Approximately 10 years ago, Nebert proposed that drug biotransformation is but a small component of a much broader set of functions that can be ascribed to “effector ligand-metabolizing enzymes” (Nebert, 1991, 1994). The central tenet of this thesis is that small (<500 Da) endogenous ligands involved in growth and differentiation (e.g., androgens, cholesterol, eicosanoids, estrogens, progestins, retinoic acid, thyroxine, and vitamin D, among others) are subject to oxidative metabolism and conjugation by the same pathways responsible for xenobiotic detoxification. Furthermore, it is quite clear that fetal development involves the coordinated actions of complex networks of genes that interact during organogenesis, as receptor systems and signal transduction networks become established and as organ function matures in preparation for perinatal survival.

For over 40 years, estriol has been recognized as the “pregnancy estrogen”. It is an important determinant of fetal growth and development as well as timing of parturition (Mesiano and Jaffe, 1997), and its biosynthetic pathway exemplifies a network of effector ligand-metabolizing enzymes. For example, SULT2A1 is responsible for dehydroepiandrosterone 3-sulfate (DHEA-S) synthesis in the fetal adrenal, and the majority of DHEA-S subsequently undergoes 16α-hydroxylation in the fetal liver (and, to a lesser extent, within the adrenal itself) (Serón-Ferré and Jaffe, 1981), an activity that has been attributed to CYP3A7 (Kitada et al., 1987). After uptake into placental syncytiotrophoblasts, DHEA-S and 16α-hydroxydehydroepiandrosterone 3-sulfate (16αDHEA-S) are deconjugated by sulfatase to generate dehydroepiandrosterone (DHEA) and 16α-hydroxydehydroepiandrosterone (16αDHEA), the primary C₁₉ steroid precursors for aromatase (CYP19)-mediated estradiol and estriol synthesis, respectively (Ryan, 1959; Baulieu and Dray, 1963; Siiteri and MacDonald, 1963).

CYP3A7 would appear to have important functions in addition to its role in estriol biosynthesis. Catalytic activity attributed to CYP3A7 has been observed in embryonic liver as early as 50 to 60 days gestation (Yang et al., 1994). Expression has been reported to increase throughout pregnancy, peaking at 2 weeks postnatal age, and declining thereafter to the low levels characteristic of adult liver (Lacroix et al., 1997), although more recent data suggest that activity declines throughout the last trimester of pregnancy (Stevens et al., 2003). Nevertheless, DHEA-S in high concentrations is associated with inhibition of cell proliferation and progesterone synthesis, leading to speculation that functional DHEA-S 16α-hydroxylation may also play a fetoprotective function (Schuetz et al., 1993). Consistent with this hypothesis, DHEA-S and other 3-S-conjugated steroids have been reported to activate the catalytic activity of CYP3A7, but not CYP3A4 (Nakamura et al., 2003). Furthermore, given its catalytic activity toward the 4-hydroxylation of 9-cis and all-trans-retinoic acid (Chen et al., 2000; Marill et al., 2000, 2002), CYP3A7 may also provide protection against the embryotoxic effects of retinoic acid, particularly from exogenous exposure following maternal ingestion of therapeutic doses. Finally, it appears that CYP3A7 also possesses the capacity to bioactivate several promutagens, including aflatoxin B₁ (Li et al., 1997).

Given these considerations and the fact that the extent of interindividual variation in fetal liver CYP3A7 activity is essentially unknown, the purpose of this investigation was to characterize the extent of variability in CYP3A7 mRNA transcripts, protein expression, and catalytic activity toward endogenous and xenobiotic substrates in a panel of fetal liver tissues. Since the DHEA 16α-hydroxylation activities reported by Stevens et al. (2003) were approximately 10-fold higher than those reported by Lacroix et al. (1997), estimates of catalytic activity and expression levels of mRNA are critically dependent upon tissue quality. Furthermore, difficulty in discriminating between CYP3A4 and CYP3A7 activities and immunoreactive protein content has been a historical problem when using conventional probes of CYP3A activity and commercially available antibodies (Stevens et al., 2003). To address these issues, the current investigation implemented strict tissue selection criteria and utilized a variety of experimental approaches to unambiguously differentiate CYP3A7-related activity from that of CYP3A4 or CYP3A5.

Materials and Methods

Materials and Reagents. Testosterone, androstenedione, 2α-hydroxytestosterone, 2β-hydroxytestosterone, 6β-hydroxytestosterone, 15α-hydroxytestosterone, glucose-6-phospate, glucose-6-phosphate dehydrogenase, NADP, and EDTA were purchased from Sigma-Aldrich (St. Louis, MO). DHEA and 16αDHEA were purchased from Steraloids (Newport, RI). All other reagents were of analytical grade. Microsomes prepared from baculovirus-infected insect cells (SUPERSOMES) expressing human P450 enzymes CYP1A1, -1A2, -1B1, -2A6, -2B6, -2C8, -2C9, -2C18, -2C19, -2D6, -2E1, -3A4, -3A5, -3A7, or control vector were purchased from BD Gentest (Woburn, MA). Criterion 4 to 15% Tris-HCl gels and Tris/glycine/SDS running buffer were purchased from Bio-Rad (Hercules, CA). Hybond C-Super nitrocellulose, ECL anti-rabbit IgG horseradish peroxidase-conjugated antibody, ECL Plus Detection Kit, and Hyperfilm were purchased from Amersham Biosciences Inc. (Piscataway, NJ).

