Abstract
Cytochrome P450-dependent desaturation of the anticonvulsant drug valproic acid (VPA) results in formation of the hepatotoxin, 4-ene-VPA. Polytherapy with other anticonvulsants which are known P450 inducers increases the flux through this bioactivation pathway. The aim of the present study was to identify specific, inducible forms of human liver P450 which catalyze terminal desaturation of VPA. Oxidized VPA metabolites formed in an NADPH-dependent manner by human liver microsomes were quantified by gas-chromatography/mass spectrometry.In vitro reaction conditions were established which reflected the product profile found in vivo. Production of 4-ene-VPA by microsomal P450s could be inhibited significantly by coumarin, sulfaphenazole and diethyldithiocarbamate, but not by triacetyloleandomycin, quinidine or furafylline. Recombinant human CYP3A4 did not form detectable levels of 4-ene-VPA and, of nine additional isoforms expressed in either HepG2 or lymphoblastoid cells which were screened for VPA desaturase activity, only CYP2C9 and CYP2A6 formed detectable levels of metabolite. Consequently, CYP3A4, the isoform usually associated with induction by anticonvulsants cannot be responsible for the enhanced 4-ene-VPA formation that occurs during polytherapy. Instead, enhanced activity in vivo likely results from induction of CYP2A6 and/or CYP2C9.
Valproic acid is a widely used, effective, anticonvulsant agent which can cause a rare, but serious hepatotoxicity (Stephens and Levy, 1992). Between 1978 and 1993, more than 100 fatalities had been attributed to this idiosyncratic reaction, 70 of them in the United States alone (Koniget al., 1994; Bryant and Dreifuss, 1996). Two retrospective studies have demonstrated that patients younger than 2 years of age who receive anticonvulsant polytherapy are at greatest risk of developing this complication (Dreifuss et al., 1987; Bryant and Dreifuss, 1996). Although the precise mechanism underlying VPA-associated liver damage remains to be established, mitochondrial dysfunction is a recognized pathology which may depend on metabolism to generate a reactive metabolite(s) of the drug (Baillie, 1988, 1992).
Oxidative metabolism of VPA by cytochrome P450 is a minor metabolic pathway for this drug, but one which generates the hepatotoxic terminal olefin, 4-ene-VPA (Rettie et al., 1987, 1988). The possibility that VPA-associated hepatic fatality could be mediated by this unsaturated metabolite was first suggested by Gerber et al. (1979). A current hypothesis is that 4-ene-VPA may cause liver damage on further bioactivation to an electrophilic diene which inhibits mitochondrial β-oxidation enzymes and/or depletes cellular glutathione (Kassahun et al., 1994; Tang et al., 1995; JurimaRomet et al., 1996).
Previously, it was shown that the metabolic flux through the 4-ene pathway in humans is elevated by coadministration of the P450 inducers, phenytoin and carbamazepine, and inhibited by stiripentol, a methylenedioxyphenyl derivative (Levy et al., 1990). However, the identities of specific human liver microsomal P450 isoforms which catalyze this process have not been established. The present study aimed to establish in vitro metabolic reaction conditions, with human liver microsomes as the enzyme source, that reflect the oxidative metabolite profile of VPA observed in vivo. These conditions were used in conjunction with chemical inhibitors and cDNA-expressed enzymes to identify specific P450 isoforms responsible for 4-ene-VPA formation.
Materials and Methods
Chemicals.
VPA and 1-methyl-1-cyclohexanecarboxylic acid were purchased from Aldrich Chemical Co. (Milwaukee, WI). VPA metabolites were obtained as described previously (Rettenmeier et al., 1986). BSTFA was purchased from Supelco Inc. (Bellefonte, PA). Resolution of (S)- and (R)-warfarin and synthesis of the hydroxylated metabolites and deuterated internal standards have been reported previously (Bush et al., 1983;Lawrence et al., 1990). Furafylline was a gift from Dr. K. Kunze (University of Washington, WA). Sulfaphenazole, TAO, quinidine HCl, coumarin, 7-hydroxycoumarin, DDC and β-NADPH were purchased from Sigma Chemical Co. (St. Louis, MO). All other chemicals and solvents used were of analytical grade.
