Interindividual variation in drug clearance rate is often responsible for toxicity or inefficacy of prescription drugs. Systemic drug clearance rate is determined by hepatic expression and activity of phase I oxidative cytochromes P450 (P450s), phase II conjugative enzymes, and transporter proteins. Expression of these metabolic enzymes is coordinately regulated by a network of transcription factors (Pascussi et al., 2004; Xu et al., 2005) exemplified in Fig. 1. The network is composed of ligand-activated nuclear receptors that recognize a variety of endogenous and xenobiotic compounds to activate transcription of metabolic enzymes involved in biotransformation and transport. Multiple nuclear receptors can recognize response elements of the same target gene (Fig. 1) and may control their own expression as well as the expression of other nuclear receptors in the pathway (Maglich et al., 2002; Pascussi et al., 2004). In addition, these regulatory proteins share a common pool of coregulators and the common heterodimeric partner, the retinoid X receptor (RXR).

The role of xenobiotic metabolism gene polymorphisms in determining clearance rate of specific prescription drugs has been extensively studied. The polymorphisms of several P450 genes can predict corresponding P450 activity level (Ingelman-Sundberg et al., 1999). However, it has become increasingly clear that trans-acting factors are also involved in determining expression and activity of xenobiotic metabolism genes. For example, the activity of CYP3A4, the most active P450 in the human liver, exhibits a high degree (at least 40-fold) of interindividual variation (Paine et al., 1997); however, CYP3A4 polymorphisms are not accurate predictors of CYP3A4 activity. Indeed, CYP3A4 activity exhibits a unimodal distribution among the population, suggesting that a multitude of factors are involved in dictating CYP3A4-mediated pathways (Wilkinson, 1996). Estimates suggest that 90% of the variation in CYP3A4 activity may be attributed to genetic factors (Ozdemir et al., 2000). Accordingly, polymorphisms of genes involved in the regulatory network affecting CYP3A4 expression or activity likely have effects upon the rate of CYP3A4-mediated metabolism.

Recent studies suggest that variation in trans-acting master regulator genes may account for differential expression of a large proportion of genes and determine complex traits (Schadt et al., 2003; Yvert et al., 2003). Indeed, complex phenotypes are often dictated by variation across entire pathways (Evans and Relling, 2004). Thus, polymorphisms in master regulator genes could serve as biomarkers to predict drug clearance rate (Carlberg and Dunlop, 2006). Genotyping for single polymorphisms predicting a complex phenotype would be more practical and cost-effective than genotyping for many individual xenobiotic metabolism genes.

Variable nuclear receptor expression level may determine target gene expression level by affecting constitutive activity in the case of some nuclear receptors or by modulating the degree of response to ligands in others. For example, CAR mRNA expression level is highly variable (at least 240-fold) among individuals (Chang et al., 2003), and livers expressing high levels of CAR have enhanced expression of CAR target genes (Finkelstein et al., 2006). Additionally, the abundance of common coregulator proteins and RXR may serve as limiting factors to nuclear receptor-mediated transcriptional control. Finally, the rate of regeneration of P450s by P450 oxidoreductase (POR), the obligate P450 regenerative enzyme in mammals (Wu et al., 2005), may dictate phase I metabolism rate. A number of isolated studies have previously identified interindividual variation in expression levels of nuclear receptors such as CAR, SXR, and hepatic nuclear factor 4α (HNF4α) and their correlations with target genes (Pascussi et al., 2001; Chang et al., 2003; Lamba et al., 2003). However, a systematic screening of the effects of variable expression of nuclear receptors and their coregulators has not been conducted. Here we hypothesize that variation in expression level of previously unidentified master regulator genes is responsible for determining basal expression level and activity of a broad scope of xenobiotic metabolism genes in the human liver.

