Antifungal Imidazoles Block Assembly of Inducible NO Synthase into an Active Dimer*

Cytokine-inducible nitric oxide synthase (iNOS) is a homodimeric enzyme that generates nitric oxide (NO) andl-citrulline from l-arginine (l-Arg) and O2. The N-terminal oxygenase domain (amino acids 1–498; iNOSox) in each subunit binds heme,l-Arg, and tetrahydrobiopterin (H4B), is the site of NO synthesis, and is responsible for the dimeric interaction, which must occur to synthesize NO. In both cells and purified systems, iNOS dimer assembly is promoted by H4B, l-Arg, and l-Arg analogs. We examined the ability of imidazole andN-substituted imidazoles to promote or inhibit dimerization of heme-containing iNOSox monomers, or to affect iNOS dimerization in cells. Imidazole, 1-phenylimidazole, clotrimazole, and miconazole all bound to the iNOSox monomer heme iron. Imidazole and 1-phenylimidazole promoted iNOSox dimerization, whereas clotrimazole (30 μm) and miconazole (15 μm) did not, and instead inhibited dimerization normally promoted byl-Arg and H4B. Clotrimazole also bound to iNOSox dimers in the absence of l-Arg and H4B and caused their dissociation. When added to cells expressing iNOS, clotrimazole (50 μm) had no effect on iNOS protein expression but almost completely inhibited its dimerization and consequent NO synthesis over an 8-h culture period, without affecting calmodulin interaction with iNOS. Thus, imidazoles can promote or inhibit dimerization of iNOS both in vitro and in cells, depending on their structure. Bulky imidazoles like clotrimazole block NO synthesis by inhibiting assembly of the iNOS dimer, revealing a new means to control cellular NO synthesis.

in primary sequence, gene chromosomal location, and activation by Ca 2ϩ (4 -6). A neuronal NOS isoform (nNOS) that is present in brain and skeletal muscle (7,8), and an endothelial NOS isoform (eNOS) expressed in the vasculature or brain (9,10) are dependent on calmodulin (CaM) binding for activity, which is reversible and occurs in response to elevated intracellular Ca 2ϩ . In contrast, a continuously active NOS isoform (iNOS) is expressed in cells exposed to inflammatory cytokines or bacterial products (11), and is neither stimulated by Ca 2ϩ nor blocked by CaM antagonists due to its containing tightlybound CaM (12). Numerous pathologies are attributed to excess NO production by iNOS (13)(14)(15)(16) and have led to a quest for specific inhibitors of this isoform. Work has focused on a broad range of molecules including substrate analogs, guanidine derivatives, thioureas, and heterocycles, with some specific inhibitors beginning to emerge (17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27).
Although the NOS isoforms differ regarding their primary sequence and mode of expression, they all are bi-domain enzymes comprised of a C-terminal reductase domain that contains binding sites for NADPH, FAD, FMN, and CaM, and an N-terminal oxygenase domain that contains binding sites for heme, tetrahydrobiopterin (H 4 B), and L-Arg (5, 28 -31). The NOS heme shares characteristics with the heme in cytochromes P-450 in that it coordinates to the protein through a cysteine thiolate (30 -32), can bind O 2 as a sixth ligand (33), and may directly participate in oxygen activation and product formation (33)(34)(35)(36)(37). Thus, as with the cytochromes P-450, the NOS heme represents a potential target for enzyme inhibition. In fact, compounds that bind directly to the NOS heme iron such as CO, NO, CN, imidazole, and N-phenyl imidazoles all inhibit NO synthesis (22, 34, 38 -40).
The NOSs are only active as homodimers (41)(42)(43). For iNOS (29) and possibly nNOS and eNOS (44 -46), only the oxygenase domains of two subunits interact to form the dimer, with the reductase domains attached as extensions that may destabilize the dimer (47). Both the oxygenase and reductase domain of each NOS isoform can fold and function independent of one another (45)(46)(47). For example, the iNOS oxygenase domain (iNOSox, amino acids 1-498) is expressed in Escherichia coli as a dimer, exhibits normal affinity toward L-Arg and H 4 B, and catalyzes NO synthesis from the reaction intermediate N-hydroxy-L-Arg either in a H 2 O 2 -supported reaction or when supplied with NADPH and its reductase domain (48 -50). A crystal structure of dimeric iNOSox with L-Arg and H 4 B bound has been published (32).