Selection of human liver groups.
The study was designed around eight human liver samples obtained from male and female kidney transplant donors between the ages of 9 and 62 years. Based on their clinical history the livers were assigned to either a control group (four livers) or induced group (four livers). The induced group of livers all had prior exposure to phenytoin, and in some cases to additional drugs, while the control group had no prior drug exposure. A brief description of the pertinent clinical data is presented in table1.
Characteristics of human liver samples
Biological material.
The source, processing and method of storage of liver samples obtained from transplant donors have been described (Rettie et al., 1989). Livers were homogenized in 10 mM potassium phosphate buffer containing 0.15 M potassium chloride, 10 mM ethylenediaminetetraacetic acid, pH 7.4 at 4°C, and microsomal pellets were prepared by differential ultracentrifugation by standard procedures. Resuspension and storage of microsomal pellets were described previously (Sadeque et al., 1992)
cDNA expressed enzymes.
Recombinant human CYP1A2, CYP2C8, CYP2C9, CYP2E1, CYP3A4 and CYP3A5 were expressed in HepG2 cells transfected with vaccina virus containing the respective human cDNAs. Vector construction and expression of these recombinant viruses has been described elsewhere (Aoyama et al., 1989; Crespiet al., 1990; Nhamburo et al., 1990). Human liver CYP2A6, CYP2B6, CYP2C9, CYP2C19, CYP2D6 and CYP3A4, co-expressed in lymphoblastoid cells with cytochrome P450 reductase (with the exception of CYP2B6), were purchased from Gentest Corporation (Woburn, MA). CYP2C9 and CYP2A6 preparations from Gentest were incubated in Tris-HCl, rather than phosphate buffer, according to the manufacturer’s recommendations.
General reaction conditions.
Unless noted otherwise, complete reactions mixtures contained 1 nmol human liver microsomal P450 or 0.1 to 0.5 nmol of each individual cDNA-expressed isoform, 100 μmol potassium phosphate, pH 7.4, and substrate in a final volume of 1.0 ml. After a 2-min preincubation at 37°C, reactions were initiated by the addition of 1 mM NADPH.
Assay for VPA metabolites.
Complete reaction mixtures containing 1 to 2 mM VPA were incubated for 30 min at 37°C and terminated by the addition of 1 ml of ice-cold 10% HCl. 1-Methyl-1-cyclohexanecarboxylic acid (93 ng) was added to each reaction as an internal standard for analysis by GC/MS. Samples were stored at 4°C overnight, centrifuged to remove precipitated proteins and extracted into ethyl acetate (2 × 3.5 ml). The organic extracts were concentrated, derivatized with BSTFA and the 4-ene, 4-hydroxy and 5-hydroxy metabolites separated on a DB-1 stationary phase and quantitated by selected-ion monitoring GC/MS, as described previously (Rettie et al., 1995).
Modulation of VPA metabolism by chemical inhibitors.
Furafylline, coumarin, sulfaphenazole, quinidine and DDC were dissolved in buffer solutions, whereas TAO was dissolved in methanol. Concentrations of these selective inhibitors were chosen to obtain the maximum inhibition for the respective CYP isozyme while maintaining adequate selectivity (Anderson et al., 1993, Chang et al., 1994). TAO (50 μM), DDC (50 μM) and furafylline (200 μM) are mechanism-based inhibitors and so were preincubated with microsomal reaction mixtures for 10 min at 37°C, in the presence of NADPH, before initiation of the reaction with the primary substrate. Sulfaphenazole (5 μM), coumarin (25 μM) and quinidine (1 μM) were co-incubated with VPA. Rates of 4-ene-VPA formation were measured as described above, and the extent of inhibition was calculated relative to control incubations containing either buffer or methanol as appropriate.
Assay for warfarin metabolites.
Complete reaction mixtures, which contained either 600 μM (R)-warfarin or 200 μM (S)-warfarin, were terminated after 30 min incubation at 37°C by the addition of 0.6 ml acetone, spiked with a mixture of deuterated internal standards and extracted into ether/ethyl acetate (2 × 3 ml). Organic extracts were evaporated to dryness and the metabolites derivatized with diazomethane and/or BSTFA before separation on a DB-5 stationary phase and quantitation by GC/MS as detailed previously (Bush et al., 1983; Lawrence et al., 1990).
