Assessment of Hepatocytes and Liver Slices as in Vitro Test Systems to Predict in Vivo Gene Expression

Bart A. Jessen*,1,2, Jennifer S. Mullins{dagger},1, Ann de Peyster{dagger} and Gregory J. Stevens*

* Pfizer Global Research and Development, 10724 Science Center Drive, La Jolla, California, 92121, and {dagger} Graduate School of Public Health, San Diego State University, San Diego, California 92182

Received April 8, 2003; accepted June 9, 2003


    ABSTRACT
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The use of in vitro systems to predict in vivo responses to chemical agents provides the benefits of requiring fewer animals, reducing variability between samples, requiring less test material, and enabling higher throughput. In the present study rat tissue slices and primary hepatocytes were compared as in vitro systems to predict in vivo changes in gene expression in response to treatment with known liver toxicants or inducers. Five compounds (phenobarbital, carbon tetrachloride, Wy-14,634, alpha-napthylisothiocyanate, and tacrine) were chosen for their established and diverse mechanisms of hepatoxicity or microsomal induction. Expression profiles from male Sprague-Dawley rats or in vitro systems treated for 24 h were measured by DNA oligonucleotide microarrays containing 8700 probe sets. Qualitative comparison of expression revealed a >80% concordance between in vivo liver and both in vitro systems; however, the responsiveness of both in vitro systems to compound-induced changes in gene expression was far less than that of in vivo. Furthermore, both in vitro systems appeared similar in their ability to reproduce compound-induced changes in gene expression observed in vivo.

Key Words: microarray; hepatocytes; liver slices; hepatotoxicity; gene expression.


    INTRODUCTION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Hepatotoxicity resulting from exposure to xenobiotics, including pharmaceuticals, continues to be of major concern. The liver is one of the first organs to be exposed to orally delivered compounds via the portal vein. Concentrations during the first-pass of compound through the liver are often much higher than peak plasma concentrations. Once distributed, the compound continues to perfuse the liver as hepatic blood flow accounts for ~30% of the cardiac output. The liver is also the major site for xenobiotic metabolism, which can lead to the formation of toxic metabolites. The high exposure and metabolic activity make the liver one of the primary targets for xenobiotic-induced toxicity.

The use of in vitro systems to predict liver toxicity provides several advantages over in vivo testing. In vitro systems require fewer animals, less test material, and are often higher throughput. Furthermore, high-throughput in vitro assays are necessary in order to keep pace with the increasingly large and diverse chemical libraries generated by the pharmaceutical industry. In order for an in vitro assay to be useful, it must be predictive of an in vivo response to similar treatment. The methods of assessing and predicting toxicity are becoming increasingly sophisticated. The ideal predictive tool is one that produces signals that precede overt signs of toxicity and are present at exposures below which overt toxicity becomes evident. Gene expression, either through expression of specific marker genes or through global expression patterns ("fingerprints"), is currently being assessed for its potential use in predicting toxicity (Gerhold et al., 2001Go). Expression patterns can also be used to compare the similarity or differences between cell lines (Achary et al., 2000Go) or disease states (Ho et al., 2001Go). In this study, changes in gene expression were assessed by DNA array technology to compare two commonly used in vitro systems, isolated hepatocytes and precision cut liver slices, to changes observed in vivo in response to several liver toxicants.

The in vitro systems chosen for this study, hepatocytes and liver slices, have been used extensively to evaluate liver function, toxicity, and metabolism (Bassir and Emafo, 1970Go; Ghantous et al., 1990Go; Henderson and Dewaide, 1969Go; Thor et al., 1978Go; Wagle, 1975Go; Weinhold, 1969Go). Since human liver slices and hepatocytes are available, they often form a bridge for comparison between animals and humans in toxicology and metabolism studies (Wrighton et al., 1995Go). Hepatocytes are generated by proteolytic digestion of liver tissue followed by centrifugation to enrich the hepatocyte cell population for adherent or suspension cultures. Precision-cut liver slices maintain the tissue architecture and contain the variety of cell types normally found in liver. Toxicological comparisons of both in vitro systems to in vivo for certain toxic agents have been previously performed (Fisher et al., 1991Go).

