Objective Pancreatic ductal adenocarcinoma (PDA) has among the highest stromal fractions of any cancer and this has complicated attempts at expression-based molecular classification. The goal of this work is to profile purified samples of human PDA epithelium and stroma and examine their respective contributions to gene expression in bulk PDA samples.
Design We used laser capture microdissection (LCM) and RNA sequencing to profile the expression of 60 matched pairs of human PDA malignant epithelium and stroma samples. We then used these data to train a computational model that allowed us to infer tissue composition and generate virtual compartment-specific expression profiles from bulk gene expression cohorts.
Results Our analysis found significant variation in the tissue composition of pancreatic tumours from different public cohorts. Computational removal of stromal gene expression resulted in the reclassification of some tumours, reconciling functional differences between different cohorts. Furthermore, we established a novel classification signature from a total of 110 purified human PDA stroma samples, finding two groups that differ in the extracellular matrix-associated and immune-associated processes. Lastly, a systematic evaluation of cross-compartment subtypes spanning four patient cohorts indicated partial dependence between epithelial and stromal molecular subtypes.
Conclusion Our findings add clarity to the nature and number of molecular subtypes in PDA, expand our understanding of global transcriptional programmes in the stroma and harmonise the results of molecular subtyping efforts across independent cohorts.
- pancreatic cancer
Statistics from Altmetric.com
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.
CM, SRH and JH contributed equally.
Contributors KPO, AC and MB: conceptualisation. HCM and JH: computational analysis. JH, PL, JZ and MB: software. HCM, SRH, JH, JG, ACI, ARS, PEO and KPO: investigation. JAC, HH, AA, TS, KPO and AC: resources. HCM, JH and KPO: visualisation. HCM, KPO and AC: funding acquisition. KPO, AC and MB: project oversight and management. HCM and KPO: wrote the manuscript with feedback from SRH, JH and AC. All authors discussed the results and commented on the manuscript.
Funding This work was supported by the National Cancer Institute (NCI) Cancer Target Discovery and Development program (U01CA217858 to AC), NCI Research Centers for Cancer Systems Biology Consortium (1U54CA209997 to AC and KPO), NCI Outstanding Investigator Award (R35CA197745-02 to AC), NCI Cancer Center Support Grant (3 P30 CA13696-40) and NCI Research Project Grant (R01CA157980 to KPO). Financial support was also provided by the Columbia University Pancreas Center. HCM. received support from a Mildred Scheel Postdoctoral Fellowship (Deutsche Krebshilfe). PEO received support from the NIH NCATS (KL2TR001874).
Competing interests AC is a founder and shareholder of DarwinHealth Inc. and a member of the Tempus Inc. SAB and shareholder. Columbia University is a shareholder of DarwinHealth Inc. KPO is a member of the SAB for Elstar Therapeutics.
Provenance and peer review Not commissioned; externally peer reviewed.
Data sharing statement All data from this study are included in the manuscript or publicly available.
Correction notice This article has been corrected since it published Online First. The funding statement and first author’s name have been corrected.
Patient consent for publication Not required.