Assay for 7-hydroxycoumarin.
Coumarin 7-hydroxylase activity was measured according to the method of Creaven et al.(1965) with slight modifications. The incubation contained 0.1 nmol P450 and 50 μM coumarin, 25 μmol Tris buffer, pH 7.4, in a final volume of 0.5 ml. The reaction was initiated by the addition of 1 mM NADPH after a 2-min preincubation. After 10 min at 37°C the reaction was terminated by adding either 0.5 ml 6% trichloroacetic acid or 0.5 ml methanol. The solution was centrifuged to remove the precipitated protein. Supernatant (0.2 ml) was added to 0.8 ml of 0.8 M Tris/glycine buffer (pH 9.0) and the fluorescence was recorded on a Perkin-Elmer spectrofluorometer (MPF-3L). The excitation and emission wavelengths of fluorescence were 351 nm and 454 nm, respectively. The amount of metabolite formed was quantitated with a standard curve generated from known amounts of 7-hydroxycoumarin added to incubation mixtures lacking NADPH.
P450 content and protein determination.
Cytochrome P450 concentrations were measured by the method of Estabrook et al. (1972). Microsomal protein was determined by the method ofLowry et al. (1951).
Results
To establish an in vitro system that reflects the oxidative metabolism of VPA in vivo as described by Levyet al. (1990), we began by identifying two groups of liver donors (table 1) from the Human Liver Bank located in the Departments of Medicinal Chemistry and Pharmaceutics at the University of Washington. Group I livers served as controls where no record of drug exposure existed. Group II livers were obtained from donors who had been exposed to the CYP3A inducer phenytoin (Watkins et al., 1985) alone, or in combination with other drugs. Total P450 and microsomal (R)-warfarin 10-hydroxylase activities were measured as indices of CYP3A induction. Although the specific content of P450 was not significantly higher in group II, CYP3A-selective 10-hydroxylation of (R)-warfarin was increased 3.5-fold (table 1).
Rates of VPA metabolite formation were measured at a substrate concentration of 1 mM which is near the therapeutically effective range found in plasma (400–700 μM). In the group II livers, 4-ene-VPA, 4-hydroxy-VPA and 5-hydroxy-VPA were elevated 2.3-fold, 2.6-fold and 5.4-fold, respectively (table 2). The ratios of the rates of formation of 4-ene/4-OH/5-OH for group I and group II livers were, 1:18:9 and 1:20:22, respectively. This product profile is in excellent agreement with the human pharmacokinetic data reported previously by Levy et al. (1990), where ratios ofin vivo formation clearances of these same three metabolites was reported to be approximately 1:20:20 in humans.
Rates of formation of VPA metabolites by human liver microsomes
The effect of several P450 isoform-selective chemical inhibitors on the rate of 4-ene-VPA formation by HL122 and HL131 at a VPA concentration of 1 mM is shown in figure 1. TAO (50 μM), quinidine (1 μM) and furafylline (200 μM) decreased metabolite formation by only 5 to 12% when either microsomal preparation served as the enzyme source. This suggests that CYP3A4, CYP2D6 and CYP1A2 are not involved in VPA oxidation. In contrast, both sulfaphenazole (5 μM) and coumarin (25 μM) inhibited formation of 4-ene-VPA from both preparations, although the effect of sulfaphenazole was more pronounced with HL122 (43% vs. 15% inhibition) and the effect of coumarin more pronounced with HL131 (43%vs. 26% inhibition). DDC (50 μM) inhibited olefin formation by 40 to 47%, which reflects catalysis by either CYP2E1 and/or CYP2A6. Similar inhibition profiles were observed for the two alcohol metabolites (data not shown).
Inhibition of human liver microsomal 4-ene-VPA formation by chemical inhibitors. Upper chart, HL122-catalyzed desaturation; lower chart, HL131-catalyzed desaturation. Reactions were conducted with 1 nmol human liver microsomal P450 and 1 mM VPA. Values are the mean of duplicate determinations.