The compounds used in this study were chosen for their diverse mechanisms of hepatotoxicity. They include enzyme inducers, such as phenobarbital (PB) and Wy-14,643 (Wy), as well as known toxicants such as CCl4, {alpha}-napthylisothiocyanate (ANIT), and Tacrine (Tac). PB enhances GABA(A) receptor activity and is prescribed as a sedative and anti-epileptic. PB is also a known inducer of drug metabolizing enzymes, such as P450 2B1/2 (Nims et al., 1993Go), UDP glucuronosyl transferase (UGT; Wishart, 1978Go), glutathione-S-transferase (Hales and Neims, 1977Go), and aldehyde dehydrogenase (Deitrich, 1971Go). Some of the gene expression changes induced by PB have been attributed to its indirect activation of the CAR nuclear receptor (Kawamoto et al., 1999Go). Studies have shown that PB induction may cause regional toxicity in vivo due to the localization of cytochrome P450 2B1/2 in the perivenous and mid-zonal regions of the liver (Hassett et al., 1989Go; Waxman and Azaroff, 1992Go). Wy is a peroxisome proliferator activated receptor (PPAR) agonist, causing direct activation of transcription of genes that mediate its hypolipidaemic properties. Wy belongs to a class of compounds known as peroxisome proliferators, and is known to induce cytochrome P450 4A, as well as other metabolic enzymes (Bojes and Thurman, 1994Go; Orton and Parker, 1982Go). Through its hyperproliferative activity, Wy exposure causes hepatic tumors in rodents (Bojes and Thurman, 1994Go; Reddy et al., 1979Go). CCl4, a common laboratory solvent, causes acute liver cell necrosis through an oxidative metabolism induced free radical mechanism (Schiaffonati and Tiberio, 1997Go). CCl4-induced cellular damage is followed by regeneration of the tissue and induced expression of heme oxygenase, Mn and Cu/Zn superoxide dismutases, and H and L ferritin subunits (Schiaffonati and Tiberio, 1997Go). The proliferative effects of CCl4 make it a nongenotoxic liver carcinogen (Uno et al., 1999Go). ANIT, once used as an insecticide, gives rise to cholangiolitic hepatitis characterized by necrosis of the bile duct epithelial cells and hepatic parenchymal cells in the liver (Fukumoto et al., 1980Go). Tac, a reversible cholinesterase inhibitor, is currently prescribed for the palliative treatment of Alzheimer’s disease (Summers et al., 1989Go). Liver toxicity is evident by elevated alanine aminotransferase (ALT) levels in about 25% of patients using Tac (Watkins et al., 1994Go) and studies have shown an association with mitochondrial dysfunction (Robertson et al., 1998Go) and glutathione depletion in hepatocytes (Dogterom et al., 1988Go), and CYP1A family induction in vivo (Sinz and Woolf, 1997Go).

The objective of this study was to measure changes in gene expression by DNA array technology in two commonly used in vitro systems, isolated hepatocytes and precision cut liver slices, as well as changes observed in vivo in response to several liver toxicants. Studies were conducted in effort to demonstrate the ability of the in vitro systems to mimic gene expression changes in the liver in vivo. These studies were also intended to differentiate the two in vitro systems by the degree to which they represent liver specific gene expression.


    MATERIALS AND METHODS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Materials.
PB, CCl4, Wy, ANIT, Tac, and vehicle control (Wesson corn oil) were purchased from Sigma (St. Louis, MO).

Animal treatment.
All animals used in both the in vivo and in vitro systems were male Sprague-Dawley rats (180–220 g) purchased from Charles River Laboratories (Kingston, NY), and were acclimated approximately one week prior to use. All animals had free access to food (Labdiet 5001, Purina Mills; St. Louis, MO) and water with a standard 12-h light: dark cycle. This work was approved by an internal Institutional Animal Care and Use Committee (IACUC) and was conducted in compliance with all applicable IACUC requirements in accordance with the NIH Guide for the Care and Use of Laboratory Animals (NRC, 1996Go).

Individual rats from each treatment group (n = 4) received a single dose of compound by either ip injection or po gavage. Dosing solutions were prepared to deliver a volume of 10 ml/kg as follows: 100 mg/kg PB (po); 1 ml/kg CCl4 (po); 100 mg/kg Wy (po); 150 mg/kg ANIT (ip); 30 mg/kg Tac (po). The control group consisted of rats treated with 10 ml/kg corn oil (Wesson). Dose levels were selected based on their association with mild hepatotoxicity and/or enzyme induction as previously reported (Bojes and Thurman, 1994Go; Fukumoto et al., 1980Go; Schiaffonati and Tiberio, 1997Go; Stachlewitz et al., 1997Go). No clinical signs of overt toxicity were observed during the study. Twenty-four hours after dosing, animals were euthanized by exsanguination under anesthesia, blood was collected by cardiac puncture, and analysis of serum biochemistry was performed (LabCorp, Inc.; San Diego, CA). Statistical analysis of the serum biochemistry involved a one-way ANOVA using Statview 5.0.1 (Abacus Concepts; Berkeley, CA). At necropsy, livers were weighed, approximately 100 mg of liver, derived from the right lobe, was placed in 1.5 ml RNAlater (Qiagen; Valencia, CA), and the remaining tissue was fixed in 10% buffered formalin for subsequent histological examination. Animals in the in vitro experiments were untreated prior to the terminal procedure.

Hepatocyte culture and treatments.
Primary rat hepatocytes were prepared from rats on three separate occasions using a two-step in situ perfusion (McQueen, 1989Go). Hepatocytes were plated on six-well collagen coated tissue culture plates in 2 ml William’s medium E at 106 cells/well. Prior to treatment hepatocytes were pre-incubated for 4 h at 37°C in a humidified incubator with 95% air and 5% CO2. After pre-incubation, medium was replaced with medium containing one of the following compounds: 0.1% DMSO (control), 0.2 mM PB, 0.3 mM CCl4, 0.5 mM Wy, 0.1 mM ANIT, or 0.05 mM Tac. The compound concentrations selected were the maximum which produced less than 20% cytotoxicity based on ATP content using the Bioluminescent Somatic Cell Assay Kit (Sigma; St. Louis, MO) analyzed with a Lumiscan Ascent luminometer (Labsystems; Vantaa, Finland). After 24 h of treatment, hepatocytes were harvested using 250 µl of Trizol (Gibco BRL; Rockville, MD) reagent per well. The contents of three wells were pooled for each sample and stored at -70°C until RNA isolation could be performed.

