Single-Cell Transcriptomics of Pancreatic Cancer Precursors Demonstrates Epithelial and Microenvironmental Heterogeneity as an Early Event in Neoplastic Progression

Abstract

Purpose: Early detection of pancreatic ductal adenocarcinoma (PDAC) remains elusive. Precursor lesions of PDAC, specifically intraductal papillary mucinous neoplasms (IPMNs), represent a bona fide pathway to invasive neoplasia, although the molecular correlates of progression remain to be fully elucidated. Single-cell transcriptomics provides a unique avenue for dissecting both the epithelial and microenvironmental heterogeneities that accompany multistep progression from noninvasive IPMNs to PDAC.

Experimental design: Single-cell RNA sequencing was performed through droplet-based sequencing on 5,403 cells from 2 low-grade IPMNs (LGD-IPMNs), 2 high-grade IPMNs (HGD-IPMN), and 2 PDACs (all surgically resected).

Results: Analysis of single-cell transcriptomes revealed heterogeneous alterations within the epithelium and the tumor microenvironment during the progression of noninvasive dysplasia to invasive cancer. Although HGD-IPMNs expressed many core signaling pathways described in PDAC, LGD-IPMNs harbored subsets of single cells with a transcriptomic profile that overlapped with invasive cancer. Notably, a proinflammatory immune component was readily seen in low-grade IPMNs, composed of cytotoxic T cells, activated T-helper cells, and dendritic cells, which was progressively depleted during neoplastic progression, accompanied by infiltration of myeloid-derived suppressor cells. Finally, stromal myofibroblast populations were heterogeneous and acquired a previously described tumor-promoting and immune-evading phenotype during invasive carcinogenesis.

Conclusions: This study demonstrates the ability to perform high-resolution profiling of the transcriptomic changes that occur during multistep progression of cystic PDAC precursors to cancer. Notably, single-cell analysis provides an unparalleled insight into both the epithelial and microenvironmental heterogeneities that accompany early cancer pathogenesis and might be a useful substrate to identify targets for cancer interception.See related commentary by Hernandez-Barco et al., p. 2027.

Figure 1:

tSNE plots of all 3,343 cells from six lesions included in this study, (A) annotated by different tissue samples (B) annotated by unique cell types characterized by gene expression (Ductal Epithelium = Ep). (C) Feature plots demonstrating expression of specified genes among clusters on the tSNE. (D) Clustered heatmap of cell cycle characteristic coefficients per cell within the subpopulations indicated by the header demonstrating a greater proliferative state of PDAC derived cells.

Figure 2:

(A) Violin plots of gene expression across lesions types confirm expression of characteristic PDAC and cystic preneoplasia related markers. (B) Heatmap of the top 20 differentially expressed genes used to identify cell phenotypes across each of the ten discrete clusters indicated by the header. (C) Sankey diagram demonstrating epithelial cells profiled from LGD-IPMNs, HGD-IPMNS, and PDAC tissue and where they reside within annotated tSNE clusters (D) Correlation heatmap of Pearson correlation coefficients of hierarchically clustered individual cells across all lesions, identified by originating lesion type and tSNE cluster.

Figure 3:

(A) tSNE plot of all stromal cells that were virtually microdissection from entire lesions. Different colors represent annotation of unique cell phenotypes. (B) Proportion of cell phenotypes enriched in each lesion (PDAC, HG IPMN, and LG IPMN), colors refer to unique cell-types in (A) (C) Heatmap of the top 20 differentially expressed genes used to identify cell phenotypes across clusters. (D) Feature Plots demonstrating expression of specific genes among clusters to identify respective cell types.

Figure 4:

Schematic representation of evolving molecular and phenotypic signatures during preneoplastic progression of pancreatic cancer.