AN EPIGENETIC CLASSIFICATION OF PANCREATIC DUCTAL ADENOCARCINOMA
Victoria G. Aveson*1,2, Rohit Chandwani1
1General Surgery, New York Presbyterian Hospital - Weill Cornell Medical Center, New York, NY; 2General Surgery, Memorial Sloan Kettering Cancer Center, New York, NY
Background
We hypothesized that a novel epigenetic interrogation of primary human pancreatic cancer could reveal a novel classification of disease predictive of clinical outcomes. Recent research suggests epigenetic changes underlie major steps in pancreatic cancer progression, however study of the epigenetics of human tumors has been limited by the low cellularity of pancreatic tumors and storage methods incompatible with most common methods of epigenetic analysis. As a result, while subtypes of pancreatic cancer have been identified based on gene expression and genomic information, epigenetic subtypes of human tumors have not been well-characterized or correlated directly with clinical outcomes.
Methods
We employed genome-wide analyses of chromatin accessibility and RNA-sequencing to archival frozen normal and pancreatic tumor tissue. Fash frozen samples from tumor and normal pancreas of patients with lesions resected at Memorial Sloan Kettering Cancer Center in 2005 were obtained. Nuclei were isolated by density gradient. Aneuploid nuclei were identified with flow cytometry to enrich for tumor DNA. Tagmentation and amplification of ATAC libraries for sequencing were performed per our optimized Omni-ATAC protocol. Further enhancement of tumor signal was achieved with computational correction with libraries from matched normal pancreas samples.
Clinical data for all patients was evaluated for association between chromatin accessibility subtype and oncologic outcomes.
Results
85 patients were included with a median age of 67 years and 37% were female. The majority of tumors (90.3%) were stage II at resection and moderately differentiated (54.2%) or poorly differentiated (31.3%). All lesions were pancreatic adenocarcinoma or intraductal papillary mucinous neoplasm. Median survival was 722 days and mean survival was 1133 days.
Nuclei were successfully isolated from flash frozen samples and viable chromatin accessibility libraries were produced from frozen tissue.
Distinct epigenetic signatures were defined, identifying subtypes of disease that do not recapitulate the Collison or Bailey schema, nor the TCGA or Moffit basal/classical subtypes. An association between subtype and risk of early recurrence was identified, suggesting that integration of chromatin accessibility data improves the clinical utility of genomic and transcriptional information
Conclusion
In our analysis, we developed and refined an epigenetic classification of pancreatic ductal adenocarcinoma using archival snap-frozen tissue. Amplification of tumor signal can be achieved with flow cytometry and computational means, which eliminates the need for costly and potentially confounding xenografts or microdissection. Our epigenetic classification offers information that, when integrated with genomic and transcriptomic characterization, can be predictive of clinical outcomes
Back to 2019 Abstracts
