Abstract
Pregnancy-induced changes in drug pharmacokinetics can be explained by changes in expression of drug-metabolizing enzymes and transporters and/or normal physiology. In this study, we determined gestational age-dependent expression profiles for all metabolic enzyme and transporter genes in the maternal liver, kidney, small intestine, and placenta of pregnant mice by microarray analysis. We specifically examined the expression of genes important for xenobiotic, bile acid, and steroid hormone metabolism and disposition, namely, cytochrome P450s (Cyp), UDP-glucuronosyltranserases (Ugt), sulfotransferases (Sult), and ATP-binding cassette (Abc), solute carrier (Slc), and solute carrier organic anion (Slco) transporters. Few Ugt and Sult genes were affected by pregnancy. Cyp17a1 expression in the maternal liver increased 3- to 10-fold during pregnancy, which was the largest observed change in the maternal tissues. Cyp1a2, most Cyp2 isoforms, Cyp3a11, and Cyp3a13 expression in the liver decreased on gestation days (gd) 15 and 19 compared with nonpregnant controls (gd 0). In contrast, Cyp2d40, Cyp3a16, Cyp3a41a, Cyp3a41b, and Cyp3a44 in the liver were induced throughout pregnancy. In the placenta, Cyp expression on gd 10 and 15 was upregulated compared with gd 19. Notable changes were also observed in Abc and Slc transporters. Abcc3 expression in the liver and Abcb1a, Abcc4, and Slco4c1 expression in the kidney were downregulated on gd 15 and 19. In the placenta, Slc22a3 (Oct3) expression on gd 10 was 90% lower than that on gd 15 and 19. This study demonstrates important gestational age-dependent expression of metabolic enzyme and transporter genes, which may have mechanistic relevance to drug disposition in human pregnancy.
Introduction
Significant physiologic changes during pregnancy are essential to support and protect the developing fetus (Carlin and Alfirevic, 2008). Changes in maternal physiology include, among others, increased renal blood flow and glomerular filtration rate (GFR) (Davison and Hytten, 1975), as well as increased portal vein blood flow (Carlin and Alfirevic, 2008). These physiologic changes alter important drug pharmacokinetics (PK) determinants such as GFR, oral absorption, plasma volume, and plasma protein binding (Anderson, 2005; Anger and Piquette-Miller, 2008; Klieger et al., 2009). Considerable data in the literature also suggest that expression and activities of drug-metabolizing enzymes and transporters are altered during pregnancy (Tracy et al., 2005; Hebert et al., 2008; Feghali and Mattison, 2011). In humans, cytochrome P450 (CYP) enzymes such as CYP2A6, CYP3A4, CYP2D6, and CYP2C9 demonstrate increased activities during pregnancy, whereas the activities of CYP1A2 and CYP2C19 are decreased (Dempsey et al., 2002; Anderson, 2005; Hodge and Tracy, 2007). These changes result in altered drug PK during pregnancy (Unadkat et al., 2007; Hebert et al., 2009).
The trends in human pregnancy regarding expression/activity of metabolic enzymes and drug transporters appear to be similar in pregnant mice. Hepatic CYP3A activity in pregnant mice, for example, is increased to a similar extent as CYP3A in humans (compared with nonpregnant controls) (Mathias et al., 2006; Zhang et al., 2008; Zhou et al., 2010). In addition, hepatic Cyp1a2 mRNA is decreased during mouse pregnancy, which is consistent with CYP1A2 activity in pregnant women (Tracy et al., 2005; Koh et al., 2011). Fewer studies have been done in humans and animals to investigate changes in activity of drug transporters in the liver, kidney, and small intestine during pregnancy. P-glycoprotein (P-gp) is perhaps the most studied of all transporters, but discrepancies exist between mice and humans. In humans, P-gp–mediated renal secretion of digoxin increased during late gestation (Hebert et al., 2008). However, during mouse pregnancy, not only were P-gp protein levels unaffected in the liver, kidney, and small intestine, but mRNA expression actually decreased in the kidney (Mathias et al., 2006; Zhang et al., 2008). Despite this difference, the role of ATP-binding cassette (ABC) transporters (i.e., P-gp and breast cancer resistance protein) in determining fetal drug exposure has been extensively studied in mouse models because of the similar efflux function of these transporters in mouse and human placenta (Smit et al., 1999; Jonker et al., 2000; Zhang et al., 2007; Zhou et al., 2008).
Current research merely provides snapshots of gene expression or protein activity during pregnancy (Zhou et al., 2008; Zhou et al., 2010). Only in the last several years has there been any attempt to understand gene or protein expression throughout gestation (Wang et al., 2006; Aleksunes et al., 2008; Zhang et al., 2008; Koh et al., 2011). This body of research targeted specific CYP isoforms, as well as important solute carrier (SLC) and ABC transporters. None of these studies, however, offer a comprehensive overview of trends in expression as gestation progresses for all mouse isoforms related to drug metabolism and disposition. After considering the similarities between pregnant mice and women, we chose the pregnant mouse as an appropriate animal model to study gestational age-dependent changes in expression of drug disposition genes. Therefore, the goal of this study was to analyze the global gene expression profiles of maternal tissues and placenta at different gestational ages using microarray approaches.
