The use of in vitro kinetic data to predict drug metabolic stability in vivo has attracted widespread interest, particularly in preclinical drug development. Most commonly, intrinsic clearance (CL_int) determined in vitro is scaled to an in vivo CL_int and hepatic clearance (CL_H) using human liver microsomes (HLMs) or hepatocytes as the enzyme source (Houston 1994; Iwatsubo et al., 1997; Obach et al., 1997; McGinnity et al., 2004). Although in vitro models and scaling approaches have been progressively refined in recent years, in vitro-in vivo extrapolation (IV-IVE) typically results in underprediction of in vivo CL_int and CL_H (and hence hepatic extraction ratio, E_H). With HLMs as the enzyme source, in vivo CL_int is underpredicted by approximately an order of magnitude for drugs eliminated by either cytochrome P450- or UDP-glucuronosyltransferase-catalyzed biotransformation (Boase and Miners, 2002; Ito and Houston, 2005). Prediction bias is improved using hepatocytes (Soars et al., 2002; Riley et al., 2005), but there remains an approximate 4-fold underprediction of the in vivo CL_int (Miners et al., 2006; Brown et al., 2007).

The reason for the underprediction of in vivo CL_int (and CL_H) based on human liver microsomal kinetic data remains unclear. Non-specific microsomal binding, variability in experimental conditions and physiological scaling factors, and inappropriate kinetic modeling all affect the reliability of in vitro-in vivo extrapolation (Miners et al., 2006), but significant underprediction remains even when these factors are taken into account. Similarly, involvement of hepatic uptake transporters potentially accounts for the improved predictivity of kinetic data generated with hepatocytes, although recent data suggest that hepatic uptake is not rate-limiting in the clearance of lipophilic drugs (Doherty et al., 2006; Halifax and Houston, 2006).

The addition of bovine serum albumin (BSA) to incubations of HLMs has been reported to increase the rate of metabolism of CYP2C9 substrates (Ludden et al., 1997; Carlile et al., 1999; Tang et al., 2002; Zhou et al., 2004) and microsomal CL_int values of UGT2B7 substrates (Uchaipichat et al., 2006; Rowland et al., 2007). After accounting for binding to albumin, K_m values for phenytoin (PHY) p-hydroxylation, tolbutamide tolymethylhydroxylation, and zidovudine glucuronidation by HLMs in incubations supplemented with 2% BSA were 6- to 20-fold lower compared with K_m values generated in the absence of BSA. Maximal velocity was generally unaffected by BSA, although Carlile et al. (1999) reported a 2-fold reduction in the V_max for tolbutamide tolymethylhydroxylation.

Consistent with this observation, long-chain unsaturated fatty acids, including oleic (C18:1n-9), linoleic (C18:2n-6), and arachidonic (C20:4n-6) acids, are known potent inhibitors of UGT2B7 (Tsoutsikos et al., 2004). We have demonstrated recently that BSA and essentially fatty acid-free human serum albumin (HSA-FAF) reduce the K_m for zidovudine glucuronidation by HLMs and recombinant UGT2B7 (expressed in HEK293 cells) by sequestration of inhibitory long-chain unsaturated fatty acids released from membranes during the course of an incubation (Rowland et al., 2007). CYP2C9 is known to catalyze the oxidation of C18:2n-6 and C20:4n-6 (Daikh et al., 1994; Rifkind et al., 1995; Draper and Hammock, 2000), and, by analogy with UGT2B7, it might be speculated that the higher K_m values for microsomal PHY and tolbutamide hydroxylation observed in the absence of BSA arise from inhibition of CYP2C9 activity by long-chain unsaturated fatty acids.

Thus, the aims of the present study were to characterize the effects of BSA, HSA, and HSA-FAF on the kinetics of PHY hydroxylation by HLMs and CYP2C9; to demonstrate inhibition of PHY hydroxylation by long-chain unsaturated fatty acids and reversal of fatty acid inhibition by albumin; and to show that the presence of BSA in incubations abolishes the metabolism of long-chain fatty acids released from HLMs during the course of an incubation. The results are consistent with the hypothesis that the “albumin effect” as it applies to CYP2C9 substrates results from sequestration of inhibitory fatty acids and suggest that under appropriate experimental conditions, HLMs may be a suitable enzyme source for the generation of kinetic data for in vitro-in vivo extrapolation.

