Intrahepatic gene expression profiles and alpha-smooth muscle actin patterns in hepatitis C virus induced fibrosis

Authors

  • Daryl T.-Y. Lau,

    Corresponding author
    1. Division of Gastroenterology and Hepatology, Department of Internal Medicine, The University of Texas Medical Branch, Galveston, TX
    2. Hepatitis Research Center, Institute for Human Infections & Immunity, The University of Texas Medical Branch, Galveston, TX
    • Division of Gastroenterology and Hepatology, 4.106 McCullough Building, University of Texas Medical Branch at Galveston, 301 University Boulevard, Galveston, TX 77555-0764
    Search for more papers by this author
    • fax: 409-772-1343.

  • Bruce A. Luxon,

    1. Department of Human Biological Chemistry & Genetics, The University of Texas Medical Branch, Galveston, TX
    Search for more papers by this author
  • Shu-Yuan Xiao,

    1. Hepatitis Research Center, Institute for Human Infections & Immunity, The University of Texas Medical Branch, Galveston, TX
    2. Department of Pathology, The University of Texas Medical Branch, Galveston, TX
    Search for more papers by this author
  • Michael R. Beard,

    1. Department of Microbiology & Immunology, The University of Texas Medical Branch, Galveston, TX
    2. Institute of Medical and Veterinary Science, Adelaide, Australia
    Search for more papers by this author
  • Stanley M. Lemon

    1. Hepatitis Research Center, Institute for Human Infections & Immunity, The University of Texas Medical Branch, Galveston, TX
    2. Department of Microbiology & Immunology, The University of Texas Medical Branch, Galveston, TX
    Search for more papers by this author

  • Potential conflict of interest: Nothing to report.

Abstract

To gain insight into pathogenic mechanisms underlying fibrosis in hepatitis C virus (HCV)-mediated liver injury, we compared intrahepatic gene expression profiles in HCV-infected patients at different stages of fibrosis and α-smooth muscle actin (α-SMA) staining patterns. We studied 21 liver biopsy specimens: 5 had no fibrosis (Ludwig-Batts stage 0); 10 had early portal or periportal fibrosis (stages 1 and 2); and 6, advanced fibrosis (stages 3 and 4). None of the patients had hepatocellular carcinoma. Transcriptional profiles were determined by high-density oligonucleotide microarrays. ANOVA identified 157 genes for which transcript abundance was associated with fibrosis stage. These defined three distinct hierarchical clusters of patients. Patients with predominantly stage 0 fibrosis had increased abundance of mRNAs linked to glycolipid metabolism. PDGF, a potent stellate cell mitogen, was also increased. Transcripts with increased abundance in stages 1 and 2 fibrosis were associated with oxidative stress, apoptosis, inflammation, proliferation, and matrix degradation, whereas transcripts increased in stages 3 and 4 were associated with fibrogenesis and cellular proliferation. Cells staining for α-SMA were detectable at all stages but infrequent in advanced fibrosis without active inflammation. A high frequency of such cells was associated with mRNAs linked to glycolipid metabolism. In conclusion, the presence of α-SMA–positive HSCs and expression of PDGF in stage 0 fibrosis suggests that stellate cells are activated early in HCV-mediated injury, possibly in response to oxidative stress resulting from inflammation and lipid metabolism. Increased abundance of transcripts linked to cellular proliferation in advanced fibrosis is consistent with a predisposition to cancer. Supplementary material for this article can be found on the HEPATOLOGY website (http://www.interscience.wiley.com/jpages/0270-9139/suppmat/index/html) (HEPATOLOGY 2005.)

Approximately 4 million people in the United States have serological evidence of infection with hepatitis C virus (HCV), of whom approximately 2.7 million have ongoing chronic infection.1 The risk of progression of liver disease in these patients is influenced by both viral and host factors,2–5 with histological features ranging from mild inflammation to severe fibrosis, as well as cancer.6, 7 Fibrosis develops as a multicellular process involving paracrine signaling between the resident liver cells and inflammatory cells.8 Central events include the activation of hepatic stellate cells (HSCs) in association with tissue necrosis and inflammation.9 In response to liver injury, human HSCs express alpha–smooth muscle actin (α-SMA), becoming “activated” and myofibroblast-like. Immunohistochemical staining for α-SMA correlates with HSC activation10 and is found in viral hepatitis as well as alcoholic liver disease.11 Conversely, a decrease in α-SMA staining correlates with treatment-related improvement in necroinflammatory lesions.12, 13 However, despite an increasing appreciation of the manner in which HSCs are regulated, the factors responsible for initiation of fibrogenesis and progression of fibrosis in chronic hepatitis C remain poorly understood.

