Introduction

Top Abstract Introduction Experimental Procedures Results Discussion References

Clinically important drug-drug interactions can occur when one agent alters the metabolism of another. These interactions can result from inhibition and/or induction of the enzymes responsible for the rate-limiting step in drug elimination (Lin and Lu, 2001). Currently, a great deal of attention has been paid to drug interactions caused by enzyme inhibition, however, enzyme induction is also very important and can be highly clinical relevant (Breckenridge and Orme, 1971). Drug-metabolizing enzyme induction can have a profound impact on drug metabolism and pharmacokinetics and can result in clinically relevant drug-drug interactions (Waxman, 1999). Recently, the biochemical processes underlying some drug-metabolizing enzyme induction have been elucidated, and several excellent reviews have been written (Blumberg and Evans, 1998; Waxman, 1999; Fuhr, 2000; Zelko and Negishi, 2000). The biochemical mechanism behind phenobarbital (Pb) induction of CYP2B6 was the first to be defined clearly and arose after the discovery of the nuclear constitutive androstane receptor (CAR) (Honkakoski et al., 1998). Pb appears to cause translocation of CAR to the nucleus where it heterodimerizes with another nuclear receptor, the 9-cis-retinoic acid receptor  (RXR) (Zelko and Negishi, 2000). This complex enhances transcription by binding to the phenobarbital response element in gene promoter regions. Another nuclear receptor, the pregnane X receptor (PXR) (Lehmann et al., 1998), has been shown to mediate some of the effects of the antibiotic rifampin on enzyme induction. This receptor is similar to CAR in that binding of drug causes dimerization with RXR, however, the PXR/RXR complex binds to a different response element within gene promoters. This response element, termed the PXR response element, has been found in the promoter for CYP3A4 (Goodwin et al., 1999).

Although the CAR/RXR and PXR/RXR complexes have been shown to bind to distinct response elements in gene promoters, recent studies have also made clear that there is cross-talk between these two systems. Xie et al. (2000) showed that PXR/RXR can bind to phenobarbital response element and induce the expression of CYP2B6; furthermore, CAR/RXR can bind PXR response element and induce the expression of CYP3A4. These results suggest that xenobiotics may induce the expression of several liver enzymes and are consistent with the fact that Pb exerts a wide range of enzymatic and other effects on hepatocytes in vivo (Zelko and Negishi, 2000). Although studies with rifampin have also shown that this drug can induce a number of different genes (Chang et al., 1997; Silva et al., 1998; Bowen et al., 2000; Rodriguez-Antona et al., 2000; LeCluyse et al., 2000; Meunier et al., 2000; Runge et al., 2000; Gerbal-Chaloin et al., 2001), most studies have focused on the regulation of a single gene or a small subset of genes already believed to be altered by rifampin. The effects of rifampin may extend beyond the few enzymes in the liver that thus far have been studied carefully.

Microarray technology is a tool useful for studying the expression of multiple genes simultaneously (Schena et al., 1995; Bartosiewicz et al., 2000; Nguyen et al., 2000). Arrays are commercially available and can be used to determine changes in the pattern of mRNA levels within cells. Primary human hepatocytes cultured in vitro appear to maintain many of the their in vivo characteristics, including the expression and inducibility of drug-metabolizing enzymes. Although some differences between these cells and hepatocytes in situ have been noted, this system is still helpful for examining the effects of xenobiotics on cytochrome P450 enzyme induction and can help in predicting enzyme induction in vivo (Li et al., 1997). Using microarrays to analyze the expression profiles of primary human hepatocyte cultures allows a more comprehensive evaluation of the effects of these drugs. One added advantage of microarrays is that a great deal of information can be obtained from a relatively small sample. This proves especially useful for studies using primary cultures of human hepatocytes because their availability is limited. Here we report our work using two different types of microarray to examine the effects of rifampin on the expression of genes in primary human hepatocyte cultures. Our goal was to provide insight into the effects of rifampin on the induction of drug-metabolizing enzymes to better understand their effects and to better predict drug interactions. Our results show that rifampin has multiple effects on the expression of hepatic genes and that microarray technology can be used reliably to measure the levels of drug-metabolizing genes in primary human hepatocyte systems.

