An atlas of the human liver diurnal transcriptome and its perturbation by hepatitis C virus infection

Abstract

Chronic liver disease and cancer are global health challenges. The role of the circadian clock as a regulator of liver physiology and disease is well established in rodents, however, the identity and epigenetic regulation of rhythmically expressed genes in human disease is less well studied. Here we unravel the rhythmic transcriptome and epigenome of human hepatocytes using male human liver chimeric mice. We identify a large number of rhythmically expressed protein coding genes in human hepatocytes of male chimeric mice, which includes key transcription factors, chromatin modifiers, and critical enzymes. We show that hepatitis C virus (HCV) infection, a major cause of liver disease and cancer, perturbs the transcriptome by altering the rhythmicity of the expression of more than 1000 genes, and affects the epigenome, leading to an activation of critical pathways mediating metabolic alterations, fibrosis, and cancer. HCV-perturbed rhythmic pathways remain dysregulated in patients with advanced liver disease. Collectively, these data support a role for virus-induced perturbation of the hepatic rhythmic transcriptome and pathways in cancer development and may provide opportunities for cancer prevention and biomarkers to predict HCC risk.

Fig. 1. The rhythmic transcriptome of human hepatocytes in vivo.

a Schematic representation for the generation of immunodeficient humanized liver chimeric mice (HLCM). Liver tissues obtained from these HLCM were used to perform RNA-seq and ChIP-seq. PHH primary human hepatocyte. The cartoon was created by the authors using images from freely available online resources. b Mean rhythmic expression pattern of core CC-oscillator genes and CC-output regulators in human and mouse liver cells of HLCM in two independent experiments (Series 1 and 2). Series 1: n = 2/timepoint, Series 2: n = 3/timepoint. Source data are provided as a Source Data file. c Mean expression pattern of genes showing rhythmicity in human and mouse hepatocytes in HLCM as predicted by dryR (n = 5 HLCM/ diurnal timepoint). dryR identified four models of rhythmic genes in HLCM. A fifth model comprising only non-cycling genes is not shown. Phase distribution of respective models is indicated by radial coordinates, where green-human, pink-mice, and gray-overlap of humans and mice. Source data are provided as a Source Data file. d HALLMARK pathways significantly (FDR <0.05) enriched for rhythmic genes in human hepatocytes in HLCM (n = 5 HLCM/ timepoint) as listed in (c) and their expression in WT mice. Similar pathways show overlapping (±1 ZT step) of peak enrichment scores comparing human and WT mice maximum enrichment scores. Source data are provided as a Source Data file. e Rhythmic expression of transcription factors (TFs) in human hepatocytes in HLCM (n = 5 HLCM/ timepoint), as predicted by dryR. Source data are provided as a Source Data file. f Examples of the mean expression pattern of dryR identified TFs as listed in (e). The Y-axis represents the DESeq2 normalized reads. Human hepatocytes (red) and residual murine (green) cells. n = 5 HLCM/timepoint. Bars represent SD. Source data are provided as a Source Data file. g ChIP-seq coverage plots indicate diurnal variations in H3K27ac levels in IRF2 in human hepatocytes and WT mice liver. Human (green) and murine (red) cells. ZT0: represented twice in (b–f) in each panel to maintain conformity.

Fig. 2. Chronic HCV infection disrupts the rhythmicity of human liver transcriptome in vivo.

a Schematic representation of the chronic HCV infection of HLCM. Liver tissues obtained from these infected HLCM were used to perform RNA-seq and ChIP-seq. Two independent infection experiments were performed. PHH primary human hepatocytes. The cartoon was created by the authors using images from freely available online resources. b Four categories of CC-disturbed and unaltered genes identified by dryR according to their expression in control and HCV-infected human liver cells in HLCM (n = 5 HLCM/group and timepoint). A fifth model comprising only non-cycling genes is not shown. Phase distribution of indicated models is indicated by radial coordinates, where green: control, pink: HCV, and gray: overlap of control and HCV. Source data are provided as a Source Data file. c Examples for mean enrichments over indicated timepoints of pathways significantly enriched (FDR < 0.05) for rhythmic genes (i.e., HM HALLMARK, GO Gene Ontology, BC BIOCARTA, RE Reactome gene sets; see Fig. 1d), shown for controls and HCV-infected livers. n = 5 HLCM/group and timepoint. Bold dots inside figure panels represent significant differences in enrichments (p < 0.05; Wilcoxon signed-rank test, two-tailed). Bars represent SD. Source data, including the exact p values are provided as a Source Data file. ZT0: represented twice in (b, c) in each panel to maintain conformity.