Full-length NOS monomers isolated from mammalian cells are devoid of H 4 B and heme, but contain bound FAD and FMN, functional NADPH and CaM binding sites, and can catalyze electron transfer to artificial acceptors such as cytochrome c at rates that match their respective dimeric forms (41,42). Work with full-length iNOS and nNOS monomers indicates that their dimerization minimally requires that heme be inserted into the protein during the dimerization reaction (41,42,51).
The heme-containing NOS monomer may be an intermediate on the path toward forming a stable dimer (52), and although it does not accumulate in mammalian cells containing H 4 B and L-Arg (41), it can be generated in vitro by dissociating purified iNOS dimers with urea (47,53). Dimerization of iNOS monomers is also stimulated by H 4 B and L-Arg (41,47,53) and is inhibited in cells by NO, which interferes with heme insertion into the monomers (54). Dimerization activates iNOS in part by enabling electron transfer between enzyme flavin and heme prosthetic groups (55), and by facilitating productive binding of substrate and H 4 B (41,49).
Because dimer assembly is critical for NO synthesis, it is a potential target for therapeutic intervention. A study of iNOS dimerization as promoted by L-Arg showed that its effect was stereospecific but not unique, because several L-Arg and guanidine analogs that bind to the iNOS dimer also promoted dimer assembly (41,56). Two substrate analogs that exhibit greater affinity toward the iNOS dimer than L-Arg, N -amino-L-Arg and L-thiocitrulline (21,24), did not promote dimer assembly, suggesting they might function as antagonists. However, they were incapable of blocking dimer assembly in the presence of excess L-Arg (56), consistent with their sharing a common binding site but being generally unable to bind to iNOS monomers (32,48,49). This suggests that substrate analogs have little inherent capacity to block iNOS dimerization.
On the basis of these considerations, we tested if imidazoles might positively or negatively influence iNOS dimer assembly. Because imidazoles bind to the NOS heme iron, their binding should not depend on NOS dimeric structure, but simply on whether heme is incorporated into the protein. In fact, the crystal structure of a heme-containing iNOSox monomer shows that two molecules of imidazole can bind within the monomer's distal heme pocket (57), one coordinating to the heme iron and the other binding to the carboxylate of Glu-371, which also binds the guanidino nitrogens of L-Arg (32,48). We therefore examined the ability of imidazole and N 1 -substituted imidazoles to promote or antagonize the dimerization of full-length iNOS and iNOSox monomers, and compared results to those obtained with L-Arg, H 4 B, or 7-nitroindazole, which are all known to stabilize the NOS dimer or promote its assembly without binding to the heme iron (56,58). We found that imidazoles can promote or inhibit iNOS dimerization depending on their structure, and act through a mechanism that involves binding to the heme iron. On this basis, we characterize an antifungal imidazole as a potent inhibitor of iNOS dimer assembly both in a purified system and in cultured cells.

EXPERIMENTAL PROCEDURES
Materials-Interferon-␥ was a gift from Genentech, South San Francisco, CA. Antibodies against iNOS and CaM were from Transduction Laboratories (Lexington, KY) and Upstate Biotechnology Inc. (Lake Placid, NY). The enhanced chemiluminescence (ECL) kit for immunodetection was from Amersham International PLC (Little Chalfon, United Kingdom). Bacterial culture materials were purchased from Difco Labs (Detroit, MI) and Becton Dickinson Microbiology Systems (Cockeysville, MD). Clotrimazole and miconazole were purchased from Sigma, while 7-nitroindazole, 1-phenylimidazole, and 2-phenylimidazole were purchased from Aldrich. Mammalian cell culture materials were purchased from Life Technologies Inc. (Gaithersburg, MD), ampicillin was from Apothecon (Princeton, NJ). Pefabloc and lysozyme were from Boehringer-Mannheim GmBH (Germany). Ni 2ϩ -nitriloacetate resin was from Novagen Inc. (Madison, WI). Gel filtration protein standards were from Bio-Rad. Clotrimazole and miconazole stock solutions were prepared in Me 2 SO, while all other imidazoles were dissolved in buffer.
