Intrinsic hepatic phenotype associated with the Cyp1a2 gene as shown by cDNA expression microarray analysis of the knockout mouse.

Several forms of cytochrome P450 (CYP) appear to metabolize principally pharmaceutical agents, as well as other dietary and plant chemicals. Other CYP forms have major roles in steroid, sterol, and bile acid metabolism. CYP1A2 expression is constitutively high in mouse liver and is well known for metabolizing several drugs and many procarcinogens to reactive intermediates that can cause toxicity or cancer. CYP1A2 is also known to carry out several endogenous functions such as uroporphyrinogen and melatonin oxidation and the 2- and 4-hydroxylations of estradiol. We have used cDNA microarray analysis of the untreated Cyp1a2(-/-) knockout mouse to search for changes in gene expression that might indicate important intrinsic roles for this enzyme. For 15 of the up- or downregulated genes, these increases or decreases were corroborated by reverse-transcription real-time polymerase chain reaction. Other than upregulation of the Hprt gene (used in the selection procedure for disrupting the Cyp1a2 gene), we found several genes upregulated that are associated with cell-cycle regulation and lipid metabolism. Besides Cyp1a2, the gene exhibiting the greatest downregulation was Igfbp1 (insulin-like growth factor binding protein-1), showing only 12% expression of that in the Cyp1a2(+/+) wild-type liver. Recurrent themes between both up- and downregulated genes include cell-cycle control, insulin action, lipogenesis, and fatty acid and cholesterol biosynthetic pathways. Histologically, the Cyp1a2(-/-) mouse exhibited an approximately 50% decrease in lipid stored in hepatocytes, and 50% increase in lipid present in interstitial fat-storing cells compared with that in the Cyp1a2(+/+) wild-type. These data suggest that the CYP1A2 enzyme might perform additional hepatic endogenous functions heretofore not appreciated.

During more than 2.5 billion years of evolution, it is likely that cytochrome P450 (CYP) genes first appeared in prokaryotes and then in early eukaryotes to carry out important roles in critical life processes; following the divergence of plants and animals some 1.8 billion years ago and especially after the radiation of innumerable phyla about 543 million years ago, animal CYP enzymes then took on the functions of metabolizing many plant products and other environmental chemicals that were consumed, inhaled, or in contact with the animal's skin (Nebert 1997;Nebert and Dieter 2000). For some CYP forms, evidence clearly indicates that their principal functions include the oxidative metabolism of endogenous molecules such as steroids, sterols, bile acids, retinoic acid, and many of the >100 eicosanoids (Nebert and Russell 2002). In addition, it is likely there are still many endogenous roles of CYP of which we remain ignorant. For instance, although CYP1A1/1A2/1B1 induction is well established in the metabolism of pro-carcinogenic polycyclic aromatic hydrocarbons and arylamines, these three enzymes are suggested to have pivotal functions in degrading the putative endogenous ligand(s) of the aryl hydrocarbon (AH) receptor and perhaps participate in apoptosis and cell-cycle regulation (Nebert et al. 2000b).
Among the few abundant CYPs that are expressed constitutively in mammalian liver, CYP1A2 carries out several known endogenous functions such as uroporphyrinogen and melatonin oxidation and the 2-and especially 4-hydroxylations of estradiol (Nebert and Russell 2002). The Cyp1a2(-/-) knockout mouse, however, has no apparent overt phenotype or problems with viability or fertility (Liang et al. 1996). These findings suggest that, in the absence of a foreign chemical, the expression of CYP1A2 might be redundant. In humans, CYP1A2-mediated activity varies >60-fold between individuals (Nebert 1997;Eaton et al. 1995;Nebert et al. 1996;Dorne et al. 2001), also with no overt phenotype; most of the variation does not depend on lifestyle or nutrition but is likely to be genetically determined (Le Marchand et al. 1997). It seems unlikely, however, that high constitutive CYP1A2 expression in mammalian liver continues without a particular purpose, whereas CYP1A1 expression occurs only under conditions of ligand-activation of the AH receptor (Nebert et al. 2000b). CYP1A2 is absent in fish but present in birds and mammals, suggesting that between 380 million and 320 million years ago this gene arose from CYP1A1 by way of a duplication event; most likely, the duplicated CYP1A2 gene "drifted" from CYP1A1 until it became involved in one or more critical life functions or in the metabolism of dietary components or environmental chemicals, such that the animal gained a reproductive or survival advantage (Heilmann et al. 1988;Nebert et al. 1991).
