Tom Luedde

Microsatellite instability determines whether patients with gastrointestinal cancer respond exceptionally well to immunotherapy. However, in clinical practice, not every patient is tested for MSI, because this requires additional genetic or immunohistochemical tests. Here we show that deep residual learning can predict MSI directly from H&E histology, which is ubiquitously available. This approach has the potential to provide immunotherapy to a much broader subset of patients with gastrointestinal cancer. A deep residual learning framework identifies microsatellite instability in histology slides from patients with cancer and can be used to guide immunotherapy.

Abstract Hepatocellular carcinoma (HCC) can have viral or non-viral causes 1–5 . Non-alcoholic steatohepatitis (NASH) is an important driver of HCC. Immunotherapy has been approved for treating HCC, but biomarker-based stratification of patients for optimal response to therapy is an unmet need 6,7 . Here we report the progressive accumulation of exhausted, unconventionally activated CD8 + PD1 + T cells in NASH-affected livers. In preclinical models of NASH-induced HCC, therapeutic immunotherapy targeted at programmed death-1 (PD1) expanded activated CD8 + PD1 + T cells within tumours but did not lead to tumour regression, which indicates that tumour immune surveillance was impaired. When given prophylactically, anti-PD1 treatment led to an increase in the incidence of NASH–HCC and in the number and size of tumour nodules, which correlated with increased hepatic CD8 + PD1 + CXCR6 + , TOX + , and TNF + T cells. The increase in HCC triggered by anti-PD1 treatment was prevented by depletion of CD8 + T cells or TNF neutralization, suggesting that CD8 + T cells help to induce NASH–HCC, rather than invigorating or executing immune surveillance. We found similar phenotypic and functional profiles in hepatic CD8 + PD1 + T cells from humans with NAFLD or NASH. A meta-analysis of three randomized phase III clinical trials that tested inhibitors of PDL1 (programmed death-ligand 1) or PD1 in more than 1,600 patients with advanced HCC revealed that immune therapy did not improve survival in patients with non-viral HCC. In two additional cohorts, patients with NASH-driven HCC who received anti-PD1 or anti-PDL1 treatment showed reduced overall survival compared to patients with other aetiologies. Collectively, these data show that non-viral HCC, and particularly NASH–HCC, might be less responsive to immunotherapy, probably owing to NASH-related aberrant T cell activation causing tissue damage that leads to impaired immune surveillance. Our data provide a rationale for stratification of patients with HCC according to underlying aetiology in studies of immunotherapy as a primary or adjuvant treatment.