Materials and Methods

PHY, 5-(4′-hydroxyphenyl)-5-phenylhydantoin [hydroxy-phenytoin (HPPH)], 5-ethyl-5-p-tolylbarbituric acid (TBA), butylated hydroxytoluene, bovine serum albumin (“crude” BSA, product no. A7906), human serum albumin (crude HSA; product no. A9511), HSA-FAF (product no. A1887), heptadecanoic acid (C17:1), linoleic acid (C18:2n-6), oleic acid (C18:1n-9), and arachidonic acid (C20:4n-6) were purchased from Sigma-Aldrich (St. Louis, MO). MS-grade ethyl acetate and acetonitrile were used for tandem mass spectrometry experiments. All other reagents and solvents were of analytical reagent grade.

HPLC was performed using an Agilent 1100 series instrument (Agilent Technologies, Palo Alto, CA) comprising a quaternary solvent delivery module, autoinjector, and UV-Vis detector. The column was maintained at 25°C. Mass spectral data were collected on an MDS Sciex 3200 Q-trap triple quadrupole mass spectrometer (Applied Biosystems, Forster City, CA).

Human Liver Microsomes and Expression of CYP2C9. Pooled human liver microsomes were prepared by mixing equal protein amounts from five human livers (H07, H10, H12, H29, and H40), obtained from the human liver “bank” of the Department of Clinical Pharmacology of Flinders University. Approval for the use of human liver tissue in xenobiotic metabolism studies was obtained from both the Clinical Investigation Committee of Flinders Medical Centre and from the donors' next-of-kin. HLMs were prepared by differential centrifugation, as described by Bowalgaha et al. (2005).

CYP2C9 and OxR cDNAs. N-terminal modifications previously shown to promote high levels of bacterial expression of human P450s were made to the wild-type CYP2C9 cDNA as previously described (Boye et al., 2004). The CYP2C9 cDNA was modified for bacterial expression by replacing the second codon with GCT (codes for Ala), deleting codons 3 to 20 and adjusting codons 21 through 26 for bacterial codon bias. The 1803-bp product (3′-untranslated region included) was ligated into the pCWori(+) bacterial expression plasmid. pCW-CYP2C9 was transformed into DH5α Escherichia coli cells and colonies screened for the correct plasmid by restriction enzyme analysis. Plasmid DNA was purified and confirmed on both strands by sequencing (ABI Prism 3100). The cDNA coding for OxR consisted of the OmpA signal sequence fused upstream of the full-length native rat OxR sequence (Boye et al., 2004). The rOxR expression construct was generated using the bacterial plasmid, pACYC.

Heterologous Expression of pCW-CYP2C9 and pACYC-rOxR. pCW-CYP2C9 and pACYC-rOxR were separately transformed into DH5α E. coli cells. Cells were cultured, and E. coli membrane fractions were separated using the method of Boye et al. (2004). CYP2C9 and OxR membrane fractions were mixed on ice to provide a P450/OxR ratio of 1:5. A CYP2C9/OxR ratio of 1:5 has been shown previously to result in optimal CYP2C9 activity (Boye et al., 2004). CYP2C9 expression was determined by carbon monoxide difference spectroscopy, whereas the rate of cytochrome c reduction was used as a measure of the OxR activity of membrane fractions.

PHY Hydroxylation Assay. Incubations, in a total volume of 500 μl, contained phosphate buffer (0.1 M, pH 7.4), HLMs (0.375 mg), albumin (0 or 2%), and PHY (0.5–100 μM). Following a 5-min preincubation, reactions were initiated by the addition of an NADPH-regenerating system (1 mM NADP, 10 mM glucose 6-phosphate, 2 IU/ml glucose 6-phosphate dehydrogenase, and 5 mM MgCl₂). Incubations were performed at 37°C in a shaking water bath for 45 min. Reactions were terminated by the addition of perchloric acid (5 μl, 70% v/v) and cooling on ice. After addition of TBA (10 μM), the assay internal standard, incubation mixtures were transferred to 15-ml culture tubes that contained sodium chloride (0.75 g) and extracted twice with 2 ml of ethyl acetate. The combined extracts were evaporated to dryness under a stream of N₂. Metabolite and internal standard extraction efficiencies, determined by comparison with a standard curve from buffer, were greater than 95%. Residues were reconstituted in 0.1 ml of the HPLC mobile phase, and a 20-μl aliquot was injected directly onto the HPLC column.