Cell Culture and Protein Preparation-RAW 264.7 macrophage cell cultures (500 ml) were grown in spinner flasks and induced to express iNOS by adding 10 units/ml interferon-␥ and 1 g/ml E. coli lipopolysaccharide as previously detailed (41). Clotrimazole was added as a 2-ml Me 2 SO solution to each culture as described below. In some cases, iNOS monomers and dimers were purified from the soluble RAW 264.7 cell fraction by sequential chromatography on 2Ј,5Ј-ADP Sepharose and Mono-Q anion exchange columns using fast protein liquid chromatography (41).
The oxygenase domain of iNOS (amino acids 1-498, iNOSox) containing a 6-hisitidine C terminus was expressed in E. coli and purified in the absence of L-Arg and H 4 B essentially as detailed in Siddhanta et al. (60), while full-length iNOS containing a 6-histidine tag at its N terminus was expressed in E. coli and purified in the absence of L-Arg and H 4 B as described in Wu et al. (61). The purified iNOSox and full-length iNOS proteins were 70 -80% dimeric as isolated and were dissociated into monomers with urea according to previous methods (47,60). Briefly, iNOSox was diluted to 3 M in 40 mM HEPES, pH 7.5, containing 10% glycerol, 3 mM DTT, and 5 M urea at 4°C. After 1 h, this solution (ϳ1 ml) was sequentially dialyzed at 4°C against 200 ml of the same buffer containing 5 M urea for 90 min, against 200 ml of the same buffer containing 2 M urea for 4 h, and against 200 ml the same buffer containing 0.1 M urea overnight. The same procedure was used for full-length iNOS except 2.5 M urea was used to dissociate the dimer and intermediate dialysis with 2 M urea was omitted. These procedures resulted in preparations that contained 60 -80% monomer and 20 -40% dimer (percentages reflect distribution of iNOS protein mass) as determined by gel filtration chromatography. The urea-generated iNOS monomers contained heme but no H 4 B (47, 53) and were used within 1 day of their preparation.
NO Synthesis Activity-The NO synthesis activity of activated cell cultures or of soluble cell supernatants was estimated using the colorimetric Griess assay for nitrite as described previously (41). Culture fluid was analyzed directly for nitrite. NO synthesis activity of cell supernatants or column fractions was determined in 100-l incubations containing aliquots of cell supernatants or column fractions and 40 mM Tris buffer, pH 7.8, 1 mM NADPH, 2 mM L-Arg, 3 mM DTT, protease inhibitors, and 4 M each of FAD, FMN, and H 4 B as described previously (41). Incubations were run for 60 min at 37°C prior to analyzing for nitrite. NO synthesis activity of purified full-length iNOS was determined using the spectrophotometric oxyhemoglobin assay. Cuvette samples (350 l) contained iNOS, 5 M oxyhemoglobin, 0.3 mM DTT, 1 mM L-Arg, 0.5 mg/ml bovine serum albumin, 1300 units/ml catalase, and 150 units/ml superoxide dismutase in 40 mM EPPS, pH 7.6. Reactions were initiated by adding 100 M NADPH and the rate of NO synthesis was measured at 401 nm, using an extinction coefficient of 38 mM Ϫ1 cm Ϫ1 .
Spectroscopy-Optical spectra were recorded at room temperature on a Hitachi U-2110 spectrophotometer as detailed in Sennequier and Stuehr (56). Spectra were collected and processed using SpectraCalc software (Galactic Industries Corp., Salem, NH). Titrations of N-substituted imidazoles were performed by adding 3-l aliquots of concentrated stock solutions (giving a final concentration range of 1-50 M and 1-20 M, respectively, for clotrimazole and miconazole) to cuvettes containing 1 ml of Bis-Tris buffer, pH 7.6, 1 mM DTT, iNOSox, and L-Arg, or H 4 B as stated in the text. Spectra were collected after each addition. Binding constants were derived from double-reciprocal plots of the absorbance difference (peak to trough in the difference spectra) versus the additive concentration.