We have carried out oligonucleotide microarray assays of percutaneous liver biopsy specimens in an effort to gain improved insights into the pathogenesis of hepatic fibrosis. The application of this technology to human liver disease has been limited, generally by the small quantity of tissue typically obtained from percutaneous biopsy, but we have adopted methods in which the sampling of cellular RNA from affected tissue is sufficient for both accuracy and consistency of microarray assays. We document differences existing in intrahepatic gene expression profiles in different fibrotic stages of chronic hepatitis C and examine the correlation of these differences with immunohistochemical staining for α-SMA as a measure of HSC activation.

Abbreviations

HCV, hepatitis C virus; HSC, hepatic stellate cell; α-SMA, alpha-smooth muscle actin; NAFLD, nonalcoholic liver disease; ROS, reactive oxygen species; IL-4, interleukin-4.

Patients and Methods

Patients and Tissue Samples.

Tissue samples were obtained by percutaneous liver biopsy from 18 patients with nonalcoholic fatty liver disease (NAFLD) and 26 patients with chronic hepatitis: 21 infected with HCV, 4 with hepatitis B virus, and 1 with hepatitis delta virus infection. The patients with hepatitis C had documented viremia before undergoing liver biopsy for evaluation for potential treatment. Only 4 of these 21 patients had received prior interferon therapy (6-24 months preceding biopsy). For additional information, see the Supplementary Materials and Methods (at http://www.interscience.wiley.com/jpages/0270-9139/suppmat/index.html).

Immunohistochemical Detection of Alpha-Smooth Muscle Actin.

We used the method described by Guido et al.10 for immunohistochemical staining of α-SMA in liver tissue. The presence of α-SMA reactive HSCs in liver biopsy samples was semiquantitatively scored by a modification of the method described by Schmitt-Graff et al.14: Score 0 = no staining or very rare cells staining postive, 1 = staining of HSCs occupying <30% of sinusoidal liver cell surface, 2 = staining of HSCs occupying 31% to 60% of sinusoidal liver cell surface, 3 = staining of HSCs occupying 61% to 90% of sinusoidal liver cell surface, and 4 = diffuse staining of more than 90% of sinusoidal liver cell surface. The periportal (zone 1), intermediate (zone 2), and perivenular (zone 3) regions of the liver lobule were scored separately. At least 5 liver acini were evaluated in each biopsy specimen, with the highest value observed for each zone taken as the final score. As with histological interpretation, scoring of the immunohistochemical studies was carried out by a blinded liver pathologist.

Microarray Analysis of Intrahepatic Transcriptional Profiles.

Complementary RNA probes were prepared from total RNA extracted from liver biopsy material and subsequently hybridized to HG-U95A Human GeneChips (Affymetrix Inc., Santa Clara, CA), each of which contains 12,625 probe sets (representing mostly full-length human genes). Detailed methods of probe preparation and analysis of the Affymetrix data may be found in the Supplementary Materials and Methods.

Statistical Methods.

Unless otherwise indicated, comparisons between two groups were carried out using 2-tailed Student t test, and between multiple groups by one-way ANOVA. Where indicated, the chi-square test was used. A P value of less than .05 was considered statistically significant unless otherwise noted.

Results

Validation of Microarray Analysis of Intrahepatic Global Transcriptional Profiles.