In this study we use linear regression analysis to identify associations between nuclear receptors and xenobiotic metabolism genes as well as to identify patterns of collinearity. We identify a high degree of transcriptional coregulation among the xenobiotic metabolism genes CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, multidrug resistance-associated protein 2 (MRP2/ABCC2), organic anion-transporting polypeptide 2 (OATP2/SLCO1B1), POR, and UDP-glucuronosyltranferase 1A1 (UGT1A1). We also identify the nuclear receptors CAR and HNF4α as extensive regulators of xenobiotic metabolism whose expression level modulates basal transcription of a broad scope of xenobiotic metabolism genes. Finally, we identify POR expression level, which is closely associated with expression level of CAR and HNF4α, to be a putative limiting factor for phase I enzyme activity.

Materials and Methods

Basal Enzyme Activities. Twenty human liver samples with variable CYP3A activity were obtained from the National Disease Research Interchange (Philadelphia, PA) and the Midwest Transplant Network (Westwood, KS). Demographics for human liver samples are listed in Table 1. Considering that CYP3A activity is probably indicative of overall liver function, using samples with variable CYP3A activity increases the statistical power of this study. Variable samples allow for the identification of significant correlations while minimizing the number of rare liver samples used. All liver samples have tested negative for human immunodeficiency virus (HIV) and hepatitis and were not suspected of harboring other infectious diseases. Liver microsomes were prepared as described previously (Obach et al., 2001) to determine P450 enzyme activity rates as described by Madan et al. (1999) using the probe drugs listed in Table 2. Incubation conditions were controlled to ensure linear metabolite formation with respect to reaction time and protein concentration.

Identification of Relevant Nuclear Receptor, Coregulator, and Xenobiotic Metabolism Genes. Nuclear receptors and coregulators were selected for the study based upon their ability to trans-activate common xenobiotic metabolism genes. Selection was performed using Ingenuity Pathways Analysis (Ingenuity Systems, Inc., Redwood City, CA) and a literature search (Pascussi et al., 2004; Xu et al., 2005). mRNA level was subsequently quantified for 10 nuclear receptors, 5 coregulators, 9 cytochrome P450 enzymes, 3 transport proteins, and the obligate regenerative enzyme for P450 enzymes, POR, as listed in Table 3.

Quantitative Real-Time PCR. Liver samples were lysed in the TRIzol reagent (Invitrogen, Carlsbad, CA) and total RNA was isolated according to the manufacturer's protocol. RNA purity was confirmed by a 260:280 nm absorbance ratio greater than 1.5, and agarose gel electrophoresis with ethidium bromide staining was used to confirm integrity of 18S and 28S ribosomal RNA bands.

Reverse transcription was carried out using MMLV Reverse Transcriptase (Invitrogen) according to the manufacturer's protocol. Primers and probes for real-time PCR amplification were designed using Primer Express 3.0 according to the manufacturer's instructions (Real-Time PCR Chemistry Guide, Applied Biosystems, Foster City, CA) or as described in the literature (Table 3). If Primer Express or the literature did not yield acceptable primer and probe sets for a given gene, predesigned Taqman Gene Expression Assays were used (Applied Biosystems; Table 3). All primer and probe sets were checked for specificity using BLAST.

Real-time PCR amplification of cDNA corresponding to 16 ng of total RNA was carried out at a total volume of 20 μl containing 1× concentration of TaqMan Universal Master Mix, a 900 nM concentration of each primer, and a 150 nM concentration of the 5-carboxyfluorescein-5-carboxytetramethylrhodamine dual-labeled probe. Amplification and fluorescence detection were carried out using the ABI Prism 7900 HT Real-Time PCR System (Applied Biosystems). Cycling conditions were as follows: initial denaturation at 95°C for 10 min followed by 40 cycles of 95°C for 15 s, and then 60°C for 1 min. Four-fold serial dilutions of reverse-transcribed RNA were amplified to create relative standard curves for quantification of each amplicon according to the manufacturer's instructions (Real-Time PCR Systems Chemistry Guide, Applied Biosystems). Gene expression values were normalized to β-actin.