The differential effects of sulfaphenazole and coumarin on microsomal metabolism catalyzed by HL122 and HL131 may simply reflect variable CYP2C9 and CYP2A6 concentrations in the two preparations. To test this we measured the isoform-selective (S)-warfarin 7-hydroxylase (CYP2C9) and coumarin 7-hydroxylase (CYP2A6) activities in the eight human liver preparations. In these eight preparations CYP2C9-dependent activity varied 2.5-fold and CYP2A6-dependent activity varied 15-fold. HL122 and HL131 exhibited the highest marker activities for CYP2C9 and CYP2A6, respectively (table 3), consistent with the observed inhibitor effects.
Human liver microsomal CYP2C9 and CYP2A6-dependent catalytic activities
The chemical inhibition studies suggest that multiple human P450 isoforms contribute to 4-ene-VPA formation. Therefore, we measured the rates of production of 4-ene-VPA by ten cDNA-expressed human liver isoforms. All the forms were not available from the HepG2 expression system and so the battery was supplemented with commercially available preparations expressed in lymphoblastoid cells. CYP3A4, expressed from either system, produced no detectable 4-ene-VPA, even at an elevated substrate concentration of 2 mM (table4). Conversely, CYP2C9 produced substantial quantities of product when expressed in both systems. The only other isoform which yielded significant quantities of the terminal olefin was CYP2A6.
4-Ene-VPA formation catalyzed by cDNA-expressed human P450s
Discussion
Several risk factors for VPA-associated hepatotoxicity have been described including young age, developmental delay, coincident metabolic disorders and polytherapy (Bryant and Dreifuss, 1996). The principal aim of the present study was to investigate how polytherapy, in particular, plays a role in this event by identifying specific human P450 isoforms which are responsible for the formation of the hepatotoxic metabolite, 4-ene-VPA. Because these studies were conductedin vitro, it was important to establish that the reaction conditions used with microsomal preparations were relevant to the clinical situation. A two-step approach was used to make this evaluation.
First, we identified a group of liver donors who had each been exposed to phenytoin and measured their rates of CYP3A4-dependent metabolism to confirm their “induction” status. As expected, this group exhibited an elevated rate of microsomal 4-ene-VPA formation. Second, we compared the oxidized metabolite profiles obtained from microsomal incubations with those which can be derived from in vivo pharmacokinetic parameters (Levy et al., 1990). In both cases, the ratios of 4-ene/4-OH/5-OH formation were about 1:20:20, which suggests that a similar complement of P450 isoforms were active, in vitroand in vivo, in the oxidative metabolism of VPA. This most likely reflects the use of a near clinically relevant substrate concentration in the microsomal reactions.
We attempted to further characterize human liver microsomal VPA desaturation by measuring kinetic constants for 4-ene-VPA formation, but were unable to obtain saturation plots for any of the oxidative metabolites generated from VPA. One reason for this problem might be that this fatty acid substrate has a detergent effect on membranes and enhances substrate access or some other parameter that increases reaction velocity at millimolar concentrations. However, additional work with both microsomal samples and cDNA-expressed enzymes is required to determine whether such a phenomenon occurs.
Because HL122 and HL131 exhibited the highest rates of production of VPA metabolites, these two microsomal preparations were chosen to evaluate the effect of a series of P450 isoform-selective chemical inhibitors on the formation of 4-ene-VPA. These experiments demonstrated that multiple forms are involved and clearly implicated CYP2C9 and CYP2A6 in this bioactivation pathway based on inhibition by sulfaphenazole and coumarin, respectively. Coumarin was a more effective inhibitor in reactions catalyzed by HL131 and sulfaphenazole was a more effective inhibitor in reactions catalyzed by HL122. The altered inhibitor profiles were consistent with a differential isoform complement in the two enzyme preparations. DDC also inhibited 4-ene-VPA formation catalyzed by both preparations. Although inhibition by DDC has often been used as an indicator of CYP2E1 involvement, Changet al. (1994) found that DDC at concentrations from 10 to 125 μM inhibited several P450 isoforms, and as little as 10 μM DDC inhibited CYP2A6 activity by 65%.