Although the focus of our study was to investigate gestational age-dependent changes in metabolic enzymes and transporters relevant to drug disposition, we could not neglect the fact that changes in Cyp, Abc, Slc, or Slco gene expression during pregnancy could potentially impact the homeostasis of endogenous substances such as bile acids and steroid hormones. Characterization of genes involved in bile acid synthesis and distribution during pregnancy may provide a physiologic basis for understanding complications that arise during pregnancy such as intrahepatic cholestasis of pregnancy. Characterization of steroid hormone production during pregnancy supports mechanistic explanations of gestational age-dependent expression of metabolic enzymes and transporters. Thus, we systematically investigated expression of all metabolic enzymes and transporters in the mouse maternal tissues and placenta across gestation, demonstrating significant changes in the expression of many genes including those important for drug, bile acid, and steroid hormone metabolism and transport. These data also provide novel insights into potential changes in drug PK during pregnancy, and support the growing foundation of evidence clinicians need to make decisions regarding dosage selection during pregnancy.
Materials and Methods
Animal Studies.
FVB wild-type mice, aged 7–10 weeks, were purchased from Taconic Farms (Hudson, NY), and cared for in accordance with the Guide for the Care and Use of Laboratory Animals published by the National Research Council. The animal protocol for this project was approved by the Institutional Animal Care and Use Committee at the University of Washington (protocol number 4035-01). Briefly, mice were maintained under 12-hour light/dark cycles, and food was provided ad libitum. Female mice, aged 7–10 weeks, were mated with male mice of the same age overnight. Gestation day (gd) 1 was defined as the presence of a sperm plug after overnight housing; gd 0 was defined as nonpregnant mice. Progression of pregnancy was monitored by visual inspection and body weight increase. On gd 0, 7.5, 10, 15, and 19, female mice (n = 5–6 per gestational age) were sacrificed under anesthesia (isoflurane) by cardiac puncture, and the maternal liver, kidney, and small intestine were collected. The placentas were collected on gd 10, 15, and 19. All tissues were rinsed with phosphate-buffered saline, snap-frozen in liquid nitrogen, and stored at −80°C until use.
Total RNA Extraction.
Total RNA was isolated from the snap-frozen mouse tissues using the miRNeasy mini kit (Qiagen, Valencia, CA) according to the manufacturer’s instructions for purification of total RNA from animal tissues. The integrity of RNA samples was assessed with an Agilent 2100 Bioanalyzer (Agilent Technologies Inc., Santa Clara, CA),which is the recognized standard in the field. RNA integrity was evaluated using the RNA integrity number and by observing distinct and sharp 18s and 28s ribosomal RNA peaks. The RNA integrity number of all samples was greater than 9. RNA quantity and purity were determined with a NanoDrop 1000 Spectrophotometer (Thermo Fisher Scientific Inc., Wilmington, DE) by measuring optical density (OD260), as well as OD260/280 and OD260/230 ratios, respectively. Samples with OD260/280 ratios greater than 1.9 and OD260/230 ratios greater than 1.6 were considered acceptable for further processing. Five RNA samples per gestation day that passed these stringent quality control measures were selected from each tissue group for further processing.
Microarray Hybridization and Data Analysis.
Processing of the RNA samples was carried out according to the AffymetrixGeneChip Whole Transcript Sense Target labeling protocol (http://www.affymetrix.com/index.affx) as reported previously (Anderson et al., 2011). The arrays were scanned with an AffymetrixGeneChip 3000 scanner. Image generation and feature extraction were performed using the AffymetrixGeneChip Command Console software. Affymetrix Mouse Gene 1.0 ST Arrays were used for this study. The output files from AffymetrixGeneChip Command Console were further processed using the Bioconductor oligo package (Carvalho and Irizarry, 2010). Raw data were normalized using quantile normalization and then summarized at the transcript level using a robust multi-array average method (Irizarry et al., 2003). For the maternal tissues, we made univariate comparisons of gd 7.5, 10, 15, and 19 to gd 0 using the Bioconductor limma package, which fits an analysis of variance model to the data and then computes individual contrasts. For the placenta, univariate comparisons of gd 10 and 15 to gd 19 were also made using Bioconductor. The limma package uses an empirical Bayes adjustment on the variance estimate (based on all genes on the array) to increase power to detect true differences (Smyth, 2004). P values were adjusted for multiplicity with the program q-value, which allows the selection of statistically significant genes while controlling the estimated false discovery rate (FDR) (Tusher et al., 2001; Dabney and Storey, 2006).