Dimer/Monomer Detection and Formation-iNOS dimer and monomer in supernatants of lysed cells or purified iNOSox samples were estimated by fractionating 100 l of sample on a Superdex 200 gel filtration column at 4°C. The column was run at 0.5 ml/min with 40 mM Bis-Tris propane buffer, pH 7.4, containing 2 mM DTT, 10% glycerol, and 200 mM NaCl. Under these conditions, dimer does not dissociate into monomer nor do monomers form dimers. 2 For samples containing iNOSox, eluted protein was monitored at 280 nm and the peaks were assigned to be dimer or monomer based on the elution volumes of protein standards and authentic iNOSox monomer and dimer (47). The dimer-monomer distribution in iNOSox samples was estimated based on relative peak heights. For cell supernatants, aliquots of each column fraction were subject to SDS-PAGE, proteins were transferred onto a Nytran membrane, and iNOS was detected using an anti-iNOS polyclonal antibody as detailed in Albakri and Stuehr (54). The SDS-PAGE used a gradient of 5-20% acrylamide which allowed us to probe the membrane for CaM (17 kDa) as well as for iNOS (130 kDa). After iNOS bands were visualized and quantitated by scanning densitometry, the membrane was stripped of the antibody directed against iNOS and reprobed using a mouse antibody directed against bovine CaM by a standard procedure (12).
Dimerization of iNOSox or full-length iNOS monomers was performed by incubating the protein (1 M) at room temperature in 40 mM HEPES buffer, pH 7.5, containing 3 mM DTT, 0.5 mg/ml bovine serum albumin and L-Arg, H 4 B, and/or compounds as noted in the text. The iNOSox dimer-monomer distribution was estimated by fractionating 100 l of the incubate at various times on the Superdex gel filtration column as noted above. To monitor dimerization of iNOSox spectroscopically, the reaction was carried out in a 300-l cuvette and optical spectra were recorded between 250 and 700 nm every 10 min. Dimerization of full-length iNOS was followed by assaying NO synthesis activity of aliquots removed at indicated times.

RESULTS
The structures of the substituted imidazoles and indazole used in this study are depicted in Fig. 1. The ability of imidazole, phenylimidazoles, and 7-nitroindazole to bind to a hemecontaining iNOSox monomer and affect dimerization was investigated by incubating monomers with each compound and monitoring their binding and dimerization over a 40-min period with UV/visible spectroscopy and gel filtration chromatography. The iNOSox preparation used here was initially ϳ70% monomeric based on gel filtration (data not shown) and in the presence of DTT exhibited a split Soret absorbance peak at 378 and 459 nm ( Fig. 2A), indicative of a DTT thiolate coordinating as a sixth heme ligand to the ferric iNOSox monomer and residual dimer (53,59). Adding imidazole to the iNOS monomer preparation caused immediate conversion to a species exhibiting a single Soret band with absorbance maximum at 429 nm ( Fig. 2A), indicating that imidazole quickly displaced DTT as a sixth ligand to the heme iron in the iNOSox monomer and residual dimer. The spectrum of imidazole-bound iNOSox did not change over the subsequent 40-min incubation period. In contrast, an iNOSox monomer incubated with L-Arg and H 4 B for 40 min underwent gradual spectral changes to generate a species with Soret maxima at 398 nm ( Fig. 2A, inset), which is the typical spectral change observed during dimerization of iNOSox monomers promoted by L-Arg and H 4 B (47), and indicates gradual displacement of DTT from the ferric heme as dimerization progresses. Incubation of iNOSox monomers with 7-nitroindazole also caused a gradual spectral change that indicated displacement of DTT, and a high degree of dimerization (data not shown).
