Immune surveillance and cancer immunoediting

The claim. Immune surveillance: lymphocytes continuously recognise and eliminate nascent transformed cells, so clinical cancer is a failure of immunity. Immunoediting (Dunn, Old and Schreiber 2002) refines this into three phases: elimination, in which most transformed cells are destroyed; equilibrium, in which immunity holds a tumour in check for years while selecting less immunogenic variants; and escape, in which edited tumours grow out by losing antigens or antigen presentation, recruiting suppressive cells, or expressing checkpoint ligands. The immune system therefore both protects against cancer and shapes the cancers that emerge.

Who and when. Ehrlich 1909; Burnet 1957 and 1970 and Thomas 1959 stated the surveillance hypothesis. Stutman's 1974 finding that athymic nude mice were no more prone to chemically induced tumours seemed to refute it. Shankaran, Old and Schreiber showed in 2001 that mice lacking lymphocytes or interferon-gamma signalling develop more tumours and that tumours from immunodeficient mice are more immunogenic, which is the signature of editing. Dunn and colleagues named immunoediting in 2002; Schreiber, Old and Smyth reviewed it in 2011. Allison's 1996 CTLA-4 blockade and Honjo's PD-1 discovery gave it its therapy; Chen and Mellman's cancer-immunity cycle (2013) is its clinical map.

Evidence for. Immunodeficient mice develop more spontaneous and induced tumours. Organ transplant recipients on immunosuppression have higher rates of many cancers, not only virus-driven ones. Tumour-infiltrating lymphocytes predict survival across cancers, and tumours show evidence of neoantigen depletion and loss of HLA consistent with editing. Checkpoint inhibitors produce durable remissions in melanoma, lung, kidney, bladder and mismatch repair-deficient cancers, and responses track with mutational burden and neoantigen load. Virus-driven tumours, rich in foreign antigens, respond well.

Evidence against and limits. Nude mice have natural killer cells, which explains Stutman's result, but the episode showed how hard surveillance is to demonstrate. Most cancers arise in people with normal immunity, and the immune system also promotes cancer through inflammation. Equilibrium is difficult to observe directly in humans. Immunotherapy fails in most patients with pancreatic, prostate and brain cancers, and the reasons are only partly understood.

Predictions that held or failed. Held: releasing checkpoints unleashes anti-tumour immunity; immunogenic tumours have lower mutational burden after editing; adoptive transfer of tumour-reactive T cells can cure. Failed: most therapeutic cancer vaccines of the 1990s and 2000s; the expectation that boosting immunity non-specifically (interferon, interleukin-2 in most settings) would be broadly effective.

Therapies that came from it. Checkpoint inhibitors (ipilimumab, nivolumab, pembrolizumab and successors), CAR-T cells, tumour-infiltrating lymphocyte therapy (lifileucel), T-cell engagers, neoantigen mRNA vaccines and the use of mutational burden and mismatch repair deficiency as predictive biomarkers. It applies clonal evolution to the immune system as predator, draws on the microenvironment view for the mechanisms of escape, and entered the hallmarks in 2011.

Status: established. Immunoediting is the accepted account of how tumours and immunity co-evolve, and it underpins the most important therapeutic advance of the last two decades.