Abstract
Noncovalent forces, other than hydrophobic interactions, are important determinants of substrate bias exhibited by some cytochromes P450. The CYP2C9 pharmacophore is proposed to include either an anionic group or hydrogen bond donor in addition to its hydrophobic groups. By constructing analogs of benzbromarone, evidence supporting the existence of a 2C9 anion-binding site was revealed. A nonsubstituted phenol analog was determined to have a pKa of 8.4 and a Ki of 414 nM whereas those with dihalogenated benzoyl phenols had pKa values between 4.2 to 5.2 and Ki values as low as 1 nM. The nonhalogenated, nonionizable analog is the poorest binder at 796 nM. The Ki range covers around three orders of magnitude with even the weakest binder being a more potent inhibitor than 2C9 substrate phenytoin. Thus, benzbromarone derivatives represent a class of molecules with the potential to reveal more structural details of the 2C9 active site.
CYP2C9 is one of the major drug oxidizing P450s inside and outside the 2C family (Miners and Birkett, 1998). As such, many studies have attempted to define the determinants of 2C9-binding molecules with analogs of known binders to examine specific chemical properties like tautomeric structure (He et al., 1999), aromaticity (Rao et al., 2000), pKa values (Mancy et al., 1995), and hydrogen bond donors (Jones et al., 1996a) and acceptors (Ekins et al., 2000). Such studies are still important even with the crystal structure of rabbit 2C5 (83% identity) and probable proprietary human 2C9 structures. This is because cytochromes P450 (P4501) can be highly dynamic and nonspecific (Ortiz de Montellano, 1995), thereby obfuscating rationale in pinpointing key active site residues and properties.
Takahashi et al. (1999) reported that the uricosuric agent benzbromarone (Bzbr) (Fig. 1) was one of the most potent 2C9 inhibitors (Ki < 10 nM). The interest in probing 2C9 developed after clinical studies in which patients receiving combination therapy of Bzbr with anticoagulant (S)-warfarin demonstrated increased bleeding (Kudo and Sudo, 1995). Such drug interactions or 2C9 polymorphisms may explain a few cases of hepatic toxicity (van der Klauw et al., 1994) associated with poor clearance of Bzbr and its analogs, such as benziodarone (Fig. 1) and benzarone (Fig. 1) (Hautekeete et al., 1995). A third analog, amiodarone (Fig. 1), diminishes warfarin clearance in vivo (O'Reilly et al., 1987). Warfarin is a 2C9 substrate, although 2C8 and 2C19 appear more important in amiodarone metabolism in vitro (Ohyama et al., 2000).
To clarify the determinants in molecules responsible for optimal binding to 2C9, we have prepared high-affinity (low nanomolar Ki) Bzbr analogs and tested their ability to inhibit 7-OH-warfarin formation in vitro. The results strengthen the arguments that optimal 2C9 binders should: 1) exist as an anion while inside the active site, 2) have a specific lipophilicity, and 3) possess a hydrophobic group(s) that can be positioned between the anionic heteroatom and the site of metabolism (Fig. 2).
Experimental Procedures
Materials. Solvents used were purchased from Fisher Scientific Co. (Pittsburgh, PA), J. T. Baker (Phillipsburg, NJ), or Sigma-Aldrich (St. Louis, MO). Starting materials were obtained from Sigma-Aldrich, except for 2-ethylbenzofuran and diisobutylamine (Lancaster Synthesis, Windham, NH), 2-n-butylbenzofuran (TCI America, Portland, OR), and anhydrous AlCl3 (J. T. Baker). Methylbenzbromarone was kindly provided by Jan L. Wahlstrom of Camitro Corp. (Redwood City, CA). (S)-Warfarin and its protio and deutero metabolites were gifts from Professor W. Trager at the University of Washington. Gentimycin and Amphotericin B were obtained from CellGro (Herndon, VA). ESF 921 insect medium was from Atlanta Biologicals (Norcross, GA). Dilauroylphosphatidylcholine was purchased from Expression Systems (Woodland, CA). Fetal bovine serum was purchased from Avanti Polar Lipids Inc. (Alabaster, AL). Other biochemicals were purchased from Sigma-Aldrich unless otherwise noted.