For reactions performed using recombinant CYP2C9, incubation mixtures contained E. coli-expressed CYP2C9 (10 pmol CYP/incubation; 0.02 mg/ml E. coli membrane protein) and OxR (50 pmol OxR/incubation; 0.57 mg/ml E. coli membrane protein) membranes in place of HLM protein, and the incubation time was increased to 90 min. Under the reaction conditions, employed PHY hydroxylation was linear with respect to incubation time and protein concentration to 90 min and 1 mg/ml, respectively, for HLMs, and 120 min and 20 pmol CYP/incubation, respectively, for recombinant CYP2C9. PHY hydroxylation by untransformed E. coli membranes was not detectable.

Quantification of HPPH Formation. HPPH formation was determined by HPLC. Analytes were separated on a Synergi Hydro RP analytical column (150 × 3.0 mm, 4 μm; Phenomenex, Torrance, CA) using a mobile phase comprising 70% 2 mM triethylamine (pH adjusted to 2.5 with perchloric acid) and 30% acetonitrile, at a flow rate of 0.6 ml/min. Column eluant was monitored by UV absorbance at 214 nm. Retention times for HPPH, TBA, and PHY were 4.3, 6.1, and 7.8 min, respectively. HPPH formation was quantified by comparison of peak area ratios with those of a HPPH standard curve prepared over the concentration range 0.05 to 5 μM. Overall within-day assay reproducibility was assessed by measuring HPPH formation in eight separate incubations of the same batch of pooled HLMs. Coefficients of variation were 4.2 and 2.6% for PHY concentrations of 1 and 50 μM, respectively. The lower limit of quantitation (assessed as 3 times background noise) for HPPH was 0.01 μM.

Quantification of the Fatty Acid Content of E. coli Membranes. The fatty acid content of E. coli membranes (1 mg/ml) was determined using a modification of the gas-liquid chromatographic method of Folch et al. (1956), as described by Rowland et al. (2007).

Binding of PHY to Albumin, HLMs, and E. coli Membranes. Binding of PHY to HLMs, E. coli membranes, and mixtures of albumin (BSA, HSA, or HSA-FAF; 2%) with each enzyme source was measured by equilibrium dialysis according to the method of McLure et al. (2000). Binding measurements were performed using a Dianorm equilibrium dialysis apparatus that comprised Teflon dialysis cells (capacity of 1.2 ml/side) separated into two compartments with a Sigma-Aldrich dialysis membrane (molecular mass cut off, 12 kDa). One side of the dialysis cell was loaded with 1 ml of PHY (1–50 μM) in phosphate buffer (0.1 M, pH 7.4). The other compartment was loaded with 1 ml of either HLMs in phosphate buffer (0.1 M, pH 7.4), E. coli membranes in phosphate buffer (0.1 M, pH 7.4) or a combination of albumin with each enzyme source in phosphate buffer (0.1 M, pH 7.4). The dialysis cell assembly was immersed in a water bath maintained at 37°C and rotated at 12 rpm for 4 h. Control experiments were also performed with phosphate buffer or albumin on both sides of the dialysis cells at low and high concentrations of PHY to ensure that equilibrium was attained. A 200-μl aliquot was collected from each compartment, treated with ice-cold methanol containing 8% glacial acetic acid (200 μl), and cooled on ice. Samples were subsequently centrifuged at 4000g for 10 min at 10°C, and an aliquot of the supernatant fraction (5 μl) was analyzed by HPLC.

Quantification of PHY Binding. Separation of PHY was achieved on a Waters NovaPak C₁₈ analytical column (3.9 × 150 mm, 4 μm; Waters, Milford, MA) using a mobile phase comprising 60% 2 mM triethylamine (pH adjusted to 2.5 with perchloric acid) and 40% acetonitrile, at a flow rate of 1 ml/min. Column eluant was monitored at 214 nm. The retention time for PHY under these conditions was 2.8 min. PHY concentrations in dialysis samples were determined by comparison of peak areas with those of a PHY standard curve prepared over the concentration range 0.25 to 50 μM. Within-day assay variability was assessed by measuring PHY (1 and 50 μM; n = 5 for each concentration) in samples containing phosphate buffer (0.1 M, pH 7.4) or BSA in phosphate buffer (0.1 M, pH 7.4). Coefficients of variation were less than 3% in all cases.

