Somatic mutation theory of cancer

The claim. Cancer is a disease of the genome of one cell. Mutations, whether from copying errors, chemicals, radiation or viruses, alter proto-oncogenes and tumour suppressor genes; a cell with enough of them escapes the controls on division and death and founds a tumour. Proliferation is not the default state of a body cell; it has to be unlocked by damage to specific genes.

Who and when. Boveri suggested in 1914 that abnormal chromosomes cause tumours; Muller showed in 1927 that X-rays cause mutations; Armitage and Doll showed in 1954 that the way cancer incidence rises with age fits a series of five or six rate-limiting steps; Knudson's 1971 study of retinoblastoma produced the two-hit model; Bishop and Varmus showed in 1976 that a viral oncogene was a captured cellular gene; Weinberg's laboratory isolated a mutated RAS from a human tumour in 1982; Fearon and Vogelstein's 1990 colorectal model ordered the steps from APC to KRAS to TP53. The name 'somatic mutation theory' was popularised by its critics, Sonnenschein and Soto.

Evidence for. Almost every carcinogen is a mutagen and mutational signatures now read the exposure history out of a tumour genome. Inherited mutations in single genes (RB1, TP53, BRCA1, APC, mismatch repair) cause hereditary cancer syndromes. Tumours are clonal and carry recurrent mutations in a few hundred genes. Drugs that block a mutated driver shrink tumours that carry it and not others (imatinib in BCR::ABL1 leukaemia, osimertinib in EGFR-mutant lung cancer). Tomasetti and Vogelstein reported in 2015 that the lifetime risk of cancer across tissues correlates with the number of stem cell divisions in each tissue, as expected if copying errors drive risk. Oncogenic viruses that disable p53 and RB directly show the same brakes can be released without mutation.

Evidence against and limits. Martincorena's sequencing of normal sun-exposed skin (2015) and normal oesophagus (2018) found them colonised by clones carrying canonical driver mutations without any cancer, so mutations are common and cancer is rare. Some cancers, especially in children, carry very few mutations (posterior fossa ependymoma has no recurrent point mutations), and tumours can be initiated by histone or chromatin changes. Transplantation experiments show malignant cells can be normalised by a normal tissue context (Mintz and Illmensee 1975). The theory says little about why incidence rises so steeply with age when mutation accumulation is roughly linear.

Predictions that held or failed. Held: matching a drug to a driver produces responses across tumour types; germline testing predicts risk in families; mutational signatures track smoking, ultraviolet light and defective repair. Failed or unfulfilled: mutation burden alone does not predict who gets cancer; the hope that sequencing would reveal one targetable driver in every tumour was not met; targeted monotherapy is rarely curative because of clonal evolution.

Therapies that came from it. The whole of targeted therapy, comprehensive genomic profiling and companion diagnostics, hereditary cancer testing and risk-reducing surgery, synthetic lethality (PARP inhibitors in BRCA-mutant cancers), and tumour-agnostic approvals by mutation. It feeds the driver and passenger model and clonal evolution, and is the theory that the tissue organisation, epigenetic, metabolic, aneuploidy and bioelectric theories were each framed against.

Status: established. It remains the working framework of oncology, read today as necessary but not sufficient: mutations start most cancers, and the epigenetic, tissue-level, evolutionary and immune theories explain why most mutated cells never become one.