	Experimental Procedures

Top Abstract Introduction Experimental Procedures Results Discussion References

Materials. NaCl, KCl, K₂PO₄, CaCl₂, Tris-HCL, MgCl₂, EGTA, bovine serum albumin (fatty acid free, low endotoxin; catalog number A-8806), dexamethasone, 5-aminolevulinic acid, rifampin, chloroform glycogen, phenol-chloroform-isoamyl alcohol (25:24:1), SDS, RNase A, and EDTA were purchased form Sigma-Aldrich (St. Louis, MO). Glucose, fetal bovine serum, fibronectin, trypan blue solution (0.4%), collagenase (hepatocyte qualified; catalog number 17103-011), 100X ITS solution (insulin, sodium selenite, transferrin, and ethanolamine; catalog number 51500-056), TRIzol reagent, proteinase K, and biotin-CTP were purchased from Invitrogen (Carlsbad, CA). Improved minimal essential medium, HEPES buffer, Waymouth's MB 752, Williams' medium E, glutamine, and nuclease-free H₂O were purchased from Biofluids Division Biosource International (Rockville, MD). Type 1 rat tail collagen was purchased from Collaborative Biomedical Products (Bedford, MA). Percoll, [-³²P]UTP, and [-³²P]dATP were purchased from Amersham Pharmacia Biotech (Piscataway, NJ). Glucagon was purchased from Calbiochem-Novabiochem Corp. (San Diego, CA). Dimethyl sulfoxide (DMSO), isopropanol, and ethanol were purchased from Fisher Scientific Inc. (Fair Lawn, NJ). Cell scrapers and tissue culture dishes were purchased from Costar Corp. (Cambridge, MA). Ethidium bromide stain was purchased from Stratagene (La Jolla, CA). RNase-free DNase was purchased from Ambion, Inc. (Austin, TX). Restriction endonucleases were purchased from New England Biolabs, Inc. (Beverly, MA). Other reagents used were provided by the manufacturers of the expression arrays.

Primary Human Hepatocytes. Primary human hepatocytes were either purchased from a commercial source (In Vitro Technologies, Inc., Baltimore, MD) or prepared from human liver provided by the Washington Regional Transplant Consortium (Falls Church, VA) using a modification of the two-step collagenase digestion method described by Gomez-Lechon et al. (1990). Briefly, small blood vessels were cannulated with 0.75-mm silicon catheters (Insyte IV catheter; Becton Dickinson, Sandy, UT), and liver samples were perfused with 37°C wash buffer and gassed with 95% air (137 mM NaCl, 2.68 mM KCl, 0.7 mM K₂PO₄, 10 mM glucose, and 0.5 mM EGTA, pH 7.4) for 15 to 20 min and then with perfusion digestion buffer (137 mM NaCl, 2.68 mM KCl, 0.7 mM K₂PO₄, 10 mM glucose, and 5 mM CaCl₂, pH 7.4) containing approximately 0.2 U/ml collagenase. The tissue was dispersed gently with a cell scraper in a sterile Petri dish containing ice-cold culture medium (improved minimal essential medium, 10% fetal bovine serum, and 20 mM HEPES, pH 7.0). The disaggregated cell suspension was strained through a 250-µm nylon filter and centrifuged at 50g for 5 min. The cell pellet was washed gently with medium, recentrifuged, and then resuspended in medium containing 25% isotonic Percoll and centrifuged (100g) for 5 min. The resulting cell pellet was washed by resuspending in culture medium and centrifugation (50g) for 5 min. The hepatocytes were resuspended in culture medium, and cell number and viability were determined by trypan blue exclusion assays. The cells were then plated onto fibronectin or collagen-coated dishes. After 1 h, unattached cells were removed by aspiration, and fresh medium was added. The cells were then cultured at 37°C in a 5% CO₂ atmosphere.

Preparation of Total RNA. A cell scraper was used to detach primary human hepatocytes from each tissue culture dish. The cells were centrifuged at 50g for 5 min, the cell pellet was suspended in TRIzol reagent, and total RNA was isolated according to the TRIzol manufacturer's instructions. Briefly, 1 × 10⁶ cells were suspended in 1 ml of TRIzol and incubated at room temperature for 5 min, 100% chloroform (0.2 ml) was added, and samples were mixed, incubated at room temperature for 3 min, and then centrifuged at 12,000g for 15 min at 4°C. The aqueous phase was transferred to a fresh tube, and 0.5 ml of isopropanol was added. The samples were mixed, incubated at room temperature for 10 min, and then centrifuged at 12,000g for 10 min at 4°C. The supernatant was then removed, and the RNA pellet washed with ice-cold 70% ethanol and recentrifuged. After removing the wash solution, the pellet was dried briefly at room temperature and then dissolved in nuclease-free H₂O. RNA yield was determined by spectrophotometry (Beckman DU 640; Beckman Coulter, Inc., Fullerton, CA) and quality assessed by calculated 260 nm/280 nm ratio and by gel electrophoresis followed by ethidium bromide staining. All samples were stored at 80°C.