Mass Spectrometric Investigation of Fatty Acid Binding to BSA. Incubation samples (500 μl), containing phosphate buffer (0.1 M, pH 7.4), HLMs (0.375 mg), and BSA (0 or 2%) were incubated using the conditions described for PHY hydroxylation (see above). Samples were acidified with glacial acetic acid then extracted with ethyl acetate (2 × 0.3 ml) containing butylated hydroxytoluene (20 mg/l) by vortex mixing. Following centrifugation (13,000g, 4 min), the combined organic extract was evaporated to dryness under vacuum, and the residue was redissolved in 1:1 acetonitrile-water (200 μl). Mass spectra were obtained by direct infusion (10 μl/min) into the electrospray ion source operated in negative ionization mode. The ion spray voltage was -3.5 kV, the declustering potential was -75 V, and the ion source temperature was set at 300°C. Curtain gas and ion source gas flow were set at 10 and 30 l/min, respectively. MS scans were collected in enhanced mode with a scan rate of 1000 amu/s. Enhanced product ion scans were used to confirm peak identity with nitrogen as the collision gas.

Data Analysis. Kinetic data are presented as the mean ± S.D. of four experiments with pooled HLMs or recombinant CYP2C9. Kinetic constants for PHY hydroxylation by HLMs and recombinant CYP2C9 in the presence and absence of albumin were generated by fitting experimental data to the Michaelis-Menten equation. Fitting was performed with EnzFitter (version 2.0.18.0; Biosoft, Cambridge, UK) based on the unbound substrate concentration present in incubations.

CL_int values for PHY hydroxylation by HLMs and recombinant CYP2C9 were determined as V_max/K_m. Microsomal CL_int was converted to whole-liver CL_int using scaling factors that correct for microsomal yield and liver weight using the equation where V_max and K_m are the maximal velocity and Michaelis constant for the microsomal reaction, respectively, MPPGL is the mass of microsomes per gram of human liver tissue (taken as 38 mg/g), and LW is the average weight of a human liver (1500 g). The result was multiplied by 0.00006 to convert microliters per minute to liters per hour. The MPPGL value of 38 mg/g corresponds to the geometric mean of the microsomal yield reported by Hakooz et al. (2006) and is in agreement with the mean MPPGL for the preparation of HLMs from livers in the Flinders Medical Centre human liver bank (J. O. Miners, unpublished data).

Similarly, CL_int determined for recombinant CYP2C9 was converted to whole-liver CL_int using scaling factors that correct for P450 abundance per milligram of microsomal tissue, microsomal yield, and liver weight using the equation (Proctor et al., 2004) where V_max and K_m are the maximal velocity and the Michaelis constant for the recombinant CYP2C9 catalyzed reaction, respectively, CYP_abundance is the abundance of CYP2C9 in 1 g of human liver microsomes (73 pmol; Rostami-Hodjegan and Tucker, 2007), and MPPGL and LW are as described previously. Again, the result was multiplied by 0.00006 to express CL_int in liters per hour.

In vivo CL_int values for PHY hydroxylation were calculated from area under the plasma concentration-time curve (AUC) data from four studies (Svendsen et al., 1975; Gugler et al., 1976; Dickinson et al., 1985; Tassaneeyakul et al., 1992) using the equation where f_m is the fraction of the PHY dose converted to HPPH (0.79; Dickinson et al., 1985), dose is the amount of drug administered (100–300 mg as the sodium salt), f_u,b is the fraction unbound in blood, calculated as the fraction unbound in plasma divided by the blood/plasma ratio (0.61; Kurata and Wilkinson 1974), and AUC is the area under the plasma concentration time curve. The mean in vivo CL_int from the four studies was determined to be 14.85 l/h.

Statistical analysis (univariate general linear model, with Tukey post hoc analysis) was performed using SPSS for Windows, release 12.0.1, 2003 (SPSS Inc., Chicago, IL). Values of p less than 0.05 were considered significant.