Statistical Analysis. SPSS version 13.0 (SPSS Inc., Chicago, IL) was used for all statistical analysis. The Wilcoxin rank-sum test was used to compare mean enzyme activity levels between demographic groups. Spearman's rho correlation coefficients between mRNA levels or enzyme activities were identified using bivariate linear regression analysis. These statistical tests were used to compare means (Wilcoxin rank-sum test) or correlations (Spearman's rho) between groups of data points with skewed distributions (for an example of data distribution, see Fig. 2). Correlations with p < 0.05 were accepted as significant. A correlation matrix excluding outliers was compared with the original correlation matrix to confirm that significant associations were not dependent upon outliers. Significant correlations were further tested using partial regression analysis to determine the role of collinearity with other factors in gene-gene or gene-activity correlations. Partial regression analysis allows for statistical control for variability of any single factor. Statistically controlling for variability in a gene whose abundance determines expression level of a group of coregulated genes would be expected to abrogate the correlations among these genes. mRNA levels of genes whose statistical correlation depends upon variability of mRNA level of a known or putative regulator gene are collinear with that regulator gene and are likely coregulated by that gene product. To screen for genes whose mRNA level accounts for the highest degree of collinearity among xenobiotic metabolism genes, we performed first order partial regression analysis controlling for variation of individual candidate master regulator genes involved in the xenobiotic metabolism regulatory network. An increase of p value above the threshold value of p = 0.05 was considered a loss of significant association. The regulatory genes were then ranked dependent upon the quantity of associations among mRNA levels of xenobiotic metabolism genes that were lost while controlling for the variability of each regulatory gene. In addition, partial correlation coefficients were identified to determine the degree of correlation among mRNA levels of putative target genes after controlling for candidate regulator genes. Partial correlation matrices excluding outliers were compared with original partial correlation matrices to confirm that associations were not influenced by outliers.

Results

Demographic Factors Had Limited Influence on P450 Activities. To determine whether demographic factors strongly affected xenobiotic metabolism in liver samples used in this study, P450 activities were compared among different demographic groups represented in Table 1. The only differences of P450 activity among demographic groups (gender, ethnicity, smoking status, and alcohol use) was that CYP2C19 activity was higher in females than in males (p = 0.002) and higher in nondrinkers than in drinkers (p = 0.034). In addition, there was no significant correlation or noticeable pattern of association between age and P450 activities of liver samples used in this study (data not shown). Thus, available demographics were not largely influential upon P450 activity rates in samples used here.

Correlations between P450 mRNA Levels and Microsomal P450 Activity. Extensive interindividual variation in microsomal P450 enzyme activity (Table 2) as well as P450 mRNA level (Table 4) was observed. To ensure that correlations identified between mRNA levels of regulator genes and that of phase I enzymes are biologically relevant, P450 mRNA levels were compared with corresponding microsomal P450 activity. P450 mRNA levels correlated significantly (p <0.05) with corresponding microsomal P450 activity for CYP1A2, CYP2A6, CYP2B6, CYP2C8, and CYP3A4 (Fig. 2). Correlations remained significant after elimination of outliers from these scatterplots (data not shown). It is important to note that in this assay, CYP3A activity represents a combination of CYP3A4 and CYP3A5 activities due to the lack of a specific probe drug; however, functional CYP3A5 is rarely expressed in human livers, and low or undetectable CYP3A5 expression was confirmed in all samples of this study (data not shown). Associations between mRNA levels and enzymatic activities of most P450 enzymes confirm previous findings that transcriptional control of most P450s dictates enzyme activity level (Rodriguez-Antona et al., 2001). No significant correlation was observed between mRNA level and microsomal P450 activity for CYP2C9, CYP2C19, CYP2D6, and CYP2E1 (data not shown). CYP2C9, CYP2C19, and CYP2D6 functional polymorphisms, which are well characterized and prevalent in the population (Ingelman-Sundberg et al., 1999), may have obscured the correlation between P450 mRNA level and activity in these samples. Regardless, within an individual carrying a hypomorphic P450 allele, a gene-gene interaction could still occur with a variant allele of a regulatory factor affecting P450 expression or activity. The discrepancy between mRNA level and activity of CYP2E1 is accounted for by the extensive post-transcriptional and post-translational regulation of this gene (Koop and Tierney, 1990).