Chronic viral hepatitis and NAFLD are both important and common causes of chronic liver disease. Although both conditions can progress to advanced fibrosis, their causes and pathogenesis are different, and one may reasonably assume that these diseases are associated with different intrahepatic gene expression profiles. To determine whether the small amounts of tissue typically provided by percutaneous liver biopsy can be used to demonstrate such differences in oligonucleotide microarray assays, we studied excess biopsy material obtained from 18 patients with NAFLD, and 26 patients with chronic hepatitis: 21 with HCV, 4 with hepatitis B virus and 1 with hepatitis delta virus infection. Probes were prepared from these 44 liver biopsy specimens, as well as a pool of RNA extracted from 5 normal human livers, and hybridized to HG-U95A Affymetrix GeneChips, which contain 12,625 oligonucleotide probe sets representing mostly full-length human mRNAs. The data, subjected to scaling and filtering analyses as described in the Supplementary Materials and Methods, demonstrated a high level of internal consistency of probe set hybridization across individual chips. The small standard deviations of trimmed means for the signal intensity of probe set hybridization (496.6 ± 0.5) and the scale factors calculated for each individual chip (1.001 ± 0.001) reflected excellent uniformity of sample preparation and probe hybridization among these 44 samples and normal liver RNA.

Scaled hybridization signal intensities for those probe sets representing RNA transcripts that were detected by microarrays in at least one specimen were analyzed by one-way ANOVA to identify genes for which differences in detectable transcript abundance (scaled hybridization signal intensity) were likely to be due to the disease state (NAFLD vs. chronic hepatitis) rather than random variation (for details, see the Supplementary Materials and Methods). Ward's hierarchical clustering method was subsequently applied to construct dendrograms and a “heat map” based on the 1,100 probe sets identified by this approach at a 99.9% confidence level (Fig. 1). Hierarchical clustering provides insight into how such subsets of genes may cluster into patterns suggesting a biologically relevant, potentially testable hypothesis. Interestingly, except for a single specimen from a patient with chronic hepatitis (1 of 26, or 4%) that clustered with the NAFLD specimens, the gene expression patterns of the other 25 chronic hepatitis and 18 NAFLD samples formed two very distinct disease-specific clusters. The resulting dendrogram demonstrated distinctly different intrahepatic gene expression profiles in chronic viral hepatitis and NAFLD, with differential expression of transcripts binding to 1,100 probe sets (or approximately 1 in 6 of the intrahepatic transcripts detected by the microarray assay), as indicated in Fig. 1. The pooled normal sample (NP-1) segregated in the same cluster as the NAFLD specimens, indicating that the intrahepatic transcript profiles in the hepatitis patients were more divergent from the normal tissue than the NAFLD specimens.

Figure 1.

Heat map produced by Ward's hierarchical clustering of intrahepatic gene expression profiles derived from Affymetrix GeneChip assays of biopsy specimens taken from patients with different types of liver disease. With only a single exception (patient HCV-12), tissue samples clustered into groups defined by disease pathogenesis: NAFLD versus chronic viral hepatitis. Transcripts were selected by analyzing for abundance across the tissue samples using ANOVA to select those that differed according to the histologic diagnosis at a 99.9% confidence level. Red color indicates an increased relative abundance of the transcript in the indicated tissue sample, whereas green indicates a decreased relative abundance. Dendrograms at the top and to the left show relatedness of the expression patterns for individual tissue samples and individual probe sets (genes), respectively. The red arrow indicates results obtained with the normal human liver RNA pool. Disease classification is shown for each patient sample at the bottom of the map: HCV, hepatitis C; HBV, hepatitis B; HDV, hepatitis delta; NAFLD, nonalcoholic fatty liver disease.

To assess the validity of the 1,100 probe sets depicted in Fig. 1, we interrogated the microarray data using an alternate method, significance analysis of microarrays (SAM), that has been suggested to produce a lower false discovery rate in selecting genes at a given level of significance15 (see Supplementary Materials and Methods). With a Δ value of 0.2335, SAM identified 1,100 probe sets for which gene expression was considered to be different in chronic hepatitis or NAFLD. Interestingly, the 1,100 probe sets identified by SAM were identical to those selected by ANOVA (Fig. 1). The estimated median number of false significant genes was 4.1 over 500 permutations of the data, resulting in a median false discovery rate of 0.37%. These results strongly support the selection of the 1,100 probe sets shown in Fig. 1, and results confirm that high-density microarrays can reliably distinguish intrahepatic gene expression patterns associated with various disease states. Low variance within replicates and high sensitivity to fluctuations in messenger copy numbers increased the accuracy of the Affymetrix system to detect differential gene profiles; thus, it becomes less essential to confirm the results with reverse transcription–polymerase chain reaction.16 Different research groups demonstrated the high degree of consistency between the Affymetrix data and the reverse transcription–polymerase chain reaction results.16–18

Intrahepatic Gene Expression Profiles in Different Stages of HCV-Related Fibrosis.