The adjusted P values or the FDR values are only interpretable when genes are ranked by unadjusted P values, from the smallest to the largest. The corresponding FDR value for an unadjusted P value of 0.05 varies for each gestational age and tissue. A given FDR value states that the group of genes listed above the corresponding unadjusted P value cut-off has a particular FDR associated with it. This FDR value represents a percentage of genes within the group list whose fold changes in expression are there by chance alone (i.e., false positive results). FDR values are therefore displayed as supplemental material for each gestation day comparison with gd 0 for maternal tissues or gd 19 for placenta (Supplemental Tables 1–14). Microarray data presented in this study were deposited in the National Center for Biotechnology Information Gene Expression Omnibus data repository under accession number GSE41438.
Filtration Criteria for Differentially Expressed Genes.
Expression patterns of differentially expressed genes were visualized by creating heatmaps of relevant genes. Ordering of genes in these plots was determined by hierarchical clustering of the gene expression values.
For the overall gene expression trends shown in heatmaps, genes with a larger than 2-fold or less than 0.5-fold change in expression compared with gd 0, an unadjusted P value less than 0.05, and a log2 average fluorescent intensity greater than 4, for at least one gestation day, were considered to be differentially expressed for the maternal liver, kidney, and small intestine. Because the placenta is an organ undergoing development as gestation progresses, the number of differentially expressed genes in this tissue, as determined using the above criteria, was much larger than the number of genes in the maternal tissues. To limit the number of differentially expressed genes in the placenta to something comparable to the maternal tissues, we implemented a more stringent statistical criteria, namely, genes with a larger than 2-fold or less than 0.5-fold change on gd 10 or 15 compared with gd 19, an unadjusted P value less than 0.00005, and a log2 average fluorescent intensity greater than 4, for at least one gestation day. Fluorescent intensity is an indication of signal strength and data integrity. Higher fluorescent intensity ensures that a detected change is not background noise and is therefore an important factor to consider before further examination of gene expression data.
For changes in gene expression shown in relevant tables, expression of metabolic enzyme and transporter genes, namely, Cyp, Ugt, and Sult enzymes as well as Abc, Slc, and Slco transporters, were filtered based on a lower fold change criteria that included genes demonstrating larger than 1.5-fold or less than 0.65-fold changes in expression, as well as unadjusted P values less than 0.05 for maternal tissues and placenta, and average log2 fluorescent intensities greater than 4, for at least one gestation day compared with gd 0 (maternal tissues) or gd 19 (placenta).
To examine Cyp, Abc, and Slc genes that are important for drug metabolism and transport, but do not meet previous filtration criteria needed for inclusion in tables, we selected specific genes known to be involved in drug metabolism and disposition (e.g., Cyp1, Cyp2, and Cyp3 isoforms) and filtered them according to unadjusted P values less than 0.01 and log2 average fluorescent intensities greater than 4, for at least one gestation day compared with gd 0 for maternal tissues and gd 19 for placenta. Fold changes in expression of these genes were smaller, but are meaningful to understand drug PK during pregnancy and were therefore presented in relevant figures.
Quantitative Real-Time Polymerase Chain Reaction Validation.
To validate microarray gene expression data, quantitative real-time polymerase chain reaction (qRT-PCR) was used. We selected three genes from the 25 maternal liver tissues (n = 5 for gd 0, 7.5, 10, 15, and 19) and seven genes from the 15 placental tissues (n = 5 for gd 10, 15, and 19), which totaled 180 microarray gene expression data for validation. The three genes selected for RNA quantification in the maternal liver were Abcc3, Cyp17a1, and Cyp2d40. The seven genes selected for RNA quantification in the placenta were Abcb1a, Abcb1b, Abcg2, Cyp2s1, Slc22a3, Slco1a4, and Slco2b1. These genes were selected because they either demonstrated large changes in the microarray data or are important for xenobiotic metabolism and transport. Fluorogenic 5′ nuclease-based assays were employed as previously described (Anderson et al., 2011; Cole et al., 2011; Stamper et al., 2011). Actb (β-actin) amplification plots derived from serial dilutions of an established reference sample were used to create a linear regression formula to calculate expression levels. Variations in β-actin gene expression throughout gestation were small (<25% in the maternal liver and <18% in the placenta) (data not shown), and β-actin was therefore selected as an internal control to normalize the data for the respective tissue.
Results
Validation of Microarray Data by qRT-PCR.
