Distinct cellular responses differentiating alcohol- and hepatitis C virus-induced liver cirrhosis

To analyze gene expression changes in cirrhotic livers associated with HCV (n = 8) or ethanol (n = 7), microarray experiments were performed comparing each type of cirrhotic liver to a reference pool of normal liver tissue. Cirrhosis severity was evaluated according to the CTP clinical scoring system (A-C, least to most severe, respectively) and patient scores are summarized in Table 1. The global gene expression profiles of all experiments are shown in Figure 1. In total, 2,965 genes were differentially expressed (≥ 2-fold, P ≤ 0.05) in at least 3 of 15 samples. The clustering algorithm demonstrates that global gene expression profiles clearly distinguish ethanol- and HCV-associated cirrhotic samples (indicated with black and blue text, respectively). Substantially more gene expression changes were observed in HCV-related samples compared to ethanol-related, ranging from 574–1622 in HCV samples as compared to 79–683 in ethanol samples. Interestingly, the gene expression patterns observed were relatively uniform among HCV-cirrhotic samples and did not appear to differ between CTP classes. In contrast, the global gene expression profiles demonstrated distinct patterns among CTP classes in ethanol-related cirrhosis.

To further classify functional pathways specifically associated with etiology, we performed a one-way error-weighted ANOVA comparison of HCV- and ethanol-associated cirrhotic liver biopsies. A set of 946 genes was identified as being significantly (P ≤ 0.05) different with respect to expression level between the two groups. We then used Fatigo software to analyze functional groups that differed between cirrhosis etiologies. Genes associated with metabolism, the immune response, and cell death were differentially regulated in HCV- and alcohol-related cirrhosis. Many of the genes specifically induced in HCV-associated cirrhotic livers were not up-regulated in ethanol-cirrhotic livers. This appears to be at least partly related to activation of the innate antiviral immune response, as many of these up-regulated genes are known to be regulated by IFN (Fig. 2A). Specifically, a subset of genes highly induced in HCV-associated cirrhotic samples were IFNα/β-inducible (IRF7, OAS2, GIP2, GIP3) or take part in its activation. Additional IFN-type I genes – OAS1, IFITM1, IFI35, IFIT1, MX1, IFI27 – were highly up-regulated in HCV-associated samples but not included in this ANOVA. A small set of genes (ISG20, IFI16, TRIM22, STAT5A) were also highly up-regulated in nearly all samples, including ethanol-cirrhotic livers, suggesting these may play a role in an inflammatory response related to cirrhosis development regardless of etiology.

Significant infiltration of lymphocytes into the liver is observed during chronic HCV infection [26]. In this study, up-regulation of gene markers for lymphocytes was indeed much more pronounced in HCV-infected liver samples than in ethanol-induced livers (Fig. 2B, Additional file 1). Specifically, RANTES, which is chemotactic for T cells in vitro was significantly up-regulated in all HCV-infected livers (1.8–4.5 fold, P ≥ 0.05). In contrast, circulating levels of natural killer cells have been shown to be suppressed under the influence of alcohol [27, 28]. Consistent with this, we observed decreased expression of natural killer cell-related genes, including CD44, IL15R, CD2 and CCL5, in alcohol-associated cirrhotic liver samples.

Whereas HCV-associated cirrhotic samples looked very similar regardless of CTP class (Fig. 1), global gene expression profiles indicated that CTP class A ethanol-associated samples exhibited unique expression patterns from CTP class B and C samples. In order to further identify which gene expression patterns were specific depending on severity of ethanol-associated cirrhosis, we performed a one-way error-weighted ANOVA (P ≤ 0.05) of less severe (CTP class A) cirrhotic samples versus more severe (CTP classes B and C) samples (Fig. 3A, Additional file 2). Expression profiles distinguished the ethanol-related CTP class A samples from ethanol-related CTP class B and C samples. In general, class A samples demonstrated up-regulation of more genes than in samples with more severe cirrhosis (Class B and C). Distinct gene expression profiles were not observed between CTP classes B and C, but this may be due to small sample size. Ingenuity analysis of groups I and II (Fig. 3A) demonstrated that the number of genes associated with various biological functions differed significantly depending on severity of alcohol cirrhosis. Of 135 annotated genes up-regulated in group II, 60 were associated with cell death as compared to only 16 in group I. High numbers of genes related to cell movement, mainly of phagocytes, and tissue development were also up-regulated in more severe ethanol cirrhosis. Many genes in group I (less severe cirrhosis) were associated with lipid metabolism, specifically the metabolism of cholesterol. Genes involved in the function of the immune and lymphatic system, such as activation of mononuclear leukocytes and proliferation of macrophages, were also up-regulated in early, less severe cirrhosis.

A number of genes that were nearly exclusively up-regulated in the CTP-A class of ethanol cirrhosis included inflammatory-related genes (Fig. 3B). Some of these – VIPR1, CCL15, PLG, CCR3, CSF1 – are involved in macrophage activation and migration. Macrophage-related genes not in this ANOVA set but also highly up-regulated in CTP class A samples included FN1, C5, LPA, and SERPINA3.