DNase-Treated RNA. Total RNA samples used for expression arrays were DNase treated before use. Briefly, 50 µg of total RNA was incubated at 37°C for 1 h in 100 µl of buffered solution (40 mM Tris-HCl, pH 7.5, 10 mM NaCl, and 6 mM MgCl₂) containing 5 units of RNase-free DNase I. Reactions were stopped by adding 10 µl of a termination mixture (0.1 M EDTA, pH 8.0, and 1.0 mg/ml glycogen) and 400 µl of nuclease-free H₂O. The RNA was extracted by adding 500 µl of phenol-chloroform-isoamyl alcohol (25:24:1), followed by vortexing and centrifugation at 12,000g for 10 min at 4°C. The aqueous phase was transferred to a clean tube and the phenol-chloroform-isoamyl step was repeated. The RNA was precipitated from the aqueous phase by adding 2 times the volume of the aqueous phase of ethanol and 1/50 volume of 5 M NaCl and then pelleted and processed as described above.

cDNA-Based Microarray Analysis. cDNA-based microarrays were purchased from CLONTECH (Palo Alto, CA). Atlas Human Stress/Toxicology Arrays (catalog number 7747-1) were used for these studies. This array contains >200 genes, many of which are involved in drug metabolism. Assays were carried out according to the manufacturer's instructions. In brief, total RNA was treated with DNase as described above to eliminate possible genomic DNA contamination, and ³²P-labeled cDNA probes were generated by reverse transcription of 5 µg of total RNA in the presence of [-³²P]dATP using a gene-specific primer mixture provided in the kit. After column purification (CHROMA SPIN-200 DEPC-H₂O), the probes were hybridized to filters overnight at 68°C with continuous agitation. Next, the filters were washed extensively to remove nonspecific binding, sealed in polyethylene sleeves to prevent drying, and placed in a phosphor screen exposure cassette for ~48 h. The screen was imaged using a PhosphorImager 445 SI (Molecular Dynamics, Inc., Sunnyvale, CA) at the highest resolution (pixel size, 88 µm). Spot intensities were quantitated using ImageQuaNT software (Molecular Dynamics, Inc.) and correlated with the amount of mRNA in the original sample. For our studies, -fold induction was calculated as the ratio of expression in rifampin-treated versus vehicle control-treated samples each normalized to the total expression of nine housekeeping genes (ubiquitin, phospholipase A2, hypoxanthine-guanine phosphoribosyltransferase, liver glyceraldehyde 3-phosphate dehydrogenase, human-specific tubulin- 1 subunit, HLA class I histocompatibility antigen C-4 -subunit, cytoplasmic -actin, 60S ribosomal protein L13A, and 40S ribosomal protein S9). Housekeeping genes code for proteins involved in the basic metabolic functions within the cell and tend not to be modulated acutely by drug treatments. However, because a given drug may in fact alter the expression of such genes, we used a panel of them so that one of several likely will not be effected by drug treatment and therefore useful for normalization.

Oligonucleotide-Based Microarray Analysis. Oligonucleotide-based microarrays were purchased from Mergen Ltd. (San Leandro, CA). The ExpressChip HO3 DNA Microarray System (catalog number HO3-001) was used for this study. This array contains >1000 genes, many of which are involved in drug metabolism. The assay was carried out according to the manufacturer's instructions. In brief, DNase-treated total RNA (20 µg) was reverse-transcribed using an oligo[(dT)₂₄ T7 promoter]₆₅ primer (consisting of the nucleotide binding sequence for the T7 RNA polymerase followed by 24 thymidine nucleotides) followed by second strand synthesis. The reaction mixture was then treated with RNase I to digest the remaining RNA. The double-stranded cDNA was phenol-chloroform extracted and used as template for in vitro transcription (T7 MEGAscript, Ambion, Inc.) to generate biotin-labeled cRNA probes. These probes were hybridized to the arrays overnight at 30°C with continuous agitation. The arrays were then washed, and hybridized probes were detected using a streptavidin/cyanine-3 fluorescent dye-conjugated antibody. Chips were imaged using a GenePix 4000A array scanner (Axon Instruments, Inc., Union City, CA) and spot intensity was quantitated using ScanAlyze Microarray Software (Michael Eisen, Stanford University, Stanford, CA). -Fold induction was calculated as the intensity in rifampin-treated versus control-treated samples each normalized to the intensity of liver glyceraldehyde 3-phosphate dehydrogenase (GAPDH). For cytochrome P450 2B6 and glutathione S-transferase A3, the level in control-treated cells was below the level of detection and was, therefore, corrected by setting the denominator to 200, which was observed as the lowest signal intensity that could be determined to be significantly above background.