The development of mice with disruptions in specific genes allows testing of hypotheses that various CYPs might have physiologic roles not yet identified (Nebert and Duffy 1997). Many knockout mouse lines show profoundly altered phenotypes from normal, including lethality during embryogenesis if the gene participates in a nonredundant component of a metabolic pathway. In one of our laboratories we have been particularly interested in the effect of knocking out members of the Cyp1a family. For example, Cyp1a1(-/-) knockout mice are fertile and healthy and show no apparent changes in the activity or expression of other genes in the [Ah] gene battery, including Cyp1a2, following administration of AH receptor ligands . Similarly, the untreated or inducer-treated Cyp1a2(-/-) null mouse shows no apparent changes in expression of other genes in the [Ah] gene battery, including Cyp1a1 (Liang et al. 1997). On the other hand, both Cyp1a1(-/-1) and Cyp1a2(-/-) lines show modified biochemical responses to foreign chemicals (Liang et al. 1996;Dalton et al. 2000a;Liang et al. 1997;Pineau et al. 1995;Buters et al. 1996;Peters et al. 1999 Kimura et al. 1999;Shertzer et al. 2002). Cyp1a2(-/-) mice metabolize zoxazolamine, caffeine, and phenacetin less extensively than Cyp1a2(+/+) wild-type mice and are more susceptible to some aspects of toxicity of these drugs (Liang et al. 1996;Buters et al. 1996;Peters et al. 1999). On the contrary, Cyp1a2-null mice are unexpectedly protected from 4-aminobiphenyl-induced tumorigenesis ) and 4-aminobiphenyl-induced methemoglobinemia (Shertzer et al. 2002). Cyp1a2(-/-) mice also appear to be completely protected from hepatic uroporphyria caused by dioxin, hexachlorobenzene, and iron overload, although metabolism of these chemicals does not seem to be involved in the mechanism of their toxicity (Sinclair et al. 1998;Sinclair et al. 2000;Smith et al. 2001); in these cases, CYP1A2 appears to be functioning in a mode that does not involve production of a reactive metabolite. Other studies suggest that CYP1A2 might have a role in bilirubin metabolism (Zaccaro et al. 2001). It has been proposed that CYP1A1 metabolizes an endogenous substrate that modulates AH receptormediated gene expression and that, in the liver, CYP1A2 can take over this role (Nebert et al. 2000b). Since there is no constitutive expression of Cyp1a1 in the liver, it would be interesting to investigate whether nonpathologic changes in hepatic metabolism might occur in the complete absence of constitutive CYP1A2. We have thus used cDNA microarrays to compare hepatic gene expression differences between the untreated Cyp1a2(-/-) knockout and the Cyp1a2(+/+) wild-type mouse.

Materials and Methods
The Cyp1a2(-/-) mouse line. Knockout mice had been generated by removing portions of exons 2 and all of exons 3-5 of the Cyp1a2 gene; insertion of the Hprt minigene cassette was used for selection in embryonic stem (ES) cell cultures (Liang et al. 1996). Although the mice were originally generated from a mixture of the C57BL/6J and 129/J inbred strains, the Cyp1a2(-/-) genotype was subsequently backcrossed into the C57BL/6J strain to a theoretical level of >99.8%; if the mouse genome contains 40,000 genes, this would mean that the Cyp1a2(-/-) line in the Nebert mouse colony should have fewer than about 80 genes that might be expected to be of 129/J origin (Nebert et al. 2000a). For this reason, C57BL/6J mice from the Jackson Laboratory (Bar Harbor, ME, USA) were used as the Cyp1a2(+/+) controls in the present experiments. We used untreated males, approximately 10 weeks of age, in the microarray studies; females and males were compared in the histologic studies.
