Genomic and transcriptional heterogeneity of multifocal hepatocellular carcinoma

Abstract

Background: Hepatocellular carcinoma (HCC) often presents with multiple nodules within the liver, with limited effective interventions. The high genetic heterogeneity of HCC might be the major cause of treatment failure. We aimed to characterize genomic heterogeneity, infer clonal evolution, investigate RNA expression pattern and explore tumour immune microenvironment profile of multifocal HCC.

Patients and methods: Whole-exome sequencing and RNA sequencing were carried out in 34 tumours and 6 adjacent normal liver tissue samples from 6 multifocal HCC patients. Protein expression of Ki67, AFP, P53, Survivin and CD8 was detected by immunohistochemistry. Fluorescence in situ hybridization was carried out to validate the amplification status of sorafenib-targeted genes.

Results: We deciphered genomic and transcriptional heterogeneity among tumours in each multifocal HCC patient including mutational profiles, copy number alterations, tumour evolutionary trajectory and tumour immune microenvironment profiles. Of note, sorafenib-targeted alterations were identified in the trunk of phylogenetic tree in only one out of the six patients, which may explain the relative low treatment response rate to sorafenib in clinical practice. Moreover, we demonstrated RNA expression patterns and tumour immune microenvironment profiles of all nodules. We found that RNA expression pattern was associated with Edmondson-Steiner grading. Based on the differential expression of 66 reported immune markers, unsupervised hierarchical clustering analysis of 34 nodules identified immune subsets: one low expression cluster with seven nodules and one high expression cluster with 11 nodules. CD8+ T cells were more enriched in nodules of the high expression cluster.

Conclusions: Our study provided a detailed view of genomic and transcriptional heterogeneity, clonal evolution and immune infiltration of multifocal HCC. The heterogeneity of druggable targets and immune landscape might help interpret the clinical responsiveness to targeted drugs and immunotherapy for multifocal HCC patients.

Keywords: clonal evolution; genomic heterogeneity; hepatocellular carcinoma; immune microenvironment; multifocal; transcriptional heterogeneity.

Figure 1.

Spatial heterogeneity of multifocal HCC based on non-silent somatic mutations. (A) The heat maps of non-silent mutations in six multifocal HCC cases. Presence (red) or absence (grey) of a non-silent mutation was indicated for each tumour. Driver somatic mutations were listed on the right. T6 of case HCC591 had no common mutations with other tumours, indicating it was of independent origin. (B) Hierarchical clustering of all HCC samples based on the mutation spectrum. T6 of case HCC591 was separated away from other foci in this patient. (C) Validation of tumour origin at the protein level by IHC staining. Higher protein expression levels of Ki67, AFP, P53 and Survivin were observed in T6 of HCC591, when compared with other nodules in this patient.

Figure 2.

Phylogenetic trees of multifocal HCC cases and validation of sorafenib-targeted alterations. (A) Phylogenetic trees were constructed using somatic mutations. The trunk, shared branches, private branches and independent branch were represented with purple, green, brown and pink, respectively. Driver alterations of HCC were mapped to the phylogenetic tree. Red: druggable alterations; purple: sorafenib-targeted alterations; black: other driver alterations. (B) BRAF amplification in T2 and T3 of case HCC591, PDGFRB amplification in T1 and T4 of case HCC642, and VEGFA amplification in T1 and T3 of case HCC642 were validated by FISH.

Figure 3.

Clone phylogenies and sample clone mixtures of multifocal HCC cases. Distinct clonal genotypes were colour-coded with numbers for each patient. The numbers indicate the timing of acquisition of these clonal genotypes. Number ‘0’ represented ancestor mutation cluster in each patient. The clonal composition was proportional to the prevalence of each clone in each sample. Samples were arrayed in the inner circle around the anatomical diagram and the outer circle was constituent clones from the overall clone phylogeny. Driver mutations were listed on the right and the number of mutations in each subclone was labelled in the box.

Figure 4.

Clustering analysis of differentially expressed genes and characterization of immune microenvironment heterogeneity using RNA-Seq data. (A) Unsupervised hierarchical clustering of gene expression across all sequenced samples identified adjacent normal liver tissues, clusters 1 and 2 of tumour tissues in this study. RNA expression pattern correlated with Edmondson–Steiner grading. (B) The absolute immune score of 34 tumours based on RNA sequencing data. (C) Representative pictures of IHC validation (left panel) and quantification (right panel) of the predicted CD8+ T cell infiltration. (D) Unsupervised hierarchical clustering of gene expression was carried out by using a curated list of 66 immune markers. One high cluster labelled in yellow box with abundant expression of immune markers and one low immune cluster in blue box with low expression of immune markers were identified. (E) Differentially enriched composition of infiltrated immune cells between the high and low expression cluster. P values were calculated by the Wilcoxon rank-sum test. The pink dotted lines on the y-axis indicated P value of 0.01. The pink dotted lines on the x-axis indicated Z score of 0. The significant enriched immune cells are labelled in red on the plot.