Riboprobe Generation. Probes to measure RNA for five clinically relevant cytochrome P450 enzymes (CYP1A2, CYP2D6, CYP2E1, CYP2C9, and CYP3A4) were generated using total RNA from primary human hepatocytes by reverse transcription and PCR amplification (GeneAmp RNA PCR Kit; PerkinElmer, Branchburg, NJ) using gene-specific oligonucleotide primers (Integrated DNA Technologies, Coralville, IA). The PCR products were cloned into the pCR2.1 vector using the TA Cloning Kit (Invitrogen) according to the manufacturer's instruction. After transformation, plasmids were purified using the Qiagen Maxi Plasmid Kit (Qiagen, Valencia, CA) and the sequences were verified by dideoxysequencing.

CYP1A2 Plasmid. A 416-bp fragment of coding sequence was amplified by RT-PCR using the specific oligonucleotide primers published by McFadyen et al. (1998) with the forward primer 5'-ACA GCA CTT CCC TGA GAG T-3' and the reverse primer 5'-TCT GGA TCT TCC TCT G-3'.

CYP2C9 Plasmid. A 356-bp fragment of coding sequence was amplified by RT-PCR using specific oligonucleotide primers published by Furuya et al. (1991) with the forward primer 5'-ACA TTG ACC TTC TCC CCA CCA GCC-3' and the reverse primer 5'-CAA ATC CAT TGA CAA CTG GAG TGG-3'.

CYP2E1 Plasmid. A 365-bp fragment of coding sequence was amplified by RT-PCR using the specific oligonucleotide primers published by Hakkola et al. (1994) with the forward primer 5'-AGC ACA ACT CTG AGA TAT GG-3' and the reverse primer 5'-ATA GTC ACT GTA CTT GAA CT-3'.

CYP2D6 Plasmid. A 332-bp fragment of coding sequence was amplified by RT-PCR using the specific oligonucleotide primers published by Hakkola et al. (1994) with the forward primer 5'-TGA TGA GAA CCT GCG CAT AG-3' and the reverse primer 5'-ACC GAT GAC AGG TTG GTG AT-3'.

CYP3A4 Plasmid. A 382-bp fragment of coding sequence was amplified by RT-PCR using the specific oligonucleotide primers published by Guillen et al. (1998) with the forward primer 5'-CCT TAC ACA TAC ACA CCC TTT GGA AGT-3' and the reverse primer 5'-AGC TCA ATG CAT GTA CAG AAT CCC CGG TTA-3'.

RNase Protection Assay. The plasmids described above were linearized by restriction digest and used to generate riboprobes using Promega System-T7 (Promega, Madison, WI) and [³²P]UTP. Briefly, 1 µg of linearized plasmid was mixed with a solution containing transcription optimized buffer (final concentration being 1 time), 10 mM dithiothreitol, 20 units of recombinant RNasin ribonuclease inhibitor, 0.5 mM the nucleotides: GTP, CTP, and ATP, 12 µM the nucleotide UTP, 20 units of T7 RNA polymerase and 50 µCi of [³²P]UTP. The samples were incubated at 37°C for 1 h then the template DNA was digested by adding 1 unit of RNase-free DNase and incubation at 37°C for 15 min. The probe was phenol-chloroform extracted as described above and radioactivity measured using by scintillation counting (Beckman LS7500; Beckman Coulter, Inc., Fullerton, CA). Riboprobes (~1.7 × 10⁶ dpm) for specific cytochrome P450 genes along with an internal control for a housekeeping gene (36B4, human acidic ribosomal phosphoprotein) (Laborda, 1991) were hybridized with 20 µg of total RNA at 50°C overnight. Samples were treated for 30 min at 25°C with 40 µg/ml RNase A to digest the unhybridized probe. The reaction was stopped by adding 20% sodium dodecyl sulfate and proteinase K (10 mg/ml). The RNA was phenol-chloroform extracted and run on 6% acrylamide Tris borate-EDTA-Urea gel (Novex, San Diego, CA) at room temperature for 1 h. Gels were dried, protected fragments were imaged using a PhosphorImager, and band intensity was quantitated using the ImageQuaNT software. The band intensity was normalized to the 36B4 level and induction was calculated as intensity of rifampin-treated versus control-treated samples.