Results

Binding of Phenytoin to HLMs, E. coli Membranes, and Albumin. We have demonstrated that the sequestration of long-chain unsaturated fatty acids released during the course of a microsomal incubation plateaus at BSA and HSA-FAF concentrations ≥ 1% (Rowland et al., 2007) and previous reports of the effect of BSA on the kinetics of CYP2C9 substrates have typically employed a BSA concentration of 2% (Ludden et al., 1997; Carlile et al., 1999; Tang et al., 2002). Thus, an albumin (BSA, HSA, HSA-FAF) concentration of 2% was selected for investigation in the present study. The binding of PHY to HLMs, E. coli membranes, and albumin was calculated as the concentration of drug in the buffer compartment divided by the concentration of drug in the protein compartment and expressed as the fraction unbound in incubations (f_u,inc). Binding of PHY to pooled HLMs and E. coli membranes was negligible (<5%) across the PHY concentration range investigated (1–50 μM), consistent with previous microsomal binding data from this laboratory (McLure et al., 2000). PHY bound to all forms of albumin. PHY binding was independent of concentration in the range 1 to 50 μM but varied with albumin form. The mean f_u,inc values for PHY binding to mixtures comprising HLMs (or E. coli membranes) with BSA, HSA, and HSA-FAF (all 2%) were 0.27, 0.51, and 0.46, respectively. Where incubations contained albumin, the concentration of PHY present in reaction mixtures was corrected for binding in the calculation of kinetic parameters.

Effect of Albumin on Phenytoin Hydroxylation by HLMs and Recombinant CYP2C9. Kinetic data for HPPH formation by human liver microsomal- and E. coli-expressed CYP2C9, in the presence and absence of albumin, were well modeled by the Michaelis-Menten equation. Kinetic parameters for PHY hydroxylation by pooled HLMs and recombinant CYP2C9, in the presence and absence of BSA, HSA, and HSA-FAF, are shown in Table 1. Respective K_m and V_max values for PHY hydroxylation by HLMs and recombinant CYP2C9 in the absence of albumin were 20.8 ± 1.5 μM, 17.8 ± 0.3 pmol/min/mg, 14.4 ± 1.3 μM, and 0.230 ± 0.006 pmol/min/pmol CYP. These data are similar to previously reported kinetic parameters for PHY hydroxylation by HLMs (Ludden et al., 1997; Carlile et al., 1999; Tang et al., 2002).

BSA (2%) and HSA-FAF (2%) increased the rate of PHY hydroxylation by both HLMs and recombinant CYP2C9, predominantly by decreasing the K_m for this pathway, with a minor, but significant, effect on V_max (Fig. 1; Table 1). Mean K_m values for PHY hydroxylation by HLMs and recombinant CYP2C9 were close in value (3.6–4.7 μM) for incubations conducted in the presence of BSA (2%) and HSA-FAF (2%). In contrast, crude HSA decreased the rate of PHY hydroxylation by HLMs by increasing the K_m for this pathway without an effect on V_max (Table 1). Crude HSA was without effect on the kinetic parameters for PHY hydroxylation by recombinant CYP2C9.

Fatty Acid Content of E. coli Membrane. The total concentration of C16 to C20 fatty acids released by hydrolysis of a suspension of E. coli membrane (1 mg/ml) was 61 μM, of which unsaturated fatty acids made up 42% (Table 2). Of the unsaturated fatty acids, heptadecanoic acid (C17:1), vaccenic acid (C18:1n-7), and linolenic acid (C18:3n-6) were observed in the highest concentrations. The fatty acid content of E. coli membrane is approximately 5-fold lower than the fatty acid content of HLMs on a w/w basis (Rowland et al., 2007).

Detection of Endogenous Fatty Acids and Their Hydroxylated Metabolites Released by Enzyme Sources in the Presence and Absence of BSA. Tandem mass spectrometry product ion scans were used to confirm the identity of peaks at each m/z corresponding to the parent fatty acids or their oxygenated metabolites (Kerwin and Torvik, 1996). Figure 2 shows the MS scans from incubations of pooled HLMs with and without the NADPH-generating system and/or BSA. Both saturated (C16:0 and C18:0) and unsaturated (C18:1n-9, C18: 2n-6, and C20:4n-6) long-chain fatty acids were detected in incubations of HLMs without generating system and BSA (Fig. 2A). There was a greater than 50% reduction in the content of C18:0, C18:1n-9, C18:2n-6, and C20:4n-6 for incubations of HLMs with the NADPH-generating system following incubation, and monohydroxylated metabolites of C16:0 and C18:1n-9 could be detected at m/z 271.3 and 297.3, respectively, with a dihydroxy metabolite of C18:2n-6 observed at m/z 311.2 (Fig. 2B). Interestingly, arachidonic acid was not observed following incubation with NADPH, presumably due to complete biotransformation. Unexpectedly, however, oxygenated derivatives of arachidonic acid were also not detected, possibly as a result of fragmentation under the MS conditions employed. When BSA was included with the NADPH-generating system in the incubation mixture, the formation of hydroxylated metabolites of long-chain fatty acids was almost abolished (Fig. 2C). It should be noted that the presence of fatty acids arises from “stripping off” fatty acids bound to BSA during extraction with ethyl acetate. Broadly similar observations where made for incubations with recombinant CYP2C9 in the absence and presence of the NADPH-generating system and/or BSA (data not shown).