Precision-cut liver slice treatment.
Precision-cut liver slices were prepared from rats on three separate occasions using a Krumdieck Tissue Slicer (Alabama Research and Development; Munford, AL) and incubated in roller culture as previously described (Smith et al., 1985Go). Briefly, each slice was transferred onto a stainless steel mesh insert, placed in a scintillation vial containing 1.7 ml of William’s Medium E, and capped with a vented teflon lid. Liver slices were maintained at 37°C in a rolling incubator with 95% oxygen and 5% CO2 and allowed to pre-incubate for 2 h prior to treatment. At this time the media was replaced with 1.7 ml of media containing the same concentrations of compounds as those described above for hepatocyte cultures. After 24 h of incubation individual slices were homogenized in 1 ml of Trizol reagent. Four to six slices were pooled for each sample and stored at -70°C until RNA isolation could be performed.

RNA isolation and probe preparation.
Total RNA was isolated from the samples using either a modified Trizol extraction or a Qiagen RNeasy protocol (Qiagen; Valencia, CA). The modified Trizol procedure, which was used for both in vitro systems, involves denaturation with guanidinium thiocyanate followed by phenol/chloroform extraction. For RNA isolation from livers of in vivo treated animals the RNeasy procedure was used. RNA quality was assessed by sample absorbance at 260 and 280 nm and by agarose gel electrophoresis. Double-stranded cDNA was prepared from the total RNA using reverse transcriptase and a T7-(dT)24 primer (Gibco BRL; Rockville, MD). Incorporation of the T7 promoter allowed for production of biotin labeled cRNA via the BioArray High Yield RNA transcript labeling kit (Enzo; Farmingdale, NY). RNA from rat heart, testes, kidney, and lung was purchased from Clontech (Palo Alto, CA) and used to generate the liver specific gene list.

Hybridization, scanning, and quantitation.
cRNA from each sample was fragmented into 35–200 bp pieces (confirmed by agarose gel electrophoresis) and hybridized to a single rat RG U34A DNA oligonucleotide Genechip array (Affymetrix; Santa Clara, CA) according to manufacturer’s specifications. Hybridized arrays were stained with R-phycoerythrin conjugated streptavidin using antibody signal amplification prior to being scanned by a GeneArray Scanner (Agilent; Foster City, CA). Array image analysis, quantification of raw gene expression values, mismatched probe background subtraction, and present/absent calls were performed using Microarray Suite software, version 5.0 (Affymetrix; Santa Clara, CA).

Gene expression analysis and data interpretation.
Comparisons of present genes, fold change determinations, statistical comparison, experiment normalizations, Venn diagrams, and experiment clustering were performed with Genespring software, version 5.0 (Silicone Genetics; Santa Clara, CA). Gene expression values for each chip were normalized to their respective median value. Experiment tree clustering was performed using the Pearson correlation. Systems were compared by linear regression of log transformed average raw expression intensities using Excel (Microsoft; Seattle, WA). Gene lists used for Venn diagram comparisons were generated by a filter requiring present call in at least 50% of the samples within a treatment group. Fold changes were generated by comparing the average normalized expression values between the control and treatment groups. Fold change filters also included the requirement that the genes be present in at least 50% of treated samples for up-regulated genes, and 50% of controls for down-regulated genes. Statistical group comparisons were performed using the Welch’s t-test. The liver-specific gene list was generated by subtracting genes found present in heart, lung, kidney, and testes from those found present in at least 50% control liver samples.


    RESULTS
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Clinical Chemistry and Histopathology
Serum samples taken 24 h postdose from the in vivo portion of the experiment were analyzed for clinical chemistry changes (Table 1Go). In the Wy treatment group there were no significant compound-related changes in the clinical chemistry parameters. The decrease in alkaline phosphatases (ALK) seen in the PB treated group is a common response to p450 inducers and is not indicative of organ damage. The minor changes in gamma glutamyl transferase (GGT) found in the Tac treated rats were consistent with a subtoxic exposure. The elevations in total bilirubin (TBIL), ALT, and aspartate aminotransferase (AST) observed in animals treated with ANIT as well as the increases in ALT in CCl4 treated animals indicated mild hepatic toxicity and were consistent with the mechanism of toxicity for these compounds. Histological examination of kidney and livers was performed for all samples. No compound-related changes were found in the kidney samples in any treatment groups nor were there compound related changes evident in the histological examinations of livers from PB-, Wy-, and Tac-treated animals. ANIT-treated animals revealed minimal levels of hepatic edema and necrosis while CCl4 treated animal tissues showed some vacuolation and centilobular necrosis (data not shown).


View this table:
[in this window]
[in a new window]
 
TABLE 1 Selected Clinical Pathology Parameters from in Vivo Treated Rats
 
Gene Expression Data
Expression profiles from the control samples of the three different systems were compared using a liver-specific gene list (Fig. 1Go). The gene list consisted of 190 genes from the RG U34A array that were expressed in rat liver but not heart, lung, testes, or kidney. Of the 190 genes, 119 genes (62%) were expressed by all three systems while 49 (26%) were not expressed in the in vitro systems. Since both in vitro systems expressed 130 (68%) of the liver-specific genes, each in vitro system expressed 11 liver-specific genes (6%) that the other in vitro system did not express.



View larger version (45K):
[in this window]
[in a new window]
 
FIG. 1. In vitro expression of rat liver specific genes. A total of 190 genes were found to be expressed in rat liver but not in rat testes, heart, kidney, or lung and were used to create the liver-specific gene list. The presence of these genes in both in vitro systems was compared to in vivo in the above Venn diagram.