Mouse EST clones and preparation of cDNA. The arrays comprised 4,246 mouse expressed-sequence-tag (EST) clones (2,783 individual Genbank clusters). Two-thirds of the clones were obtained from the I.M.A.G.E. collections held at the MRC Human Gene Mapping Project (http://www.hgmp.mrc.ac.uk/). The remaining one-third of the EST clones were obtained from Research Genetics (RG9 set; http://www.resgen.com). All clones described in this article were verified by sequence analysis. cDNA from the EST was obtained via polymerase chain reaction (PCR) amplification using plasmid-specific primers. The PCR products were separated by electrophoresis on agarose gels to ensure that only a single product was obtained for each clone. The reaction products were precipitated and prepared for array, using methods described (DeRisi et al. 1997;Eisen and Brown 1999).
Printing of the arrays. Arrays were printed on poly-L-lysine-coated slides, UVcross-linked, and blocked prior to use (DeRisi et al. 1997;Eisen and Brown 1999;Turton et al. 2001). The arrays were printed using an arrayer built essentially according to the Stanford designs (cf. http://www.le.ac.uk/cmht/microarray_lab/ Home.htm). The center-to-center distance of the features was 210 µm, and each feature was 90-100 µm in diameter.
Labeling and hybridizations. Total RNA was prepared from mouse liver by sedimentation through CsCl. The RNA of five individual Cyp1a2(+/+) wild-type mice and five individual Cyp1a2(-/-) knockout mice were each separately labeled with both Cy3 dUTP and Cy5 dUTP. RNA labeling was carried out essentially as described (DeRisi et al. 1997;Eisen and Brown 1999;Turton et al. 2001). Priming was achieved with the oligo dT (25) , using 4 µg of the oligo with 50 µg of total RNA. After denaturation at 70°C for 8 min, annealing was allowed to occur as the temperature fell to 42°C over 30 min. At this point, dNTPs (Pharmacia/Amersham, Bucks, UK) were added to final concentrations of 0.5 mM, with the exception of dTTP, which was at 0.2 mM. The desired Cy-labeled dUTP (Pharmacia/Amersham) was then added to a final concentration of 0.1 mM. We used the 1X first-strand buffer (Gibco/ Invitrogen, Paisley, Scottland). RNAsin (20 U) was added to the reaction. Transcription was initiated by addition of 100 U of Superscript II (Gibco/Invitrogen) and allowed to proceed for 1 hr at 42°C before addition of a second 100 U of Superscript II and another 1-hr incubation at 42°C. RNA was removed from the synthesized cDNA by addition of NaOH/ EDTA/sodium dodecyl sulfate (SDS) to final concentrations of 0.195 M/10 mM/0.22%, respectively, and incubated at 70°C for 10 min. The reaction was neutralized by addition of HCl and buffered to pH 7.5 by the addition of Tris-HCl. The reaction products were purified by passage through a Centri-Sep column (Princeton Separations, Inc., Adelphi, NJ, USA), dried, and resuspended in hybridization buffer. Prior to hybridization, the samples were heated to 100°C for 2 min, then at 42°C for at least 30 min.
Analysis of fluorescence and data processing. The fluorescence of all the features on the slides was measured using the GenePix software (version 3.0.0.85; Axon Instruments, Union City, CA, USA). Feature sizes were determined using the inbuilt automated parameters in the first instance and then adjusted manually where appropriate. The fluorescence of each pixel within the feature was determined, and the median fluorescence of these pixel measurements was taken as the measure of fluorescence for the whole feature. The local background fluorescence was measured using the default GenePix parameters. The raw feature data for each channel were globally centered by reference to the median fluorescence of the whole feature set for that channel. The changes in gene expression obtained are shown as means ± standard deviations of the ratio between the Cy3 and Cy5 channels for 10 pairs of hybridizations.