Liver Samples. A total of 54 liver samples 76 days to 32 weeks estimated gestational age (EGA) were obtained through the National Institute of Child Health and Human Development-supported tissue retrieval programs, 26 from the Central Laboratory for Human Embryology at the University of Washington (Seattle, WA), and 28 from the Brain and Tissue Bank for Developmental Disorders at the University of Maryland (Baltimore, MD). Available demographic data are presented in Table 1. Tissues were stored at –70°C prior to preparation of subcellular fractions. The post-mortem interval was 2 h or less for 49 samples; five had post-mortem intervals of 3 h (n = 2), 4 h (n = 1), 5 h (n = 1), and 6 h (n = 1). The use of these tissues was approved by the University of Missouri-Kansas City Pediatric Health Sciences Review Board.

Preparation of Fetal Liver Microsomes. Microsomes were prepared from 51 of the 54 human fetal livers by differential centrifugation according to the method of Lu and Levin (1972). Isolated microsomal pellets were removed from centrifuge tubes with a Teflon-coated spatula, transferred to a low-volume glass mortar (tapered, 4 ml), manually resuspended in 0.25 M sucrose with a tapered Teflon pestle, and stored at –70°C until use. Protein concentrations were determined using the Micro BCA Protein Assay kit (Pierce Chemical Co., Rockford, IL).

RNA Extraction and Quantitative Reverse Transcriptase (QRT)-PCR Analysis. Frozen liver tissues (20 to 30 mg) were homogenized and total RNA extracted according to the QIAGEN RNeasy protocol (QIAGEN, Valencia, CA) with an on-column DNase I treatment. The quality of the isolated RNA was assessed by agarose gel electrophoresis, and RNA quantity was determined spectrophotometrically. For one-step QRT-PCR reactions, RNA was diluted to 5 ng/μl, and 25 ng were used in a reaction volume of 20 μl. Primer pairs designed to specifically bind in exon 1 and exon 4 of each of the four CYP3A isoforms (Table 2) were used at a final concentration of 250 nM. Reactions were performed in triplicate with the QuantiTect SYBR Green one-step RT-PCR kit (QIAGEN) on a DNA Engine Opticon 2 instrument (MJ Research, Watertown, MA). Serial dilutions of PCR amplicons for each of the four specific CYP3A isoforms were used to generate standard curves ranging from 100 to 10⁷ copies. CYP3A mRNA transcript numbers were calculated from linear regression analysis of the respective standard curves. Data were normalized to 18S rRNA using the TaqMan Ribosomal RNA Control Reagent kit (Applied Biosystems, Foster City, CA).

Specificity of the assays for the four CYP3A isoforms was confirmed by sequencing the amplicons and further determined by comparative QRT-PCR experiments, wherein each CYP3A isoform-specific primer set was tested for specificity against 10⁶ molecules of CYP3A4, CYP3A5, CYP3A7, and CYP3A43 templates. When the CYP3A7 primer set was used under the experimental conditions described above, a cycle threshold (Ct) difference of >14 cycles was observed between the CYP3A7 template and the other three templates (i.e., Ct values were >14 cycles greater for CYP3A4, CYP3A5, and CYP3A43 than that observed with the CYP3A7 template). Using the CYP3A4, CYP3A5, and CYP3A43 primer sets, differences in Ct values were greater than 13 cycles between the cognate and alternative targets.

Preparation of Anti-CYP3A7/4 Antibody. The primary antibody was raised in rabbits against the C-terminal pentapeptide of CYP3A4 and -3A7, Thr-Val-Ser-Gly-Ala, that had been conjugated to keyhole limpet hemocyanin using sulfosuccinimidyl 4-[N-maleimidomethyl]cyclohexane-1-carboxylate chemistry via a cysteine residue added to the N terminus of the pentapeptide. Pathogen-free New Zealand White rabbits (HsdOkd:NZW) were immunized with 1 mg of peptide-keyhole limpet hemocyanin antigen emulsified in Freund's Complete Adjuvant followed by secondary immunizations on days 28, 56, and 84. Production bleeds (∼25 ml) were obtained on day 98 and day 105 followed by complete exsanguination on day 112. Antigen preparation and antibody production was carried out by Harlan (Indianapolis, IN).

