Gene Expression Analysis Offers Unique Advantages to Histopathology in Liver Biopsy Evaluations

Abstract

Liver diseases that induce nonuniform lesions often give rise to greatly varying histopathology results in needle biopsy samples from the same patient. This study examines whether gene expression analysis of such biopsies could provide a more representative picture of the overall condition of the liver. We utilized acetaminophen (APAP) as a model hepatotoxicant that gives a multifocal pattern of necrosis following toxic doses. Rats were treated with a single toxic or subtoxic dose of APAP and sacrificed 6, 24, or 48 hours after exposure. Left liver lobes were harvested, and both gene expression and histopathological analysis were performed on biopsy-sized samples. While histopathological evaluation of such small samples revealed significant sample to sample differences after toxic doses of APAP, gene expression analysis provided a very homogeneous picture and allowed clear distinction between subtoxic and toxic doses. The main biological function differentiating animals that received sub-toxic from those that had received toxic doses was an acute stress response at 6 hours and signs of energy depletion at later time points. Our results suggest that the use of genomic analysis of biopsy samples together with histopathological analysis could provide a more precise representation of the overall condition of a patient’s liver than histopathological evaluation alone.

Introduction

The diagnostic procedure for most severe liver diseases, if not prohibited by impaired hemostasis due to advanced liver damage, includes histopathological analysis of small biopsy material. Diagnostic uncertainty and sampling variability has been reported in multiple studies examining liver biopsies in patient populations affected by liver diseases that induce an uneven phenotype across the organ (Dutt et al., 1983; Maharaj et al., 1986; Bedossa et al., 2003; Siddique et al., 2003; Ratziu et al., 2005). We were interested in exploring whether genomic analysis of liver biopsy samples might serve as a better predictor for toxic exposure than histopathological analysis of equivalent samples. We chose acetaminophen (APAP) as a model hepatotoxicant due to its well described patchy pattern of necrosis observed in response to toxic doses in rats (McLean and Day, 1975). To compare genomic with histopathological analysis, we exposed animals to a range of acute oral doses, from subtoxic to toxic, and sampled their livers at terminal necropsies after 6, 24, and 48 hours—bridging times before liver damage is manifest (as seen histologically and by measuring serum enzyme elevations), over peak manifestation to early recovery.

Comparison of both analysis approaches done on randomly selected liver samples, equivalent in size to samples that would be obtained by needle biopsy, revealed significant advantages of genomic analysis compared to histopathological evaluation. Gene expression patterns created a more homogeneous picture that was less influenced by random sampling differences. Based on alterations of the trascriptome, we were able to classify animals as having received a toxic dose at early time points at which histopathological alterations were not yet visible.

Materials and Methods

Animals and Animal Care

Male F344/N rats, 8–12 weeks old, were obtained from Taconic Laboratories, Inc., Germantown, NY, acclimated for 14 days, and observed for absence of disease. Rats were housed in groups of 3 in polycarbonate cages (Lab Products, Inc., Maywood, NJ) with Sani-Chips (P.J. Murphy Forest Products Corp., Montville, NJ) bedding. The animal rooms were maintained at 21–22°C and 48–53% relative humidity with a 12-hour dark-light cycle with lights coming on at 6 AM and room air changes of 10/hour. Cages were changed twice per week. NIH-07 diet and tap water was provided ad libitum.

Study Design

Three male rats each received a single dose of APAP in 0.5% ethyl cellulose by gavage at doses of 0 (vehicle only), 50, 150, 1500, or 2000 mg/kg. Body weights were determined prior to dosing and dose was calculated based on individual body weights. Animals were not fasted before treatment. The rats for the 6-hour time point were dosed between 5:30 AM and 7:30 AM and the 24- and 48-hour time point animals were dosed between 9:30 and 10:30 AM. All animals were observed twice daily for signs of toxicity. Experiments were performed according to the guidelines established in the NIH Guide for the Care and Use of Laboratory Animals (Council, 1996) and an approved Animal Study Protocol was on file prior to initiation of the study.