Correlations among Microsomal P450 Activities. There was a substantial degree of correlation among microsomal P450 enzyme activity rates (Table 5). P450 activities with the highest degrees of significant correlation with that of other P450s were CYP2C9 (six correlations), CYP2A6 (five correlations), CYP2B6 (five correlations), CYP2C8 (four correlations), and CYP3A4/5 (four correlations).

POR mRNA Level Significantly Correlates with P450 Activities and Determines Associations among Them. POR mRNA level significantly correlated with microsomal activity of CYP2A6, CYP2B6, CYP2C9, and CYP2E1 (Table 5). Partial regression analysis was carried out to determine the correlations abrogated among P450 activity levels when controlling for POR mRNA level. Partial regression analysis controlling for POR mRNA level abrogated the correlations between CYP2C9 and CYP2A6, CYP2B6, CYP2D6, and CYP2E1 as well as the correlation between CYP2A6 and CYP2E1, accounting for 5 of the 14 significant correlations present among P450 enzyme activity levels (data not shown). As such, controlling for POR mRNA level variability affected associations of CYP2E1 (abrogating two of the two preexisting correlations), CYP2C9 (four of six), CYP2D6 (one of two), CYP2A6 (two of five), and CYP2B6 (one of five), suggesting that POR expression level serves as a limiting factor for the activity of these P450s. Although there were substantial correlations between mRNA levels and enzyme activity among most P450s, there was no correlation between mRNA level of nuclear receptors or coregulators and P450 enzyme activity (data not shown).

Correlations among mRNA Levels of Phase I and Phase II Enzymes and Transporters. Extensive interindividual variation of mRNA levels of most genes was observed (Table 4). There was a substantial degree of significant correlations among P450 mRNA levels (Table 6). P450s with the highest degree of association with other P450s at the mRNA level were CYP2C8 (six correlations), CYP2A6 (five correlations), CYP2B6 (five correlations), CYP2C9 (four correlations), CYP2C19 (four correlations), and CYP3A4 (three correlations). There was a strong pattern of association among the mRNA levels of CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, MRP2, OATP2, POR, and UGT1A1 (Table 7).

Collinearity of Xenobiotic Metabolism Gene mRNA Level with mRNA Level of Nuclear Receptors and Coregulators. Correlations present between mRNA levels of nuclear receptors/coregulators and xenobiotic metabolism genes suggest that the abundance of these regulatory proteins may determine the expression level of target genes. Notably, CAR, HNF4α, and SXR expression levels correlated with the highest proportion of xenobiotic metabolism genes (10 of 14) quantified in this study. The expression of all three nuclear receptors correlated with that of CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, CYP2D6, MRP2, OATP2, POR, and UGT1A1 (Table 7). Candidate regulatory genes were ranked by importance dependent upon the quantity of associations among mRNA levels of xenobiotic metabolism genes that were lost while statistically controlling for the mRNA level variability of each regulatory gene as described under Materials and Methods (Table 8). Genes that were responsible for collinearity of the highest number of target genes may function as master regulators and will be subject to further investigation for functional mechanism and the basis for regulation. Variability in HNF4α mRNA level was necessary to account for 25 correlations among mRNA levels of the xenobiotic metabolism genes studied here. Variability of CAR mRNA level was necessary to account for 22 significant correlations, and SXR mRNA level variability was necessary for 15 significant correlations. All other nuclear receptors and coregulator genes studied here (listed in Table 3) accounted for a lower degree of collinearity of xenobiotic metabolism genes quantified in this study. An example of linear correlations present before, but not after, partial regression analysis controlling for a nuclear receptor (HNF4α) is given by comparing Tables 7 and 9.