	Results

Top Abstract Introduction Experimental Procedures Results Discussion References

Effects of Rifampin on mRNA Expression in Primary Human Hepatocytes Assessed by cDNA-Based Microarrays. CLONTECH Atlas Human Stress/Toxicology Arrays were used to measure the overall effects of rifampin on the level of mRNA expression in primary human hepatocytes. A series of hepatocyte cultures were treated with vehicle control (<0.1% DMSO) or 33 µM rifampin for 3 days, and the mRNA levels in each group were compared. Figure 1 shows the results from one representative hepatocyte preparation. Figure 1A is an image showing subsections of the filters, which contain a portion of the genes induced by rifampin. On this array the cDNAs have been spotted in duplicate and increased spot intensities related directly to the mRNA levels in the sample. The spots corresponding to CYP2A6 in the control-treated cells were barely detectable while the rifampin-treated cells clearly show marked level of expression. Also shown are spots for the CYP2C family of genes. Again the intensities of the spots in rifampin-treated cells are greater than those in control-treated cells. Furthermore, rifampin-treated cells express higher levels of the gastrointestinal form of glutathione peroxidase (GSHPX-GI) and UDP-glucuronosyltransferase 1-1 precursor (UGT-1A). Shown for reference purpose are six sets of spots corresponding to housekeeping genes, the expression of which was not altered by rifampin. The spot intensities were quantitated, and rifampin induction was calculated as the ratio of the intensity in rifampin-treated cells to control cells and normalized by dividing the signal intensities by a sum of the intensities of the nine housekeeping genes. The results for cytochrome P450 genes and other genes associated with drug metabolism are shown as bar charts in Fig. 1B. Rifampin induced CYP2A6 15.4-fold and the CYP2C family 3.0-fold. In addition, a 2.2-fold induction of CYP1A1 and CYP1A2 was seen while little to no effect was seen on the rest of the CYP genes on this array. This experiment was repeated using additional hepatocyte preparations treated with the same concentration of rifampin. They showed considerable variation in the absolute -fold induction, however, the pattern of gene induction remained consistent. A full list of genes on the arrays induced >2-fold by rifampin is shown in Table 1 as the average and standard deviation of three hepatocyte preparations. This list contains many genes involved in drug metabolism including many phase I and phase II enzymes.

View larger version (43K): [in this window] [in a new window]   Fig. 1.   Induction of genes by rifampin as determined by cDNA microarrays. A, image showing subsections of cDNA-based microarray filters. Areas shown contain genes induced and housekeeping genes for comparison. Human hepatocytes were treated with vehicle control (<0.1% DMSO) or 33 µM rifampin for 3 days. Total RNA was isolated and 32P-labeled cDNA probes generated by reverse transcription of 5 µg of total RNA in the presence of [-32P]dATP. Probes were hybridized to filters overnight and washed, and images were acquired using a PhosphorImager. Spot intensity was quantitated using ImageQuaNT and these results are shown in the bar graph (B). -Fold induction was calculated as the normalized ratio of expression in rifampin-treated versus vehicle control-treated cells.