Chemical

APAP (APAP, 99% pure) was obtained from Sigma Chemical Company (St. Louis, MO). Dosing solutions were prepared daily and protected from light. The dose formulations were prepared by mixing APAP with 0.5% aqueous ethyl cellulose (USP/FCC grade; Fisher Scientific Company, St. Louis, MO) suspension to give the required concentrations.

Necropsy Procedures

Animals were euthanized with carbon dioxide (CO₂/O₂ mixture) from a regulated source 6, 24, or 48 hours following dosing. Each treatment group had a concurrent vehicle-treated control group. A mid-sagittal section was taken from the left lobe of the liver. The remainder of the left lobe was quickly cubed and frozen in liquid nitrogen for differential gene expression analysis. The time from drawing of the blood sample to freezing of the liver was generally less than 90 seconds. Tissues were stored at −80°C until processed for RNA extraction.

Histopathology

The mid-sagittal sections of the left liver lobe taken at necropsy were fixed in 10% neutral-buffered formalin for 24 hours, dehydrated in 70% ethanol for 48 hours and embedded in paraffin. Sections (6 μm) were stained with hematoxylin and eosin (H&E) and evaluated by a study pathologist. A second pathologist reviewed the diagnosis. Slides with differences were resolved by a Pathology Working Group review (Boorman and Eustis, 1986).

Point Count for Degree of Necrosis

Frozen liver cubes (average size: 0.5 cm × 0.5 cm × 0.5 cm) were placed in 10% neutral-buffered formalin. Multiple H&E stained sections were made and coded. The liver sections were photographed and printed in color on 8″ × 11″ paper. A grid with points 25 mm apart was placed on each photograph and the points falling on the liver section and those falling on areas of necrosis were counted. A dimensionless location on the points was selected to avoid bias (Mouton, 2002). The number of points falling on areas of necrosis versus the number of points falling on the tissues was used to determine the percent necrosis.

The estimated tissue area examined was determined using the formula A_obj = ∑P_obj X a(p) where A_obj is the reference area, ∑P _obj is the sum of points (P) hitting necrotic regions and a(p) is the area per point (Mouton, 2002). The resulting value was divided by the square of the magnification. The point count was done twice on each micrograph. The histological slide was available and examined at higher magnification to confirm the presence or absence of necrosis.

RNA Isolation

Total hepatic RNA was isolated from individual rat livers using QIAGEN RNeasy Maxi Kits (QIAGEN, Valencia, CA) (〈http://dir.niehs.nih.gov/microarray/rnaprep.htm〉). Equal amounts of RNA from each of the control animals at every dose and time period were pooled for control gene expression and compared with individual rats at each dose and time period. The samples were hybridized in duplicate for each individual rat.

cRNA Labeling

Amplification and labeling of cRNA was done according to the instructions of the manufacturer (Agilent Technologies, Palo Alto, California). Briefly, one μg of RNA per sample was converted to cDNA with reverse transcriptase and then amplified using T7 RNA polymerase while labeling with either Cy3-dUTP or Cy5-dUTP. The labeled cRNAs from treated individuals and control pools were hybridized together twice, with Cy3 and Cy5 dye swap.

Microarray Analysis

The labeled cRNA samples from individual treated animals and control pools were mixed and hybridized for 16 hours at 60°C to the oligonucleotide probes on the Agilent Rat Oligo Microarray (Agilent #G4130A) that contains approximately 22,000 features coding for known genes and ESTs. Chips were scanned with an Agilent G2565 AA Scanner and processed with the Agilent G2566AA Feature Extraction Software. Detailed protocols are available at 〈http://dir.niehs.nih.gov/microarray/methods.htm〉.