Subsequent analysis focused on the 21 liver biopsy samples from patients with HCV infection (Table 1). Each of these patients had HCV RNA detectable in serum branched DNA assay, with the majority infected with genotype 1 HCV. None had elevated α-fetoprotein or radiological evidence of carcinoma. Liver biopsies were carried out before anti-viral therapy, and scored histologically using the Ludwig-Batts system,19 resulting in three groups based on hepatic fibrotic scores: 5 with no fibrosis (stage 0), 10 with early portal or periportal fibrosis (stages 1-2), and 6 with advanced bridging fibrosis or established cirrhosis (stages 3-4). Those with cirrhosis had compensated liver disease. Whereas more subjects with stage 0 and stage 1-2 fibrosis had mild inflammation (P = .003 by chi-squared test), there were no significant differences in frequency of moderate or severe fibrosis among the three groups (P > .10). The proportion of patients with hepatic steatosis was between 20% and 40% and not statistically different between groups.

Table 1. Clinical Features of 21 Patients With Chronic HCV
 Fibrosis Stage 0 (n = 5)Fibrosis Stages 1–2 (n = 10)Fibrosis Stages 3–4 (n = 6)
Age (mean)41 years50 years48 years
Male3 (60%)5 (50%)4 (66%)
Race   
 Caucasian245
 African American241
 Mexican American120
HCV genotype 14 (80%)7 (70%)6 (100%)
Inflammation   
 Mild5 (100%)8 (80%)1 (17%)
 Moderate02 (20%)3 (50%)
 Severe002 (33%)
Steatosis2 (40%)2 (20%)2 (33%)

We compared the transcriptional profiles in these liver samples and the pooled control sample of five normal subjects. A total of 157 genes were selected by ANOVA based on a 99% confidence interval, after filtering to remove probe sets with a less than twofold difference in the hybridization signal intensities compared with the normal control. SAM, using the same three groups based on fibrotic scores, identified 142 of these 157 probe sets with Δ set to generate an estimated false discovery rate of 14.9%. Because there was no substantive difference between the probe sets selected by ANOVA and SAM in terms of hierarchical clustering or apparent biological context (see below), further analysis focused on the 157 probe sets identified by ANOVA.

Hierarchical clustering by Ward's method defined three distinct clusters of tissue samples based on groups of genes for which transcripts were increased in abundance (increasing shades of red in the heat map shown in Fig. 2). These groups showed distinct associations with different stages of hepatic fibrosis (Fig. 2), with all 5 patients with no fibrosis (stage 0) clustering into a single group (Fig. 2, group A). The only other sample clustering into this group had stage 3-4 fibrosis. Eight of the 10 tissue samples with stage 1-2 fibrosis clustered into a second group (Fig. 2, group B), whereas the remaining 2 stage 1-2 samples and 5 of the 6 stage 3-4 samples clustered into a third (Fig. 2, group C). The normal pooled liver RNA sample showed a transcript profile most closely related to, but distinct from, that of group B. Thus, the microarray assay appears to distinguish transcriptional profiles correlating with fibrosis stage in chronic hepatitis C.

Figure 2.

Heat map produced by Ward's hierarchical clustering of intrahepatic gene expression profiles in patients with chronic hepatitis C. Transcripts were selected by ANOVA (99.9% confidence level) for differences among samples with different stages of fibrosis, after filtering to exclude probe sets for which transcripts were not increased greater than two-fold in abundance over the normal RNA sample. At the top of the figure are pie charts showing the distribution of histological diagnoses in the threegroups (clusters) defined by the hierarchical analysis. Transcripts with relative increases in abundance in tissue groups defined by the hierarchical analysis are shown in boxes within the heat map for each group. See also the legend to Fig. 1.