Protein Expression and Purification. Budded recombinant h2C9 baculovirus generated by the Bac-to-Bac methodology (Invitrogen, Carlsbad, CA) was provided by the A. E. Rettie Laboratory at the University of Washington. Insect cell culture and infection, and 2C9 purification, were carried out as previously reported except for minor differences (Haining et al., 1996). Sf9 cells (Invitrogen) were grown with 10 μg/ml gentimycin and 2.5 μg/ml amphotericin B. Postinfection, 500 μmol of δ-aminolevulinic acid and 100 μmol of ferric citrate were added twice per liter of culture. Expression and purification of human P450 reductase was carried out as previously described (Rock and Jones, 2001), followed by Triton X-100 exchange for 0.1% sodium cholate on hydroxyapatite type I. Human cytochrome b5 (cyt b5) in the pTrcHis B vector (Invitrogen) was expressed in Escherichia coli TOP10 (Invitrogen) in Terrific broth containing 1 mM δ-aminolevulinic acid and 1 mM ferric citrate. Cultures were shaken at 250 rpm for 24 h at 28°C. Purification from 30 ml of pelleted cells was performed with a 5-ml Pharmacia Hi-Trap nickel chelate column (Pharmacia, Piscataway, NJ) as recommended by the manufacturer. Cyt b5 eluted above 40 mM imidazole in the presence of 500 mM NaCl.
Enzyme Incubations. Incubations were carried out in duplicate or triplicate. Dilauroylphosphatidylcholine in chloroform was rotovapped, hydrated in 50 mM KPi buffer overnight at 4°C, and extruded through a 100-nm pore membrane (Avanti Polar Lipids Inc.) to make unilamellar vesicles. CYP2C9 was added to 2 Eq of reductase. After 30 min on ice, 0.2 μg of dilauroylphosphatidylcholine was added per picomole of 2C9. Cyt b5 (1 Eq) finished the enzyme mixture. Stocks of ionizible inhibitors were prepared in water with 1 Eq of KOH and gentle heat. Compounds 2, 3, and 4 were prepared in methanol/water. The methanol volume did not exceed 0.1% of the total incubation volume. Reconstituted 2C9 (20 pmol) was incubated with 1000 U of catalase and (S)-warfarin at concentrations between 2.0 and 40 μM at three concentrations of inhibitor (below, at, or above the Ki) in 1 ml of 50 mM KPi, pH 7.4, in a shaking incubator at 37°C. NADPH (1 μmol) was used to initiate the reaction and 0.6 ml of acetone to quench it after 30 to 40 min. d5-7-OH-warfarin (0.1 nmol) was added as an internal standard before extraction, and derivatization was carried out as before (Bush et al., 1983). The Km agreed with published values at 4.2 ± 2.6 μM (Rettie et al., 1992). Vmax was calculated with 7-OH- and 6-OH-warfarin metabolites using standard curves and was 22.6 ± 0.9 pmol/min/pmol of 2C9. A Thermo Finnigan 3500 (San Jose, CA) was used for gas chromatography/mass spectrometry analysis of the incubations (Bush et al., 1983).
NMR Spectroscopy. Spectra were collected on a 300 MHz Varian mercury NMR spectrometer (Varian, Inc., Palo Alto, CA).
Synthesis of 2-alkyl-3-(4-hydroxybenzoyl)-benzofurans. Friedel-Crafts acylation of benzofurans was carried out with 1 Eq of p-anisoyl chloride and 1.2 Eq of AlCl3 in chlorobenzene at 4°C. The reaction was quenched with ice water and extracted with ether. Following silica chromatography with 10% ether in hexanes (60% yield), methyl ether cleavage was carried out on the yellow oil with 2 to 3 Eq of AlCl3 in chlorobenzene at 80°C. Chromatography with 30% ether in hexanes gave white to light brown solid phenols in 60% yields that were recrystallized with <5% methylene chloride in hexanes.