To minimize nonspecific binding, the antipeptide antibody was affinity purified using the PinPoint Xa Protein Purification System (Promega, Madison, WI). Oligonucleotides (sense, 5′-AGCTTTGCACCGTGAGCGGCGCGTAAG-3′ and antisense, 5′-GATCCTTACGCGCCGCTCACGGTGCAA-3′) encoding the CYP3A4/7 pentapeptide were synthesized (Sigma-Genosys, The Woodlands, TX), allowed to anneal, phosphorylated with T4 polynucleotide kinase, and ligated into the PinPoint Xa-3 vector between the HindIII and BamHI restriction sites to create an in-frame fusion with the biotinylated target sequence encoded by the vector. Clones were sequenced to confirm the presence of the oligonucleotide insert. The pentapeptide epitope was expressed in Escherichia coli as the C terminus of a biotinylated fusion protein. An affinity column was prepared by immobilization of the biotinylated fusion protein on TetraLink Tetrameric Avidin Resin (Promega). Antipeptide antibodies were allowed to bind to the fusion protein solid phase, the column washed with >5 bed volumes of buffer, and bound antibodies eluted with 100 mM glycine, pH 2.8. The antibody was specific for CYP3A4 and -3A7 and did not react with heterologously expressed CYP1A1, -1A2, -2A6, -2B6, -2C8, -2C9, -2C19, -2D6, -2E1, or -3A5.

Immunoquantitation of Microsomal CYP3A7/4 Protein. Fetal liver microsomes were diluted in sample buffer (10% glycerol, 141 mM Tris base, 106 mM Tris-HCl, 2% SDS, 0.51 mM EDTA, 0.22 mM SERVA Blue G250, and 0.175 mM Phenol Red) containing fresh 50 mM dithiothreitol and heated for 10 min at 70°C. Microsomal proteins, 0.5 μg per lane, were separated on 4 to 15% Criterion 26-well gels in running buffer (25 mM Tris, 192 mM glycine, and 0.1% w/v SDS, pH 8.3) at 200 V for 45 min. Proteins were transferred to nitrocellulose membranes using a Bio-Rad Semi-dry transfer unit with transfer buffer containing 39 mM glycine, 48 mM Tris, 0.0375% SDS, and 20% methanol. Membranes were blocked overnight at room temperature on a rocker in 10 mM Tris, 150 mM NaCl, 0.2% Tween 20, pH 8.0 (TNT) containing 4% skim milk powder (M-TNT). Blocked membranes were incubated for 1 h with affinity-purified anti-CYP3A7/4 antibody diluted 1:20,000-fold in M-TNT and washed with TNT (5 min × 6). To detect bound antibody, blots were incubated with horseradish peroxidase-conjugated donkey anti-rabbit antibody (1:50,000) in M-TNT for 30 min, washed (5 min × 6), incubated with ECL Plus chemiluminescence reagents according to the manufacturer's directions (Amersham Biosciences Inc.), and exposed on Hyperfilm ECL. Films were scanned using a flat-bed scanner, and densitometric analysis of immunoreactive protein was conducted using Kodak Digital Science 1D Image Analysis Software, version 3.6 (Eastman Kodak Company, Rochester, NY). Standard curves (0.025 to 0.3 pmol per lane heterologously expressed CYP3A7; BD Gentest) were present on each membrane. Standard curves were linear over the range of standards, and coefficients of determination (r² values) ranged from 0.972 to 0.999. Back-extrapolated values for the standards were within 3.6 to 8.8% of the nominal values between 0.3 and 0.05 pmol, and within 29.9% of the nominal value for the 0.025 pmol standard. Values presented are the mean of duplicate determinations.

Incubation Conditions for Catalytic Activity Assays. In vitro enzyme assays were performed in 96-well microtiter plates. Standard incubation reactions (100 μl) contained microsomes prepared from fetal human liver (5–50 μg of microsomal protein) or insect cell microsomes containing baculovirus-expressed cytochrome P450 enzymes (1–5 pmol) coexpressed with P450 reductase, potassium phosphate buffer (50 mM, pH 7.4), MgCl₂ (3 mM), EDTA (1 mM), and substrate (250 μM testosterone; 100 μM DHEA). Reactions were initiated by the addition of an NADPH-generating system consisting of NADP (1 mM), glucose 6-phosphate (1 U/ml), and glucose-6-phosphate dehydrogenase (5 mM), placed in a Thermo Forma (Marietta, OH) Benchtop Orbital Shaker incubator at 37 ± 0.1°C, and terminated after 10 to 30 min by the addition of 100 μl of ice-cold methanol. Protein was precipitated by centrifugation at 10,000g_max for 10 min. An aliquot (20–100 μl) of the supernatant was analyzed by reversed-phase HPLC via direct injection. Under these conditions, metabolism of the parent compounds did not exceed 20%, and rates of metabolite formation were proportional to incubation time and protein concentration. Duplicate determinations were performed for each sample.

