Genetics of P450 oxidoreductase: Sequence variation in 842 individuals of four ethnicities and activities of 15 missense mutations
- Departments of *Pediatrics and
- †Biopharmaceutical Sciences, University of California, San Francisco, CA 94143
-
Communicated by Melvin M. Grumbach, University of California School of Medicine, San Francisco, CA, December 11, 2007 (received for review November 1, 2007)
Abstract
P450 oxidoreductase (POR) is an electron-donating flavoprotein required for the activity of all microsomal cytochrome P450 enzymes. We sequenced 5,655 bp of the POR gene in a representative population of 842 healthy unrelated individuals in four ethnic groups: 218 African Americans, 260 Caucasian Americans, 179 Chinese Americans, and 185 Mexican Americans. One hundred forty SNPs were detected, of which 43 were found in ≥1% of alleles. Twelve SNPs were in the POR promoter region. Fifteen of 32 exonic variations altered the POR amino acid sequence; 13 of these 15 are previously undescribed missense variations. We found eight indels, only one of which was in the coding region. A previously described variant, A503V, was found on 27.9% of all alleles with some ethnic predilection (19.1% in African Americans, 26.4% in Caucasian Americans, 36.7% Chinese Americans, and 31.0% in Mexican Americans). We built cDNA expression vectors for the 13 previously undescribed missense variants, expressed each protein lacking 27 N-terminal residues in Escherichia coli, and assayed the apparent K m and V max of each in four assays: reduction of cytochrome c, oxidation of NADPH, 17α-hydroxylase activity of P450c17, and 17,20 lyase activity of P450c17. The catalytic activities of several missense mutants differed substantially in these assays, indicating that each POR mutant must be assayed separately with each potential target P450 enzyme. The activity of A503V was reduced to a modest but statistically significant degree in all four assays, suggesting that it may play an important role in interindividual variation in drug response.
P450 oxidoreductase (POR) is a single protein containing both flavin adenine dinucleotide (FAD) and flavin mononucleotide (FMN) moieties that transfers electrons from NADPH to microsomal (type II) cytochrome P450 enzymes (for review, see ref. 1) (Fig. 1). Cytochrome P450 enzymes (http://drnelson.utmem.edu/CytochromeP450.html) are heme-containing proteins that catalyze a broad range of oxidative reactions. The human genome contains 57 genes for cytochrome P450 enzymes; 7 encode Type I (mitochondrial) P450s and 50 encode Type II P450s, found in the endoplasmic reticulum (2). Among the 50 human microsomal P450 enzymes, approximately 20 are involved in the biosynthesis of steroids, sterols, fatty acids, and eicosanoids, approximately 15 participate in hepatic drug metabolism, and approximately 15 are “orphan” P450 enzymes whose catalytic roles are unclear (3).
Relationship of POR to a microsomal cytochrome P450 enzyme. POR contains two flavins, a flavin adenine dinucleotide (FAD) and a flavin mononucleotide (FMN) in two separate lobes of the POR protein. Electrons from reduced nicotinamide adenine dinucleotide phosphate (NADPH) are taken up by the FAD moiety, then transferred to the FMN moiety. The acidic POR domain containing the FMN moiety then associates with the basic redox-partner binding site of a cytochrome P450 by electrostatic interactions, and the electrons pass from the FMN group to the heme iron of the P450, permitting catalysis. Both POR and the P450 are bound to the endoplasmic reticulum (ER), but most of the protein surface is exposed to the cytoplasm (CYTO). Adapted from ref. 4.
POR transfers electrons to three steroidogenic enzymes: P450c17 (17α-hydroxylase/17,20 lyase), P450c21 (21-hydroxylase), and P450aro (aromatase) (4). Individuals with POR deficiency have a broad range of steroidogenic disorders, extending from infants with ambiguous genitalia and skeletal malformations to women with the polycystic ovary syndrome (5–8). Finding severe POR mutations in human patients was unexpected, because POR knockout mice die during embryonic development (9, 10). By contrast, liver-specific POR knockout mice are morphologically normal, reproductively competent and have normal life spans, but their hepatic drug metabolism is severely impaired (11). Because most drugs used in clinical practice are metabolized by hepatic microsomal (Type II) P450 enzymes that require POR (12, 13), one would predict that human POR mutants would impair drug metabolism and/or bile acid biosynthesis. Disorders of these systems have not yet been reported in patients, but POR mutants Y459H and V492E, found in patients (5) disrupt binding of FAD and have impaired activity with CYP4A4 (14). Sequence variations in some Type II P450 enzymes can cause variations in drug responses, both among individuals and among ethnic groups (12, 13). However, the potential importance of POR sequence variations in drug responses has not been investigated. As the first step in determining whether POR sequence variants contribute to variations in P450-mediated reactions, we sought to determine the degree to which there are POR sequence variants in human populations and to assess the effects of missense variants on POR-mediated catalysis.