 
Average raw gene expression values from control samples from all three systems were log transformed and plotted against each other in Figure 2Go. Linear regression analysis showed that the correlation coefficient for hepatocytes versus in vivo (Fig. 2AGo) and slices versus in vivo (Fig. 2BGo) were virtually indistinguishable (r = 0.90 and 0.89, respectively). The correlation coefficient for the two in vitro systems plotted against each other was 0.93, indicating common changes in gene expression due to in vitro culturing. The correlation coefficients for the untransformed data for the hepatocytes versus in vivo comparison, slices versus in vivo comparison and hepatocytes versus slices comparison were 0.78, 0.76, and 0.87, respectively (data not shown).



View larger version (26K):
[in this window]
[in a new window]
 
FIG. 2. Correlation of expression among the three systems. Log transformed average raw expression data was plotted to generate correlation coefficients for hepatocytes versus in vivo (A), slices versus in vivo (B), and hepatocytes versus slices (C). Data were derived from three independent experiments.

 
Expression profiles from treated samples were then compared using several other gene lists, or subsets of genes from the U34A array (Table 2Go). The All Genes List refers to all of the 8799 probe sets on the U34A array. The Toxicology Gene List was taken from the 1031 rat probes sets that make up the rat toxicology array (Affymetrix; Santa Clara, CA). The Predictive Toxicology Genes were derived from eight rat orthologs of mouse genes reported to be 100% predictive of hepatoxicity (Thomas et al., 2001Go). Each gene list was further refined to include only those genes present on the U34A array and expressed in the in vivo system for each treatment.


View this table:
[in this window]
[in a new window]
 
TABLE 2 Comparison of the Number of Genes Expressed among the Three Systems
 
When compared using the All Genes List, hepatocyte cultures expressed slightly fewer genes than in vivo samples, but more genes than slices, in all but the ANIT and Wy treatment groups. All samples expressed between 2500 and 3600 genes evaluated by the 8799 probe sets represented on the array. A similar pattern was seen when toxicology genes were evaluated. Hepatocytes expressed fewer toxicology genes than the in vivo system but slightly more than slices in all but the ANIT treatment group. Of the six predictive toxicology genes expressed in vivo, five were expressed in both in vitro systems from the Wy, ANIT, and Tac groups. Five predictive toxicology genes were also expressed in hepatocyte cultures undergoing vehicle, PB, or CCl4 treatment, while the slices expressed four predictive toxicology genes under these treatment conditions.

The in vitro systems were also compared to the in vivo system based on treatment-induced changes in expression as compared to control (Table 3Go). The same gene lists were used as in Table 2Go, but only those genes whose expression levels changed by >=2-fold in the in vivo system with respect to controls were compared among the three systems. The fold change criteria greatly increased the stringency of the comparison. The majority of genes whose expression changed in vivo in response to treatments were not reflected in vitro, regardless of the gene list used. When comparing from the All Genes List, hepatocytes better approximated the in vivo expression changes with ANIT and CCl4 treatments, while PB-treated slices better approximated the in vivo changes in gene expression. The two in vitro systems performed nearly identically when subjected to Tac and Wy treatments. When changes in toxicology gene expressions were compared, the most dramatic differences were found in the Wy and PB treatments in which hepatocytes and slices, respectively, shared more in vivo gene changes. Only one in vitro treatment condition, Wy treated slices, caused changes in more than one of the predictive toxicology genes. ANIT treatment was the only condition where hepatocytes showed changes in expression in the predictive toxicology category. Comparisons of statistically significant differences (p < 0.05) in gene expression were also made using the All Genes List. In most cases there were substantially fewer genes that were statistically significantly changed than were changed by >=2-fold. There was also a smaller proportion of gene expression changes in the in vitro systems that were statistically significantly altered as compared to those that were changed >=2-fold.


View this table:
[in this window]
[in a new window]
 
TABLE 3 Comparison of the Number of Genes Altered among the Three Test Systems
 
Average gene expression data from all of the treatment conditions and all three test systems were subjected to experiment tree-clustering analysis using the Pearson correlation (Fig. 3Go). In Figure 3AGo, raw expression data was clustered for a comparison of the expression levels between the three systems under the various treatments. In most cases the samples cluster by system, with both in vitro systems clustering equidistant from in vivo. The one exception was the in vivo ANIT treatment, which clustered with slices. Within the in vivo sample groups, the Tac treated group clustered closest to control, followed by CCl4, Wy, PB, and ANIT. Within the hepatocyte groups, PB clustered closest to control, followed by Tac, ANIT and CCl4, which clustered together, and Wy, which clustered furthest from the control. Within the slice groups, Tac clustered closest to control, CCl4 and ANIT clustered together, while Wy and PB clustered furthest from controls. Based on this clustering diagram, the slices appear to better represent the affects of treatment on gene expression observed in vivo.



View larger version (46K):
[in this window]
[in a new window]
 
FIG. 3. Experiment tree clustering using Pearson correlation. Treatment and control groups were clustered using the Pearson correlation of average expression data (A) or average ratios (sample group/control group) (B).

 
In Figure 3BGo, expression ratios (group average/control average) were clustered using the Pearson correlation. Clustering of the normalized data caused the correlations to be based on compound induced changes in gene expression. Since the sample groups are normalized to controls, the controls cluster together. With the exception of the Wy treated in vivo group, which clustered away from the other treatment groups, the in vivo samples clustered with hepatocytes and separate from slices. Within the all three groups Tac and ANIT cluster closest together. In hepatocytes and in vivo, CCl4 clustered next to the ANIT/Tac branch. In slices and in vivo, the Wy group clustered furthest from the other treatments. With normalized data, the hepatocytes appear to better represent in vivo treatment related gene expression changes.