A significant portion of imidazole-bound iNOSox monomer assembled into a dimer over the 40-min incubation period (Fig.  2B). The amount of dimer present in the sample incubated with imidazole exceeded the amount of dimer present in the sample incubated with DTT alone, 3 and was comparable to that ob-tained for monomers incubated with L-Arg and H 4 B (Fig. 2B). Spectral and dimerization results obtained using 1-phenylimidazole were identical to those obtained with imidazole (data not shown), indicating that a phenyl substituent at the N 1 position has no adverse effect on either heme binding or dimerization. In contrast, incubating iNOSox monomers with 2-phenylimidazole did not result in spectral change or dimerization (data not shown), consistent with 2-phenylimidazole binding poorly to heme proteins due to a steric interaction between the phenyl ring and porphyrin (22).
The rate of iNOS dimerization in the presence of imidazole was investigated using urea-generated heme-containing monomers of full-length iNOS and assaying recovery of NO synthesis, which is a dimer-specific activity used to measure iNOS dimerization (41,47,53). As shown in Fig. 3, there was an increase in NO synthesis activity within the first 20 min of incubation followed by a more gradual increase. This time course is similar to that observed for iNOS dimerization in response to L-Arg and H 4 B (47,49,53), and suggests that dimerization follows binding of imidazole to the heme iron, which occurs within seconds after mixing. Recovery of NO synthesis was also associated with an increase in the proportion of dimer from 30 to 70% over the course of the imidazole incubation as determined by gel filtration chromatography of the 0 and 180-min samples (data not shown). We conclude that imidazole and small N 1 -substituted imidazoles can promote dimer assembly after binding to the heme iron. This differs from L-Arg, H 4 B, or 7-nitroindazole, which promote dimer assembly without binding to the heme iron.
We then investigated if larger N 1 -substituted imidazoles would bind to iNOSox monomer and affect dimer assembly. Addition of clotrimazole (30 M) or miconazole (15 M) led to an immediate spectral change identical to that obtained when imidazole was added (data not shown), indicating that these bulky imidazoles bind rapidly to the monomer heme iron. The affinity of the iNOSox monomer heme for clotrimazole and miconazole was determined by perturbation difference spectrophotometry (shown for clotrimazole in Fig. 4A). The binding constants derived from double-reciprocal plots of spectral change versus wavelength were 7.3 Ϯ 1.0 M for clotrimazole (Fig. 4B) and 12 Ϯ 0.4 M for miconazole (data not shown). The constant for clotrimazole binding to iNOSox is in agreement with its inhibitory constant for NO production by stimulated hepatocytes (13 M) (62). We conclude the iNOSox monomer heme can bind large N 1 -substituted imidazoles with high affinity, consistent with the monomer containing a relatively open distal heme pocket (57).
Gel filtration analysis of iNOSox monomers after incubation with either clotrimazole or miconazole for 40 min showed that it remained monomeric in both cases (Fig. 5, top horizontal  row), indicating that these bulky imidazoles did not promote dimer assembly. We therefore examined their ability to antagonize L-Arg-and H 4 B-promoted dimerization. Monomers were preincubated for 10 min either with buffer alone or containing various concentrations of clotrimazole or miconazole, and then exposed to 1 mM L-Arg and 10 M H 4 B to promote dimerization, which was assayed after an additional 10 and 40 min incubation. In the absence of clotrimazole and miconazole, significant dimer assembly occurred within 10 min of incubation with L-Arg and H 4 B and generated ϳ90% dimer by 40 min (Fig. 5,  first vertical row). Clotrimazole or miconazole inhibited dimer assembly as promoted by L-Arg/H 4 B in a concentration-dependent manner (Fig. 5), with miconazole being somewhat more effective than clotrimazole. In cases where little or no dimerization occurred (30 M clotrimazole or 15 M miconazole), spectral change characteristic of L-Arg-and H 4 B-promoted dimerization (see Fig. 2A, inset) was not observed, whereas in cases of partial dimerization (for example, at 10 or 20 M clotrimazole) we observed a time-dependent increase in absorbance centered at 396 nm, consistent with some H 4 B-and L-Argpromoted dimerization and displacement of clotrimazole (Fig.  6).