Characterization of a Hierarchical Regulatory Pathway. Partial regression analysis demonstrated a strong statistical influence of mRNA level of the nuclear receptors CAR, HNF4α, and SXR upon correlations among mRNA levels of coregulated xenobiotic metabolism genes (Table 8). Therefore, to determine independent effects, if any, of these three nuclear receptors, correlations among xenobiotic metabolism gene mRNA levels with one another as well as with CAR, HNF4α, or SXR were categorized based upon collinearity with CAR, HNF4α, or SXR expression level (Fig. 3). The Venn diagram (Fig. 3) represents collinearity with more than one nuclear receptor as overlapping sections. For example, section D refers to correlations that are abrogated by partial regression controlling for either CAR or HNF4α, and section G refers to correlations that are abrogated by controlling for variability of any of the three nuclear receptors. Notably, statistically controlling for variability of HNF4α abrogates the associations between CAR and all coregulated xenobiotic metabolism genes studied here (Fig. 3, sections B and E). On the other hand, statistically controlling for CAR variability abrogates the associations between HNF4α and CYP2A6, CYP2B6, CYP2C8, CYP2C19, POR, and UGT1A1 (Fig. 3, sections A and F). SXR mRNA level variability was not independently required for associations among any of the xenobiotic metabolism genes studied here (Fig. 3, section C).

Identification of a strong linear relationship between CAR and SXR mRNA levels (Spearman's ρ = 0.860, p < 0.001) led us to believe the lack of associations dependent upon SXR variability may reflect the statistical effect of CAR mRNA level rather than that of SXR. This is expected due to coregulation of CAR and SXR by GR (Pascussi et al., 2000). However, partial regression analysis did not identify a single regulatory gene whose variable expression was required for the correlation between CAR and SXR mRNA level (data not shown). The role of SXR expression level in determining basal expression level of xenobiotic metabolism genes merits further investigation.

Analysis of independent effects of CAR and HNF4α allows for the characterization of a hierarchical pathway mapping the relative effects of the expression of these nuclear receptors upon basal expression of xenobiotic metabolism genes (Fig. 4). HNF4α controls the expression of CAR (Ding et al., 2006). Thus, common target genes that no longer associate with HNF4α expression level after controlling for CAR variability (Fig. 3, sections A and F) are placed downstream of CAR, which is itself under the control of HNF4α, because associations between these genes are dependent upon variability of both CAR and HNF4α. On the other hand, genes collinear with HNF4α that still correlate with HNF4α expression after controlling for CAR variability (specifically, CYP2C9, MRP2, and OATP2; see Fig. 4) are placed downstream of HNF4α only. For example, the association between CAR and CYP2C19 is dependent upon HNF4α expression level variability (Fig. 3, section E), and the association between HNF4α and CYP2C19 is dependent upon CAR variability (Fig. 3, section A). Thus, CYP2C19 is placed under the control of CAR, which is itself under the control of HNF4α, to demonstrate that basal CYP2C19 expression is determined by expression of both CAR and HNF4α (Fig. 4). On the other hand, the correlation between CAR and CYP2C9 is dependent upon HNF4α (Fig. 3, Section E), but the correlation between HNF4α and CYP2C9 does not depend upon CAR variability (Fig. 3). Therefore, CYP2C9 is placed under control of HNF4α only (Fig. 4). This manner of systematic analysis of closely associated xenobiotic metabolism genes was used to construct Fig. 4. Pathways were classified as being direct or indirect targets of the nuclear receptors according to the literature (see Fig. 4 legend).

Discussion

Gene expression profiling of human liver samples allows for the identification of patterns of gene expression among individuals, most notably the correlations present among mRNA levels. Liver tissue samples from 20 individuals with variable CYP3A activity were profiled for mRNA levels of genes involved in xenobiotic metabolism and the regulation thereof.

Identification of Coregulated Xenobiotic Metabolism Genes. There was a strong pattern of association among the expression levels of CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, MRP2, OATP2, POR, and UGT1A1. Identification of common response elements in the upstream regulatory regions of these genes, with the exceptions of CYP2A6 and POR, provides a mechanistic basis for coregulation (Pascussi et al., 2004). Although the mechanisms responsible for CYP2A6 and POR induction have not been characterized in the human liver, both enzymes have been found to be induced by phenobarbital and rifampicin in primary human hepatocytes (Maglich et al., 2002; Madan et al., 2003). A system of transcriptional coregulation of POR and P450 enzymes would be advantageous considering the obligatory nature of POR for P450 activity (Wu et al., 2005).