The complete data set is deposited at GEO (〈http://www.ncbi.nlm.nih.gov/geo/〉) and CEBS (〈http://cebs.niehs.nih.gov〉, investigation with accession number 001-00002-0010-000-4)

Identification of Differentially Expressed Genes (DEGs)

Seventy-two gene expression data files were loaded into Rosetta Resolver database (build 5.1.0.1.23, Rosetta Inpharmatics, Agilent Technologies, Palo Alto, California). There they were merged according to fluor-flip hybridization pairs to generate weighted-averaged ratio values (computed from the normalized and background subtracted pixel intensity values). DEGs were extracted from each merged hybridization pair according to Rosetta Resolver’s error model with a p-value equal or less than 0.001 and an absolute fold change of equal or more than 2-fold. Features that had an intensity of less than 300 in either channel were excluded from the analysis. The lists of DEGs originating from the single animals were analyzed for their consistency between animals in the single treatment/time point groups and only genes that were differentially expressed in three out of three animals were used for further analysis.

The total of 996 identified DEGs (Supplemental Table 1) and subsets of those genes were analyzed in Rosetta Resolver by principal component analysis (PCA) and hierarchical clustering. PCA is a data reduction method that allows capturing of degrees of variation within a data set (for detailed explanation of the mathematical modeling see p. 92, Causton et al., 2003).

Biological Function Analysis

Genes that contributed the most information to PC1 and PC2 were analyzed for their biological function in the Ingenuity Pathway Analysis tool (Ingenuity Systems, Inc., Redwood City, California).

Results

We undertook this study to test the hypothesis that genomic analysis of a liver sample equivalent in size to a biopsy sample is more informative than histopathological evaluation. This study was designed to address the great regional differences in the degree of necrosis after exposure to a toxic dose of APAP and its potential negative impact on accurate diagnosis of liver biopsies.

To perform these analyses, we exposed animals to either a subtoxic (0, 50 or 150 mg/kg) or toxic (1500 or 2000 mg/kg) dose of APAP and collected liver tissue 6, 24, and 48 hours after dosing. We randomly selected samples for both histopathological analysis and gene expression analysis, and compared the performance of both methods.

Clinical Chemistry Results

In order to create a measurement for the response of the study animals to treatment, independent of either histopathology or gene expression, we obtained blood samples from every animal in the study and determined clinical chemistry parameters (blood urea nitrogen, creatinine, total protein, albumin, and total bile acid concentrations, and activities of alanine aminotransferase, alkaline phosphatase, creatine kinase, and sorbitol dehydrogenase). In general, no clinical chemistry parameters were significantly altered in groups that received 50 or 150 mg/kg APAP. At 24 and 48 hours post dosing, acetaminophen induced marked increases in alanine aminotransferase, aspartate aminotransferase, sorbitol dehydrogenase and 5′-nucleotidase activities and total bile acid concentrations in the 1500 and 2000 mg/kg group (Supplemental Table 2). In general, the increases were more prominent at 24 hours and appeared to be ameliorating by 48 hours postdosing. This would be consistent with the liver necrosis observed morphologically at 24 hours and the suggestion of regenerative/reparative changes observed at 48 hours postdosing. The increases in serum biomarkers of liver injury seen at the high dose were expected and have been reported previously (Echard et al., 2001; Ruepp et al., 2002; Schiodt et al., 2002; Heinloth et al., 2004).

Histopathology Alterations

Livers from animals receiving 50 or 150 mg/kg appeared normal and were not distinguishable from control animals. Livers from animals that were treated with 1500 mg/kg exhibited mild to moderate centrilobular hepatocellular necrosis and inflammatory lesions that were most prominent at 24 hours and appeared to be ameliorating at 48 hours following APAP administration. Livers from animals after exposure to 2000 mg/kg displayed no histopathological alterations at 6 hours, after 24 hours they showed minimal to moderate centrilobular hepatocellular necrosis and inflammatory lesions which were still persistent at 48 hours, but at this point they also had mild to moderate signs of liver cell regeneration. Overall, the animals exposed to toxic doses of APAP showed great regional differences in the degree of necrosis displayed (Figure 1) which has been described before (McLean and Day, 1975).