Results
Genetic Variations in the POR Gene.
To identify POR amino acid sequence variations in the normal population, we sequenced the 15 protein-coding exons and at least 50 bp of each intron's splice donor and splice acceptor site adjacent to each exon from 842 individuals. To identify potential regulatory variants or splicing variants, we also sequenced the first untranslated exon 1U, which is 38.8 kb upstream from protein-coding exon 1 (15) and 274 bp of DNA 5′ to exon 1U, in 701 of these 842 individuals. Approximately 5,655 bp of DNA were sequenced in each individual; the data can be found at www.pharmgkb.org. We found 140 distinct nucleotide variations, with a frequency of approximately 3.1 single-nucleotide polymorphism (SNPs) per kb of sequence. Only 37 of 140 variable sites had been identified in previous studies; the remainder were previously undescribed. Of these 140 variants, 32 were in the protein-coding regions, 2 were in untranslated exon 1U, 12 were in the 5′ flanking DNA, and 94 were in the segments of introns directly adjacent to exons. Of the 140 sites, 43 were polymorphic, found at allele frequencies of >1% in any population. Eighteen of the polymorphisms were population-specific, found in only one ethnic group. Of these polymorphisms, 13 were found in African Americans, 1 in Asians, 2 in Mexican Americans and 2 in European Americans. We found a total of 8 indels, of which only one, a 3-bp in-frame deletion, was in the coding region.
Fifteen of the 32 sequence variations in the protein-coding regions changed the encoded amino acid; 13 of these 15 changes were previously undescribed, most were found in exons 8–14, and all were found as heterozygotes with the wild-type allele (Fig. 2). Among the 13 previously uncharacterized missense variants, 12 were found on only once, and one was found twice. However, the amino acid sequence variant A503V [listed as rs17859083, rs17846082, or rs1057868 in the National Center for Biotechnology Information (NCBI) SNP database] was found in 463 of 1,670 alleles (27.9%), including 83 of 434 African American alleles (19.1%), 137 of 518 Caucasian American alleles (26.4%), 130 of 354 Chinese American alleles (36.7%), and 113 of 364 Mexican American alleles (31.0%). These 463 alleles were found in 398 individuals; 65 were A503V homozygotes and 333 were heterozygous with the wild-type allele. Others have also found A503V in the normal population (16).
Location of protein-altering variants within the 16 exons of POR. The first exon, 1U, contains the basal promoter and untranslated region of POR; all other exons are protein-coding. The downward facing arrows show the location of the variants identified in this study. The longer downward facing arrow (in gray), shows the location of the polymorphic variant, A503V. The two upward facing arrows show the location of two variants that are commonly associated with Antley–Bixler syndrome, A287P and R457H. Variants are indicated by letters, as follows: a, delE53, b, P55L; c, D211N; d, G213E; e, P284L; f, P284T; g, E300K; h, R406H; i, P452L; j, A462T; k, V472M; l, A485T; m, A503V; n, R600W; o, Y607C.
Population Genetics.
The population genetics of the coding region of POR is presented in Table 1. Although an equal number of synonymous and nonsynonymous sites were identified, the nucleotide diversity at various sites (πS) was considerably greater than that at nonsynonymous sites (πNS). The ratio of πNS/πS, used as a measure of selective pressure, was 0.19 for POR, a value much less than one and indicative of selective pressure on the gene (17, 18). Selective pressure on POR is comparable with other genes resequenced in large SNP genotyping projects (17, 19). Values of πNS did not vary much among the four ethnic groups studied (Table 1). Although African Americans had the greatest number of variable sites (n = 76), Asians had the greatest number of nonsynonymous variants and the greatest value of πNS; over half of the nonsynonymous variants identified were in Asian samples (n = 9).
Population genetics statistics for variation in POR
Functional Analysis of the Previously Undescribed POR Missense Variations.
To determine the functional activities of the 13 previously undescribed sequence variants, the common A503V mutation, and the E300K polymorphism (rs11540674) found in the database, we assessed their activities in four enzymatic assays: the reduction of cytochrome c, the oxidation of NADPH, and their ability to support 17α-hydroxylase and 17,20 lyase activities of human P450c17. The classical assay for POR activity measures its capacity to support the reduction of cytochrome c or the oxidation of NADPH (20). Human POR lacking 27 N-terminal residues (N-27 POR) associates with membranes, as does wild-type POR (8). Therefore, we expressed each mutant in N-27 POR and measured the K m, V m, and V m/K m for the reduction of cytochrome c and the oxidation of NADPH (Table 2). The V m/K m values, an estimate of catalytic efficiency, shown as percentages of the wild-type values, reveal a broad range enzymatic impairment. There is general agreement between the two assays, with some variation (e.g., P55L) (Table 2).