Inhibition of Phenytoin Hydroxylation by Fatty Acids in the Presence and Absence of Albumin. Inhibition of PHY hydroxylation, at an added concentration of 5 or 15 μM (the respective approximate K_m values in the presence and absence of albumin), by heptadecanoic acid (C17:1), oleic acid (C18:1n-9), linoleic acid (C18:2n-6), and arachidonic acid (C20:4n-6) was measured in the presence and absence of albumin using recombinant CYP2C9 as the enzyme source. When added at a concentration corresponding to 1/20 of the known content in either HLMs (C18:1n-9, C18:2n-6, C20:4n-6) or E. coli membranes (C17:1), each fatty acid inhibited PHY hydroxylation. The magnitude of inhibition increased with increasing degree of unsaturation; heptadecanoic (C17:1; 3 μM), oleic (C18:1n-9; 3 μM), linoleic (C18:2n-6; 3 μM), and arachidonic (C20:4n-6; 1.5 μM) acids inhibited PHY hydroxylation by 13, 22, 31, and 61%, respectively. In contrast, rates of PHY hydroxylation differed from the control values by <4% for incubations performed in the presence of BSA (2%).

The effect of a combination of fatty acids (comprising 3 μM C18:1n-9, 3 μM C18:2n-6, and 1.5 μM C20:4n-6) on the kinetics of PHY hydroxylation by recombinant CYP2C9 was characterized in the presence and absence of BSA (2%). In the absence of BSA (2%), the combination of fatty acids caused a 2-fold increase in the K_m for PHY hydroxylation, from 14.4 to 28.2 μM, without a significant effect on V_max (0.246 versus 0.225 pmol/min/pmol CYP) (Fig. 3). In the presence of BSA (2%), the combination of fatty acids had no effect on the kinetics of PHY hydroxylation; respective K_m and V_max values in the presence and absence of the fatty acid mixture were 4.0 and 3.6 μM and 0.242 and 0.246 pmol/min/pmol CYP.

Kinetics of Arachidonic and Heptadecanoic Acid Inhibition of PHY Hydroxylation by HLMs and Recombinant CYP2C9. Arachidonic and heptadecanoic acids inhibited PHY hydroxylation by human liver microsomal- and E. coli-expressed CYP2C9 in a competitive manner. The K_i values for arachidonic acid inhibition of HLM- and E. coli-expressed CYP2C9 were 3.8 and 1.6 μM (Fig. 4), respectively, whereas the K_i values for heptadecanoic acid inhibition of HLM- and E. coli-expressed CYP2C9 were 21.9 and 25.7 μM, respectively.

Prediction of Phenytoin Clearance Based on Kinetic Values Determined in the Presence and Absence of Albumin. Predicted whole-liver intrinsic clearance (CL_int.liver) values for PHY are shown in Table 1 for human liver microsomal and recombinant CYP2C9 kinetic data generated in the absence and presence of albumin. In the absence of albumin, the predicted CL_int.liver values derived from incubations with HLMs and recombinant CYP2C9 were 2.93 ± 0.16 and 4.01 ± 0.24 l/h, respectively. The use of kinetic data generated in the presence of BSA and HSA-FAF resulted in higher predicted CL_int.liver values generated with HLMs (14.85 ± 0.87 and 15.39 ± 0.32 l/h, respectively) and recombinant CYP2C9 (17.06 ± 0.75 and 15.36 ± 0.43 l/h, respectively) as the enzyme sources. The use of microsomal and recombinant CYP2C9 kinetic data obtained in the presence of BSA and HSA-FAF predict CL_int.liver values that agreed well with the known in vivo mean CL_int for PHY of 14.85 l/h (data from Svendsen et al., 1975; Gugler et al., 1976; Dickinson et al., 1985; Tassaneeyakul et al., 1992).