Analytical Methods. HPLC analyses were performed with a Hewlett Packard HP1100 HPLC system with programmable 1100 series diode array and fluorescence detectors (Hewlett Packard Instruments, Santa Clara, CA). All data were collected and integrated with Hewlett Packard Chemstation V A.0401 software. Testosterone and its 2α-, 2β-, and 6β-hydroxylated metabolites were resolved by reversed-phase HPLC via direct injection according to the method of Purdon and Lehman-McKeeman (1997), as modified by Usmani et al. (2003). Under these conditions, 6α-, 15α-, 6β-, 2α-, and 2β-hydroxytestosterone, androstenedione, and testosterone had retention times of 14.8, 15.9, 16.4, 20.8, 21.6, 25.0, and 28.0 min, respectively. The solvent program was modified slightly such that 16αDHEA eluted at 17.0 min and DHEA eluted at 27.2 min. The analytical methods demonstrated linearity (r² = 0.999) over the range of standards evaluated (5 to 1250 pmol of testosterone metabolites and 5 to 1500 pmol of 16αDHEA injected).

Genomic DNA Isolation and Genotyping forCYP3A7*1B, -*1C,and-*1EAllelic Variants. Genomic DNA was isolated from 5 to 25 mg of tissue using a DNeasy Tissue Kit (QIAGEN). CYP3A7*1C genotyping was adapted from Burk et al. (2002) with the following modifications: a 2468-bp long fragment was amplified from genomic DNA in a 15-μl reaction using JumpStart REDTaq DNA polymerase (Sigma-Aldrich), the buffer provided with the enzyme, and 0.2 μM each primer (listed in Table 2). After PCR amplification, 12 μl of a mixture containing 2.5 units of SspI and NEB digestion buffer “U” (New England Biolabs, Beverly, MA) was added to each reaction and incubated overnight at 37°C. Restriction fragments were separated on an agarose gel by electrophoresis and documented with a Kodak 440 CF Image Station (Eastman Kodak Co., New Haven, CT). A CYP3A7*1/*1C sample was included as a positive control.

To confirm the CYP3A7*1C genotype and to genotype for CYP3A7*1B, -*1D, and -*1E alleles, genomic DNA was amplified with JumpStart REDTaq DNA polymerase (Sigma-Aldrich) and 0.2 μM oligonucleotide primers 3A7 197291-F (5′-CACCTCTGCTAAGGGAAACAGGCC) and 3A7 198150-R (5′-GCCAGCCTGAACATCCTTTTTGCTA). Specificity of both PCR primers was achieved by aligning the four CYP3A genes and the two pseudogene sequences CYP3AP1 and CYP3P2 using the LAGAN algorithm (Brudno et al., 2003). The 859-bp amplicon encompassing 680 bp of the proximal promoter, exon 1 and 109 bp of intron 1 was treated with ExoSAP-IT [USB (U.S. Biochemical Corp.), Cleveland, OH] and subsequently sequenced with the DY-Enamic ET Dye Terminator Cycle Sequencing Kit for MegaBACE (Amersham Biosciences Inc.). Unincorporated fluorescent dye-terminators were removed from the reactions by solid-phase magnetic CleanSEQ beads (Agencourt, Beverly, MA) according to the manufacturer's recommendations prior to analysis on a MegaBACE 500 DNA Analysis System. The data were analyzed with the Phred/Phrap/Consed software package (University of Washington, Seattle, WA).

Genotyping forCYP3A7*2by Extra-Long-PCR Sequencing. To detect the g.26041C>G (Thr409Arg) variant, a CYP3A7-specific 6.51-kb long PCR amplicon including exons 11 to 13 was generated with Platinum Taq DNA Polymerase High Fidelity (Invitrogen, Carlsbad, CA) and primers 3A7 223378-F (5′-CAATAATCCTTGTCGCACAGAGAATTTG) and 3A7 229891-R (5′-ATAATTTGGAGTTATCATTTGGAGGGTCT). Reaction conditions were as follows: denaturing, 94°C for 20 s; annealing, 64°C for 30 s; and extension, 68°C for 6 min, for 40 cycles. The GenBank entry NG_000004.2 was used to design the amplification primers. The PCR reaction was subsequently cleaned up as described above, and a 0.8-μl aliquot was sequenced with primers 3A7 223735-F (5′-TCAACAGTACTACATGGACTG) and 3A7 224241-R (5′-GGGACTGTGACTGGCTATAG).