Interestingly, severe ethanol cirrhosis differed from less severe ethanol cirrhosis in expression of genes associated with fatty and bile acid metabolism (Fig. 3C). Many of these genes were up-regulated during less severe cirrhosis but down-regulated by end-stage ethanol cirrhosis. Sterol 27-hydroxylase (CYP27A1) oxidizes cholesterol in the alternative pathway of bile acid biosynthesis; a defect in this enzyme leads to excessive accumulation of sterols [29]. We observed significant up-regulation of this gene in livers of all patients with early ethanol cirrhosis and, as with other genes in this group, down-regulation during late (end-stage) cirrhosis (Figure 3C). Fatty acids are substrates for cytochrome CYP2E1 (P450 2E1) [5], which is a key enzyme in the microsomal ethanol oxidation system and plays a role in alcohol detoxification and nutritional support [30]. It is also a major source of oxidative stress [6, 5], and both protein and mRNA levels increase dramatically in actively drinking patients [31]. Therefore it was interesting that significant down-regulation of this gene was also observed in high grade (CTP class B and C) cirrhotic livers (Figure 3C). A similar pattern was observed for genes encoding alcohol dehydrogenases (ADH4, ADH1B, ADH1C).

CTP class A alcohol-associated samples appeared to have unique expression profiles compared to all other samples regardless of etiology (Fig. 1). We therefore performed a one-way error-weighted ANOVA (P ≥ 0.05) of less severe (CTP class A alcohol-associated) versus more severe ethanol-related and HCV-associated cirrhotic samples to identify genes with expression levels unique to CTP class A alcohol. Of 607 total genes in this set, a cluster of 273 were highly up-regulated in all samples but ethanol CTP class A (Fig. 4, group I), and a cluster of 109 were highly up-regulated specifically in class A (Fig. 4, group II); see Additional file 3 for more information on gene names. Functional analysis (Ingenuity) revealed that genes associated with cell death, movement, and the immune response were much more abundant in group I than in group II. Most immune response genes up-regulated in group I were generally involved in proliferation and migration of leukocytes, while those in group II were again related to activation and infiltration of macrophages. The majority of genes up-regulated in group II of early ethanol cirrhosis are involved in the modification of lipids. The response to oxidative stress was also significantly different in early ethanol cirrhosis as related to all other samples. Two genes in particular were highly up-regulated in CTP class A livers – PRDX2 (up to 4.7 fold) and APOE (up to 17.5 fold) – and either less up-regulated or even down-regulated in the majority of the remaining samples.

Cirrhotic liver samples from 7 alcoholic and 8 HCV-infected patients were obtained from liver disease patients at the University of Barcelona, Spain who underwent liver transplant or hepatic resection. None of the HCV-infected patients in this study were undergoing anti-viral treatment at the time the liver sample was obtained. Alcohol consumption information was based on data obtained in patient interviews. The severity of cirrhosis was clinically graded according to Child-Turcotte-Pugh (CTP) scores (summarized in Table 1). Normal, uninfected liver tissue from 8 patients who underwent tumor resection due to metastases of non-liver origin was collected at the same clinic (non-tumor material was used). These uninfected samples were pooled to create a standard normal liver reference (pNL) that was used for all microarray experiments. All patients gave informed prior consent to protocols approved by the University of Barcelona, Spain.

Frozen liver tissues were disrupted in Trizol Reagent (Invitrogen, Carlsbad, CA) using a Polytron homogenizer (PowerGene 700, Fisher Scientific), and total RNA was isolated according to the Trizol protocol. All total RNA samples were double amplified using the RiboAmp RNA Amplification kit (Arcturus, Mountain View, CA). The quality of amplified RNA was evaluated by capillary electrophoresis using an Agilent 2100 Bioanalyzer (Agilent Technologies, Palo Alto, CA).

Microarray format, protocols for probe labeling, hybridization, slide treatment, and scanning were performed as previously described [35] and are also available at our website [51]. Human cDNA set 1 expression arrays carrying 13,026 unique cDNA clones were purchased from Agilent. A single experiment comparing two samples, one from a liver with infection and one from the normal liver reference, was performed with four replicate arrays using the dye label reverse technique as described before, thus providing mean ratios between the expression levels of each gene in the analyzed sample pair, standard deviations, and P values [35, 52‐54]. All data were entered into a custom-designed database, Expression Array Manager, and then uploaded into Rosetta Resolver System 4.0 (Rosetta Biosoftware, Kirkland, WA) and Spotfire Decision Suite 7.1.1 (Spotfire, Somerville, MA). Data normalization and the Resolver System Error model specifically developed for this array format are described on the web site [51]. This web site is also used to publish all primary data in accordance with the proposed standards [55]. The annotation sites Fatigo [56] and Ingenuity [57] were used to classify genes into functional groups.