Effects of Rifampin on mRNA Expression in Primary Human Hepatocytes Assessed by Oligonucleotide-Based Microarrays. The results from CLONTECH expression arrays showed that rifampin selectively induced a number of genes involved in drug metabolism. Because the CLONTECH array is based on hybridization to cDNA, we wished to expand on these studies using oligonucleotide-based arrays, which may confer increased specificity. In particular, the cDNA-based arrays we used were not designed to measure differences within the CYP2 family of genes. Therefore, we wanted to determine whether oligonucleotide-based arrays could be used to measure differences between genes with high sequence homology. We chose the ExpressChip HO3 DNA Microarray System (Mergen Ltd.) for these studies. This array has the added advantage that it contains a more comprehensive list of drug-metabolizing enzymes and consists of >1000 gene-specific oligonucleotides (~30 mers) spotted on to a glass slide. The array test for the expression of a number of genes in common with the CLONTECH array. As a result, we were able to compare results obtained by each. To do so, we analyzed the same RNA sample shown in Fig. 1. This system utilizes fluorescent detection and the intensity of the spot directly correlates with the amount of mRNA in the sample. The fluorescent scans can be converted into pseudocolor images that represent the fluorescein dye used for detection, and Fig. 2A is one such image showing subsections of the arrays that contain a number of spots that represent different cytochrome P450 enzymes. Each gene is represented by an individual spot. The one representing CYP3A4 is notably greater in intensity in the rifampin-treated cells compared with the control-treated cells, whereas spots for CYP2E1 appear to be equally intense in each array. In contrast to the cDNA-based array, the oligonucleotide-based array is able to distinguish between three of the CYP2C family members: CYP2C8, CYP2C9, and CYP2C18. From the pseudocolored image it is clear that the CYP2C9 and CYP2C8 are induced by rifampin, whereas CYP2C18 is not. The spot intensities were measured using the ScanAlyze software and induction calculated as the ratio of the intensities in rifampin-treated cells/control-treated cells after normalizing to the intensity of a housekeeping gene (GAPDH). The bar graph in Fig. 1B shows the results calculated for the cytochrome P450 genes. Rifampin selectively induces cytochrome P450 genes, with CYP3A4 being the most highly induced at 55.1-fold. Also induced were the other CYP3A family members CYP3A5 and CYP3A7 at 5.0- and 27.7-fold, respectively. Also shown is the differential induction within the CYP2C family, with CYP2C8 being the highest induced at 6.5-fold followed by CYP2C9 at 3.7-fold, whereas CYP2C18 was not induced. Table 2 shows all genes that were induced >2-fold as determined by the oligonucleotide-based arrays. The list includes a number of cytochrome P450 genes and two glutathione S-transferases. These results are in agreement with those obtained using the cDNA-based arrays.

View larger version (31K): [in this window] [in a new window]   Fig. 2.   Analysis of genes induced by rifampin as determined by oligonucleotide-based microarrays. A, areas of glass slide containing cytochrome P450 enzymes are shown. Human hepatocytes were treated with vehicle control (<0.1% DMSO) or 33 µM rifampin for 3 days. Total RNA was isolated, and biotin-labeled cRNA was generated and hybridized to chip. After washing, hybridized probe was detected by incubating with streptavidin followed by Cy3-conjugated antibodies. Chips were imaged using a GenePix 4000A array scanner. Spot intensity was quantitated using ScanAlyze Microarray Software and these results are shown in the bar graph (B). -Fold induction was calculated as the normalized ratio of expression in rifampin-treated versus vehicle control-treated cells. *, intensity in control-treated cells was below level of detection and corrected for as discussed in the text.

Validation of Expression Arrays Using RNase Protection Assay. We generated probes suitable for RNase protection assays to confirm and validate the data we obtained from the expression arrays. We chose this method because it is a robust, sensitive, and specific method of measuring mRNA levels (Nass and Dickson, 1995). The genes chosen for validation were based on their clinical relevance and on their differential induction by rifampin as determined by the arrays. The results are shown in Fig. 3. The undigested fragments run slightly higher than the protected fragments due to the presence of residual noncoding nucleotides from the plasmid, which is important because it allows undigested probe to be distinguished from the hybridization signal. When the probes are hybridized to the sample RNA and digested, the coding sequence is protected and the intensity of the protected fragment correlates directly with the amount of mRNA in the original sample. One can compare the level of expression in two samples by measuring the intensity of the bands after normalizing to a housekeeping gene. This normalization controls for any loss of sample during processing. For our studies we used a probe for 36B4 (human acidic ribosomal phosphoprotein, GenBank accession number M17885) to normalize our message because its expression has been shown not to change by a number of drugs including estradiol and tetradecanoylphorbol acetate (Laborda, 1991). As shown, the intensity of the protected fragments for CYP2D6 and CYP2E1 are similar in RNA samples from control-treated cells and rifampin-treated cells. However, rifampin-treated cells show an increase in expression for CYP1A2, CYP2C9, and CYP3A4. The band intensities were measured, and -fold induction was calculated as intensity in rifampin-treated versus control-treated cells after normalizing for 36B4 signal. The ratios of these results are shown as numeric values in the table in Fig. 2B.