Discussion

The addition of BSA (2%) to incubations has previously been demonstrated to increase the rate of PHY hydroxylation by HLMs (Ludden et al., 1997; Carlile et al., 1999; Tang et al., 2002; Zhou et al., 2004). More recently, we reported that BSA and HSA-FAF, but not crude HSA, decreased the K_m values for zidovudine glucuronidation by approximately an order of magnitude, with both HLMs and recombinant UGT2B7 (expressed in HEK293 cells) as the enzyme source (Rowland et al., 2007). The effect of BSA and HSA-FAF was concentration dependent between 0.05 to 1% but plateaued from 1 to 4%. This work further showed that the mechanism of the albumin effect involved sequestration of long-chain unsaturated fatty acids (e.g., C18:1n-9, C18:2n-6, and C20:4n-6) that were released from the microsomal membrane during the course of an incubation and acted as potent competitive inhibitors of UGT2B7 (Tsoutsikos et al., 2004; Rowland et al., 2007). The results of the present work similarly demonstrate that the decrease in the K_m for HPPH formation by both HLM- and E. coli-expressed CYP2C9 in the presence of BSA and HSA-FAF arises from sequestration of inhibitory long-chain unsaturated fatty acids present in incubations. Importantly, kinetic parameters generated for HPPH formation by HLMs and recombinant CYP2C9 in the presence of BSA and HSA-FAF accurately predicted PHY hepatic CL_int and CL_H in vivo, indicating that this experimental system may be of value for IV-IVE for drugs cleared by cytochromes P450 and UDP-glucuronosyltransferase.

As in the previous investigation of zidovudine glucuronidation BSA and HSA-FAF, but not crude HSA, decreased the K_m for HPPH formation by HLMs and recombinant CYP2C9, from 20.8 to 4.7 (HLMs) and 14.4 to 3.6–4.1 (CYP2C9) μM based on the unbound substrate concentration present in incubations. There was also a small but statistically significant increase in V_max. The addition of crude HSA to incubations either increased (HLMs) or had no effect on the K_m for PHY hydroxylation. Crude HSA has significant amounts of bound long-chain unsaturated fatty acids that limit further binding and may allow desorption of fatty acids into the incubation medium (Rowland et al., 2007).

C18:1n-9, C18:2n-6, and C20:4n-6 are among the most abundant long-chain unsaturated fatty acids present (as phospholipids) in HLMs (Waskell et al., 1982; Rowland et al., 2007). In contrast, the present work shows that heptadecanoic (C17:1) accounts for approximately 60% of the long-chain unsaturated fatty acid content of E. coli membranes, with lower proportions of C18 and C20 unsaturated fatty acids. Consistent with the postulated inhibitory effect of fatty acids on CYP2C9 activity, addition of C17:1, C18:1n-9, C18:2n-6, and C20: 4n-6 to incubations of E. coli-expressed CYP2C9 at a concentration corresponding to approximately 1/20 of the known content of these fatty acids present in either HLMs or E. coli membranes, decreased PHY hydroxylation activity due to an increase in K_m without a change in V_max, indicative of a competitive mechanism. Further experiments with C17:1 and C20:4n-6 confirmed that the individual fatty acids are competitive inhibitors of CYP2C9. K_i values for C20:4n-6 inhibition of PHY hydroxylation by HLMs (3.8 μM) and CYP2C9 (1.6 μM) were lower than the K_i values for C17:1 (21.9–25.7 μM). It should be noted that these values are almost certainly overestimates since inhibition of PHY hydroxylation by long-chain unsaturated fatty acids “released” by the microsomal and E. coli membranes during the course of an incubation will occur simultaneously. In all cases, addition of BSA reversed the inhibition. The lower K_m value observed for incubations of recombinant CYP2C9 compared with HLMs presumably reflects the lower content of inhibitory fatty acids in the expression system.

Mass spectral studies provided direct confirmatory evidence for a role of endogenous long-chain unsaturated fatty acids in the inhibition of PHY hydroxylation by HLMs. Several P450 enzymes, including CYP2C9, are known to oxidize long-chain unsaturated fatty acids (Daikh et al., 1994; Rifkind et al., 1995; Draper and Hammock, 2000). In the absence of the NADPH-generating system, molecular ions with m/z values corresponding to deprotonated C18:1n-9, C18:2n-6, and C20:4n-6 were detected in incubations of HLMs without exogenous substrate. Abundance of these ions increased during the course of an incubation (data not shown). Addition of the NADPH-generating system to incubation mixtures decreased the abundance of ions corresponding to the parent fatty acids, and, in general, there was a corresponding increase in the content of mono- or dihydroxylated metabolites. Further addition of BSA to incubations essentially abolished formation of the hydroxylated metabolites presumably since the sequestered long-chain unsaturated fatty acids were not available for biotransformation. Although the semiquantitative nature of these experiments is acknowledged, the data are nevertheless consistent with the inhibition studies and with the previous observation that addition of BSA to incubations of HLMs supplemented with UDP-glucuronic acid prevents the glucuronidation of endogenous long-chain unsaturated fatty acids released during an incubation (Rowland et al., 2007).