Kinetics of cytochrome c reduction or NADPH oxidation activities supported by POR variants
Cytochrome c, a soluble mitochondrial protein, is not a physiologic substrate for POR in the endoplasmic reticulum (20), and we have shown that POR assays based on P450c17 activity provide a better correlation between the phenotype and genotype of disease-causing POR mutants (5, 8). Therefore, we supplemented the cytochrome c reduction/NADPH oxidation assays with assays that directly measure the capacity of POR to support the activities of P450c17. These assays combined membranes containing N-terminally modified full-length human P450c17 expressed in bacteria (21) and a membrane fraction containing human N-27 POR mutants expressed in Escherichia coli (22). Cytochrome b 5, which acts as an allosteric factor to facilitate the 17,20 lyase activity of P450c17 (23), was added to support the lyase reaction, and the K m, V m, and V m/K m for the 17α-hydroxylase and 17,20 lyase activities of P450c17 were determined by Lineweaver–Burke analysis (Table 3). The K m and V m for the wild-type POR are in very good agreement with our previous data (8), although the V m values in the present study are pmol/pmol P450c17/min. There is broad agreement between the assays based on cytochrome c (Table 2) and those based on P450c17 (Table 3), although the activities of several mutants, especially P284L and P452L differed substantially between the two types of assays. The activity of the common A503V variant was consistently 56–67% of wild type in the cytochrome c assays and 58–68% in the P450c17 assays; these differences were significantly different from the control values (cytochrome c reduction, P = 0.011; NADPH oxidation, P = 0.0110; 17α-hydroxylase, P = 0.0115; 17,20 lyase, P = 0.0062).
Kinetics of P450c17 activities supported by POR variants
The activities of these missense variants correspond well with their locations in the POR protein. The structure of rat POR has been determined by x-ray crystallography (24), permitting the modeling of the corresponding regions of human POR, which is 94% identical to the rat sequence (8). Because the crystallized rat protein and corresponding human model lack 65 N-terminal residues, the delE53 and P55L mutants are not modeled. This model previously showed that disease-causing mutants retaining <50% of activity in both P450c17-based assays mapped close to the POR sites that bind FAD (14), FMN, or NADPH, whereas mutants that retained >50% activity mapped near the surface of the protein (8). Consistent with these results, the model shows that the only two mutants identified with <50% activity in the P450c17 assays, P284T and R600W, mapped very close to the sites binding FAD and NADPH, respectively, whereas all other mutants mapped near the surface of the protein, at a distance from the FAD-, FMN-, and NADPH-binding sites (Fig. 3).
Model of human POR, showing the locations of the sequence variants identified. The model has been described (8) and is based on the crystal structure of nonfunctional rat POR lacking 65 N-terminal residues (24); therefore, the delE53 and P55L variants are not shown. The α-carbon backbone is depicted as a narrow ribbon, the FAD and FMN moieties (magenta) and NADPH (cyan) are depicted as ball-and-stick models, and the variant amino acids are depicted by packed sphere images of different colors, corresponding to their V max/K m for 17,20 lyase activity: red, <25%; blue, 25–50%; green, 51–115%; yellow, >115%. There are two variants at P284; P284T, which retained 19% activity, is shown and P284L (82% activity) is not. The figure was created with Molsoft ICM.
Discussion
Pharmacogenetics seeks to identify mechanisms that underlie individual variations in responses to drugs, with the aim of optimizing the choice and dose of drugs based on individual genetics (25, 26). To date, most efforts in pharmacogenetics have concerned genes encoding drug transporters and drug-metabolizing enzymes (27). Most drugs in clinical use, as well as most xenobiotics, are metabolized by one or more microsomal cytochrome P450 enzymes, all of which require electron donation from POR, but the genetics of POR in drug metabolism have not been evaluated. Our analysis of the POR gene sequence in 842 individuals from four ethnic groups identified 108 noncoding SNPs, 32 coding SNPs, 15 of which altered the encoded amino acid sequence, and eight indels. Most of these SNPs were rare, but 36 noncoding and 7 coding SNPs, one of which altered the coding sequence, were found at allele frequencies of ≥1% in at least one of the ethnic groups. A complete list of these SNPs and their distributions in the four ethnic groups is presented in supporting information (SI) Table 4. These data may underestimate the genetic differences among the ancestral populations, because of the genetic mixing of ethnic groups in the United States. Because our approach emphasized changes in coding regions and splice sites, substantially more SNPs may be found in the unsequenced regions of introns.