Halogenation of 2-alkyl-3-(4-hydroxybenzoyl)-benzofurans. Halogenations gave white solids in yields typically over 60% after purification. Bromination utilized 2 Eq of Br2 and triethylamine in MeOH at 4°C (Pearson et al., 1967). I2 and NaI (2 Eq), and 4 Eq of NaOH in water at 65°C gave the diiodinated product after 24 h (Sion, 1985). Sulfuryl chloride (2 Eq) with 0.1% molar equivalent of diisobutylamine in toluene at 60°C gave the dichlorinated product after 2 h (Gnaim and Sheldon, 1995). All compounds were characterized by 1H NMR. 1H NMR (CDCl3, 300 MHz) δ: 0.92–2.92 (alkyl), 6.35 (s, 1H, OH), 7.23–7.50 (m, 4H, ArH), 7.98 (s, 2H, ArH). Gas chromatography/mass spectrometry provided expected molecular ions and fragments of 2, 6, and 7.
pKa Measurement. Data were collected on a diode array HP 845 UV-visible system (Hewlett Packard, Palo Alto, CA). The pKa values of Bzbr and analogs were determined by adding 10 or 100 nmol of the respective compound into 1 ml of 0.1 M KPi prepared at pH values between 2.0 and 12.0. As the phenols were converted to phenolate at higher pH, peaks emerged with λmax between 355 and 360 nm. Absorbance was plotted versus pH and fit to a sigmoidal function using GraphPad Prism v2.0 (GraphPad Software, Inc., San Diego, CA). Results agreed with ACD calculated values (data not shown).
Measurement of Ki Values. Warfarin metabolism data were plotted in hyperbolic form and fit best to the competitive model of inhibition using nonlinear regression with EnzFitter v2.0 (Biosoft, Ferguson, MO). Although the best fit to any simple inhibitor model is for competitive inhibition, the reciprocal plots do not intersect on the y-axis owing to the large error associated with some Ki values (Table 1). This could result from reconstitution conditions or even non-Michaelis-Menton behavior, which is increasingly documented in 2C9 studies (Hutzler et al., 2002). If this phenomenon truly originates from 2C9, a more complex model such as partial noncompetitive or mixed inhibition may prove to be more accurate. More experiments are underway to confirm these hypotheses.
Calculations. The predicted log P, log D, and pKa values for Bzbr and its analogs were obtained from the ACD database through SciFinder Scholar (American Chemical Society, Columbus, OH) (Table 1).
Results and Discussion
Features of 2C9 previously implicated in small molecule binding include hydrophobic regions (Jones et al., 1996b; Ekins et al., 2000; De Groot et al., 2002), H-bond accepting or donating capabilities (Jones et al., 1996a; Ekins et al., 2000), and a somewhat enigmatic anion binding region (Mancy et al., 1995; Jones et al., 1996b; De Groot et al., 2002). After learning that benzbromarone has the highest reported affinity for 2C9, we synthesized a small series of analogs. All of them proved to be potent inhibitors, the weakest having Ki values comparable with warfarin and diclofenac and the strongest surpassing sulfaphenazole (Mancy et al., 1995).
It was thought that with a larger side chain at the 2-position, the resulting increase in lipophilicity would produce a better binder than Bzbr (Ki = 19 nM). This was not the case. Substitution at the 2-position of the benzofuran to a larger n-butyl group (5) decreased affinity (Ki = 91 nM). It is plausible that a decrease in entropy upon binding or increased steric hindrance is responsible for the change in Ki. The n-butyl group may have reduced rotational degrees of freedom when bound, or the active site boundaries of the protein collide with the butyl group. This could further prohibit optimal anion and aromatic interactions proposed to be important for 2C9 (Jones et al., 1996b; Mancy et al., 1996). On the other hand, downsizing the ethyl group to a methyl group (7) enhanced affinity (Ki = 9 nM). Although the affinity change is much less pronounced with the methyl group substitution, it appears that the 2-position indeed plays a role in 2C9 binding via an entropic or steric effect.
Substitution of halogens adjacent to the phenol oxygen greatly increased binding affinity. Our findings indicate that this 3,5-dihalogenation of Bzbr is largely responsible for producing the tightest group of 2C9 binders reported to date. In solution, these compounds exist as phenolate ions at physiological pH with pKa values between 4.1 and 5.2. One of the most significant changes in affinity of Bzbr analogs for 2C9 are seen between the halogenated phenols (1, 5-8) and compound 3 with no halogens, whose pKa is at least three log units higher at 8.3. The Ki value of nonhalogenated 3 decreased roughly 40-fold when brominated to give 7. This validates proposals for the presence of a 2C9 anion-binding site first proposed by Smith and Jones (1992). It was noted in this review that 2C9 substrates diclofenac, naproxen, ibuprofen, and tolbutamide are all ionized at physiological pH. Mancy et al. (1995) corroborated the anion hypothesis with novel anionic tienilic acid derivatives. The primary metabolite of Bzbr is also opposite the acidic phenol, occurring at position 6 (Fig. 1) of the benzofuran heterocycle (De Vries et al., 1993). As with the nonsteroidal anti-inflammatory drugs, this places the anion of the molecule near a proposed cationic site on the protein (Jones et al., 1996b) and the metabolized position adjacent to the heme. In fact, when methylbenzbromarone (4) is used as a substrate, 2C9 O-demethylates this analog to produce the parent Bzbr (T. Lin, J. Wahlstrom, K. Korzekwa, and T. Carlson, manuscript in preparation). This indicates 4 can bind in the opposite orientation, with expected lower affinity, when the substrate lacks ionic character. The existence of distinct binding modes is also documented with warfarin, which is metabolized at the corresponding 7 and 4′ positions by 2C9 (Fig. 1).