Of the 157 genes that define the clusters shown in Fig. 2, 129 have known functions. Among the genes for which transcript abundance appears to be increased in a fashion related to advancing stages of fibrosis, 75 (58%) have associations with cellular mechanisms that may possibly be related to liver fibrosis. These genes could be further classified into 7 categories based on known functional associations: glycolipid metabolism, apoptosis, oxidative stress responses, inflammation, cellular proliferation, contractility, matrix degradation, collagen synthesis, and tumor-associated. All but 3 of these genes were identified by the SAM analysis (see Supplementary Tables A-C published online). Figure 3 shows the categorical distribution of these 75 genes within the 3 groups defined in the clustering analysis depicted in Fig. 2. Tissue samples clustering into group A had a significantly greater number of transcripts with increased abundance associated with glycolipid metabolism (9 of 45, or 20%), compared with group B (2 of 46, or 4%) or group C (0 of 38) (P < .001 by chi-squared test). Four of the genes involved in glycolipid metabolism identified in group A also have functions related to oxidative stress responses (Supplemental Table A).

Figure 3.

Functional classes of genes for which transcripts differed in abundance according to fibrotic stages (see Fig. 2). The figure shows the number of genes falling into each functional class that were increased in abundance within groups A, B, or C that were defined by the hierarchical clustering analysis.

Transcripts that were increased in abundance in group B (stage 1-2 fibrosis, Fig. 2) were broadly distributed across functional categories, including oxidative stress response (7%), apoptosis (7%), inflammation (7%), cellular proliferation (11%), contractility (7%), and matrix degradation (13%) (Fig. 3 and Supplementary Table B). In contrast, among transcripts increased in abundance in group C (mostly stage 3-4 disease; Fig. 2, group C), 10 (26%) of 38 were associated with collagen synthesis pathways or carcinogenesis.

We next examined individual genes that had a greater than 1.5-fold difference in mean hybridization signals between tissues with various degrees of fibrosis. We compared scaled hybridization signal intensities (which are directly related to transcript abundance) for representative genes in patients with different degrees of fibrosis, as shown in Fig. 4A-C. This approach generally confirmed the clustering results shown in Fig. 2. In stage 0 fibrosis, transcripts for all but 2 genes associated with glycolipid metabolism and oxidative stress response (Supplementary Table A) had ≥1.5-fold increases in signal intensity compared with more advanced fibrosis (Fig. 4A). Interestingly, in stage 1-2 fibrosis, most transcripts with increased abundance were associated with inflammation, matrix degradation, and proliferation (Fig. 4B). For advanced stage 3-4 fibrosis, 8 of 10 genes associated with collagen synthesis or carcinogenesis demonstrated significant increases in abundance compared with those with milder degrees of fibrosis. The hybridization signal intensities were particularly high for osteopontin and collagen type 3 (Fig. 4C).

Figure 4.

Hybridization signal intensities of representative transcripts that differed by greater than 1.5-fold between tissue samples with various stages of fibrosis. (A) Transcripts with relative increases in abundance in stage 0; (B) transcripts with relative increases in abundance in stage 1-2; (C) transcripts with relative increases in abundance in stage 3-4 fibrosis. Results are shown as the mean hybridization signal intensities obtained with tissues samples grouped by stage of fibrosis as determined by histological examination; error bars represent standard deviation. ACADSB, acyl-coenzyme A dehydrogenase; ADH6, alcohol dehydrogenase 6; GBE1, glucan (1,4-alpha-) branching enzyme; CYP2C19, human cytochrome P450 2C19; ALDH5A1, aldehyde dehydrogenase 5 family member 1; AGL, amlo-1,6-glucosidase; FOXM1, Forkhead Box M1; ART1, ADP-ribosyltransferase 1; NTRK3, neutrotrophic tyrosine kinase type 3; IL-4, interleukin 4; CD72, CD72 antigen; ADAM20, metalloproteinase domain 20; RBBP1, retinoblastoma-binding protein 1; SPP1, osteopontin; CD24, CD24 antigen; COL3A1, collagen type 3 alpha1; COL1A2, collagen type1 alpha2; KRT7, Keratin 7. *P < .005, +P < .01.

α-SMA Immunohistochemical Patterns in Fibrosis.

α-SMA staining was scored semiquantitatively in 19 of the 21 liver biopsy specimens by a blinded pathologist. α-SMA–positive HSCs were detectable in all stages of hepatic fibrosis (Fig. 5). The presence of α-SMA–positive HSCs in stage 0 disease suggested that HSCs were already activated, even in the absence of detectable fibrosis. Interestingly, α-SMA–positive cells concentrated mostly in the perivenular region (zone 3) with extension to the intermediate lobule (zone 2), regardless of the severity of fibrosis. The periportal region (zone 1) was relatively spared. An example of a perivenular α-SMA stain pattern in Stage 0 fibrosis is shown in Fig. 5A. Figure 5B shows more diffuse perivenular and intermediate zonal staining in a patient with stage 1-2 fibrosis. Patients with advanced stage 3-4 fibrosis, but without active severe inflammation, tended to have a paucity of α-SMA–positive cells (Fig. 5C).