The differences in POR SNP frequencies among ethnic groups may contribute to differences in the metabolism of drugs by hepatic cytochrome P450 enzymes. SNPs may be in linkage disequilibrium with as-yet unidentified polymorphisms in gene regulatory regions, contributing to different levels of POR expression, or they may contribute to variations in RNA splicing efficiency, mRNA stability, or the efficiency of translation. For example, the rare silent exon 1 SNP 15A→G, which may be associated with altered POR RNA splicing (28), seems to confer a modestly increased risk for breast cancer in African-American women (16). SNPs in coding regions may contribute directly to variations in catalysis. One amino acid sequence variant, A503V, was common, being found in 27.9% of all alleles, varying from 19.1% in African Americans to 36.7% in Chinese Americans. Although this variant contains a highly conservative change in a region of the protein not associated with a known function, the A503V variant showed modestly but statistically significantly decreased catalytic activity in all four assays used.
Analysis of the enzymatic activities of all coding sequence variants confirmed our previous observation that different assays of POR activity will yield different results (5, 8). POR is typically assayed by its capacity to reduce cytochrome c, reflecting its initial identification as a cytochrome c reductase (29). This assay is rapid, reproducible and directly assays electron transfer, but cytochrome c is not a physiologic substrate for POR. The structure of rat POR, determined crystallographically, indicates that the electron-donating FMN domain interacts with the redox-partner binding site of a P450 by electrostatic interactions (24). Modeling of these interactions with cytochrome c (24) and with P450 enzymes (30) suggests that the POR residues that interact with the electron recipient may vary among various P450s. This variation in POR/P450 contact sites would suggest that POR missense mutants that alter POR conformation might impair activity to differing degrees, depending on the electron recipient used to assay activity. In comparing the capacity of POR mutants to reduce cytochrome c, oxidize NADPH, and support the two activities of P450c17, we identified substantial, but highly reproducible differences in the activity revealed by these four assays (5, 8). Similar differences have been described in comparing the capacity of a single POR mutant to support the activities of P450c17 and P450c21 (31) and in the capacity of our initially described mutants to support the activity of aromatase (P450aro) (30). Thus, multiple studies show that the activity of a specific POR missense mutant as assayed with one P450 cannot be extrapolated to infer its activity with other P450s. This variation in activity will be of substantial importance in characterizing the effects of POR sequence variants such as A503V and missense mutants with the principal hepatic drug-metabolizing P450 enzymes, and may contribute to individual variations in drug response.
Materials and Methods
DNA Sequencing.
DNA was collected from 842 unrelated individuals in the San Francisco Bay Area as part of the Studies of Pharmacogenetics in Ethnically Diverse Populations (SOPHIE) project (32). Subjects identified themselves as African American, Caucasian American, Chinese American, or Mexican American by having all four grandparents of the stated ethnicity. DNA was sequenced as described (8, 15). Primers (SI Table 5) were designed by using Primer3 (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3_www.cgi) to span the first untranslated exon (exon 1U), 274 bp of 5′ flanking DNA, the 15 protein-coding exons and at least 100 bp of intronic DNA adjacent to each exon. Genomic DNA samples (10 ng) were used as templates for PCR amplification in a 5 μl reaction. PCR amplification of exons 1U, 1, 2, 3, 4, 5, 6, 7, 8–9, 10–11, 12–13, and 14–15 of POR were performed by using Platinum Taq (Invitrogen, Carlsbad, CA), under high touchdown cycling conditions, starting at 95°C for 5 min, followed by 14 touchdown cycles of 94°C for 20 s, 65°C to 58.5°C for 20 s (the annealing temperature for each subsequent cycle was decreased by 0.5°C), and 72°C for 1 min. The touchdown PCR was followed by 35 cycles of PCR amplification at 94°C for 20 s, 58°C for 20 s, and 72°C for 1 min; the final extension was held at 72°C for 10 min, followed by +4°C to stop the reaction. The sizes and locations of the PCR products are listed in SI Table 6.