Reverse-transcription real-time PCR using SYBR Green. For an alternative estimation of changes in gene expression, the relative mRNA levels of genes of interest were determined by comparison with a mouse endogenous control gene, β-actin (Actb). The primers used (Table 4) were designed to cross exon-exon boundaries to eliminate the detection of any contaminating genomic DNA.
In the first step, cDNA was synthesized from total RNA. A 1,450-µL master mix was made up as follows: 200 µL PCR buffer, (catalog no. Y02028; Gibco/ Invitrogen); 100 µL MgCl 2 (50 mM); 20 µL each of dATP, dCTP, dGTP, dTTP (all 100 mM); 19.8 µL random hexamers (catalog no. 27-2166-01; Amersham; 90 OD U/mL); dithiothreitol (Gibco 400147); and 1,030.2 µL water. For cDNA synthesis, a 40-µL reaction contained: 29 µl of the master mix; 8 µL total RNA (400 ng); 40 U RNAsin (catalog no. N211B; Promega, Southampton, UK); and 400 U SuperScript II reverse transcriptase (Gibco/ Invitrogen). The mixture was heated for 10 min at 23°C, then for 30 min at 42°C, and finally for 10 min at 99°C. In the second step, 1 µL of the product of the cDNA synthesis (from 10 ng RNA), or a non-template control, was incubated with 24 µL SYBR Green PCR Master Mix (part no. 4309155; Applied Biosystems, Warrenton, UK) containing 900 nM forward primer and 300 nM reverse primer in an ABI PRISM 7700 Sequence Detection System. The thermal-cycler protocol was stage one, 50°C for 2 min; stage two, 95°C for 10 min; stage three, 40 cycles at 95°C for 15 sec and 60°C for 1 min. For every sample of Cyp1a2(+/+) and Cyp1a2(-/-) liver, the relative level of each gene examined was compared with that for β-actin. A comparison was then made for the expression of the gene between the wild-type and the knockout animal.
Histologic analysis of liver. Liver was prepared for routine light-and electronmicroscopic histology and morphometry ). Twenty-eight Cyp1a2(-/-) and 28 Cyp1a2(+/+) mice were used, with equal numbers of males and females (8-10 weeks of age). Phasecontrast microscopy of toluidine bluestained 1.5-micron-thick plastic sections was used to quantify the relative amounts of parenchymal and interstitial cells and to determine the volume density of hepatocyte and of interstitial fat, glycogen pools, and necrosis. A grid of 75 intersections was visualized over the light-microscopic image using a Zeiss Photomik lightmicroscopy (Carl Zeiss GmbH, Vienna, Austria) and camera lucida, and the number of positive intersections lying over hepatocytes, interstitial cells, lipid, and glycogen was counted. The volume density (Vd) of each parameter was determined by dividing those values by the total number of positive intersections lying over the entire tissue.
Statistical analysis of microarray data was performed using the two-tailed paired t-test, taking the reference ratio for the population as 1.0 and comparing this with the ratio change. Means and standarderrors-of-the-mean were obtained from morphometric data, using the General Linear Model of SAS 6.1 (SAS Institute, Inc., Cary, NC, USA). A p <0.05 value was regarded as statistically significant.

cDNA expression microarray.
To investigate potential biochemical phenotypic differences, we compared constitutive hepatic gene expression of Cyp1a2(-/-) mice with that of aged-matched Cyp1a2(+/+) mice, using cDNA microarrays. Labeling of samples with both Cy3 and Cy5 dyes (reverse-labeling) was performed to take into account any methodologic bias. Our information was thus based on 10 separate hybridizations from five mice. Although it is often customary in array work to use an arbitrary cut-off at a 2-fold change in expression, we believe that <2-fold alterations can be important in critical-lifeprocess pathways and therefore chose to list all expressions that had changed significantly as assessed statistically at p <0.05 (Tables 1, 2). Only a relatively few genes were detected that exhibited significant upor downregulation. With the particular array used, we found a greater number of downregulated than upregulated genes in the Cyp1a2(-/-) mouse.