Gene expression levels determined by microarray analysis correlated well with mRNA levels determined by qRT-PCR (Pearson’s R coefficients of 0.99 and 0.95 for the liver and placenta, respectively) (Supplemental Fig. 1). A direct comparison of the approaches is shown in Figs. 1 and 2. Cyp17a1, Cyp2d40, and Abcc3 (Mrp3) expression in the maternal liver is shown in Fig. 1; placental expression of Cyp2s1, Abcb1a (Mdr1a), Abcb1b (Mdr1b), Abcg2 (Bcrp1), Slc22a3 (Oct3), Slco1a4 (Oatp4c1), and Slco2b1 (Oatp2b1) is shown in Fig. 2. In both tissues, the absolute magnitude of change in expression was consistently greater for the qRT-PCR data compared with the microarray data. Both approaches also agreed on the direction of change (upregulation or downregulation), except for Cyp2d40 on gd 7.5 in the maternal liver as well as Abcg2 and Slco1a4 on gd 15 in the placenta where changes in gene expression were small. Taken together, these results establish the accuracy and reliability of our microarray analysis.
Overall Gestational Age-Dependent Trends in Gene Expression.
Hierarchical one-dimensional clustering of differentially expressed genes for each tissue is shown in Fig. 3. Gene expression levels in the liver and small intestine were similar between replicate samples within the same gestational age group. Changes in expression in the liver and small intestine (upregulation or downregulation) were more robust as gestation progressed, peaking on gd 15 and/or gd 19 (Fig. 3, A and C). Interestingly, the kidney exhibited a bimodal pattern of gene expression that peaked on gd 10 and gd 19 (Fig. 3B). This apparent trend is the result of replicate samples gd 10_3 and gd 10_6. In addition, the expression on gd 15 in the placenta was not very different from term placenta (gd 19); however, large differences were visible on gd 10 compared with gd 19 (Fig. 3D). Figure 3D also shows that the pattern of expression between placenta sample replicates was very similar, except for gd 19_1, which appeared to be more robust in comparison with the other replicates in the gd 19 group. Although expression in the placenta was lower on gd 10, this is in comparison with gd 19 placenta levels, indicating that gene expression in the placenta is actually increasing throughout the course of pregnancy. All tissues demonstrated a preference for upregulation during pregnancy; however, the pattern of changes in gene expression throughout gestation was uniquely different for the kidney.
The filtration criteria used to generate heatmaps in Fig. 3 were also used to count the number of differentially expressed genes for each gestation day. In the liver, 56, 103, 216, and 138 genes were differentially expressed on gd 7.5, 10, 15, and 19, respectively. In the kidney, there were 17, 163, 38, and 74 genes differentially expressed on gd 7.5, 10, 15, and 19, respectively. The small intestine had 37, 55, 47, and 75 differentially expressed genes on gd 7.5, 10, 15, and 19, respectively. In the placenta, 856 genes were differentially expressed on gd 10, but only 8 genes were differentially expressed on gd 15. As a whole, this study identified 1883 differentially expressed genes in the maternal tissues and placenta.
Drug Metabolism during Pregnancy.
Major gene families involved in drug metabolism (Cyp, Ugt, and Sult) were analyzed for gestational age-dependent changes in expression. Raw microarray data from these gene families were filtered based on three criteria. First, on at least one gestation day, fold changes in expression (compared with gd 0 for maternal tissues and gd 19 for placenta) had to be greater than 1.5 or less than 0.65. Second, these fold changes had to have a corresponding unadjusted P value that was less than 0.05. Third, for a particular fold-change that was significant, the associated average log2 fluorescent intensity had to be larger than 4. The resulting genes are shown in Table 1 and Supplemental Tables 15–17 for genes in the maternal liver, kidney, small intestine, and placenta, respectively.
In the maternal liver, the Cyp and Slc gene families were most significantly altered during pregnancy; however, only a few genes had larger than 2-fold changes in expression. By far the largest upregulation occurred for Cyp17a1, which was increased 3.3-fold, 8.2-fold, and 10.4-fold on gd 7.5/10, gd 15, and gd 19, respectively. Cyp4a12a/b was decreased by 70%–80% as gestation progressed, which was the largest downregulation in Cyp gene expression. Cyp2b13, Cyp2c50, Cyp2c54, Cyp2c55, Cyp2d9, and Cyp4a12 expression were also decreased throughout gestation. Many of the Cyp isoforms thought to be involved in drug metabolism, such as the Cyp1, Cyp2, and Cyp3 families, did not demonstrate large enough changes in gene expression to be included in Table 1. Therefore, time-dependent trends in the expression of Cyp1, Cyp2, and Cyp3 families are shown in Fig. 4, which was generated with only two filtration criteria, namely, the unadjusted P value associated with a fold-change had to be less than 0.01 and the average log2 fluorescent intensity had to be greater than 4 for at least one gestational age. These criteria allowed us to examine with confidence smaller, but significant, changes in gene expression related to drug metabolism. The data in Fig. 4 indicate that many genes from the Cyp1, Cyp2b, Cyp2c, and Cyp2d families are downregulated as gestation progresses (Fig. 4, A–C). This pattern is also true for Cyp3a11 (Fig. 4A). Whereas the majority of Cyp genes were downregulated during pregnancy, Cyp2d40, Cyp3a16, Cyp3a41a, Cyp3a41b, and Cyp3a44 were upregulated (Fig. 4D). Overall, changes in hepatic gene expression of Cyp isoforms important for drug metabolism were rather small (downregulated by less than 50% or upregulated less than 1.5-fold).