PCR cycling and sequencing were done on ABI Geneamp PCR system 9700 thermocyclers (Applied Biosystems, Foster City, CA). The sequencing was done with ABI BigDye terminator version 3.1 and displayed on ABI 3730 × 1 DNA Analyzer. Genomic DNA from a normal individual was used as internal controls, and NCBI genomic sequence NC_000007.12 was used as the reference sequence. Sequence variations were analyzed with DNA Sequencher 5.2 (Gene Codes, Ann Arbor, MI). All sequence variations were confirmed by repeating the PCR amplification and sequencing the opposite strand. Sequence variants in protein-coding regions are numbered according to NCBI reference sequence NM_000941.2 (A of the initiation codon ATG is denoted +1), and protein sequence mutations are numbered according to NCBI reference sequence NP_000932.3. Nucleotide variations located in the intronic region are numbered as described in ref. 33.
Expression of POR and P450c17 in Bacteria.
Expression vectors for the POR sequence variants lacking 27 N-terminal residues (22) were generated as described in ref. 8, by PCR-based, site-directed mutagenesis using the primers in SI Table 7. These plasmids were used to express the mutant forms of POR, which were prepared from bacterial membranes as described in ref. 8. Mutant protein was quantitated by Western blotting compared with a standard curve of wild-type POR as described in ref. 8, except that blots were incubated with goat anti-rabbit infrared (IR)-labeled secondary antibody at a dilution of 1:10,000, signals were detected with the green fluorescent channel (700 nm) on an Odyssey Infrared Imaging System (LI-COR Bioscience, Nebraska), and quantitated with Odyssey software. All POR levels were normalized against wild-type POR.
Full length human P450c17 cDNA (34) with an N-terminal modification (21) and a C-terminal 4xHis tag was cloned into bacterial expression vector pCW and transformed into E. coli JM109. P450c17 expression was induced and membranes were prepared as described in ref. 35. P450c17 content in the membrane preparation was determined by carbon monoxide-difference spectra (36).
POR Assays Based on Cytochrome c.
The ability of bacterially expressed POR to reduce cytochrome c or oxidize NADPH was assayed as described in ref. 8, except that assays were performed with bacterial membranes containing 2 pmol of POR. Kinetic parameters, maximum velocity (V m) and apparent Michaelis constant (K m), were determined by nonlinear Michaelis-Menten plots using Graph Pad Prism 3. Results are mean ± SEM of at least three independent experiments.
POR Assays Based on P450c17.
The ability of bacterially expressed POR to support 17α-hydroxylase and 17,20 lyase activities of P450c17 was assayed by using human P450c17 expressed in bacteria. Membrane fractions containing 10 pmol of P450c17 were combined with membranes containing 20 pmol of POR, 20 μg of 1,2-didodecanoyl-sn-glycero-3-phosphocholine (Sigma) in 100 mM K phosphate, and 20 μg of 3-[(3-cholamidopropyl)dimethylammonio]-1-propanesulfonate (Sigma) and sonicated in an ice bath for 10 min. The reaction mixture was completed by adding 6 mM K acetate, 10 mM MgCl2, 1 mM reduced glutathione, 20% glycerol, and radiolabeled substrate, in a total volume of 200 μl. Substrates were [14C]progesterone for the 17α-hydroxylase reaction and [3H]17α-hydroxypregnenolone for the 17,20 lyase reaction, with 50 pmol of cytochrome b 5 added as an allosteric facilitator (23). Reactions were started by adding 4 μl of 100 mM NADPH, incubated at 37°C for 2 h, and stopped by adding 450 μl of ethyl acetate:iso-octane (1:1) to extract the steroids. Steroids were analyzed by TLC as described in ref. 37, and quantitated by phosphorimaging. Data were analyzed as Lineweaver–Burk plots using Graph Pad Prism 3 (GraphPad Software) and expressed as mean ± SD of at least three independent experiments.
Acknowledgments
We thank Hee Jae, Melanie Dela Cruz, and Elaine J. Carlson from the University of California at San Francisco Genomics Core Facility for the POR sequencing, Doug Stryke for help with data analysis, and Ms. Izabella Damm for excellent technical assistance. This work was supported by National Institutes of Health Grant R01 GM073020 (to W.L.M.).
Footnotes
- ‡To whom correspondence should be addressed at: Department of Pediatrics, HSE 1401, 513 Parnassus Avenue, University of California, San Francisco, CA 94143-0978. E-mail: wlmlab{at}ucsf.edu
-
Author contributions: W.L.M. designed research; N.H. and V.A. performed research; N.H., V.A., K.M.G., and W.L.M. analyzed data; and K.M.G. and W.L.M. wrote the paper.
-
The authors declare no conflict of interest.
-
This article contains supporting information online at www.pnas.org/cgi/content/full/0711621105/DC1.
- © 2008 by The National Academy of Sciences of the USA