Figure 5.

Immunohistochemical staining for α-smooth muscle actin (α-SMA) in patients with chronic hepatitis C. At the top are show representative α-SMA staining patterns in tissues with (A) stage 0, (B) stages 1-2, and (C) stages 3-4 fibrosis (all original magnifications ×80). (D) Distribution of α-SMA staining in different regions of the liver lobule in liver tissues showing different stages of fibrosis. α-SMA–positive hepatic stellate cells were present in tissues with all stages of fibrosis and were most evident in zones 2 and 3 of the liver acinus.

Mean α-SMA scores, 3.4 ± 1.1 and 3.4 ± 3.2, were similar for patients with stage 0 and stage 3-4 fibrosis, respectively. Specimens with moderate stage 1-2 fibrosis had higher scores (4.2 ± 1.8), but this was not statistically different from the other 2 groups (P = .72). The zonal distribution of the α-SMA scores in the different fibrotic stages is shown in Fig. 5D. For stage 0 and stage 1-2 disease, the α-SMA score was relatively low (between 0 and 1) in the periportal zone, with higher scores (between 1 and 3) in the perivenule and intermediate zones. For stage 3-4 fibrosis, patients with mild to moderate inflammation had very low total α-SMA scores (between 0 and 1); however, two patients with severe inflammation had high total scores of 6 and 7.

Correlation of α-SMA Scores With Intrahepatic Gene Expression.

Further systematic clustering analyses were carried out in an effort to correlate the α-SMA scores with intrahepatic gene expression profiles. Efforts to correlate transcriptional profiles with HSC activation were complicated by the multiplicity of cell types in the liver and the fact that HSCs represent only a small percentage of cells. One hundred fifteen genes were selected by ANOVA based on 99% CI and a greater than two-fold difference compared with the control RNA. Two distinct tissue clusters were defined by hierarchical clustering using Ward's method, one representing samples with α-SMA scores of 1 to 2, and the other scores of 3 to 7 (Fig. 6).

Figure 6.

Heat map generated by Ward's hierarchical clustering analysis of transcripts identified by ANOVA (99.9% confidence interval) to vary according to overall tissue α-smooth muscle actin (α-SMA) score. A total of 115 probe sets were identified in the Affymetrix assays after filtering to remove those that differed in abundance from the normal control RNA sample by less than twofold. These transcripts clustered into two groups, one with low α-SMA scores (1-2) and the other with high α-SMA scores (3-7). See also the legend to Fig. 1.

Among the 37 genes for which transcripts were increased in abundance in tissues with the higher α-SMA scores, 12 belonged to well-defined molecular pathways (Supplementary Table D). Interestingly, 5 (42%) of these 12 genes are associated with either fatty acid or glycogen metabolism. In contrast, none of the 21 genes with known functions that were increased in abundance in tissues with low α-SMA scores were associated with fatty acid or glycogen metabolism (P = .001 by chi-square test).

Discussion

Hepatic fibrosis is a major feature of the liver injury that accompanies chronic HCV infection and in many patients leads to cirrhosis and end-stage liver disease. HSC activation, occurring in response to tissue injury and oxidative stress, is believed to be the central event in fibrosis.20, 21 Activation comprises two major phases: (1) initiation, and (2) perpetuation.9 Initiation is associated with paracrine-mediated changes in gene expression, as cells become responsive to cytokines and other stimuli. Perpetuation results from the maintenance of signals that lead to further increases in cytokine secretion and accelerated extracellular matrix remodeling. However, the exact molecular events leading to HSC activation are not well understood.

The 157 probe sets included in the heat map shown in Fig. 2 represent those that were identified as varying in hybridization signal intensity according to the histological classification of fibrosis in liver tissue from patients with chronic hepatitis C. Interferon-stimulated genes are not prominently represented among these genes, reflecting the specific question for which the data set was interrogated. These results, however, indicate that interferon-stimulated gene expression levels, although possibly different in normal and HCV-infected tissue, do not correlate with fibrosis stage.