Genes with expression changes >=2-fold with treatment in all three test systems are shown in Table 4Go. The Wy compound altered the greatest number of genes across the three systems with 159 genes, consistent with its mechanism as a PPAR{alpha} nuclear hormone receptor agonist. ANIT altered 53 genes while the well-established liver enzyme inducer, PB, altered 40 genes. Tac and CCl4 each altered the expression of 38 and 21 genes, respectively. The vast majority of gene changes showed an increase (I) as opposed to a decrease (D) in expression relative to untreated controls. Specifically, 89% of ANIT, 62% of CCl4, 97% of PB, 84% of Tac, and 93% of Wy treatment-related fold changes represented increases rather than decreases in expression. The functions of the genes affected were predominantly organized into growth and differentiation (e.g., growth factors), immune response (e.g., acute phase, cytokines, and immunoglobulins), and signal transduction (e.g., receptors, kinases, and phosphatases), regardless of the treatment group. Genes involved in xenobiotic metabolism were also induced by several treatments. These included cytochrome p450s CYP 1A1 by ANIT and Tac, and CYP 2B family enzymes by PB. PB also induced UDP-glucuronosyl transferase 2B and aldehyde dehydrogenase.


View this table:
[in this window]
[in a new window]
 
TABLE 4 Gene Expression Affected by >=2-Fold in All Three Test Systems
 

    DISCUSSION
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
The use of expression profiling to compare isolated hepatocytes and precision cut liver slices to in vivo did not reveal a dramatic difference between the two in vitro systems. In untreated hepatocytes and slices, both in vitro systems were similar in gene expression profile to that observed in whole liver. Both in vitro systems expressed approximately 80% of the genes expressed in vivo under similar treatment conditions (Table 2Go). Hepatocytes and slices both exhibit expression profiles that correlate to a similar degree with those in vivo. The higher correlation between the two in vitro systems, compared to the correlation between either in vitro system to in vivo (Fig. 2Go), indicate that similar changes in expression occur with both culturing systems. These changes possibly reflect de-differentiation implicated by the fact that both in vitro systems lose expression of 32% of liver-specific genes over the course of isolation and in vitro incubation (Fig. 1Go).

Of the small set of eleven liver-specific genes found in only one in vitro system (Fig. 1Go), a subset of genes showed dramatic differences in expression levels. CYP 1A1 and 1A2 were among those genes expressed in hepatocytes but not slices (data not shown). The expression of these genes in hepatocytes was one-half the expression level of in vivo and > 10 times that of slices. The expression of CYP 3A1, steroid 5 {alpha}-reductase, and insulin-like growth factor binding protein in hepatocytes was greater than twice that of slices but about one-fifth that of in vivo. The expression differences of the remaining seven genes were minimal. Among those genes expressed in slices but not hepatocytes were k-kininogen, N-hydroxy–2-acetylaminofluorene sulfotransferase, alcohol sulfotransferase, apolipoprotein B, and hydroxysteroid sulfotransferase. These genes were expressed at levels in slices that ranged from three to twenty times that of hepatocytes and from 40% to three times that of in vivo. Four more genes, granulocyte colony stimulating factor 3 (GCSF3), insulin-like growth factor II, 20-{alpha}-hydroxysteroid dehydrogenase, and CYP 2C40, were expressed about twice as high in slices as in hepatocytes. The remaining two genes displayed only minor differences in expression levels. Many of these gene products take part in metabolism of xenobiotics or constituents of the culture medium. These metabolic differences could affect the responses of the in vitro systems to chemical treatments. At least two genes found in slices, GCSF3 and k-kininogen, are involved in inflammatory response and could contribute to the mechanism of some xenobiotic induced toxicities.

The degree to which the in vitro systems mimicked the xenobiotic-induced in vivo changes in gene expression could influence their utility as predictive in vitro tools for measuring changes in gene expression. When the All Genes List was used for this comparison (Table 3Go), less than 30% of the genes whose expression changes by 2-fold or more in vivo, were also altered by 2-fold in the in vitro systems. Subsets of genes were viewed in an attempt to find a more predictive set of genes. The Toxicology Gene List consisted of 1031 probes sets that make up the Affymetrix rat toxicology array. Nearly the same fraction of genes (<35%) was up or down regulated in vitro compared to the corresponding treatments in vivo. The Predictive Toxicology Genes List consisted of six rat orthologs of mouse genes that have been proposed as predictive of hepatotoxicity in vivo (Thomas et al., 2001Go). Between two and six of these genes were altered at least 2-fold by the various treatments in vivo, but only one of the genes was altered under a single treatment condition in hepatocytes. Two of the treatment conditions caused no changes in the predictive toxicology genes and under no conditions were more than two of these genes altered in slices. Although a small percentage of the genes affected by the toxicants were altered in vitro, there were still between 97 and 383 genes that were changed at least 2-fold in either in vitro system. It is possible that, with a more extensive database of compound treatments, these genes could be measured in vitro to predict in vivo toxicity. The experiment tree clustering using raw expression data did little to differentiate the in vitro systems as they clustered equidistant from in vivo. The slices better reflected the compound related branching observed with in vivo. However, when normalized to control groups, the hepatocytes clustered closer to in vivo, indicating that they better reflect compound-induced changes in gene expression.

