Abstract C034: Bioinformatical identification of cancer-associated-fibroblasts with tumor cells pro-migratory crosstalks depending on activin

Pancreatic ductal adenocarcinoma (PDAC) aggressiveness results from its high metastatic and chemoresistance potential. Pancreatic cancer cells are immersed in an exuberant tumor microenvironment (or stroma), which represents up to 80% of PDAC tumor volume, and is hijacked by cancer cells for their own survival and metastasis. Additionally, PDAC tumors are heterogeneous, e.g. transcriptomic (RNAseq) analyses of human PDAC tumors have allowed patients to be stratified into two tumor subgroups (basal and classical) with different prognoses. Therefore, we aimed to understand which tumor-stroma crosstalks are critical for PDAC aggressiveness in the context of its heterogeneity, and how they can be pharmacologically targeted, to then propose more effective personalized therapeutic strategies. By performing bioinformatical analysis on RNAseq data from PDX (Patient Derived Xenograft), a hybrid tumor model from which human-derived tumor cell sequences are distinguishable from murine-derived stromal cell sequences, we searched for different stromal behaviors linked to the tumor aggressiveness, to be then validated in published PDAC patient databases. According to NMF (Non-negative matrix factorization) and GSEA (Gene Set Enrichment Analyses), we identified two components, which classify the samples according to a stromal gradient. Interestingly, these components are characterized by distinct functional signatures of aggressiveness and they are highly prognostic (correlated with survival). In addition, these components stratify patients independently on the described molecular tumor (basal or classical) classification. Importantly, we validated those results on other PDAC published databases. Moreover, ligand-receptor bioinformatics analyses identified novel players of the TGFβ superfamily in the most aggressive stromal component. Indeed, we have highlighted the involvement of the ligand activin a member of this large superfamily. Using in situ hybridization coupled with immunofluorescence, we were able to demonstrate that activin is only expressed by stromal cells, and more specifically by cancer associated fibroblasts (CAF), with however a strong heterogeneity (inter and intra-tumor). In vitro, activin strongly increases the migratory capacity of fibroblasts but also of specific cancer cells, this effect being effectively blocked pharmacologically with an activin receptor-targeting drug that is now tested in a pre-clinical PDAC murine model. Therapeutic targeting o