We next investigated whether clotrimazole or miconazole could dissociate an iNOSox dimer 4 that had or had not been preincubated with L-Arg plus H 4 B. Dimer not exposed to L-Arg and H 4 B bound clotrimazole (15 M) or miconazole (30 M) rapidly, as judged by an immediate spectral change from the split Soret peak to a single peak centered at 429 nm (data not shown; identical to that observed for imidazole in Fig. 2A). Their binding was associated with subsequent conversion of dimer to monomer over a 40-min incubation as determined by gel filtration (Fig. 7, traces A-C). In contrast, iNOSox dimer that had been preincubated with L-Arg and H 4 B was unable to bind clotrimazole or miconazole, as judged by a lack of spectral change (data not shown), and also did not significantly dissociate into monomer over a 40-min incubation (Fig. 7, trace D). Thus, clotrimazole and miconazole only bound to (and dissociated) dimeric iNOSox in the absence of L-Arg plus H 4 B.
On the basis of the above data, we tested if clotrimazole would block assembly of dimeric iNOS in the RAW 264.7 macrophage cell line. In a typical experiment, a 500-ml spinning cell culture was made 50 M in clotrimazole, incubated for 30 min, then induced to express iNOS by adding lipopolysaccharide and interferon-␥ and incubated for an additional 8 h before harvesting and cell lysis. Control cultures received Me 2 SO alone but were otherwise identically activated and processed. The nitrite concentration in the control culture increased from 2 to 30 M over the 8-h induction period (Fig. 8), consistent with induced expression of active dimeric iNOS, but the nitrite level did not increase in the culture given clotrimazole. Cell supernatants prepared from representative matched control and clotrimazole-treated cultures (500 ml each) contained 28 and 23 mg of soluble protein, respectively, indicating that clotrimazole caused little or no cell lysis under these conditions. Supernatant prepared from the clotrimazoletreated culture displayed a NO synthesis specific activity that was 6-fold lower than the control supernatant (0.38 versus 2.3 nmol of nitrite/min/mg of protein, respectively), but contained a similar amount of iNOS protein as the control as judged by Western analysis (data not shown), consistent with corresponding observations in activated hepatocytes (62). Thus, the clotrimazole-treated culture contained iNOS in a predominantly inactive form.
Inactive iNOS could arise from clotrimazole inhibiting dimer assembly, or alternatively, by clotrimazole inhibiting CaM binding to iNOS (63,64), which would likely prevent its proper folding (61). We differentiated between these two possibilities by determining iNOS dimer-monomer ratios in the cell lysates and checking whether CaM was associated to iNOS. The control activated supernatant (Fig. 9, panel A) contained both dimeric and monomeric iNOS in relatively equal proportion, consistent with previous reports (41,54), whereas the clotrimazole-treated supernatant (Fig. 9, panel B) predominantly contained monomeric iNOS. Scanning densitometry estimated the dimer-monomer ratios to be 45:55 and 15:85 in each case, respectively. Reprobing the Western blots with an antibody directed against CaM showed that CaM was present in the same fractions that contained the iNOS dimer and monomer, with the majority running at an apparent molecular mass of 17 kDa (shown in Fig. 9, panel C, for the cell lysate from clotrima-zole-treated cells), as is typically observed for iNOS in denaturing gels (12,61). Measurement of NO synthesis activity in each gel filtration fraction (Fig. 9, panel D) confirmed that much less iNOS dimer was present in the supernatant from clotrimazole-treated cells. We also checked for iNOS-associated CaM by determining the cytochrome c reductase activity of iNOS monomer purified from clotrimazole-treated cells. Monomer reductase activity in the absence or presence of added Ca 2ϩ /CaM was 2.0 Ϯ 0.3 and 3.1 Ϯ 0.5 mol of cytochrome c reduced/min/mmol of iNOS, respectively, similar to the value reported for iNOS monomers purified from conventional cultures (2.5 Ϯ 0.4 mol of cytochrome c reduced/min/mmol; Ref. 41). We conclude that clotrimazole inhibited iNOS dimer assembly in the cells without significantly preventing CaM binding to iNOS. DISCUSSION Our data show that imidazoles can positively or negatively regulate iNOS dimer assembly. Small molecules like imidazole and 1-phenylimidazole promoted iNOS dimerization, whereas bulky imidazoles like clotrimazole and miconazole did not, and instead inhibited iNOS dimerization both in a purified system and in cultured cells. Their ability to block iNOS dimer assembly was unexpected and may suggest new strategies to control NO synthesis in cells.