A priori one might have expected that slices would better reflect in vivo gene expression since they maintain much of the tissue architecture and are comprised of the several diverse cell types that make up the liver. The fact that hepatocytes reflect in vivo gene expression at least as well as slices, if not better, could be due in part to the fact that hepatocytes comprise approximately 78% of the liver parenchymal volume (Blouin et al., 1977Go). It is therefore likely that the majority of mRNA from all three systems originates from hepatocytes.

There are many possible reasons for the diminished expression response of in vitro systems. Among the 10–20% of genes whose expression was lacking in vitro were receptors, kinases, and transcription factors, (data not shown) all of which could be involved in the compound induced changes in gene expression found in vivo. The full battery of in vivo expression changes may require interactions with extrahepatic cell types. In particular, interactions with immune cells in response to injury may contribute to the expression changes seen in vivo. There is also difficulty in normalizing exposure to the test compounds across the different test systems. In the in vivo system compound levels would be expected to rise sharply and fall under first order kinetics. In both in vitro systems, the concentration of compound in the media was held constant for the treatment duration, depending on the degree of in vitro metabolism. Further, while the monolayer of hepatocytes was likely well perfused, the exposure to cells within the slices may have varied from the interior to the exterior of the slice (Worboys et al., 1996Go).

Genes that were altered >=2-fold in all three systems were shown (Table 4Go) as possible markers of specific toxicant exposure or as potential direction for future mechanistic studies. A 2-fold selection criterion was used since it was at the limit of quantitative discrimination for popular methods of expression measurement (e.g., RT-PCR) and was considered biologically relevant given the small number of replicates. Several of the genes listed in Table 4Go were previously reported to be induced by the compounds used in the current study. PB is a know inducer of cytochrome P450s (CYP) of the 2B family (Nims et al., 1993Go), UDP-glucuronosyl transferase 2B (Wishart, 1978Go), and aldehyde dehydrogenase (Deitrich, 1971Go), while Tac has been shown to induce CYP 1A1 (Sinz and Woolf, 1997Go). Wy has been associated with DNA replication and fatty acid metabolism, pathways that are both represented by the array results (Glauert et al., 1984Go; Lalwani et al., 1981Go). Interestingly, CYP4A1, induced in vivo by Wy, was not among those induced >=2 fold in all three test systems. Closer inspection of the data reveals that while CYP4A1 was up regulated 9.3-fold in vivo and 2.4-fold in slices, it was virtually unaffected (1.3-fold) in hepatocytes (data not shown). This result could be explained by reduced responsiveness of gene induction in the presence of insulin in the culture medium (Woodcroft and Novak, 1999Go), with hepatocytes potentially having greater exposure than slices. Many genes listed in Table 4Go are not commonly associated with their respective treatment. These genes could represent novel markers related to the specific mechanism of compound induced toxicities. The large number of expression changes found in all three systems with ANIT treatment was surprising since the compound induces cholistasis and is thought to target the biliary epithelium with indirect effects to the hepatocyte. The lack of induction of CCl4-related genes, like TGFß 1 (Jeon et al., 1997Go), may be due to the minimally toxic doses that were chosen for the three test systems.

While expression profiling was unable to confirm the superiority of either in vitro system with respect to mimicking the in vivo response, they both yielded robust expression profiles. Greater than 80% of the in vivo expressed genes were found in both in vitro systems under several different treatment conditions and more than 60% of the liver specific genes were present in both in vitro systems under untreated conditions. Many of the previously reported compound-induced changes in expression were also seen in both in vitro systems. The choice of which in vitro system to use ultimately depends on the experimental objective and hypothesis being tested. Based on gene expression profiling either system appears suitable as long as appropriate controls are included. While slices may be technically less challenging, hepatocytes are more amenable to high-throughput assays.

One question that remains to be answered is whether gene expression measured in either in vitro system can be used to predict hepatic, or in some cases, extra-hepatic toxicity. Mechanisms in which gene expression changes are directly related to the specific xenobiotic-induced effects, such as the case with Wy and PB will most likely translate well from in vitro to in vivo settings. Gene expression changes that are not clearly related to toxicity but occur consistently may also be useful as predictive markers. However, in cases where in vivo compound-induced gene expression changes are predominantly driven by repair mechanisms or interactions with extrahepatic tissues, in vitro systems may not be predictive.


    NOTES
 
1 Both authors contributed equally to this work. Back

2 To whom correspondence should be addressed. Fax: (858) 678-8276. E-mail: bart.jessen{at}pfizer.com. Back


    REFERENCES
 TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Achary, M. P., Jaggernauth, W., Gross, E., Alfieri, A., Klinger, H. P., and Vikram, B. (2000). Cell lines from the same cervical carcinoma but with different radiosensitivities exhibit different cDNA microarray patterns of gene expression. Cytogenet. Cell Genet. 91, 39–43.[CrossRef][ISI][Medline]

Bassir, O., and Emafo, P. O. (1970). Oxidative metabolism of aflatoxin B 1 by mammalian liver slices and microsomes. Biochem. Pharmacol. 19, 1681–1687.[CrossRef][ISI][Medline]

Blouin, A., Bolender, R. P., and Weibel, E. R. (1977). Distribution of organelles and membranes between hepatocytes and nonhepatocytes in the rat liver parenchyma. A stereological study. J. Cell. Biol. 72, 441–55.[Abstract]