The >5-fold elevation in Hprt gene expression (Table 1) (Liang et al. 1996), or any other gene (van der Lugt et al. 1991), in ES cells. The Gadd45g gene, the Sprlc1 gene, and the Cyp2a4/2a5 gene(s) were upregulated 3.2-, 2.4-and 1.9-fold, respectively. In addition to the marked decrease in Cyp1a2 gene expression (Table 2), as expected, several other genes were downregulated by 2.5fold or more. These included the Igfbp1, G0s2, Fasn, Cyp4a14, Scd1, Hmgcr, Gk and Fabp2 genes. Recurrent functional themes among both these up-and downregulated genes include cell-cycle control, insulin action, lipogenesis, and fatty acid and cholesterol biosynthetic pathways. Besides Cyp1a2, there were at least three other mouse Cyp genes upregulated (Table 1) and two others downregulated ( Table 2). We also documented in our microarray that the expression of at least nine other mouse Cyp genes was not significantly altered ( Table 3). Expression of genes encoding the AH receptor and both forms of heme oxygenase was also not significantly changed (Table 3).

Reverse-transcription real-time PCR.
To prove the accuracy of the cDNA microarray, we chose 15 of the genes that had exhibited significant differences in expression between the genotypes (in Tables 1, 2) for further analysis by reversetranscription real-time PCR, and we used Actb expression as a reference "housekeeping" gene. For the 6 upregulated and the 9 downregulated genes investigated (Table 4), there was good agreement with those changes that had been observed using the cDNA arrays. The data in Table 4 thus confirm the robustness of the cDNA expression microarray approach.
The Vd of hepatic glycogen pools was slightly but not significantly greater in the Cyp1a2(-/-) mouse. The increase in Cyp1a2(-/-) interstitial fat accounted for a small increase in the Vd of interstitium overall, which had a p value of 0.089. Total lipid stores were significantly (p = 0.035) decreased in the Cyp1a2(-/-) liver compared with that in Cyp1a2(+/+) liver. This decrease principally reflected a Vd of hepatocyte-containing lipid droplets in Cyp1a2(-/-) liver that was approximately 55% of that seen in Cyp1a2(+/+) liver. On the other hand, there was a 45% increase in the Vd of lipid found within the Cyp1a2(-/-) interstitial fat-storing cells compared with that in Cyp1a2(+/+) liver. Gender influenced the response, with females showing a greater increase than males in fat stored in interstitial cells; in retrospect, therefore, it might have been worthwhile in the microarray experiments to include comparisons between males and females.
There were no significant differences in hepatocyte necrosis, inflammatory infiltrate, binucleated hepatocytes, or number of apoptotic hepatocytes that could be attributed to the loss of CYP1A2. The mitotic index in Cyp1a2(-/-) liver was slightly elevated but not statistically significant compared with that in Cyp1a2(+/+) liver (p = 0.16).

Discussion
The data in this study strongly suggest that mouse hepatic CYP1A2 is involved in a number of previously unrecognized endogenous functions. These findings are in contrast to popular opinion that CYP1A2 exists solely for the metabolism of pharmaceuticals and other environmental chemicals.
Toxicogenomics | Smith et al. 858 VOLUME 111 | NUMBER 6 | May 2003 • Environmental Health Perspectives Table 3. Hepatic Cyp and other possibly relevant genes whose expression is not significantly different between the Cyp1a2(-/-) and Cyp1a2(+/+) mice.   (Yeagley et al. 2001;Joseph et al. 2002). The expression of these five genes was significantly decreased in Cyp1a2(-/-) mouse liver ( Table 2). The 1.6-fold increase in Cyp7b1 (Table 1), and the downregulation of fatty acid-binding protein-2 (Fabp2) and apolipoprotein A-IV (Apoa4) ( Table 2), provide further evidence of the possible involvement of CYP1A2 in fatty acid and cholesterol pathways. This possible perturbation of fatty acid and cholesterol pathways might be reflected in the histologic assessment (Table 5) in which the Cyp1a2(-/-) mouse liver has less volume density of lipid within its hepatocytes, concomitant with increases in lipid in the fat-storing cells in its interstitium compared with that in the Cyp1a2(+/+) wild-type liver. Cyp gene expression. As expected, the signal for Cyp1a2 gene expression was markedly decreased by >99% in the Cyp1a2(-/-) compared with that in the Cyp1a2(+/+) mouse (Table 2). Since the Cyp1a2 gene is constitutively quite highly expressed in wild-type mice, it was of interest to see whether there might be compensatory expression of any other CYP enzymes. The upregulation of Cyp3a11 (Table 1) and the downregulation of Cyp4a10 and Cyp4a14 (Table 2) might be related to alterations in the arachidonic acid cascade or other physiologic homeostasis (Nebert and Dieter 2000;Nebert and Russell 2002) in the absence of CYP1A2. The ESTs for each of the two Cyp4a genes would have detected either cDNA and, thus, these two could not be properly distinguished--although differences in degree of upregulation were seen.