Cyp expression in the maternal kidney was very dynamic during pregnancy (Supplemental Table 15). A few genes, namely Cyp2d9, Cyp2d12, and Cyp3a41a/b, were altered in a gestational age-dependent manner. Cyp3a41a/b was highly upregulated by approximately 12-fold on gd 19. In the small intestine, only Cyp1a1, Cyp2a5, Cyp2c55, and Cyp2j6 displayed larger than 1.5-fold changes in expression during pregnancy (Supplemental Table 16). The Cyp isoforms that were most notably changed in the placenta during pregnancy are likely responsible for xenobiotic detoxification (Cyp2s1), arachidonic acid metabolism (Cyp4a14), bile acid synthesis (Cyp7b1), steroid biosynthesis (Cyp11a1, Cyp11b1, and Cyp17a1), and retinoic acid and vitamin D regulation (Cyp26a1 and Cyp2r1) (Supplemental Table 17). In contrast to the liver, all Cyp enzymes in the placenta, except for Cyp2s1 and Cyp7b1, were upregulated during mid-gestation (gd 10) and decreased to term expression levels by mid to late gestation (gd 15). Placental Cyp1, Cyp2, and Cyp3 gene families were further evaluated for gestational age-dependent trends in expression (Fig. 4E). All Cyp isoforms in the placenta were upregulated on gd 10 compared with gd 19, with two exceptions. Cyp1a1 peaked on gd 15 rather than gd 10, and Cyp3a57 was downregulated on gd 10 (rather than upregulated) compared with gd 19.
Across tissues, Sult and Ugt expression were relatively unaffected by pregnancy. In the maternal liver, Sult1d1 was decreased by as much as 60% on gd 15 (Table 1). In the kidney and small intestine, only a few Sult and Ugt isoforms were altered during pregnancy, but their relevance to drug metabolism is not well understood (Supplemental Tables 15–17). In the placenta, Sult1a1, Sult5a1, and Ugt1a9 were only affected on gd 10 during pregnancy (Supplemental Table 17). Specifically, Sult1a1 was downregulated by 50% on gd 10 compared with gd 19, and Sult5a1 and Ugt1a9 were upregulated more than 2-fold. Overall, the number of Sult and Ugt genes in the maternal tissues and placenta affected by pregnancy was small, but the magnitude of changes in expression for these genes was quite large.
Drug Transport during Pregnancy.
Gestational age-dependent expression of important drug transporter genes such as Abc, Slc, and Slco transporters were characterized using the same fold change, unadjusted P value, and fluorescent intensity filtration criteria as the metabolic enzymes. Abc transporter expression in maternal tissues appeared relatively unaffected by pregnancy compared with nonpregnant controls (gd 0), except for Abcc3 (Mrp3) and Abcb1a/1b (mouse P-gp or Mdr1a/1b) (Table 1; Supplemental Tables 15 and 16). In the maternal liver, expression of Abcc3 and Abcb1a was decreased by approximately 70% and 40%, respectively, on gd 15. When the fold change filtration criteria was removed and the unadjusted P value criteria was lowered to less than 0.01, Abcc6 (Mrp6) was downregulated by 20–30% on gd 10 and 15 (Fig. 5A). In the maternal kidney, Abcb1a expression was constantly decreased by 30–40% throughout mid to late gestation (Fig. 5C; Supplemental Table 15). Abcc4 expression in the maternal kidney was also decreased by 20%–30% on gd 10 and 15 and returned to nonpregnant levels by term (Fig. 5C). Changes in transporter expression in the small intestine were minimal compared with the liver and kidney (Supplemental Table 16). There were many changes, however, in placental Abc transporters (Fig. 5D; Supplemental Table 17). Specifically, Abcb1a and Abcb1b were 70% and 80% lower, respectively, on gd 10 compared with gd19. Abcc5 (Mrp5) expression was 3-fold higher on gd 10 compared with gd 19. It appears that most Abc transporters in the placenta, with a few exceptions, tend to be suppressed during mid to late gestation (gd 10 and 15) and then return to the control levels on gd 19.