View larger version (26K): [in this window] [in a new window]   Fig. 3.   RNase protection assay for cytochrome P450 enzymes CYP2D6, CYP2E1, CYP1A2, CYP2C9, and CYP3A4. Human hepatocytes were treated with vehicle control (<0.1% DMSO) or 33 µM rifampin for 3 days. Total RNA was isolated and hybridized with [-32P]dUTP riboprobes specific for each CYP. Single-stranded RNA was digested, and resulting protected fragments were run on 6% acrylamide Tris borate-EDTA-Urea gels. Gels were dried then imaged using a PhosphorImager, and band intensity was quantitated using ImageQuaNT software. A, image of gels. B, table showing -fold induction calculated as the normalized ratio of expression in rifampin-treated versus vehicle control-treated cells.

	Discussion

Top Abstract Introduction Experimental Procedures Results Discussion References

We have shown that rifampin selectively induces a number of genes involved in drug metabolism and that the genes induced by rifampin include many in addition to those that code for the cytochrome 450 enzymes. Furthermore, the effects of rifampin appear selective within CYP families. Rifampin induced the expression of CYP2C8 and CYP2C9 but had no effect on the message level for CYP2C18. These results are in agreement with those recently published by Gerbal-Chaloin et al. (2001) who used RNase protection assays to showed that rifampin induces CYP2C8 and CYP2C9 but not CYP2C18. In addition, their results show that CYP2C8 was induced to a greater extent than CYP2C9, which is consistent with our results. Specificity of effect within the CYP2C family is probably the result of differences in the promoters of these genes. Indeed, studies have shown that the promoter for CYP2C18 is different from the CYP2C8 and CYP2C9 promoters that exhibit very high (75%) sequence homology (de Morais et al., 1993).

In contrast to the CYP2C family of genes, all three members of the CYP3A family were induced by rifampin, although to different degrees. Others have reported that the promoter regions of CYP3A4 and CYP3A7 are very similar and contain PXR response elements in the proximal and distal promoter regions that are required for rifampin inducibility (Goodwin et al., 1999; Bertilsson et al., 2001). From our results, it is interesting to speculate that the promoter for CYP3A5 has this same arrangement. Besides the cytochrome P450 enzymes, rifampin induced other phase I enzymes including FMO4 and FMO5. The effects of rifampin were selective within the monoamine oxidase family of enzymes, with the induction of MAO-B compared with no effect on MAO-A. Differences within the promoter regions of these genes have been shown, but the effects of rifampin on RNA expression have not been reported previously (Shih et al., 1993).

In addition to these phase I enzymes, our data show that rifampin brought about a broad increase in the expression of phase II enzymes including certain glutathione-S-transferases and UDP-glucuronosyltransferases. It makes intuitive sense that a xenobiotic would induce the expression of enzymes responsible for eliminating the reactive intermediate generated by phase I reactions. To our knowledge this is the first report showing simultaneous induction of both phase I and phase II enzymes by rifampin. These data suggest, but do not prove, that rifampin may induce phase II enzymes by a PXR mediated mechanism; however, it is possible that other regulatory pathways exist. PXR may be involved mechanistically in the induction of other genes induced by rifampin namely, FMO4, FMO5, MAO-B, UGT-1A, and GSHPX-GI. In fact, Sugatani et al. (2001) recently characterized a CAR response element in the UGT-1A promoter and because PXR/RXR complexes have been shown to bind CAR response elements, this may be the mechanism by which rifampin induces the expression of this gene. A careful comparison of the promoter regions of genes effected by rifampin may identify response elements common to rifampin inducible genes and may help identify new PXR response elements.