GenotypingCYP3A5Allelic Variants. PCR-RFLP-based assays for CYP3A5*3, CYP3A5*4, and CYP3A5*5 were adapted from published methods (van Schaik et al., 2002) or developed de novo (CYP3A5*2, CYP3A5*6, and CYP3A5*7). Complete assay details are presented in Table 3. Assays for CYP3A5*2, CYP3A5*3, and CYP3A5*7 were conducted using genomic DNA with 40 cycles of amplification. Improved assay performance for the CYP3A5*4, CYP3A5*5, and CYP3A5*6 alleles was achieved using an 8152-bp long CYP3A5-specific template generated with Platinum Taq DNA Polymerase High Fidelity according to the manufacturer's recommendations (Invitrogen) with 10-s annealing at 70°C and 9-min extension at 68°C for 35 cycles (primers listed in Table 3). All other PCR reactions were conducted with JumpStart REDTaq DNA polymerase (Sigma-Aldrich) using the buffer provided with the enzyme except for the CYP3A5*7 product, which was generated using Fail-Safe buffer C (Epicenter Technologies, Madison, WI). After PCR amplification, an equal volume (8 μl) containing the relevant restriction enzyme (New England Biolabs and Amersham Biosciences Inc.) and digestion buffer was added to each reaction and incubated for at least 2 h. PCR-RFLP fragments were separated by gel electrophoresis (3% agarose gels containing Synergel; Diversified Biotech, Boston, MA) and documented as described above.

Statistical Analysis. Results are reported as the mean ± S.D. Measures of linkage disequilibrium were determined using Haploview version 3.11 (Barrett et al., 2005). Univariate linear regression and ANOVA with Tukey's post hoc analysis were conducted using SPSS version 12 for Windows (SPSS Inc., Chicago, IL). p < 0.05 was accepted as a statistically significant difference.

Results

Variability inCYP3AGene Transcription in Human Fetal Liver. Expression of CYP3A mRNAs is reported as transcripts per nanogram total RNA corrected for the expression of 18S rRNA as recommended by Koch et al. (2002). Although RNA quality demonstrated some variability across the panel of specimens, prominent 18S and 28S bands were evident in all samples (not shown), and values for 18S rRNA expression varied only 3.9-fold in this sample set (mean ± S.D., 4.97 ± 1.55 × 10⁷ transcripts/ng total RNA; range, 2.48 × 10⁷ to 9.73 × 10⁷ transcripts/ng total RNA).

CYP3A7 transcripts were the most abundant of the individual CYP3A isoforms (Fig. 1) and expression varied 634-fold in this panel of human fetal liver samples. Mean ± S.D. (range) expression was 14,200 ± 15,000 transcripts/ng total RNA (151 to 95,700 transcripts/ng total RNA; Table 4). A subset of samples had markedly reduced levels of CYP3A7 mRNA, protein, and activity (vide infra). When these five samples were excluded from the analysis, variability in CYP3A7 mRNA expression decreased from 634- to 72-fold. On an individual basis, CYP3A5 mRNA expression was, on average, 100-fold lower than CYP3A7 expression and varied 123-fold (254 ± 321 transcripts/ng total RNA; range 13.5 to 1670 transcripts/ng total RNA) across the full panel. CYP3A4 mRNA expression was approximately 1000-fold lower than CYP3A7 expression and varied 10.3-fold (15.1 ± 8.9 transcripts/ng total RNA; range 4.7 to 48.2 transcripts/ng total RNA). Expression of CYP3A43 mRNA was similar to that observed for CYP3A4. There was no relationship between the expression of CYP3A7 and either CYP3A5 (r² = 0.103) or CYP3A4 (r² = 0.080) nor was the expression of any CYP3A mRNA associated with EGA (Fig. 1).

Variability in CYP3A7/4 Immunoreactive Protein. Immunoreactive CYP3A7/4 protein content was determined for 51 of the 54 fetal liver samples and averaged 234.8 ± 123.1 pmol CYP3A7/4 protein/mg microsomal protein (Table 4). A representative immunoblot (standard curve: r² = 0.972) illustrating a range of immunoreactive protein contents is presented in Fig. 2A. CYP3A7 remained undetectable in one sample (CMM1153) even when the amount of protein loaded was increased 4-fold to 2 μg/lane. For samples with detectable protein, CYP3A7/4 content varied 21.6-fold from 20.6 to 439.9 pmol/mg microsomal protein. Although samples with the lowest CYP3A7 mRNA expression also had the lowest immunoreactive CYP3A7/4 protein expression, the overall correlation was poor (r² = 0.129). Likewise, immunoreactive CYP3A7/4 protein across all samples was poorly correlated with EGA (r² = 0.149; Fig. 2B). Significant correlations were, however, observed between protein content and both testosterone 2α-hydroxylation (T2αH; r² = 0.541, p < 0.01; Fig. 2C) and DHEA 16α-hydroxylation (DHEA16αH; r² = 0.585, p < 0.01; Fig. 2D) activities.