Fig. 3. Chronic HCV infection impairs the remodeling of the temporally regulated enhancers in vivo.

a Density of H3K27ac peaks at different timepoints across the gene bodies in control and HCV-infected human liver cells in HLCM. TSS transcription start site. Source data are provided as a Source Data file. b Density of TSS-associated H3K27ac peaks at different timepoints in control and HCV-infected human liver cells in HLCM. TSS transcription start site. Source data are provided as a Source Data file. c Quantitation of H3K27ac peak numbers at TSSs from (b) CTR control (green), HCV HCV-infected (red), p = 0.002, Mann–Whitney test, two-tailed, n = 5/timepoint. The box plots show the median (line), the 25th and 75th percentiles (box), and values within 1.5 times the interquartile range (whiskers). Source data are provided as a Source Data file. d Density of H3K9ac peaks at different timepoints across the gene bodies in control and HCV-infected human liver cells in HLCM. TSS transcription start site. Source data are provided as a Source Data file. e Density of TSS-associated H3K9ac peaks at different timepoints in control and HCV-infected human liver cells in HLCM. TSS transcription start site. Source data are provided as a Source Data file. f Quantitation of H3K9ac peak numbers at TSSs from (e), CTR control (green), HCV HCV-infected (red) p = 0.026, Mann–Whitney test, two-tailed, n = 5/timepoint. The box plots show the median (line), the 25th and 75th percentiles (box), and values within 1.5 times the interquartile range (whiskers). Source data are provided as a Source Data file. g Promoter-enriched H3K27-acetylation for HCV host factors XBP1 and RAF1 in control (green) and HCV-infected (red) human liver cells in HLCM. h Percentages of enhancer/TSS-enriched peaks with gain (orange) or loss (green) of H3K27-acetylation that overlap with transcription factor binding sites in gene targets as listed in the Jaspar database. Only the 50 top hits are shown for each direction. Source data are provided as a Source Data file.

Fig. 4. Chronic HCV infection perturbs the expression of rhythmic pathways in patients and cancer risk.

a Mean enrichments of Hallmark pathways enriched for human rhythmic genes as identified in Fig. 1d and the PLS in the liver of HCV-infected HLCM and HCV-infected patients. Humanized mice: HCV-infected HLCM (Fig. 2; HCV vs. controls), Patients: HCV-infected vs. controls. Pool: bulk analysis of all samples. Red=significant (FDR <0.05) enrichment, blue=significant (FDR <0.05) negative enrichment, white=no significant enrichment/unchanged. Source data are provided as a Source Data file. b Spearman correlation between the induction of the CC-oscillator signatures and clinical demographics and the PLS in 216 early-stage (i.e., Child-Pugh class A) HCV cirrhosis patients. The signatures were defined as genes with >4-fold over-expression in either HCV-infected or control hepatocytes in early or late timepoints in each of the four representative models (i.e., loss, gain, altered, and unaltered). All pair-wise correlations between the variables (correlation matrix) as shown in rows and columns. The correlation matrix was clustered to depict groups of variables sharing similar patterns of correlation. The deep blue, light blue, and brown color bars indicate the presence of the three major correlation clusters, a two-tailed test. Source data are provided as a Source Data file. c Classification of the 216 early-stage cirrhosis patients by the CC-oscillator signatures associated with clinical features related to liver disease severity (Supplementary Table 2), magnitude of gene signature/set modulation is shown in each patient (in each column) in the cohort. The induction/suppression of the circadian clock gene sets (shown in the bottom half of the panel) was determined by GSEI. For the PLS, normalized expression levels or the PLS member genes were used for the NTP analysis^,. AFP alpha-fetoprotein, ALT alanine aminotransferase, AST aspartate aminotransferase, PLS prognostic liver signature, SVR sustained virologic response. Source data are provided as a Source Data file. d Association of the disease severity-related HCV_CLOCK signature-based classification with overall survival in the 216 early-stage HCV cirrhosis patients. p = 0.03, log-rank test, two-tailed. Source data are provided as a Source Data file. e The Hallmarks of Cancer and perturbed rhythmic pathways in the humanized liver (HCLM), which were altered by HCV infection, predisposing toward HCC.