How small molecules influence NOS dimer assembly is not clear. Outside of a few cell culture studies (54,65,66), our current understanding is based on work with purified NOS proteins or subcellular preparations. Some general features have emerged. For example, various derivatives of L-Arg and guanidine, various pteridines, ethylisothiourea, and 7-nitroindazole all promote NOS subunit dimerization (41,47,51,56,67) or stabilize the dimer (42,43,58). This suggests structural constraints are minimal, and the promoting effect is not limited to molecules acting as substrates or cofactors for NO synthesis. The active molecules noted above all bind to dimeric NOS isoforms with moderate to good affinity, but appear to bind poorly or not at all to NOS monomers, suggesting their binding sites are incompletely formed in the monomer. In addition, while most of these molecules seem to bind near the distal heme pocket of NOS, they do not ligate directly to the heme iron, and instead bind through interactions with the protein.
When molecules of this type are added to heme-containing iNOS monomers, they cause a gradual shift to five-coordinate high-spin heme that proceeds according to the tempo of iNOS dimerization, as determined independently by monitoring protein fluorescence and catalytic changes associated with dimerization (47,58). Together, these features support a mechanism in which NOS subunits first interact reversibly to form a loose dimer before the promoter molecules bind. 5 Subsequent promoter molecule binding then stabilizes the dimeric structure and drives the monomer-dimer equilibrium toward stable dimer (42,52).
Our current results suggest that imidazoles may not function by this mechanism, but instead affect dimerization by binding directly to the iNOS monomer heme iron. This possibility is supported by the immediate and complete binding of imidazole when it was added to iNOS monomers, the rate at which dimerization occurred when imidazole was present at a concentration approximately 100 times K d , and the fact that imidazole remained bound to the heme after the dimer had formed. Its binding to the monomer heme appears critical, because dimerization is not promoted by imidazoles that cannot bind to heme (i.e. 2-phenylimidazole). Thus, two means for promoting iNOS dimerization appear possible: one involves subunit dimeric interaction as an initial step followed by promoter binding to stabilize the dimeric structure (for example, H 4 B and L-Arg); while the other may involve promoter binding as an initial event followed by productive subunit interaction (imidazole). Both events may cause a common change in iNOS structure that stabilizes the dimeric form.
Clotrimazole bound immediately to an iNOSox monomer or to a L-Arg-and H 4 B-free dimer, but it was unable to bind to a H 4 B-and L-Arg-saturated dimer. This provides further evidence that the distal heme pockets of an iNOS monomer or H 4 B-and L-Arg-free dimer are similar in being exposed to the solvent and able to accommodate large heme iron ligands like clotrimazole or DTT. Bound H 4 B and L-Arg excludes these bulky ligands. Because clotrimazole could only dissociate iN-OSox dimers with which it formed a heme-iron complex, it likely affects iNOS quaternary structure after binding to the heme iron, like the smaller imidazoles.
A scheme that incorporates clotrimazole (and miconazole) inhibition into a current model for iNOS dimer assembly (52) is illustrated in Fig. 10. Heme-free iNOS monomers (iNOS M ) must incorporate heme (Fe) in order to dimerize (41,42). Heme-containing iNOS partitions in a monomer-loose dimer equilibrium (iNOS M Fe N iNOS D Fe). H 4 B and L-Arg, if present, can bind to the dimer and drive formation of stable, active iNOS dimers (iNOS D Fe(Arg, H 4 B)). Clotrimazole reversibly binds to the heme iron of a monomer and prevents its dimerization; it can also reversibly bind to the heme in a loose dimer and cause it to dissociate. Both processes lead to monomer accumulation. Occupancy of the distal heme pocket by clotrimazole versus H 4 B plus L-Arg appears to be mutually exclusive. This equilibrium binding model is consistent with clotrimazole slowing iNOSox dimerization in response to L-Arg and H 4 B (Fig. 5), rather than making it completely impossible.