Coregulation of P450s at the Enzyme Activity Level by POR. There was a substantial degree of correlations among microsomal P450 activities. This finding suggests that a post-transcriptional mechanism may be responsible for coregulation of P450 enzyme activity. POR mRNA level correlated with enzymatic activity of CYP2A6, CYP2B6, CYP2C9, and CYP2E1. In addition, CYP2A6, CYP2B6, CYP2C9, CYP2D6, and CYP2E1 were collinear with POR to some extent. This is the first study (to our knowledge) identifying a linear correlation between POR expression level and phase I enzyme activity level. Our observations suggest that POR expression level serves as the limiting factor for the activity of many P450s and thus dictates phase I metabolism rate. Polymorphisms in coding sequences of POR have been well characterized as common causes of a disease termed disordered steroidogenesis (Pandey, 2006). In addition, the HapMap project has identified polymorphisms in the upstream and intronic noncoding regions of the POR gene that may provide a genetic basis for variable POR expression.

Identification of HNF4α as an Extensive Regulator of Xenobiotic Metabolism. Of all the nuclear receptors and coregulators studied here, HNF4α expression level was responsible for the highest degree of collinearity among xenobiotic metabolism gene expression levels, most notably among CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, MRP2, OATP2, POR, and UGT1A1.

HNF4α was originally identified as an orphan nuclear receptor but has been found to constitutively bind fatty acids in the ligand binding pocket, using the ligand as a structural cofactor rather than an activation signal (Wisely et al., 2002). Accordingly, HNF4α is constitutively active, and its expression level determines its activity.

HNF4α is necessary for formation of the structural architecture of the liver and expression of a normal liver phenotype (Parviz et al., 2003). ChIP-on-Chip analysis (Agilent Technologies, Palo Alto, CA) has implicated HNF4α in binding regulatory regions of more than 1500 genes including GR, SXR, MRP2, CYP1A2, CYP2B6, CYP2C8, CYP2D6, and CYP2E1 (Odom et al., 2004). In addition, Kawashima et al. (2006) demonstrated binding of HNF4α to the CYP2C9 promoter with an immunoprecipitation from human liver. Although recent microarray data did not detect the same correlations between expression levels of HNF4α and target genes as the current study, this effect is likely accounted for by the low detection level of HNF4α by the microarray used (Slatter et al., 2006). HNF4α has been shown to exert a dose-dependent effect upon mRNA expression level of CYP2A6, CYP2B6, CYP2C9, CYP2D6, CYP3A4, and CYP3A5 in primary human hepatocytes when HNF4α expression was gradually reduced using antisense RNA (Jover et al., 2001). This finding suggests that HNF4α expression level is a limiting factor in the transcription rate of phase I xenobiotic metabolism genes. Here we present the novel finding that HNF4α expression level is likely a dominant regulator of basal xenobiotic metabolism gene expression in the human liver. The HapMap project has identified polymorphisms in the upstream and intronic noncoding regions of the HNF4α gene that may provide a genetic basis for variable HNF4α mRNA level.

Identification of CAR as a Xenobiotic-Responsive Regulator of Basal Expression of Xenobiotic Metabolism Genes. CAR mRNA level also correlates with and accounts for collinearity of CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, MRP2, OATP2, POR, and UGT1A1, although its abundance exerts fewer independent effects than that of HNF4α in this descriptive statistical analysis.

CAR is widely recognized as a general xenobiotic receptor, recognizing a diverse array of xenobiotics and activating transcription of biotransformation and transport genes (Pascussi et al., 2004; Xu et al., 2005). CAR exhibits a substantial level of constitutive activity in the absence of ligand (Honkakoski et al., 1998), providing support that its expression level may account for basal expression of target genes.

