Gastroenteropancreatic neuroendocrine neoplasms (GEP-NENs) are a rare and heterogeneous subgroup of tumors with a challenging management because of their extremely variable biological and clinical behaviors. Due to their different prognosis, there is an urgent need to identify molecular markers which would enable to discriminate between grade 3 neuroendocrine tumors (NETs) and neuroendocrine carcinomas (NECs), despite both being diagnosed mainly on the basis of proliferation index and cell differentiation. DLL3, a negative Notch regulator, is a promising molecular target highly expressed in several tumors with neuroendocrine features. We conducted a retrospective analysis of DLL3, RB1, and PD-L1 expression by immunohistochemistry (IHC), in formalin-fixed, paraffin-embedded (FFPE) samples from 47 patients with GEP-NENs. Then, we correlated the results with patients' clinical features and outcome. The absence of DLL3 expression in 5 well-differentiated GEP-NETs with high-grade features (G3 NET), and the presence of DLL3 in 76.9% of poorly-differentiated NECs (G3 NEC), highlights DLL3 expression as a marker of G3 NECs (p = 0.007). DLL3 expression was correlated with RB1-loss (p < 0.001), negative 68 Ga-PET/CT scan (p = 0.001), and an unfavorable clinical outcome, with important implications for treatment response and patient's follow-up. Median progression-free survival (PFS) and overall survival (OS) were 22.7 months (95% CI 6.1-68.8) and 68.8 months (95% CI 26.0-78.1), respectively, in patients with DLL3-negative tumor compared with 5.2 months (95% CI 2.5-18.5) and 9.5 months (95% CI 2.5-25.2), respectively, in patients with DLL3-positive tumor (PFS p = 0.0083, OS p = 0.0071). Therefore, combined with morphological cell analysis, DLL3 could represent a valuable histological marker, for the diagnosis of poorly differentiated NECs. The high percentage of DLL3 expression in NEC patients also highlights a potential opportunity for a DLL3 targeted therapy in this tumor subset.
Keywords: DLL3; GEP-NEN; NEC; Prognostic marker.