Bojes, H. K., and Thurman, R. G. (1994). Peroxisomal proliferators inhibit acyl CoA synthetase and stimulate protein kinase C in vivo. Toxicol. Appl. Pharmacol. 126, 233–239.[CrossRef][ISI][Medline]

Deitrich, R. A. (1971). Genetic aspects of increase in rat liver aldehyde dehydrogenase induced by phenobarbital. Science 173, 334–336.[ISI][Medline]

Dogterom, P., Nagelkerke, J. F., and Mulder, G. J. (1988). Hepatotoxicity of tetrahydroaminoacridine in isolated rat hepatocytes: effect of glutathione and vitamin E. Biochem. Pharmacol. 37, 2311–2113.[CrossRef][ISI][Medline]

Fisher, R., Hanzlik, R. P., Gandolf, A. J., and Brundel, K. (1991). Toxicity of ortho-substituted bromobenzenes in rat liver slices: A comparison to isolated hepatocytes and the whole animal. In Vitro Toxicol. 4, 173–186.

Fukumoto, Y., Okita, K., Kodama, T., Noda, K., Harada, T., Mizuta, M., and Takemoto, T. (1980). Studies of {alpha}-napthylisothiocyanate-induced hepatic disturbance. Hepatogastroenterology 27, 457–464.[ISI][Medline]

Gerhold, D., Lu, M., Xu, J., Austin, C., Caskey, C. T., and Rushmore, T. (2001). Monitoring expression of genes involved in drug metabolism and toxicology using DNA microarrays. Physiol. Genom. 5, 161–70.[Abstract/Free Full Text]

Ghantous, H. N., Fernando, J., Gandolfi, A. J., and Brendel, K. (1990). Toxicity of halothane in guinea pig liver slices. Toxicology 62, 59–69.[CrossRef][ISI][Medline]

Glauert, H. P., Reddy, J. K., Kennan, W. S., Sattler, G. L., Rao, V. S., and Pitot, H. C. (1984). Effect of hypolipidemic peroxisome proliferators on unscheduled DNA synthesis in cultured hepatocytes and on mutagenesis in Salmonella. Cancer Lett. 24, 147–156.[CrossRef][ISI][Medline]

Hales, B. J., and Neims, A. H. (1977). Induction of rat hepatic glutathione S-transferase B by phenobarbital and 3-methylcholanthrene. Biochem. Pharmacol. 26, 555–556.[CrossRef][ISI][Medline]

Hassett, C., Luchtel, D. L., and Omiecinski, C. J. (1989). Hepatic expression of rat P450 mRNA assessed by in situ hybridization to oligomer probes. DNA 8, 29–37.[ISI][Medline]

Henderson, P. T., and Dewaide, J. H. (1969). Metabolism of drugs in isolated rat hepatocytes. Biochem. Pharmacol. 18, 2087–2094.[CrossRef][ISI][Medline]

Ho, L., Guo, Y., Spielman, L., Petrescu, O., Haroutunian, V., Purohit, D., Czernik, A., Yemul, S., Aisen, P. S., Mohs, R., et al. (2001). Altered expression of a-type but not b-type synapsin isoform in the brain of patients at high risk for Alzheimer’s disease assessed by DNA microarray technique. Neurosci. Lett. 298, 191–194.[CrossRef][ISI][Medline]

Jeon, Y. J., Han, S. H., Yang, K. H., and Kaminski, N. E. (1997). Induction of liver-associated transforming growth factor beta 1 (TGF-beta 1) mRNA expression by carbon tetrachloride leads to the inhibition of T helper 2 cell-associated lymphokines. Toxicol. Appl. Pharmacol. 144, 27–35.[CrossRef][ISI][Medline]

Lalwani, N. D., Reddy, M. K., Qureshi, S. A., and Reddy, J. K. (1981). Development of hepatocellular carcinomas and increased peroxisomal fatty acid beta-oxidation in rats fed [4-chloro-6-(2,3-xylidino)-2-pyrimidinylthio] acetic acid (Wy-14,643) in the semipurified diet. Carcinogenesis 2, 645–650.[ISI][Medline]

Kawamoto, T., Sueyoshi, T., Zelko, I., Moore, R., Washburn, K., and Negishi, M. (1999). Phenobarbital-responsive nuclear translocation of the receptor CAR in induction of the CYP2B gene. Mol. Cell Biol. 19, 6318–6322.[Abstract/Free Full Text]

McQueen, C. A. (1989). Hepatocytes in monolayer culture: An in vitro model for toxicity studies. In In Vitro Models in Toxicology: Model Systems and Methods, pp. 131–155. Telford Press, Cranford, NJ.

National Research Council (NRC) (1996). Guide for the Care and Use of Laboratory Animals (N. Grossblatt, Ed.). National Academy Press, Washington, DC.