Testosterone Biotransformation. Since published data indicated that human CYP3A isoforms demonstrated regiospecific patterns of testosterone biotransformation (Usmani et al., 2003), we investigated the potential for testosterone 6β-hydroxylation (T6βH) and T2αH activities to determine the relative contributions of CYP3A isoforms toward catalytic activities in human fetal liver. Product formation using heterologously expressed rCYP3A4, rCYP3A5, and rCYP3A7 (and experimental conditions identical for those to be employed for incubations containing fetal liver microsomal protein) are presented in Table 5. The ratio of T2αH and T6βH activities (T2αH/T6βH ratio) appeared to be particularly useful for discriminating among the three CYP3A isoforms since the rCYP3A7 value of 1.19 was an order of magnitude greater than the rCYP3A5 value of 0.079, which was approximately 10-fold greater than the ratio of 0.007 determined with rCYP3A4 (Table 5).

In the fetal liver microsomal samples, the T2αH/T6βH ratio was 1.71 ± 0.32 with a minimum value of 0.54 and a maximum value of 2.14 (Fig. 3). The value of 1.19 associated with rCYP3A7 was exceeded in 49 of 51 samples, the exceptions being samples with values of 0.54 and 1.17, respectively. The lowest ratio, 0.54, was observed in a sample with very low T2αH and T6βH activities of 6.4 and 11.9 pmol/min/mg protein, respectively. To address potential contributions from CYP3A5, the samples were genotyped for the CYP3A5 intron 3 polymorphism (g.6986A>G). ANOVA revealed a statistically significant inverse relationship between the T2αH/T6βH ratio and the number of functional g.6986A alleles (reference sequence as defined by the Human Cytochrome P450 Nomenclature Committee; http://www.imm.ki.se/CYPalleles/cyp3a5.htm; F = 7.01, p = 0.002; not shown). The presence of functional CYP3A5 protein has been confirmed in these samples by immunoblot analysis and midazolam hydroxylation (data not shown; manuscript in preparation). Although considerable overlap among groups was observed, post hoc analysis with Tukey's honestly significant difference confirmed that at least one g.6986A allele resulted in significantly lower ratio values compared with two g.6986G alleles. It should be noted that both CYP3A5*1 and -*6 alleles genotype as g.6986A, although the CYP3A5*6 allele is associated with reduced activity due to an additional downstream splicing defect (Rogan et al., 2003). For the sample with the lowest T2αH/T6βH ratio, the contribution of CYP3A5 likely is minimal since this sample was genotyped as CYP3A5*3/*6, and neither immunoreactive CYP3A5 protein nor midazolam hydroxylation activity was detected (not shown).

T2αH activity varied approximately 175-fold from 3.6 to 642.9 pmol/min/mg (323.1 ± 171.1 pmol product formed/min/mg microsomal protein); the majority (46 of 51) of samples fell within a 6.7-fold range with five samples displaying markedly reduced activities. Similarly, T6βH activity varied 167-fold (186.6 ± 105.5 pmol/min/mg; range 2.6 to 439.4 pmol/min/mg) and decreased to 9.1-fold in the absence of the low activity subgroup. A strong correlation was observed between T6βH and T2αH activities (r² = 0.859; Fig. 4A) with the greatest deviations from the fitted model parameters occurring in those samples with the highest T2αH activities. For those livers in which CYP3A5 protein theoretically should be expressed (at least one g.6986A allele; solid symbols, Fig. 4A), the coefficient of determination was essentially identical to that observed with the entire data set (r² = 0.861). Moreover, in the subset of samples genotyped as CYP3A5*3/*3 (n = 28), T2αH and T6βH activities were almost exclusively catalyzed by a single enzyme (r² = 0.974; open symbols, Fig. 4A). In Fig. 4A, those samples genotyped as CYP3A5*3/*6 tended to associate more with the CYP3A5*3/*3 samples, whereas the sole CYP3A5*6/*6 specimen fell close to the line of best fit for the g.6986A group. A weak (r² = 0.194) but statistically significant (F = 11.83, p = 0.0012) relationship was observed between T2αH activity and EGA. However, removal of the set of samples with extremely low activity abolished the statistical significance.

DHEA16αH Activity. Fetal liver microsomes were more active toward DHEA as a substrate than toward testosterone. Mean ± S.D. DHEA16αH activity was 7.30 ± 3.61 nmol product formed/min/mg microsomal protein. Values varied approximately 250-fold ranging from 0.059 to 14.91 nmol/min/mg when all samples were included in the analysis but varied only 6.2-fold when the low activity samples were excluded. DHEA16αH activity was significantly correlated with T2αH activity (r² = 0.856, p < 0.0001) but was not affected by CYP3A5 genotype (Fig. 4B). A weak (r² = 0.256) but statistically significant (F = 15.96, p = 0.0002) correlation was observed between DHEA16αH activity and EGA, but as observed for T2αH activity, this relationship was dependent upon the presence of the set of samples with extremely low activity (not shown). There was no relationship between DHEA16αH activity and the total length of time that the tissues were stored prior to isolation of microsomes (r² = 0.002).

