Multi-omics analysis of intra-tumoural and inter-tumoural heterogeneity in pancreatic ductal adenocarcinoma

Abstract

The poor prognosis of pancreatic ductal adenocarcinoma (PDAC) is associated with the tumour heterogeneity. To explore intra- and inter-tumoural heterogeneity in PDAC, we analysed the multi-omics profiles of 61 PDAC lesion samples, along with the matched pancreatic normal tissue samples, from 19 PDAC patients. Haematoxylin and Eosin (H&E) staining revealed that diversely differentiated lesions coexisted both within and across individual tumours. Whole exome sequencing (WES) of samples from multi-region revealed diverse types of mutations in diverse genes between cancer cells within a tumour and between tumours from different individuals. The copy number variation (CNV) analysis also showed that PDAC exhibited intra- and inter-tumoural heterogeneity in CNV and that high average CNV burden was associated poor prognosis of the patients. Phylogenetic tree analysis and clonality/timing analysis of mutations displayed diverse evolutionary pathways and spatiotemporal characteristics of genomic alterations between different lesions from the same or different tumours. Hierarchical clustering analysis illustrated higher inter-tumoural heterogeneity than intra-tumoural heterogeneity of PDAC at the transcriptional levels as lesions from the same patients are grouped into a single cluster. Immune marker genes are differentially expressed in different regions and tumour samples as shown by tumour microenvironment (TME) analysis. TME appeared to be more heterogeneous than tumour cells in the same patient. Lesion-specific differentially methylated regions (DMRs) were identified by methylated DNA immunoprecipitation sequencing (MeDIP-seq). Furthermore, the integration analysis of multi-omics data showed that the mRNA levels of some genes, such as PLCB4, were significantly correlated with the gene copy numbers. The mRNA expressions of potential PDAC biomarkers ZNF521 and KDM6A were correlated with copy number alteration and methylation, respectively. Taken together, our results provide a comprehensive view of molecular heterogeneity and evolutionary trajectories of PDAC and may guide personalised treatment strategies in PDAC therapy.

Keywords: cancer evolution; heterogeneity; immunotherapy; molecular targeted therapy; multi-omics analysis; pancreatic ductal adenocarcinoma.

FIGURE 1

Specimens and workflow. (A) The locations of samples acquired from each patient. A total of 61 lesion samples, together with the matched normal pancreatic tissue samples, were obtained from 19 patients with pancreatic ductal adenocarcinoma (PDAC). (B) The morphological heterogeneity of representative samples. P, patient, L, lesion and N, matched normal tissue. Scale bar = 600 μm. (C) A schematic diagram of multi‐omics analysis

FIGURE 2

Somatic mutations of driver genes and copy number variations (CNVs) in PDAC. (A) The classification of somatic mutations. (B, C) Comparison of TMB and CNV burden within and across individual tumours. TMB, tumour mutation burden

FIGURE 3

Tumour evolution analysis. (A) Phylogenetic trees were plotted to show the clonal evolution of each sample using Revolver. The grey circle denotes the cluster without driver gene mutations. The coloured circle denotes the cluster with one or more driver gene mutations. The number in the circle represents the quantity of single nucleotide variants. (B) The proportion of trunk/branch mutations in each patient based on phylogenetic trees. (C) The probability density of driver gene fold enrichment among trunk and branch mutations in pan‐cancer and PDAC driver genes. (D) Distribution of most frequently mutated driver genes on phylogenetic trees. GL, germline

FIGURE 4

Timing of somatic events in PDAC evolution. The somatic mutations and chromosome‐arms are represented by bars indicating whether the events are clonal or subclonal. Clonal somatic mutations, chromosome‐arms and mutational signatures are further classified as early (before genome duplication) or late (after genome duplication). The frequency of somatic mutations and chromosome‐arm calculated by (late clonal + subclonal)/total are indicated on the right side of the bars. The pie charts show the proportion of each signature. GD, genome doubling

FIGURE 5

Transcriptional and epigenetic heterogeneities of PDAC. (A) The number of differentially expressed genes (DEGs), which were defined as genes with an absolute GFOLD value greater than 2. The matched normal tissues were used as control. (B) A heatmap of all the downregulated and upregulated DEGs. Red represents upregulated DEG, and blue represents downregulated DEG. (C) KEGG pathway enrichment analysis of the shared DEGs. The DEGs shared by all the lesions within the same tumour were defined as shared DEGs. (D) The heatmap of 22 differentially expressed TME‐related immune markers. Absolute immune cell abundance was calculated using a set of 22 immune cell reference profiles (LM22) on CIBERSORT website to analyse the heterogeneity of the immune microenvironment in each tumour sample. TME, tumour microenvironment. (E) Phyloepigenetic trees. (F) A heatmap of differentially methylated regions in 26 lesions from 7 patients with PDAC. Blue represents hypomethylation, and red represents hypermethylation. (G) Heatmaps show the heterogeneity with the top 2000 hypermethylated regions in the tumour. Yellow represents high similarity. Blue represents low similarity

FIGURE 6

The association of genomic alterations with RNA expression. (A) Correlation between CNV and mRNA expression in all samples. (B) The RNA expressions of some genes were significantly correlated with the gene copy numbers. Blue dot represents driver gene, and red dot represents non‐driver gene. (C) A heatmap of mutations, CNV and their associations with RNA expression of the most frequently mutated PDAC driver genes

FIGURE 7

Correlation of mRNA expression of some genes with CNV or methylation status. (A) Correlation between mRNA expression and methylation status of ZNF521. (B) Correlation between mRNA expression and CNV of KDM6A. (C) The distribution of KDM6A and ZNF521 expression, copy number and methylation status across samples. (D) KEGG pathway enrichment analysis of key genes. Hyper/hypo represents the degree of methylation. –1 and 1 represent hypomethylated and hypermethylated states, respectively. 0 indicates no obvious change in methylation state. Gain/loss represents the change of copy number, –1 and 1 represent the deletion or amplification of copy number respectively, and 0 represents the insignificant change of copy number

FIGURE 8

Correlations of CNV burden with patients’ survival and characterisation of KDM6A and ZNF521 in PDAC patients from the The Cancer Genome Atlas (TCGA) database. (A, B) Kaplan–Meier survival curves of patients with low or high average CNV burden. Patients with average CNV burden in the top quartile and the rest were segregated into high and low CNV burden groups, respectively. Disease‐free survival (DFS) was defined as the time from surgery to locoregional recurrence or distant metastasis. Overall survival (OS) was defined as the time from surgery to death from any cause or last follow‐up (censored patient). (C, D) mRNA expression and promoter methylation of ZNF521 in normal tissue and primary tumour tissue from TCGA database. (E and F) Correlations of ZNF521 mRNA expression with disease‐free survival and overall survival of patients from TCGA database. (G, H) mRNA expression and promoter methylation of KDM6A in normal tissue and primary tumour tissue from TCGA database. (I, J) Correlations of KDM6A mRNA expression with disease‐free survival and overall survival of patients from TCGA database. The dotted lines in E, F, I and J represent the error bars of 95% CI. PAAD, pancreatic adenocarcinoma