The ambiguities that can arise during the interpretation of data generated with such nonselective inhibitors as DDC can be compounded if the reaction of interest is catalyzed by an isoform for which no diagnostic inhibitor has been established (e.g., CYP2C8). Therefore, we also screened ten human cDNA-expressed isoforms for their ability to form 4-ene-VPA. CYP1A2, CYP2B6, CYP2C8, CYP2C9, CYP2E1, CYP3A4 and CYP3A5 expressed in HepG2 cells and CYP2A6, CYP2B6, CYP2C9, CYP2C19, CYP2D6 and CYP3A4 expressed in lymphoblastoid cells were used. This study confirmed the ability of CYP2C9 and CYP2A6 to form 4-ene-VPA and excluded participation of CYP2E1 which might have been inferred from the chemical inhibitor experiments.
Numerous weakly acidic compounds such as phenytoin, warfarin, diclofenac and several other nonsteroidal anti-inflammatory drugs are preferentially metabolized by CYP2C9, and current concepts of this enzyme’s active-site architecture highlight potential electrostatic interactions between substrate carboxylate functionalities and positively charged groups on CYP2C9 (Mancy et al., 1995;Jones et al., 1996). Therefore, the finding that VPA is a substrate for CYP2C9 finds a rationalization within this evolving model. Although CYP2A6 metabolizes coumarin itself (Yamano et al., 1990), unlike CYP2C9, it does not accept bulkier coumarin derivatives such as warfarin or dicoumarol as substrates (Pearceet al., 1992). With the exception of VPA, few, if any, other substrates selectively metabolized by both CYP2C9 and CYP2A6 have been identified. Therefore, a comparison of the kinetics of VPA metabolism by these two isoforms, as well as an analysis of CYP2C9 and CYP2A6-dependent prochirality of VPA side-chain oxidation (Shirleyet al., 1993) would be required before any conclusions can be reached regarding similarities in the active-site features of these two isoforms that promote metabolism of small, acidic molecules like VPA.
In summary, the present study excludes CYP3A isoforms and identifies CYP2C9 and CYP2A6 as the principal human liver microsomal P450s involved in the formation of the hepatotoxic metabolite, 4-ene-VPA. This is significant because CYP3A4 is the isoform usually associated with induction by anticonvulsants (Watkins et al., 1985;Pichard et al., 1990). Nonetheless, our data suggest that it is not involved in mediating the enhanced formation of the terminal olefin in patients receiving anticonvulsant polytherapy. Rigorous comparisons of the inducibility of human hepatic CYP2A6 and CYP2C9 by anticonvulsants have not been performed, but it is clear that CYP2A6 is a phenobarbital-inducible form in cynomologus monkeys (Bullock et al., 1995), and catalytic activities attributable to CYP2A6 can vary by 30-fold in human liver microsomes (Wrighton et al., 1993). In contrast, CYP2C9 activities vary by less than 5-fold (Wrighton et al., 1993; Hall et al., 1994). Therefore, we suggest that CYP2C9 catalyzes the majority of constitutive VPA 4-ene-desaturation and that CYP2A6 plays a more prominent role during anticonvulsant polytherapy. Finally, the use of two different cDNA expression systems limits quantitative comparisons of the relative terminal VPA desaturase efficiencies of the isoforms which are implicated here. Kinetic studies are underway with purified, recombinant human CYP2 isoforms to address this issue.
Footnotes
-
Send reprint requests to: Allan E. Rettie, Department of Medicinal Chemistry, University of Washington, Seattle, WA 98195.
-
↵1 This study was supported in part by NIH grants GM32165 and GM49054. MBF was supported by NIH Training Grant GM07750.
-
↵2 Present address: Center for Clinical Pharmacology, Scaiffe Hall, University of Pittsburgh Medical Center, Pittsburgh, PA 15217.
- Abbreviations:
- VPA
- valproic acid
- DDC
- diethyldithiocarbamate
- TAO
- triacetyloleandomycin
- GC/MS
- gas chromatography/mass spectrometry, BSTFA, bis(trimethylsilyl)trifluoroacetamide
- Received April 28, 1997.
- Accepted July 29, 1997.
- The American Society for Pharmacology and Experimental Therapeutics