Association ofCYP3A7Allelic Variants with Measures ofCYP3A7Expression. Since DHEA16αH activity did not appear to be influenced by CYP3A5 genotype, subsequent phenotype-genotype correlations were evaluated using this measure. No CYP3A7*1B, -*1C, or -*1D alleles were observed in this sample set, whereas three African-American and one Caucasian sample were heterozygous for the CYP3A7*1E allele (F = 0.037). No relationship between CYP3A7*1E and DHEA16αH activity was apparent (values of 0.09, 4.4, 9.0, and 12.4 nmol/min/mg).

The frequency of the CYP3A7*2 allele in this set of fetal samples was 0.378 and was in linkage disequilibrium with CYP3A5*3 (D′ = 0.90 with 95% confidence interval 0.76 to 0.93; r² = 0.74). No association was observed between DHEA16αH activity and the number of CYP3A7*2 alleles by ANOVA (F = 0.326, p = 0.724; Fig. 5).

Discussion

Historically, CYP3A7 has been considered a “fetal-specific” form of cytochrome P450, but little is known concerning the extent of “normal” variation in its expression or the genetic and environmental determinants underlying that variability. Stevens et al. (2003) have identified several critical technical and logistical issues contributing to this knowledge deficit including 1) limited numbers of samples within targeted developmental stages; 2) difficulty acquiring tissues of sufficiently high quality to generate meaningful data in in vitro studies; and 3) an inability to accurately and reliably ascertain the relative contributions of CYP3A4 and -3A5 to observed CYP3A7 activity using currently available substrates and reagents.

Although 170- to 250-fold variability in testosterone and DHEA hydroxylation and 634-fold variability in CYP3A7 mRNA expression was observed across the entire panel, most of this variability could be attributed to the presence of a subset (n = 5) of samples with extremely low values. Variability in catalytic activity was 10-fold or less for the remaining samples (Table 4). It is unlikely that poor sample quality was a major factor contributing to the observed interindividual variability for several reasons. The post-mortem interval was2hor less for 49 of the 54 samples. Furthermore, 18S rRNA values demonstrated minimal (3.9-fold) variation, consistent with the observations of Koch et al. (2002) in adult liver, and no correlation was observed between either the post-mortem interval or 18S rRNA value and any measure of CYP3A7 expression.

A major concern in characterizing the extent of variability in CYP3A7 activity has been ascertaining the contribution of even minimal amounts of CYP3A4 (or CYP3A5) present in a sample due to the overlapping activities toward probe substrates by these closely related CYP3A isoforms and the lack of specificity of antibody reagents used for immunoquantitation (Stevens et al., 2003). In the set of fetal liver samples investigated for this report, expression of CYP3A4 appeared to be minimal based on relative mRNA expression and catalytic activity toward testosterone. The average T6βH activity in this panel of fetal liver microsomes was 186.6 pmol/min/mg protein with a maximum value of 439.4 pmol/min/mg, markedly lower than activities of approximately 800 to 14,000 pmol/min/mg reported in adult liver microsomes (Pearce et al., 1996). Observed regiospecific patterns of testosterone hydroxylation also were more consistent with a pattern of catalysis dominated by CYP3A7 activity rather than by CYP3A4 activity (Usmani et al., 2003). For example, T2αH activity measured in fetal liver microsomes exceeded T6βH activity in all samples but one, and in 49 of 51 samples, the observed T2αH/T6βH ratio was greater than the value of 1.186 established with rCYP3A7. Moreover, the lowest observed ratio in fetal liver microsomes (0.535) was considerably greater than the values of 0.079 and 0.007 observed with rCYP3A5 and rCYP3A4, respectively. Finally, visual inspection of HPLC chromatograms from analyses of DHEA biotransformation failed to reveal any peaks that would correspond to 7β-hydroxyDHEA formation by fetal liver microsomes, a product that is highly specific to CYP3A4 (Stevens et al., 2003).

In contrast, CYP3A5 may be more likely to confound estimates of CYP3A7 expression since CYP3A5 immunoreactive protein has been detected in approximately 50% of fetal liver microsomal samples (Hakkola et al., 2001; Stevens et al., 2003). In our panel, genotyping for CYP3A5 g.6986A>G identified a subset of samples where the expression of CYP3A5 protein was predicted to be negligible (CYP3A5*3/*3). The high correlation between T2αH and T6βH activities (r² = 0.974) for CYP3A5*3 homozygotes strongly suggests that a single enzyme, CYP3A7, was primarily responsible for both activities. In contrast, DHEA16αH activity was not affected by CYP3A5 genotype providing support that it is a more specific marker of CYP3A7 activity than testosterone in fetal liver microsomes, i.e., under conditions where CYP3A4 expression is minimal or absent. The fact that CYP3A5 genotyping segregates testosterone hydroxylation activity, but not DHEA16αH activity, into two groups suggests that CYP3A5 may contribute to the biotransformation of some substrates but not others in fetal liver. A more detailed characterization of CYP3A5 activity in this panel is currently in preparation.