Expression of many Slc and Slco transporter genes in the maternal liver and kidney was significantly altered throughout gestation (Table 1; Supplemental Table 15). In the liver, a large portion of these Slc transporters were nutrient transporters, rather than xenobiotic transporters. Gestational age-dependent trends were evident, however, for Slco1b2 (Oatp1b2, ortholog of human OATP1B3), Slc10a1 (Ntcp), and Slc47a1 (Mate1), which were constantly downregulated by approximately 25% on gd 15 and gd 19 (Fig. 5B). In the maternal kidney, many Slc and Slco transporters demonstrated only moderate gestational age-dependent changes in expression (Supplemental Table 15). Slco4c1 (Oatp4c1), on the other hand, showed decreased expression (approximately 60–70%) across gestation (Fig. 5C). In the small intestine, several Slc transporter genes demonstrated gestational age-dependent changes in expression; however, their role in xenobiotic transport is not known. For example, Slc6a9, Slc10a5, and Slc12a6 were downregulated by 20–30% on gd 10 and/or gd 15, whereas Slc25a43, Slc31a1, Slc35c2, and Slc36a1 expression increased as much as 2.3-fold during pregnancy (Supplemental Table 16). In the placenta, Slc22a3 (Oct3) expression was 80% lower on gd 10 compared with gd 19 (Supplemental Table 17). In contrast, Slc22a5 (Octn2) and Slc6a2 (Net) were both upregulated more than 2.5-fold on gd 10 (Fig. 5F). Overall, relatively few Slc and Slco transporter genes known to be involved in drug transport were affected during pregnancy. Changes in expression of Slco1b2 in the liver, Slco4c1 in the kidney, and Slc22a3 in the placenta, however, may affect drug pharmacokinetics during pregnancy.
Metabolism and Transport of Endogenous Substances during Pregnancy.
Gestational age-dependent trends were also observed for genes involved in the metabolism and transport of endogenous substances such as bile acids and steroid hormones. According to our data, bile acid production may be altered during pregnancy, particularly in late gestation. For example, Cyp7a1 and Cyp7b1, the hepatic enzymes responsible for the rate-limiting step in bile acid synthesis, were significantly induced as much as 1.5-fold across gestation. The expression of bile acid phase I metabolic enzyme, Cyp3a11, decreased by approximately 30% in a gestational age-dependent manner (Table 1; Supplemental Table 18). The expression of bile acid transporters also changed. Abcb1a (Mdr1a), a canalicular bile acid efflux transporter, decreased by 40% on gd 15 (Supplemental Table 18), and Abcc3 (Mrp3), the primary transporter responsible for basolateral efflux of bile acids from hepatocytes into the systemic circulation, decreased by 30%, 70%, and 60% on gd 10, 15, and 19, respectively (Fig. 5A; Supplemental Table 18). Expression of bile acid hepatic uptake transporters Slc10a1 (Ntcp) and Slco1b2 (Oatp1b2) was decreased slightly by approximately 20–30% on gd 15 and 19 (Supplemental Table 18).
Fold changes in steroidogenic enzyme expression were quite large in the maternal liver and placenta; however, the number of genes that changed throughout gestation was low. In the maternal liver, Cyp17a1, which oxidizes progesterone and pregnenolone to their 17α-OH metabolites, was induced 3- to 10-fold across gestation (Table 1). In the placenta, expression of Cyp11a1, which converts cholesterol to pregnenolone, was 2.8-fold higher on gd 10 compared with gd 19, yet 30% lower than term levels on gd 15 (Supplemental Table 17). Likewise, Cyp17a1 was 2.7-fold higher on gd 10 compared with gd 19 (Supplemental Table 17). Therefore, both Cyp11a1 and Cyp17a1 in the placenta were downregulated during pregnancy, but their expression returns approximately to control levels by gd 15. The effects of pregnancy on hepatic and placental expression of aromatase (Cyp19a1), an enzyme that converts androstenedione to estrogens, were very small (data not shown). Murine orthologs of other important human steroidogenic enzymes, such as 3β-hydroxysteroid dehydrogenase (3β-HSD) and 17β-hydroxysteroid dehydrogenase (17β-HSD), are not known. Several mouse isoforms of Hsd genes were affected in a statistically significant way during pregnancy (Supplemental Tables 13 and 14).
Discussion
In the present study, we have performed the first comprehensive gestational age-dependent analysis of metabolic enzyme and transporter gene expression in the maternal liver, kidney, small intestine, and placenta of pregnant mice using microarray analysis. Ten genes were selected for validation by qRT-PCR based on their robust changes in expression and/or involvement in drug metabolism or transport. Fold changes in the qRT-PCR data were consistently higher than those in the microarray data, but the direction of changes was similar between the two approaches (Figs. 2 and 3). These differences in magnitude are possibly due to higher detection sensitivity of qRT-PCR (Evans et al., 2002). We did not expect differences in the reliability/accuracy of the microarray data across tissues because sample preparation was the same for all tissues; therefore, qRT-PCR validation for the liver and placenta was considered adequate. As a secondary validation, we confirmed our results with previously well characterized genes affected during pregnancy. These included induction of Cyp3a16 (Zhang et al., 2008) and Cyp8b1 (Aleksunes et al., 2012) in the liver and downregulation of Abcc3 in the liver (Aleksunes et al., 2012), Abcb1a in the kidney (Zhang et al., 2008), and Abcc5 in the placenta (Aleksunes et al., 2008) (Table 1; Supplemental Tables 15–17).