The ionic interaction would also appear to place the Bzbr benzoyl ring near specific hydrophobic residues known to impact warfarin (Haining et al., 1999) and nonsteroidal anti-inflammatory drug binding (Tracy et al., 2001). It is possible that this Bzbr aromatic ring π-stacks with Phe114 as proposed in the case of a cyclocoumarol library (Rao et al., 2000). Another study suggests involvement of the nearby region encompassing Phe476 and that the role of Phe114 is more substrate-dependent or that it may help orient Arg108 for an electrostatic interaction (De Groot et al., 2002). Regardless, tienilic acid and the hydrophobic groups of sulfaphenzole, noted to be important for activity, occupy the same region (Mancy et al., 1996; Poli-Scaife et al., 1997). A common anionic site and substrate oxidation site also align in these studies.
Still, it is difficult to ascertain whether a substrate remains anionic in the 2C9 active site to form an ion pair or whether the acid form of a substrate is acting as a hydrogen bond donor. Refined 2C9 homology models (Rao et al., 2000; De Groot et al., 2002) suggest Arg residues play the role of cation. Possibilities include Arg105 (Rao et al., 2000) and Arg108 (Ridderstrom et al., 2000; De Groot et al., 2002) with the latter position slashing diclofenac metabolism 100 times when mutated to alanine. As is being revealed from 2C5 structures, another interaction may be that of anion with solvent. A network of H-bonded active site water is still present with diclofenac bound to 2C5 (Johnson et al., 2002). Hence, more than one line of evidence can demonstrate why anionic Bzbr analogs 5 to 8 are clearly favored over H-bond donor 3 by 2C9. We, therefore, suggest that 1 and 5 to 8 exist as anions even in the lower dielectric of the 2C9 active site. The participation of a small H-bond donating effect is not ruled out as the explanation for the enhanced affinity of 3 over its aryl ether (2).
The fact that nonionizable methylbenzbromarone (4) is a worse binder than Bzbr, but similar to 3, suggests a role for halogen-induced lipophilicity. Methylbenzbromarone is the most lipophilic of the analogs analyzed having a log D7.0 value two units higher than most of the other compounds (Table 1). Although there could be additional factors, one obvious explanation is that lipophilicity appears to compensate for lack of an anionic group. This is not surprising considering hydrophobic residues make up much of the 2C5 active site (Williams et al., 2000). Steric properties probably prevent 2-n-butyl derivatives from benefiting from increased lipophilicity.
In summary, a small series of Bzbr analogs has been used to test fundamental chemical features of 2C9-binding molecules (Fig. 2). It is remarkable that the Bzbr Ki values of the analogs spanned around 3 orders of magnitude, and all proved to be at least as potent as molecules previously considered some of the tightest binders for 2C9. Even the weakest benzarone analog, 2, produced a Ki value below that of phenytoin and diclofenac. Hence, there appears to be an entire class of tight-binding benzofuran derivatives similar to Bzbr, such as benzarone and possibly amiodarone or its metabolites, amenable to structure-function studies with 2C9. Further studies are currently underway to test the effect of other substitutions at both ends of the Bzbr skeleton.
Acknowledgments
We thank Leslie Dickmann for invaluable help with tissue culture.
Footnotes
-
↵1 Abbreviations used are: P450, cytochrome P450; Bzbr, benzbromarone; cytb5, cytochrome b5.
-
Supported by National Institutes of Health Grants GM032165 and ES009122
- Received February 24, 2003.
- Accepted March 19, 2003.
- The American Society for Pharmacology and Experimental Therapeutics