Abstract Pancreatic ductal adenocarcinoma (PDAC) is a lethal disease with a 5 year survival rate of 11%. PDAC arises from precursor lesions, the most common of which is pancreatic intraepithelial neoplasia (PanIN). Molecular analysis of human PanINs is critical to understand early pancreatic tumorigenesis, which could inform risk stratification, early detection, and cancer prevention approaches. Because PanINs are microscopic, it is challenging to determine their full extent and anatomic relationships using two dimensional histological sections. To determine the density and connectivity of PanINs, we utilized a novel machine learning algorithm (CODA) for three-dimensional (3D) reconstruction and cellular quantification. Large blocks of grossly normal pancreas were harvested from 39 surgical pancreatectomy specimens, followed by formalin fixation, paraffin-embedding, and complete serial sectioning. 3D reconstruction using CODA revealed striking multifocality of PanINs within the pancreata of most analyzed patients, with more than 600 spatially PanINs being modeled. Next, to determine whether these multifocal PanINs arose as independent neoplasms or via intraductal spread of a single PanIN, we assessed their clonal relationships using somatic multi-region DNA next generation sequencing (NGS). To do so, 8 additional blocks of grossly normal pancreatic tissue were harvested and 3D modeled, yielding 37 anatomically distinct PanINs for genomic analysis. Each spatially unconnected PanIN was separately microdissected in five different regions to assess both intra-PanIN and inter-PanIN genetic heterogeneity. 99 samples from the 37 PanIN lesions underwent targeted NGS using a panel covering major drivers of pancreatic ductal neoplasia. For PanINs with sufficient DNA, whole exome sequencing and deep sequencing of KRAS was also performed. Across the 8 blocks, 10 PanINs shared no mutations with other PanINs in the same block, indicating independent clonal origin. In addition, 8 PanINs shared only KRAS hotspot mutations with numerous other unshared mutations, likely indicating independent PanINs that shared hotspot mutations by chance. Six spatially separate low grade PanINs shared both driver and passenger mutations with at least one other spatially unconnected PanIN, demonstrating the ability of PanIN cells to dissociate from one lesion and establish unconnected, genetically related PanINs nearby. Furthermore, 5 PanINs harbored multiple KRAS mutations within a single PanIN, suggesting a polyclonal origin. One PanIN lacked any driver gene mutations. The genetic origins of the remaining 7 spatially separate PanINs could not be resolved due to a lack of discrete mutations found by targeted sequencing. Our 3D genomic analysis of anatomically distinct PanINs demonstrates that the unexpectedly large number of PanINs in normal pancreas most often arise independently, providing new foundations for our understanding of early pancreatic tumorigenesis. Determining the mechanisms for this multifocal neoplasia is an important direction for future research. Citation Format: Alicia M. Braxton, Ashley Kiemen, Rachel Karchin, Yuchen Jiao, Pei-Hsun Wu, Ralph H Hruban, Denis Wirtz, Laura D. Wood. Three-dimensional genomic analysis of human pancreatic intraepithelial neoplasia (PanIN) reveals striking multifocality and genetic heterogeneity [abstract]. In: Proceedings of the AACR Special Conference on Pancreatic Cancer; 2022 Sep 13-16; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2022;82(22 Suppl):Abstract nr PR007.
Read full abstract