In the kidney, a bimodal pattern of gene expression on gd 10 and 19 was produced by two kidney samples, gd 10_3 and gd 10_6. This interanimal variability appears to be specific for the kidney because the liver, small intestine, and placenta collected from the same mice did not exhibit such a pattern. The reason for such an observation is not known; however, recent computational studies do predict that cooperative binding of transcription factors to DNA promotes a bimodal gene expression response in eukaryotic systems (Gutierrez et al., 2012). Thus, the same might explain the bimodal pattern of gene expression in the mouse kidney observed in this study.
The data from the small intestine and placenta should be interpreted with caution for the following reasons. The entire small intestine was used to isolate total RNA, resulting in a mixing of RNA from smooth muscle tissues and mucosal epithelial cells. Our data therefore do not reflect changes in gene expression in the more functionally relevant enterocytes (from a drug disposition perspective), nor do they reflect changes in specific sections along the small intestine. Consequently, we placed less of an emphasis on the small intestine data. The formation of placenta is different between mice and humans, which complicates translation of the placental data to human pregnancy. For example, development of the prominent chorionic villus structure occurs quickly in human pregnancy (by day 21 of 270), but not until gd 11.5 in the mouse. This may explain why the mouse placenta had more differentially expressed genes on gd 10 versus gd 15, and challenges whether it is best to analyze gene expression before gd 11.5. Interpretation of drug transport across the mouse placenta on gd 15 and 19 is still relevant to human placenta, because both mice and humans use syncytiotrophoblasts to modulate the exchange of nutrients, hormones, and xenobiotics between the mother and fetus.
Our study is the first to characterize gestational age-dependent changes in gene expression for all metabolic enzymes and transporters. Zhang et al. (2008) reported that although Cyp3a16, Cyp3a41, and Cyp3a44 were induced in the maternal liver during pregnancy, Cyp3a11, Cyp3a13, and Cyp3a25 were downregulated. This is consistent with our results (Fig. 4). Koh et al. (2011) examined gestational age-dependent expression of hepatic Cyp1a2, Cyp2a5, Cyp2b10, Cyp2c37, Cyp2d22, Cyp2e1, Cyp3a11, and Cyp3a41 using qRT-PCR. Again, their observations mirror our data (Fig. 4). In the placenta, Cyp upregulation during mid-gestation is also consistent with previous studies (Hakkola et al., 1996). We are the first, however, to examine Cyp gene expression in the maternal kidney. Interestingly, Cyp3a41a and Cyp3a41b were markedly induced in late gestation. Implications of Cyp3a41a/b induction are not known, but might include altered xenobiotic metabolism, increased bioactivation of nephrotoxic pathways, or increased sodium retention to control blood pressure (Clore et al., 1992).
We also used our microarray data to explore CYP-mediated steroid hormone biosynthesis during pregnancy. CYP17A1 is a critical enzyme for steroid hormone biosynthesis. The large induction of hepatic Cyp17a1 (approximately 3- to 10-fold) suggests that production of 17α-hydroxyprogesterone and 17α-hydroxypregnenolone may increase during mouse pregnancy, which is consistent with 17α-OH progesterone increases in humans during pregnancy (Dörr et al., 1989). Increased levels of 17α-OH metabolites could ultimately increase estrogen production in the maternal liver; however, downregulation of placental Cyp17a1 throughout gestation favors progesterone formation, not estrogen. Therefore, our microarray data support the general pattern of steroid hormone biosynthesis during pregnancy, that is, although progesterone and estrogen are increasingly produced as gestation progresses, the majority of progesterone is made by the placenta in late gestation (Raunig et al., 2011). Nevertheless, interpretation of placental steroidogenesis in mice should be done with caution because one study has shown that the pattern of gestational age-dependent expression of some steroidogenic enzymes can be opposite between mice and humans (Malassiné et al., 2003).
Our results revealed that Abc transporter gene expression was decreased during pregnancy, particularly Abcc3 in the liver, Abcb1a in the kidney, and Abcc5 in the placenta. Similarly, several Slc or Slco transporters important for drug disposition were moderately decreased during pregnancy (such as Slco1b2 in the liver and Slco4c1 in the kidney). Slc22a3 (Oct3) and Slc6a2 (Net) in the placenta were downregulated and upregulated, respectively, to a larger extent. The small number of transporters affected by pregnancy may still be enough, however, to explain reported increases in hepatic bile acid concentrations in late pregnancy in mice (Milona et al., 2010; Aleksunes et al., 2012). Consistent with these previous studies, our data support increased bile acid synthesis (Cyp7b1) and decreased basolateral and canalicular bile acid efflux (Abcc3 and Abcb1a) during mouse pregnancy. Decreased expression of hepatic uptake transporters in the maternal liver may therefore be an adaptive response to manage increased intracellular bile acid concentrations.