The hierarchical clustering of these 157 genes defined three major groups of tissue samples that correlated well with the histological diagnosis of fibrosis. Approximately 20% of the genes for which abundance was increased in the tissue samples in group A (Fig. 2, predominantly stage 0 disease) relative to those in other groups have functions related to glycolipid metabolism. This finding is consistent with hepatic microarray analyses of acutely infected chimpanzee liver, which demonstrated a prominent association of lipid metabolism genes with the onset of viremia.22

Hepatocytes are a potent source of fibrogenic lipid peroxides in liver injury and inflammation. Expression of the HCV core protein in cultured cells or HCV structural proteins (core, E1, and E2) in mice leads to increased reactive oxygen species (ROS).23, 24 The source of the HCV core protein–induced ROS is likely to be mitochondrial.25 The release of lipid peroxidation products into the extracellular environment could have a direct effect on stellate cell activation and collagen synthesis.20 In response to inflammation-mediated injury, both Kupffer cells and locally recruited neutrophils are also likely to be important sources for ROS that are not present in these in vitro and animal model systems. Furthermore, activated HSCs can potentiate these pathways and produce endogenous ROS via angiotensin II and NADPH oxidase–mediated pathways.26 We found α-SMA–positive cells in stage 0 fibrosis, suggesting that stellate cell activation occurs early in the disease process. In contrast, α-SMA–positive cells decreased with advanced fibrosis in the absence of inflammation (Fig. 5). HSCs may return to an inactive state once fibrosis is well established, especially in the absence of ongoing hepatic inflammation. This conclusion is strengthened by the clustering analysis, which suggested that the transcript profiles of those with stage 0 and advanced stage 3-4 fibrosis are closely related, and distinct from those in stage 1-2 fibrosis. A striking number of the RNA transcripts increased in abundance in association with high α-SMA scores were related to mitochondrial function and lipid metabolism. Collectively, these observations strengthen the hypothesis that oxidative stress, resulting from hepatic and mitochondrial injury with increased lipid metabolism, contributes to the initiation of fibrosis with HSC activation and expression of α-SMA in chronic hepatitis C.

In chronic hepatitis C, both Th1 and Th2 subsets of lymphocytes are important in regulating host responses via cytokine production.27, 28 In general, Th1 cells produce cytokines such as interferon-gamma and tumor necrosis factor that promote inflammation and cell-mediated immunity in an attempt to control infection. Th2 lymphocytes produce cytokines, especially interleukin 4 (IL-4), that favor fibrogenesis in liver injury to a greater extent than Th1 lymphocytes.29–31 The increased abundance of IL-4 message identified in stage 1-2 fibrosis is therefore likely to contribute to the overall fibrogenic process.

A number of the genes for which transcripts were increased in abundance in advanced stage 3-4 disease are responsible for collagen deposition and fibrogenesis. A high level of collagen gene expression is consistent with advanced liver fibrosis in these patients, but was not correlated with α-SMA staining patterns. The large number of tumor-related gene transcripts with increased abundance in tissue with advanced fibrosis in the absence of overt carcinoma also may be of relevance. These genes included osteopontin, osteonectin, RAS oncogene, retinoblastoma-binding protein, and CD24 antigen, which is associated with small cell lung carcinoma. Increased expression of osteopontin and osteonectin are well documented in human hepatocellular carcinomas and liver metastatic tissues.32, 33 Stage 3-4 disease, in particular, was associated with very high level of osteopontin expression.

In conclusion, the differential regulation of various intrahepatic genes at different stages of fibrosis suggests a significant role for cellular lipid metabolism in the “early” stages of HCV-mediated liver injury and the initiation of HSC activation. Additional interactions between the host and the virus likely play a critical role in the perpetuation of HSC activation and progression of fibrosis. High levels of “oncogenic” gene expression observed in advanced fibrosis could explain the predisposition of hepatocellular carcinoma in cirrhosis. The results presented document the value of microarray technology and suggest that it will be instrumental in identifying the genes that regulate specific steps in the process of fibrosis.

Acknowledgements

The authors are grateful for expert technical assistance provided by Dr. Tom Woods and Michelle Guigneaux of the UTMB Molecular Genetics Core Laboratory.

Ancillary