How do bulky imidazoles like clotrimazole prevent iNOS dimerization at the molecular level? That clotrimazole binds to iNOSox monomer is consistent with crystal structures that reveal an open heme pocket in the monomer (57). Computer modeling of a clotrimazole molecule bound to the heme of an iNOSox monomer or dimer predicts extensive collisions between clotrimazole's phenyl moieties and iNOSox ␤-strands 8b and 9b, which are located above the heme. Modeling also shows that the phenyl groups do not extend far enough out of the pocket to directly contact the dimer interface region. 6 Thus, distortion of iNOS ␤-strands 8b and 9b is likely to occur upon binding clotrimazole, which may perturb nearby structural elements that participate in forming the dimer interface.
The function we propose for clotrimazole and miconazole in Fig. 10 differs from their other known effects. For example, clotrimazole and miconazole inhibit cytochrome P450 catalysis (69), notably in the course of oxidation of N -hydroxy-L-Arg (70). This occurs because clotrimazole and miconazole are hydrophobic enough to enter the P450 active site, which allows 6 B. R. Crane and J. T. Tainer, personal communication.  their imidazole nitrogen to bind to the heme iron and prevent substrate access and O 2 binding. A similar mechanism cannot operate in an active iNOS dimer, which by virtue of its bound H 4 B and L-Arg excludes bulky imidazoles from the active site. Clotrimazole and miconazole are also known to inhibit CaM binding to receptor proteins (63,64) and to nNOS at higher concentrations (22). However, our dimerization experiments utilized iNOSox, which does not have a CaM-binding site.
Significantly, clotrimazole also inhibited dimerization of fulllength iNOS in cells that were actively engaged in iNOS protein synthesis, leading to accumulation of iNOS monomers in the clotrimazole-treated cells and a lack of NO synthesis. The effect of clotrimazole on cellular iNOS dimer assembly is striking if one considers that blocking NO synthesis during iNOS expression normally increases the proportion of dimer from 50 to 90% of the total iNOS (54). Our data argue for a heme-based mode of action for clotrimazole in the cells. For example, the clotrimazole-treated cells grew normally and expressed a normal amount of protein, confirming it is not toxic at the concentration used (62,71). Immunochemical and catalytic data indicated that iNOS monomers from clotrimazole-treated cells still contained a normal amount of bound CaM. This is consistent with clotrimazole being unable to dissociate CaM from purified iNOS dimer or inhibit its NO synthesis when added to the assay system (12,72). The high affinity of iNOS toward CaM (12) likely constrains clotrimazole to interact primarily in a heme-dependent manner. Our findings are in apparent conflict with Bogle and Vallance (71), who report to have purified CaM-free iNOS from cells treated with econazole, a structural relative of miconazole. Whether structural differences allow econazole to function through a CaM-based mechanism deserves further study.
Although clotrimazole appears to inhibit cellular iNOS assembly through a heme-based mechanism, the exact point where inhibition occurs is still unclear. For example, clotrimazole could possibly affect events upstream from dimerization that control heme availability or insertion into iNOS monomers. Indeed, iNOS monomers which accumulated in the clotrimazole-treated cells were heme-free, suggesting clotrimazole affects heme incorporation into iNOS. Heme-free iNOS monomers also accumulate in cells actively generating NO, because NO apparently blocks heme insertion (54). In any case, our current data indicate that cellular assembly of active iNOS is remarkably sensitive to clotrimazole, and in principle may be sensitive to any bulky imidazole that can bind to the monomer heme iron.
Clotrimazole will likely inhibit iNOS dimerization in animal cells only when it is present during active iNOS protein synthesis and dimer assembly. This is because clotrimazole cannot bind to the heme iron of an assembled iNOS dimer that has already incorporated H 4 B and L-Arg, and almost all animal cells and cell lines synthesize H 4 B and take in L-Arg (73). Nevertheless, cytokine induction of iNOS is long lasting (74), tissues like lung epithelium continuously express iNOS (75), and cellular assembly of dimer is sufficiently slow (66) to make iNOS susceptible to clotrimazole inhibition. It remains to be seen whether cellular assembly of constitutive NOS isoforms is also blocked by bulky imidazoles, or if structural analogs with specificity toward a particular isoform can be developed.