Molecular mechanisms involved in the interaction effects of alcohol and hepatitis C virus in liver cirrhosis

Abstract

The mechanisms by which alcohol consumption accelerates liver disease in patients with chronic hepatitis C virus (HCV) are not well understood. To identify the characteristics of molecular pathways affected by alcohol in HCV patients, we fit probe-set level linear models that included the additive effects as well as the interaction between alcohol and HCV. The study included liver tissue samples from 78 patients, 23 (29.5%) with HCV-cirrhosis, 13 (16.7%) with alcohol-cirrhosis, 23 (29.5%) with HCV/alcohol cirrhosis and 19 (24.4%) with no liver disease (no HCV/no alcohol group). We performed gene-expression profiling by using microarrays. Probe-set expression summaries were calculated by using the robust multiarray average. Probe-set level linear models were fit where probe-set expression was modeled by HCV status, alcohol status, and the interaction between HCV and alcohol. We found that 2172 probe sets (1895 genes) were differentially expressed between HCV cirrhosis versus alcoholic cirrhosis groups. Genes involved in the virus response and the immune response were the more important upregulated genes in HCV cirrhosis. Genes involved in apoptosis regulation were also overexpressed in HCV cirrhosis. Genes of the cytochrome P450 superfamily of enzymes were upregulated in alcoholic cirrhosis, and 1230 probe sets (1051 genes) had a significant interaction estimate. Cell death and cellular growth and proliferation were affected by the interaction between HCV and alcohol. Immune response and response to the virus genes were downregulated in HCV-alcohol interaction (interaction term alcohol*HCV). Alcohol*HCV in the cirrhotic tissues resulted in a strong negative regulation of the apoptosis pattern with concomitant positive regulation of cellular division and proliferation.

Figures

Figure 1

Unsupervised cluster analysis. An unsupervised analysis was conducted by first filtering the data set by retaining only the most variable probe sets, for which the range of each probe set was calculated and probe sets having a range among the top 1% were retained. Thereafter, agglomerative hierarchical clustering performed using the Ward method was applied to the filtered data set with the 1-Pearson correlation as the dissimilarity measure. Blue light boxes, normal liver tissues; black boxes, alcohol-cirrhotic liver tissues; pink boxes, HCV-cirrhotic liver tissues; green boxes, alcohol-HCV cirrhotic liver tissues.

Figure 2

Venn diagram. For each sample, RNA was extracted and after being converted to biotin-labeled cRNA was hybridized to an Affymetrix HG-U133Av2 GeneChip. After image analysis, probe-set expression summaries were calculated by using the robust multiarray average. Thereafter, probe-set–level linear models were fit such that probe-set expression was modeled by HCV status, alcohol status, and the interaction between HCV and alcohol. Of particular interest were those probe sets for which all three terms (HCV, Alcohol, and HCV*Alcohol) were significant. To control for multiple hypothesis tests, probe sets were significant at an α level of 0.0001.

Figure 3

The top-scoring network of interactions among the 3370 probe sets identified as significantly differentially expressed when we compared HCV with normal samples. The probe sets were subsequently analyzed by using the Ingenuity pathway analysis software (https://analysis.ingenuity.com). This software is designed to identify dynamically generated biological networks, global canonical pathways and global functions. At interconnections of significant functional networks, protein nodes appeared in different shades of red and green or white depending on being upregulated and downregulated or no-change, respectively, in HCV-cirrhotic samples.

Figure 4

Supervised cluster analysis for response to virus genes. Gene symbols for all Affymetrix probe sets were obtained by using the Bioconductor annotation package. Thereafter, all probe sets annotated to interrogate specific genes in the pathways of interest (response to virus) were retained, and agglomerative hierarchical clustering was performed by using the Ward method with 1-Pearson as the dissimilarity measure was applied to the pathway. Blue light boxes, normal liver tissues; black boxes, alcohol-cirrhotic liver tissues; pink boxes, HCV-cirrhotic liver tissues; green boxes, alcohol-HCV cirrhotic liver tissues.

Figure 5

Significant interactions. For CASP1, BCL2, TRADD and BAD genes, interaction plots were constructed to enable visualization of the expression changes based on the HCV status (HCV positive versus negative) or alcohol status (yes versus no). In each plot one line represents alcohol = yes (solid line) and the other line represents alcohol = no (dashed line). Each line is formed by connecting two points: the group mean expression for HCV-negative patients to the group mean expression for HCV-positive patients.