In agreement with our data, a recent microarray-based study also identified highly significant (p < 0.0005) Spearman's rho correlations between CAR mRNA level and that of CYP2A6, CYP2B6, CYP2C8, CYP2C9, CYP2C19, OATP2, MRP2, and UGT1A1, among other xenobiotic metabolism genes, in 75 human livers (Slatter et al., 2006). The lack of a correlation between CAR and POR in this microarray study may be an artifact of an extremely high or saturated level of POR detection. In addition, Finkelstein et al. (2006) identified enriched expression of these same genes, excluding UGT1A1, in three livers of Hispanics with enriched CAR mRNA expression whose transcriptome closely resembled that of phenobarbital (a CAR activator)-treated primary human hepatocytes. In addition, CAR has previously been implicated in regulating CYP2A6 (Madan et al., 2003), CYP2B6 (Honkakoski et al., 1998), CYP2C8 (Ferguson et al., 2005), CYP2C9 (Al-Dosari et al., 2006), CYP2C19 (Chen et al., 2003), MRP2 (Kast et al., 2002), OATP2 (Staudinger et al., 2003), POR (Maglich et al., 2002), and UGT1A1 (Sugatani et al., 2001), among other xenobiotic metabolism genes.

Our findings confirm the central role of CAR expression level in regulating basal expression level of xenobiotic metabolism enzymes in the human liver. Using a highly sensitive detection technique (Taqman Real-Time PCR), we were able to characterize associations among expression of CAR and target genes at a more detailed level than in previous studies. Polymorphisms in the upstream and intronic noncoding regions of the CAR gene, which may provide a genetic basis for variable CAR expression, have been described by the Hap-Map project.

Hierarchical Control of Xenobiotic Metabolism Gene Expression. Identification of a functional HNF4α binding site in the CAR upstream regulatory region (Ding et al., 2006) provides mechanistic support for our observation that both HNF4α and CAR account for collinearity among a similar group of xenobiotic metabolism genes (Fig. 3). Analyzing the independent statistical effects of either CAR or HNF4α (Fig. 3) as well as mechanistic studies from the literature, we identified a hierarchical organization of the effects of variable expression of the two regulator genes (Fig. 4). The pattern of HNF4α expression level taking precedence over CAR expression level in functioning as a central regulator of basal xenobiotic metabolism gene expression is a novel observation. Although the roles for CAR and SXR functioning as master regulators of xenobiotic metabolism in drug-induced livers have been well characterized (Pascussi et al., 2004; Xu et al., 2005), the central role for HNF4α as a dominant regulator of basal expression of xenobiotic metabolism genes merits recognition and further study.

Study Limitations. A major limitation of any descriptive study is the inability to perturb the system. For example, the close linear association between CAR and SXR mRNA levels in the liver samples of this study prevent us from isolating the effects of SXR from those of CAR. Indeed, any associations characterized here should be analyzed mechanistically to isolate effects of single genes. Characterization of the effect of in vivo reduction or enhancement of POR, HNF4α, and CAR expression would be central to the understanding of their roles in dictating drug metabolism rate.

In addition, the current study focused upon a representative set of phase II metabolism and transporter genes; expansion thereof would be beneficial to studying the effects of master regulator proteins.

Summary. Expression levels of xenobiotic metabolism genes are coordinately regulated to couple the processes of xenobiotic transport and biotransformation as well as P450 regeneration. HNF4α expression level influences the degree of overall liver function by affecting the expression level of liver-specific genes, as is consistent with its role in development. CAR expression level dictates basal transcription level of xenobiotic metabolism genes and likely modulates the degree of sensitivity to xenobiotic exposure over one's lifetime. POR expression level, which is strongly associated with the expression level of CAR and HNF4α, likely serves as the limiting factor for the activity of many P450s and thus dictates phase I metabolism rate. Our results suggest that these three genes, which are expressed at variable levels among individual human livers, exert dose-dependent effects upon the expression (CAR and HNF4α) or activity (POR) of xenobiotic metabolism enzymes. It would be highly beneficial to identify genotypes linked with variable expression of CAR, HNF4α, and POR. Ultimately, detection of these genotypes could help to predict an individual's clearance rate of a wide variety of prescribed drugs in order to design personalized drug dosing regimens in a cost-effective manner.