AUC data from four studies (Svendsen et al., 1975; Gugler et al., 1976; Dickinson et al., 1985; Tassaneeyakul et al., 1992), which employed single doses of PHY sodium (100–300 mg), were used to calculate in vivo CL_int. The mean CL_int from the four studies was 14.85 l/h. Data from multiple dose studies was excluded to avoid the effect of autoinduction, as were data from studies that employed doses of ≥400 mg to ensure, as much as possible, that dosage was within the region of linear kinetics (Carlile et al., 1999). CL_int values generated with HLMs and recombinant CYP2C9 in the absence of albumin underpredicted CL_int.liver 4- to 5-fold. In contrast, kinetic data generated with HLMs and recombinant CYP2C9 in the presence of BSA or HSA-FAF predicted mean CL_int.liver within ± 15% (Table 1), with exact correspondence of the extrapolated human liver microsomal CL_int obtained from incubations containing BSA. As alluded to above, however, the saturable kinetics of PHY in vivo may impact on the accuracy of IV-IVE for this substrate.

The K_m for PHY hydroxylation by human hepatocytes appears not to have been determined. However, as with K_m and K_i values generated for UGT2B7 substrates and inhibitors (Uchaipichat et al., 2006; Rowland et al., 2006, 2007), available data indicate that the K_m values for PHY hydroxylation obtained with HLMs and recombinant CYP2C9 in the presence of BSA and HSA-FAF (namely 3.6–4.7 μM) reflect hepatocellular K_m. The mean in vivo unbound K_m for PHY hydroxylation is 3.4 μM (discussed in Carlile et al., 1999).

The effect of BSA (2%) on CL_int.liver prediction for PHY has been investigated previously by Carlile et al. (1999). These authors reported that the presence of BSA in incubations improved predictivity, but mean CL_int.liver was underestimated by approximately 50%. It is noteworthy that V_max values reported by Carlile et al. (1999) are somewhat lower than those determined here and by Ludden et al. (1997). The effects of BSA and HSA-FAF on the kinetics of tolbutamide metabolism, another CYP2C9 substrate (Miners et al., 1988), have also been investigated. Carlile et al. (1999) reported that BSA (2%) increased the microsomal CL_int almost 6-fold, arising from a 20-fold reduction in mean K_m but halving of V_max. The higher microsomal CL_int generated in the presence of BSA tended to overestimate CL_int.liver. However, the study of Carlile et al. (1999) included a very limited number of substrate concentrations, especially at concentrations below K_m. In contrast, Wang et al. (2002) reported that the inclusion of HSA-FAF (0.5%) in microsomal incubations only halved the K_m for tolbutamide hydroxylation (without affecting V_max), with remaining underprediction of CL_int.liver. The relatively low HSA-FAF concentration employed by Wang et al. (2002) may be suboptimal for complete fatty acid sequestration. Further work with tolbutamide under standardized experimental conditions is warranted to resolve these differences. In this regard, studies are underway with a range of P450 substrates to determine the universality of the albumin effect and to identify alternative fatty acid sequestering agents to albumin, given the current requirement to measure unbound fraction in incubations.

Metabolism by CYP2C9 is the primary clearance mechanism for a large number of clinically used drugs (Miners and Birkett, 1998). The kinetics of most CYP2C9 substrates have been characterized in vitro, with HLMs and/or recombinant CYP2C9 as the enzyme source. Based on the effects of BSA and HSA-FAF on PHY and tolbutamide hydroxylation, published K_m values for CYP2C9 substrates are likely to be overestimations. Taken together with the recently published data for UGT2B7 substrates and inhibitors (Uchaipichat et al., 2006; Rowland et al., 2006, 2007), the present study further suggests that the addition of BSA or HSA-FAF to incubations of HLMs and recombinant expression systems is likely to decrease the K_m values for substrates of any P450 or UGT enzyme inhibited by unsaturated long-chain fatty acids. In turn, wherever metabolism involves an enzyme inhibitable by unsaturated long-chain fatty acids, it would be expected that prediction of in vivo CL_int (or CL_H) is likely to be significantly improved when kinetic data are generated in the presence of BSA or HSA-FAF.