No single characteristic defines a tumour-suppressor gene, but important classical features include: loss-of-function mutations accompanied by loss of heterozygosity (or gene inactivation by epigenetic mechanisms such as methylation); mutation in inherited syndromes that predispose to cancer; somatic mutation in spontaneous tumours; and the ability to inhibit the growth of transformed cells in vitro.

Another optional, but popular, criterion in the era of homologous recombination is that mice with null mutations in the putative tumour suppressor show a predisposition to cancer that mirrors a human cancer syndrome. Haber and Harlow9 have suggested a simpler, operational definition of a tumour suppressor, which requires only that the gene often sustains loss-of-function mutations in the development of cancer. But this means that inactivating mutations must be unequivocally demonstrated within the gene itself — the deletion of a large chromosomal region harbouring a candidate suppressor gene is not enough.

At this early stage of the game, p73 has not yet lived up to either the classical standards for tumour suppressors or the kinder and gentler criteria of Haber and Harlow. But should an imprinted gene be exempted from these rules? The answer is, probably not. The current litmus test for any candidate tumour-suppressor gene is the demonstration of intragenic mutations within expressed alleles.