The greatest difference (1.94-fold upregulation) was observed for Cyp2a4 (Table 1). Although this gene encodes steroid 15α-hydroxylase in the synthesis of testosterone and estradiol, both CYP2A4 and CYP2A5 have been shown to be modulated by circadian rhythm (Lavery et al. 1999;Akhtar et al. 2002). CYP2A5 has coumarin 7-hydroxylase activity and metabolically activates many chemicals such as nitrosamines and aflatoxin that are known hepatic carcinogens in mice (Negishi et al. 1989;Camus-Randon et al. 1996). CYP2A5 induction differs from most because it seems to occur by a variety of agents, including not only drugs and chemicals such as pyrazole, phenobarbital and cobalt but also viral and parasitic inflammation and in hepatic neoplasia (Camus-Randon et al. 1996;Wastl et al. 1998   induction might be associated with a subtle form of liver injury, although nothing was seen histologically. Moreover, mouse Cyp2a5 has been found to possibly be involved in perturbation of the cell cycle and apoptosis (Pelkonen 2002). There was no significant change in the expression of heme oxygenase-1 (inducible form) (Table 3), however, which is often associated with CYP heme turnover and oxidative stress. In addition, there was no detectable elevation in the transcription of Alas1, coding for 5-aminolevulinate synthase, which can also occur in situations of heme insufficiency (unpublished data). The expression of nine other CYPs was detected but was not significantly different between Cyp1a2(-/-) and Cyp1a2(+/+) mice (Table  3). Interestingly, this includes the Cyp2e1 gene, encoding an enzyme that is constitutively quite highly expressed in mouse liver and known to be induced by ethanol and involved in small-molecule intermediary metabolism, diabetes mellitus and ketosis, as well as in the metabolism of drugs and dietary components.
Hprt gene expression. The unexpected observation that the Hprt gene is elevated more than 5-fold (Table 1) can most likely be explained by use of the Hprt minigene cassette in the selective elimination of the Cyp1a2 gene in ES cells (Liang et al. 1996). In all probability, Hprt expression is being driven by the nearby Cyp1a2 5´-flanking regulatory region and promoter. We are unable to conclude with certainty that none of the changes observed (Tables 1, 2) are the downstream consequence of this striking elevation in Hprt expression.

Conclusions
In summary, absence of the mouse Cyp1a2 gene appears to lead to changes in expression of some genes related to cell growth as well as to a downregulation of genes in energy production mediated by insulin. These endogenous functions would appear to be separate from any activity involving the metabolism of environmental chemicals. It is possible that this Cyp1a2 gene knockout reflects a general disturbed metabolic activity and an intrinsic function (Nebert and Dieter 2000) of CYP1A2 in liver and energy metabolism, which results in the decreased ability of lipid to accumulate within the hepatocyte cytoplasm. It is also not possible at this stage to deduce if there is any phenotypic consequence of Hprt overexpression resulting from the use of this gene in developing the knockout mouse line. It will be easy enough in the near future, however, to disrupt this Hprt selection gene and re-examine the Cyp1a2(-/-) knockout mouse in the absence of Hprt.