Histopathological Evalutions of Biopsy-Sized Samples Reveals Great Diagnostic Uncertainty

To test in the rat model the diagnostic accuracy of liver biopsy samples, we chose 3 randomly selected liver cubes from each of the animals that had received 1500 mg/kg APAP and that were sacrificed after 24 or 48 hours, one randomly selected cube from each of the 6-hour animals in this dose group and each control animal (as there was no necrosis detectable in liver sections). From those cubes, H&E slides were prepared and necrotic areas were quantified as described in materials and methods. No necrosis was detected in control animals or animals having received 1500 mg/kg APAP at 6 hours. After 1500 mg/kg at 24 and 48 hours, the amount of necrotic area differed widely between the three cube samples from individual animals (Table 1).

Gene Expression Analysis Allows Clear Distinction between Toxic and Subtoxic Exposure to Acetaminophen

To test the potential diagnostic accuracy of gene expression analysis performed on liver biopsy samples, we performed microarray analysis on all animals in the study with RNA prepared from randomly selected liver cubes. PCA of all differentially expressed genes showed clear separation of animals that received a toxic dose of APAP (1500 and 2000 mg/kg) from those that received subtoxic treatment (50 and 150 mg/kg) (Figure 2). At 6 hours, this separation was along Principal Component 2 (PC2), while at 24 and 48 hours Principal Component 1 (PC1) was separating subtoxic from toxic samples.

The top 100 genes driving separation along PC 1 in positive direction were strongly down-regulated in response to toxic doses of APAP after 24 and 48 hours, less so after 6 hours, while being predominantly unchanged after subtoxic exposure (Figure 3A). This gene set contained several phase 1 biotransforming enzymes (cytochrome p450s, Supplemental Table 3) and members of the fatty acid biosynthesis and metabolism pathway. The top 100 negative separators along PC1 showed significant over-representation of genes involved in cell cycle regulation and hypoxia signaling. Those genes were strongly up-regulated 24 and 48 hours after exposure to toxic doses of APAP, while being predominantly unchanged at earlier times and lower doses (Figure 3B).

Analysis of the top 100 genes driving PC2 in positive direction revealed that all the biological pathways significantly overexpressed in this group were involved in amino acid metabolism (e.g., phenylalanine, tyrosine and tryptophan metabolism, cysteine metabolism, arginine and proline metabolism, etc., complete listing in Supplemental Table 4). These genes showed strong induction at 6 hours after toxic exposure to APAP but were less induced or down-regulated at later time points and subtoxic doses (Figure 4A). The top 100 genes driving PC2 in the negative direction showed significant overrepresentation of genes involved in sterol biosynthesis, down-regulated at 6 hours after 1500 and 2000 mg/kg APAP, and apoptosis, up-regulated after 48 hours in animals treated with 1500 and 2000 mg/kg APAP (Figure 4B).

Top 20 Genes From PC1 and PC2 Allow Clear Separation of Subtoxic And Toxic Doses of Acetaminophen

For clinical use, a PCR-based analysis procedure is preferable to a microarray-based approach. Therefore, we reduced the list of discriminating genes based on PC1 and PC2 to the top 10 from each of those lists. These 20 genes still resulted in good separation of the animals that had received toxic treatment (1500 or 2000 mg/kg) from those that had received subtoxic treatment (50 or 150 mg/kg) along PC1 (Figure 5), although the amount of variation in the data set captured in PC1 went down to 13% from being 40% with the larger set of genes (Figure 2).

Discussion

In this study we demonstrate the high variability of necrosis after APAP intoxication diagnosed by histopathology when based on biopsy-sized samples while equivalent samples produced a much more uniform picture when analyzed for gene expression alterations. Our histopathological results are in agreement with observations by McLean et al. (McLean and Day, 1975) describing patchy appearance of liver necrosis after APAP intoxication. To our knowledge our study is the first to address the effect of patchy manifestation of liver necrosis on diagnoses based on liver biopsy-sized samples and compared those to genomic analyses. Our results are important in that they (a) address the limitations of the evaluation of a small sample in the histopathological diagnosis of the extent of hepatocellular necrosis after APAP intoxication and (b) explore genomic analysis of liver biopsy samples as an alternative diagnostic approach.