Translation of our findings to human pregnancy is not straightforward. One reason is that gene expression and protein levels are also not always correlative (Schwanhäusser et al., 2011). In the placenta, Abcb1a/1b expression increases; however, its protein expression has been shown to decrease as gestation progresses (Zhang et al., 2008). Microarray data therefore may not entirely reflect pregnancy’s effect on protein expression/activity. Clinical studies indicate that activity of CYP2C9, CYP2D6, and CYP3A4 increases during pregnancy, but activity of CYP1A2 and CYP2C19 decreases (Dempsey et al., 2002; Anderson, 2005; Hodge and Tracy, 2007). In addition, the number of Cyp1, Cyp2, and Cyp3 isoforms in mice is greater than that in humans, and there are potential species differences in substrate specificity and/or enzymatic activity. All of these factors make it challenging to predict drug metabolism during human pregnancy using the mouse data. If our microarray data do predict CYP activity, then xenobiotic substrates of CYP1 or CYP2 isoforms may need to be given at smaller doses during pregnancy. Prediction for CYP3A substrates is less certain, because Cyp3a16, Cyp3a41a/b, and Cyp3a44 are upregulated during pregnancy, yet Cyp3a11 and Cyp3a13 are downregulated. In addition, fold changes in Cyp3a gene expression do not reflect the magnitude of increase that we and others have shown for in vivo CYP3A activity in pregnant mice and humans (Mathias et al., 2006; Hebert et al., 2009; Zhou et al., 2010). Therefore, the Cyp isoform(s) that best represent in vivo CYP3A activity should be carefully evaluated.
Mouse Cyp3a11 is most closely related to human CYP3A4 with 76% amino acid identity and may be the dominant contributor to CYP3A activity in mouse liver (Martignoni et al., 2006). However, CYP3A11 is likely not the mouse CYP3A isoform responsible for increased CYP3A activity in pregnant mice, because its gene expression was downregulated over gestation. Cyp3a13 and Cyp3a16 appear less important because hepatic Cyp3a13 levels are 5- to 10-fold lower than Cyp3a11 in nonpregnant female mice, and Cyp3a16 is considered a fetal isoform (Martignoni et al., 2006). On the other hand, Cyp3a41a/b and Cyp3a44 are female-specific isoforms expressed solely in the liver, and they are known to be induced by growth hormone, glucocorticoids, and estradiol (Sakuma et al., 2002; Sakuma et al., 2004; Jarukamjorn et al., 2006). Indeed, maternal estradiol plasma concentrations triple from gd 10 to gd 17 (Barkley et al., 1979), which parallels the trend of hepatic induction of Cyp3a41a/b and Cyp3a44 during gestation (Fig. 4D). Clearly, isoform-specific quantitation of protein levels and activity is needed to further evaluate which isoforms are responsible for the in vivo CYP3A activity in pregnant mice.
In summary, we have determined global gene expression profiles in the maternal tissues and placenta of pregnant mice, providing a genetic basis for understanding gestational age-dependent physiologic and pharmacological changes during pregnancy. These data also offer a means to study the mechanisms of regulation behind such changes. On the basis of our findings, mid to late gestation (gd 10–15) may be the best time to study the full impact of mouse pregnancy on drug metabolism and disposition.
Acknowledgments
The authors greatly thank Drs. Thomas Easterling and Sibylle Heidelberger for reading and editing this manuscript.
Authorship Contributions
Participated in research design: Mao, Thummel, Hebert, Farin.
Conducted experiments: Shuster, Tsai.
Contributed new reagents or analytic tools: MacDonald, Beyer, Bammler.
Performed data analysis: Shuster, MacDonald, Beyer, Bammler.
Wrote or contributed to the writing of the manuscript: Shuster, Mao, Tsai, Farin, Bammler, Beyer, Thummel, Hebert.
Footnotes
This study was supported by the Eunice Kennedy Shriver National Institute of Child Health and Human Development [Grant U10HD047892]. This study was also supported in part by the National Institutes of Health National Institute of Environmental Health Sciences [Grant P30ES007033] and by the National Institutes of Health Pharmacological Sciences Training Grant [Grant T32GM007750]. The content is solely the responsibility of the authors and does not necessarily represent the official views of the Eunice Kennedy Shriver National Institute of Child Health and Human Development or the National Institutes of Health.
↵This article has supplemental material available at dmd.aspetjournals.org.
Abbreviations
- ABC
- ATP-binding cassette
- CYP
- cytochrome P450
- FDR
- false discovery rate
- gd
- gestation day
- GFR
- glomerular filtration rate
- P-gp
- P-glycoprotein
- PK
- pharmacokinetics
- qRT-PCR
- quantitative real-time polymerase chain reaction
- SLC
- solute carrier
- UGT
- UDP-glucuronosyltranserases
- Received October 16, 2012.
- Accepted November 21, 2012.
- Copyright © 2013 by The American Society for Pharmacology and Experimental Therapeutics