Nims, R. W., Syi, J. L., Wink, D. A., Nelson, V. C., Thomas, P. E., Jones, C. R., Diwan, B. A., Keefer, L. K., Rice, J. M., and Lubet, R. A. (1993). Hepatic cytochrome P450 2B-type induction by ethyl/phenyl-substituted congeners of phenobarbital in the rat. Chem. Res. Toxicol. 6, 180–187.[ISI][Medline]

Orton, T. C., and Parker, G. L. (1982). The effect of hypolipidemic agents on the hepatic microsomal drug-metabolizing enzyme system of the rat. Induction of cytochrome(s) P-450 with specificity toward terminal hydroxylation of lauric acid. Drug Metab. Dispos. 10, 110–115.[Abstract]

Reddy, J. K., Rao, M. S., Azarnoff, D. L., and Sell, S. (1979). Mitogenic and carcinogenic effects of a hypolipidemic peroxisome proliferator, [4-chloro-6-(2,3-xylidino)-2-pyrimidinylthio]acetic acid (Wy-14, 643), in rat and mouse liver. Cancer Res. 39, 152–161.[ISI][Medline]

Robertson, D. G., Braden, T. K., Urda, E. R., Lalwani, N. D., and de la Iglesia, F. A. (1998). Elucidation of mitochondrial effects by tetrahydroaminoacridine (tacrine) in rat, dog, monkey and human hepatic parenchymal cells. Arch. Toxicol. 72, 362–371.[CrossRef][ISI][Medline]

Schiaffonati, L., and Tiberio, L. (1997). Gene expression in liver after toxic injury: Analysis of heat shock response and oxidative stress-inducible genes. Liver 17(4), 183–191.[ISI][Medline]

Sinz, M. W., and Woolf, T. F. (1997). Characterization of the induction of rat microsomal cytochrome P450 by tacrine. Biochem. Pharmacol. 54, 425–427.[CrossRef][ISI][Medline]

Smith, P. F., Gandolfi, A. J., Krumdieck, C. L., Putnam, C. W., Zukoski, C. F., III, Davis, W. M., and Brendel, K. (1985). Dynamic organ culture of precision liver slices for in vitro toxicology. Life Sci. 36, 1367–1375.[CrossRef][ISI][Medline]

Stachlewitz, R. F., Arteel, G. E., Raleigh, J. A., Connor, H. D., Mason, R. P., and Thurman, R. G. (1997). Development and characterization of a new model of tacrine-induced hepatotoxicity. J. Pharmacol. Exp. Toxicol. 282, 1591–1599.

Summers, W. K., Tachiki, K. H., and Kling, A. (1989). Tacrine in the treatment of Alzheimer’s disease. A clinical update and recent pharmacologic studies. Eur. Neurol. 29(Suppl. 3), 28–32.[ISI][Medline]

Thomas, R. S., Rank, D. R., Penn, S. G., Zastrow, G. M., Hayes, K. R., Pande, K., Glover, E., Silander, T., Craven, M. W., Reddy, J. K., et al. (2001). Identification of toxicologically predictive gene sets using cDNA microarrays. Mol. Pharmacol. 60, 1189–1194.[Abstract/Free Full Text]

Thor, J., Moldeus, P., Hermanson, R., Hogberg, J., Reed, D. J., and Orrenius, S. (1978). Metabolic activation and hepatotoxicity. Toxicity of bromobenzene in hepatocytes isolated from phenobarbital-and diethylmaleate-treated rats. Archiv. Biochem. Biophys. 188, 122–129.[CrossRef][ISI][Medline]

Uno, Y., Matsuura, K., Miyagawa, M., Takasawa, H., Tanifuji, H., Abe, K., Akimoto, A., Asanoma, K., Baba, K., Daigo, H., et al. (1999). Rat liver in vivo replicative DNA synthesis test for short-term prediction of nongenotoxic (Ames-negative) hepatocarcinogenicity: A collaborative study of the Nongenotoxic Carcinogen Study Group of Japan. Toxicol. Lett. 109, 105–114.[CrossRef][Medline]

Wagle, S. R. (1975). Interrelationship of insulin and glucagon ratios on carbohydrate metabolism in isolated hepatocytes containing high glycogen. Biochem. Biophys. Res. Commun. 67, 1019–1027.[CrossRef][ISI][Medline]

Watkins, P. B., Zimmerman, H. J., Knapp, M. J., Gracon, S. I., and Lewis, K. W. (1994). Hepatotoxic effects of tacrine administration in patients with Alzheimer’s disease. JAMA 271, 992–998.[Abstract]

Waxman, D. J., and Azaroff, L. (1992). Phenobarbital induction of cytochrome P450 gene expression. Biochem. J. 281, 577–592.[ISI][Medline]

Weinhold, P. A. (1969). Phospholipid metabolism in the developing liver. I. The biosynthesis of choline glycerophosphatides by liver slices from fetal, newborn and adult rats. Biochim. Biophys. Acta 187, 85–93.[ISI][Medline]

Wishart, G. J. (1978). Demonstration of functional heterogeneity of hepatic uridine diphosphate glucuronosyltransferase activities after administration of 3-methylcholanthrene and phenobarbital to rats. Biochem. J. 174, 671–672.[ISI][Medline]

Woodcroft, K. J., and Novak, R. F. (1999). Insulin differentially affects xenobiotic-enhanced, cytochrome P-450 (CYP)2E1, CYP2B, CYP3A, and CYP4A expression in primary cultured rat hepatocytes. J. Pharmacol. Exp. Ther. 289, 1121–1127.[Abstract/Free Full Text]

Worboys, P. D., Bradbury, A., and Houston, J. B. (1996). Kinetics of drug metabolism in rat liver slices. II. Comparison of clearance by liver slices and freshly isolated hepatocytes. Drug Metab. Dispos. 24, 676–681.[Abstract]

Wrighton, S. A., Ring, B. J., and VandenBranden, M. (1995). The use of in vitro metabolism techniques in the planning and interpretation of drug safety studies. Toxicol. Pathol. 23, 199–208.[ISI][Medline]