Genomic analysis of biopsy-sized samples allowed discrimination between animals that had received subtoxic (50 and 1500 mg/kg) or toxic (1500 and 2000 mg/kg) doses of APAP at any time point examined. This is especially exciting at the early 6-hour time point after dosing at which no histopathological alterations in any of the liver samples were detectable. Analysis of the genes driving differentiation between subtoxic and toxic doses at that time point revealed a major influence of exposure to toxic doses of APAP on amino acid metabolism. Genes belonging to those pathways were significantly enriched in the list of top discriminatory genes and were overexpressed in animals after exposure to toxic levels of APAP (Supplemental Figure 1). This observation can be interpreted as part of an acute stress response.

Activation of amino acid metabolism is a pre-requisite for both glutathione (GSH) (Wu et al., 2004) and nitric oxide (NO) synthesis (Morris, 2006), besides others. The importance of GSH in the detoxification of APAP has been well described (Jaeschke et al., 2003; James et al., 2003; Lauterburg, 2002). NO is a precursor of peroxynitrite, which is a strong oxidant and interacts with multiple macromolecules; thereby, altering cellular signaling pathways (Denicola and Radi, 2005). Peroxynitrite is one of the radicals detoxified by GSH (James et al., 2003) and its formation in response to toxic doses of APAP has been described (Jaeschke et al., 2003; James et al., 2003).

At the later time points in our study (24 and 48 hours postdosing) the most discriminating genes between animals that received toxic or subtoxic doses of APAP showed an overrepresentation of genes belonging to the class of phase 1 biotransforming enzymes and genes involved in fatty acid metabolism. Genes in those classes were down-regulated relative to control animals which might be an indication for a general down-regulation of energy demanding cellular processes. This observation is consistent with previous reports of ATP depletion and energy deficits after exposure to APAP (Heinloth et al., 2004; Jaeschke and Bajt, 2006; Masubuchi et al., 2005).

Interestingly, the analysis not only provided clear separation between subtoxic and toxic exposure, but also between the single time points after toxic doses of APAP. This separation was mainly along PC2, and one of the major biological processes discriminating between the 24- and 48-hour time point is the up-regulation of genes involved in apoptosis at 48 hours after dosing. There are controversial reports regarding apoptosis and necrosis as predominant forms of cell death after intoxication with APAP in the literature (Ray et al., 1996; Gujral et al., 2002; El-Hassan et al., 2003). Malhi et al. (2006) recently summarized the importance of ATP content in the cells that determines if cells die via apoptosis or necrosis after exposure to toxic levels of APAP. This data, combined with the histological evidence of necrosis, suggests both necrosis and an activation of apoptotic pathways on the transcriptional level after toxic doses of APAP.

One of the limitations of the presented approach to improve the diagnostic value of biopsy samples is the technical and financial effort connected with microarray analysis. Therefore, we evaluated if it is feasible to attain similar separation with a more limited gene set that would lend itself to PCR multiplexing as a technical alternative to microarray analysis allowing high throughput analysis in a clinical setting. Our data show this is possible in our model system. In summary, we were able to prove that gene expression analysis of biopsy-sized samples in rats exposed to APAP allowed higher diagnostic accuracy than histopathological evaluation alone of such samples.

We hypothesize this higher accuracy is caused by a generalized gene expression response within the liver to toxic doses of APAP independent of the microscopic phenotype observed in the analyzed sample. Therefore, we propose that gene expression analysis of biopsy samples would be a very valuable addition to traditional histopathological evaluation after acetaminophen intoxication. Future studies will be needed to examine if this finding can be extended to liver pathologies of other etiologies such as chronic hepatitis C and nonalcoholic fatty acid liver disease. Clinical trials will also be needed to determine the usefulness of this methodology in the diagnosis of liver disease in the human patient.