Immune landscape of pancreatic neuroendocrine tumours (PanNETs)

Abstract

Background: To date little is known about the immune landscape of PanNETs and if immunotherapy could play a role in their treatment. We previously identified 3 molecular subtypes in PanNETs: Metastases like primary (MLP), intermediate and insulinoma like tumours (PanNETassigner signature, Sadanandam Cancer Discovery 2015). Here we sought to profile the immune architecture of 48 PanNET patient samples across these subtypes. Methods: Patients were recruited by the ARC-Net Research Centre Verona, within an ethically approved protocol. Quality RNA was isolated from fresh frozen samples for immune profiling using microarrays, nCounter platform (Nanostring Technology) and RNAseq (Illumina). CIBERSORT analysis was performed to assess immune cell enrichment. Results: 48 PanNET samples were classified using the PanNETassigner gene signature. Based on immune expression profile analysis, tumours were divided into two categories, immune high or immune dormant. The majority of the MLP subtype were immune high, whereas most of the insulinoma and intermediate samples were immune dormant. A small number of insulinoma samples were immune high reflecting the heterogeneity of this tumour. Within the MLP subtype there was increased expression of CD8B, LAG3, CD38, CXCL10, CXCL9, CCL19, CD28 and CD27 compared to the other subtypes. Some of these genes are associated with chronic infection (CD38, CXCL10) whilst others are markers of T cell exhaustion (LAG3). This pattern is consistent with CIBERSORT analysis conducted using microarray data on an overlapping cohort of PanNET samples, where the MLP subtype was associated with increased levels of infiltrating T cells but also an increase in exhausted CD8+ve T cells. PD1 was highly expressed in 2/15 MLPs. PDL1 expression was heterogeneous in MLP but high in 7/13 insulinomas. FOXP3 was highly expressed in a subset of the MLP samples (7/16). Conclusions: We have demonstrated the differential expression of immune related genes across 3 known PanNET subtypes. The MLP subtype appears to be associated with an immune profile similar to that seen in chronic infection with increased T-cell exhaustion. Such detailed profiling is essential to inform patient selection approaches for immunotherapy and rational immunotherapy combinations for panNETs in the future.