Our estimate of average CYP3A7 content in fetal liver microsomes is in agreement with one published report (Stevens et al., 2003) but higher than that reported by others (Lacroix et al., 1997). Stevens et al. exploited differential patterns of DHEA 16α- and 7β-hydroxylation by CYP3A7 and CYP3A4 in conjunction with a validated nonlinear multivariate regression model to calculate CYP3A4 and CYP3A7 protein in a subset of pre- and postnatal livers. Using this innovative but indirect measure, they observed an average CYP3A7 content of 311 pmol/mg microsomal protein in the second trimester, 201 pmol/mg in the third trimester, and 158 pmol/mg in premature birth samples, suggesting that increasing gestational age may also contribute to the observed variability in CYP3A7 expression (Stevens et al., 2003). Published data are conflicting in this regard (Lacroix et al., 1997), and only a weak association between EGA and CYP3A7 activity was observed across the entire panel of samples in our investigation due, most likely, to the limited number of late term samples. Nevertheless, it is likely that factors other than EGA are responsible for the extremely low values of CYP3A7 activity observed in our study since Stevens et al. (2003) reported only a 50% decrease in CYP3A7 content between second trimester samples and samples from prematurely born infants.

Since CYP3A5 genotype did not have any discernable effect on DHEA16αH activity, this measure was considered to best reflect variability in CYP3A7 expression. We speculate that the extreme range observed in the current study does not reflect normal variability in fetal liver CYP3A7 activity since variability was only 6.2-fold when the subset of low samples was excluded. Furthermore, three of the samples in the “low activity” group were obtained from fetuses with severe congenital anomalies—anencephaly, porencephaly, and hydrops fetalis. Although hepatic dysfunction generally is not considered to be associated with either anencephaly or porencephaly, hypothalamic-pituitary function is disrupted in anencephalic fetuses such that adrenocorticotropin levels are insufficient to support fetal adrenal development as inferred from dramatically reduced maternal estrogen concentrations (Mesiano and Jaffe, 1997). Although regulation of estrogen biosynthesis is severely dysregulated under these conditions, the effects on CYP3A7 activity are less clear. Milewich et al. (1986) observed that 16α-hydroxylase activity toward estrone sulfate and estradiol sulfate was similar in anencephalic fetuses (gestational ages not reported) compared with normal fetuses, although activities were markedly lower in one of a pair of twin anencephalic fetuses. However, the relative paucity of available information makes it impossible to know whether the extremely low CYP3A7 activity was due to fetal pathology or poor tissue quality (subject to the considerations discussed above). The 6- to 7-fold variation in activity observed in the majority of samples more likely represents true variability within the viable fetal population.

Persistence of CYP3A7 mRNA expression postnatally has been associated with the presence of the CYP3A7*1C allele (Kuehl et al., 2001). No CYP3A7*1C alleles were observed among our 54 fetal samples, although we have observed an allele frequency of 2.3% in an African-American control population (n = 87). Thus, the CYP3A7*1C allele does not contribute to the observed variability in mRNA expression or catalytic activity observed in fetal liver. Genotyping for the previously described CYP3A7*2 allele (Rodriguez-Antona et al., 2005) also could not account for the variation observed in any of the in vitro phenotypic measures investigated. Although Rodriguez-Antona et al. (2005) reported moderately higher DHEA hydroxylation for the CYP3A7*2 allele, the tendency of this allele to demonstrate higher DHEA16αH activity among our fetal panel did not achieve statistical significance. However, genotype-phenotype correlations may be obscured by the fact that CYP3A7 activity has been reported to be activated by steroid sulfate conjugates, such as pregnenolone 3-sulfate, 17α-hydroxypregnenolone 3-sulfate, and DHEA-S, but not the corresponding unconjugated forms (Nakamura et al., 2003). Thus, activity/phenotype of the wild-type and variant proteins may be differentially affected by the specific substrate concentration used in the in vitro analyses.

The overall variability in CYP3A7 activity observed in this study may seem inconsistent with an important physiological function, but exclusion of a low activity subgroup resulted in only 6- to 7-fold variability in catalytic activity. Comparative analyses revealed that in vitro clearance of prototypic CYP3A substrates is generally 5- to 10-fold lower for CYP3A7 compared with CYP3A4 (Williams et al., 2002), implying a primarily physiological role for CYP3A7 in fetal liver. Given the importance of CYP3A7 to estriol biosynthesis during pregnancy and its potential to provide protection against exogenous sources of retinoic acid esters, a thorough understanding of allelic variation, alternative splicing, and developmental expression of trans-acting factors potentially involved in regulating CYP3A7 expression, such as pregnane X receptor, constitutive androstane receptor, retinoid X receptor, and the vitamin D receptor, is warranted, and relevant investigations currently are in progress.