For these studies we used two different commercially available expression microarrays. These arrays have many overlapping genes and we were therefore able to compare results obtained by each. We found that the pattern of induction caused by rifampin was generally consistent among hepatocyte preparations, however, the absolute -fold induction varied considerably. Large variability in the level of induction between hepatocyte preparations may be caused by many possible factors, including differences in liver condition, time to preparation, and culture conditions (Silva et al., 1998; LeCluyse et al., 1999, 2000; Meunier et al., 2000). Within our study, different extracellular matrix conditions were used (fibronectin and collagen), which could have influenced induction (Lindblad et al., 1991). To compare accurately the results from the microarrays that we used, we analyzed the same RNA sample with each array. To validate the array results, the same sample was further analyzed using RNase protection. The five probes generated for RNase protection were based on their clinical relevance and their differential induction by rifampin. The results obtained from the two array systems were in very close agreement. For example, each showed that rifampin does not cause a significant induction in CYP2E1 or the housekeeping genes (ubiquitin, phospholipase A2, GAPDH, -tubulin, and 40S ribosomal protein) but does induce the CYP2C enzymes.

The induction of the CYP2C family determined with the cDNA-based method was not specific for individual family members. The oligonucleotide-based method however, was able to measure specific differences with in the CYP2C family of genes. Results from RNase protection validate the array results with the calculated -fold induction being remarkable similar.

The methods of RNA detection we tested were not in agreement for CYP1A2. Both the RNase protection and the cDNA-based array detected a >2-fold induction of this gene. However, the oligonucleotide-based method did not agree. Although there are many potential explanations for this discrepancy, one possibility is that a manufacturing defect lead to less oligonucleotide being present in the spot for CYP1A2 on one of the slides. One of the advantages of the CLONTECH cDNA-based array is that each gene is represented by duplicate spots. This helps mitigate a potential source of artifact. CYP2D6 was not detected by the oligonucleotide-based method, however the mRNA was clearly present based on results from the radioactive methods of detection. We also found that several other genes were not detected using the fluorescent-based system, and this may speak to the general relative sensitivities of these assays.

Although RNase protection assay is a reliable and accurate method of RNA detection it is not suitable for looking at multiple genes simultaneously; in addition, the amount of sample needed for analysis is relatively high. For example, 20 µg of total RNA was used to measure the expression of a single gene by RNase protection assay, whereas 20 and 5 µg of total RNA was used for the oligonucleotide-based array and the cDNA-based arrays, respectively, to measure expression of multiple genes. Therefore, we found the RNase protection assay is not particularly well suited for our studies and was used only for validation purposes.

It is clear that each method that we tested had a distinct set of advantages and disadvantages. The cDNA-based array required the use of radioactivity with all of the issues of safety and expense that are associated with the use of radio-isotopes. On the other hand, these arrays proved to be generally more sensitive than the oligonucleotide-based array, and the duplicate spots for each gene helped to eliminate loss of data due to artifacts. In addition, it is possible to use these filters several times, which can result in significant cost reductions. The oligonucleotide-based array has a significant advantage in that it was able to discriminate between the members of closely related gene families (CYP2C and CYP3A). This is consistent with the manufacturer's claim that a set of rigorously controlled criteria are used to design oligonucleotides that can discriminate between genes with high sequence homology. However, validating the expression of each gene within the families by additional techniques would provide better proof of specificity. These arrays have the additional advantage that they did not require the use of radioactivity. However, these arrays were found to be rather less sensitive, and a dedicated fluorescent array reader is required to use this technology.

Primary human hepatocytes are useful for examining the effects of xenobiotics on cytochrome P450 enzyme induction and can help predict clinical drug-drug interactions. We are aware that our studies examined effects only at the level of transcription and so we must be cautious in extrapolating these results to alterations in protein levels. Although the effects of rifampin on the transcription of CYP3A4 have been correlated with protein levels, activity, and clinical consequences, clinical consequences resulting from the induction of other drug-metabolizing enzymes remains to be determined. Therefore, the microarray data we present generates a large number of hypotheses that need to be addressed by further studies. Our data suggest that microarray analysis can be used to compare the pattern of effects caused by different xenobiotics, and this may prove a useful screening system for drugs that are believed to effect liver enzyme levels.

This study shows that rifampin is able to induce the expression of a large number of pharmacologically important enzymes, but that it also has a selective effect. It is clear that it is no longer reasonable to presume that the inducing effects of rifampin are the result of a single effect on one enzyme system, and clinicians and researchers need to be alert to the possibility of multiple effects of the drug. Rifampin is likely able to exploit a common biochemical mechanism to bring about a specific pattern of gene expression and other drugs may do so also. Our data suggest a number of important areas for further investigation that include comparison of the molecular mechanism of induction of the cytochrome P450, FMO, MAO, GST, and UGT enzyme families.