40: Spatial Transcriptomics Reveals Molecular Subtypes of Pancreatobiliary IPMN

Intraductal papillary mucinous neoplasms (IPMN) occur in 10% of the population, but only a small minority progress to pancreatic duct adenocarcinoma (PDAC). The lack of accurate predictors of high-risk disease can lead to missed diagnoses of PDAC and unnecessary surgeries. Previously, digital spatial RNA profiling (DSP-RNA) of IPMN showed that pancreatobiliary (PB) subtype closely mirrored PDAC. Here, we present a spatial transcriptomic characterization of the progression of PB-IPMN from normal-appearing ducts to carcinoma.


Methods:

NanoString GeoMx profiling of patients with PB-IPMN was conducted using Whole Transcriptome Atlas (WTA) probes (Figure 1A). Epithelial (PanCK+) regions of interest (ROIs) were annotated as normal duct (NL), low-grade dysplasia (LGD), high-grade dysplasia (HGD), or invasive carcinoma (INV). Illumina sequencing and post-processing produced count data that was analyzed using R/Bioconductor.


Results:

Principal component analysis (PCA) distinguished ROIs by specimen pathology (HGD/INV versus LGD/NL) on the first PC and isolated a group of NL ROIs on the second PC (Figure 1B). Unsupervised clustering identified three clusters (C1, C2, C3): C1 exclusively contained ROIs from INV specimens, C2 was enriched with LGD ROIs, and C3 predominantly contained NL ROIs. Differential expression analysis yielded cluster-specific marker genes: C1 featured known PDAC genes including tripartite motif containing 29 (TRIM29), C2 markers included progastricsin (PGC) and Mucin 6 (MUC6), and C3 contained normal ductal cell markers such as cysteine rich secretory protein 3 (CRISP3) (Figure 1C). Notably, C2 and C3 exhibited high levels of exocrine enzymes, suggesting a connection to acinar-ductal metaplasia. An overlay of marker genes CRISP3, PGC, and TRIM29 confirmed distinct patterns among the clusters (Figure 1D). Expression patterns were concordant with IPMN clusters and PDAC tumor subtypes: C1 expressed the “basal-like” program, C1/C2 both expressed the “classical” program, and C2/C3 expressed genes from the “exocrine” subtype (Figure 1E).


Conclusions: Here, DSP-RNA analysis of PB-IPMN revealed molecular heterogeneity within regions of similar morphology. Subtyping PB-IPMN into three groups based on expression profiles more clearly explained this variability, suggesting a hypothetical model of PB-IPMN progression (Figure 1F). Our proposed PB-IPMN subtypes strongly correlated with well-studied PDAC tumor subtypes. Our results motivate future work to understand the polarization of PB-IPMN into indolent and high-risk tumors.