacterial CYP51, we can suggest that the region between the BЈ and C helices plays key roles in the azole specificity and particularly hydrophobicity in the C helix being a crucial factor. However, structural and functional characterization of the mutant proteins in which the residues in the region between the BЈ and C helices were exchanged in the two CYP51s is further required to confirm the suggestion. The detailed characterization of the interactions of the azoles with the region between the BЈ and C helices of CYP51s in the target bacteria and host would pave the way for new azole inhibitors with high specificity and reduced side effects.