Abstract A042: Clinical classification of cancer-associated fibroblast subtypes predicts prognosis and treatment response

Abstract Currently, it remains unclear how the knowledge of cancer-associated fibroblast (CAF) subtypes in pancreatic adenocarcinoma (PDAC) may apply to patients. Using a combination of sc- and bulk RNAseq methods, we present two CAF subtypes with permissive (permCAF) and restraining (restCAF) characteristics in patients that can be classified by a clinically usable single-sample classifier, DeCAF. DeCAF was trained in 4 independent datasets using penalized logistic regression and validated in 7 independent datasets with high reproducibility (AUC=0.961). We showed that DeCAF genes (n=18) were expressed in only CAF cells, but not immune cells or tumor cells using scRNAseq data, supporting that DeCAF is fibroblast specific. Patients with permCAF tumors had a median overall survival (OS) of 18.5 months compared to 30.2 months for restCAF tumors (p<0.001, log-rank test). To determine the association of DeCAF subtypes with the existing understanding of tumor-intrinsic subtypes in PDAC, we used PurIST to call the tumor subtypes in the same datasets. We found that a significantly greater proportion of basal-like tumors had permCAF subtypes (p<0.001, Fisher’s exact). While DeCAF (permCAF/restCAF) and PurIST (basal-like/classical) subtypes were independently associated with OS (p<0.001 & p=0.005, Cox regression), patients with basal-like and permCAF tumors showed the worst outcome with a median OS of 11.2 months; patients with classical and restCAF tumors showed the best outcome with a median OS of 30.2 months (p<0.001, log-rank test). Thus, we hypothesized that basal-like tumors may provide a favorable environment for tumor permissive fibroblasts (permCAF) or vice versa. To evaluate tumor-CAF crosstalk, we examined cell-cell communications mediated by ligand-receptor interactions using scRNAseq . We found that among all interactions between tumor and CAF cells, the highest frequencies of interactions were predicted between permCAF and basal-like cells. Furthermore, we found high permCAF enrichment surrounding basal-like enriched carcinoma loci in the spatial transcriptomics data. To determine if the CAF-tumor environment may favor different immune cells, we determined the immune landscape of bulk tumors using CIBERSORT. We found that the immune landscape of permCAF tumors were more immunosuppressive with enrichment of myeloid cells, and restCAF tumors were enriched in CD8+ T cells. In agreement with these findings, we found that patients with permCAF tumors had a better response to CCR2 inhibitor and patients with restCAF tumors had a better response anti-PD-L1 antibody approaches. In summary, DeCAF subtypes were independently prognostic and associated with treatment response. DeCAF provides a powerful tool to understand the tumor microenvironment landscape of bulk tumors in patients and may facilitate the design of future clinical trials. Citation Format: Xianlu L Peng, Elena V Kharitonova, Ian C McCabe, Changfei Luan, Ryan T Zhao, Priscilla S Chan, Arthi Hariharan, John R Leary, Ashley B Morrison, Joseph F Kearney, Hannah E Trembath, Gabriela Herrera, Ashley Cliff, Hong Jin Kim, Brian A Belt, Roheena Z Panni, David C Linehan, Jeffrey S Damrauer, Alina C Iuga, William Y Kim, Naim U Rashid, Jen Jen Yeh. Clinical classification of cancer-associated fibroblast subtypes predicts prognosis and treatment response [abstract]. In: Proceedings of the AACR Special Conference in Cancer Research: Advances in Pancreatic Cancer Research; 2024 Sep 15-18; Boston, MA. Philadelphia (PA): AACR; Cancer Res 2024;84(17 Suppl_2):Abstract nr A042.