A tumor suppressor gene is identified by genetic changes that inactivate the function of the gene. The Adenomatous polyposis coli (APC) gene was first localized by the heterozygous Herrera deletion carried in a human familial polyposis kindred.1 Molecular analyses in the human and in animal models are summarized by the “gatekeeper” metaphor: loss of function of both alleles of the autosomal APC/Apc locus is necessary for colon cancer. In our recent quantitative study in murine models, published in The Proceedings of the National Academy of Sciences of the United States of America, both genetic and epigenetic changes have been identified, inactivating the Apc gatekeeper.2 A number of investigators are actively addressing issues of the stability, reversibility and environmental sensitivity of the genetic and epigenetic mechanisms affecting colonic and other cancers. We limit this commentary to identifying issues arising from our study. The Min mouse and Pirc rat genetic models for early familial colon cancer display greater genomic stability than many advanced human colon cancers.3 As in many human cases, the loss of function of the wild-type Apc allele from the mutant heterozygote involves irreversible genetic change: intragenic mutation in the wild-type Apc allele or conservative somatic recombination.4,5 By contrast, epigenetic changes involve a spectrum of mechanisms, each reversible in principle. Developmental “programmed” epigenesis is commonly implemented by the spatial and temporal elaboration of transcription factors that act in trans to control the activity of both alleles of target genes, including the synthesis and modification of specific chromatin proteins.6 “Stochastic” epigenetic change, such as the initiation of X inactivation in diploid eutherian females, acts in cis to control the activity of only one of the alleles of a target gene, chosen at random. For both programmed and stochastic epigenetic change, fundamental issues include mitotic transmission, reversibility and causative role in the hierarchy of biological change. These issues are illuminated by studies of gene silencing in S. cerevisiae, an organism deficient in cytosine methylaton.7 Single-cell monitoring demonstrates monoallelic silencing. Both monoallelic and biallelic silencing show measurable rates of reversion. The issues of mitotic transmission and reversibility come into sharp focus as the spectrum of molecular epigenetic mechanisms becomes defined. For example, X inactivation is observed over a series of molecular events: successive modification of the inactive X-chromosome by the non-coding RNA XIST, histone modification and symmetric methylation of CpG dyads.8 Female embryonic stem cells expressing XIST show random but unstable monoallelic X inactivation with finite reversion rates.9 We have used a quantitative Pyrosequencing assay to detect changes in the tumor genome. Analysis of allelic ratios in cDNAs from F1 hybrid genetic constructs expands the dynamic range to detect monoallelic silencing regardless of mechanism. Precise measurement of allelic ratios permits a quantitative assessment of the extent to which a tumor contains cells with a range of genetic or epigenetic states. Our study identified four ways in which the Apc gatekeeper function is affected in colonic tumors: the conservative pathway, with mitotic recombination usually limited to the chromosome carrying the Apc locus; a haploinsufficient pathway, in which the Apc locus is expressed at normal total levels but with half of the transcripts carrying the germline mutation in Apc; a stochastic silencing pathway, in which only the mutated Apc allele is expressed; and biallelic silencing of the Apc locus (Fig. 1). In principle, the haploinsufficient pathway must involve further events elsewhere in the genome or epigenome, since only a finite number of tumors form in the heterozygous animals. Use of F1 analysis of the tumor genome and epigenome complements that provided by mutation detection to create a comprehensive understanding of the above four ways by which the Apc gatekeeper function is affected in colon cancer. Figure 1 Familial colon cancer in humans and in the Pirc rat and Min mouse kindreds arises in heterozygotes for a mutation (red star) in the APC/Apc gatekeeper gene. Tumors arise when the remaining wild-type allele of APC/Apc loses its function. This can involve ... The cDNA assay for silencing can detect states that may be less stable than that involving the methylation of CpG dyads. It is interesting to ask, what level of mitotic stability of an epigenetic state is necessary to support the sub-exponential growth of a neoplasm? What environmental signals affect this stability? Does the estimated level of admixture in early tumors represent polyclonality,10 or does it represent a spectrum of epigenetic states in dynamic equilibrium?8 If stochastic silencing occurs over the entire genome, including genes for which biallelic silencing is detrimental to the tumor, can over-silencing lead to tumor stasis or regression? Recent quantitative studies of the growth profiles of early adenomas in Min mice show that a tumor can go through successive cycles of growth, stasis and partial regression.11 Does the longitudinal fate of a colonic tumor depend on the way in which the gate is opened? Overall, this report raises fundamental biological questions and provides a strategy to begin addressing these questions for the tumor, its genome and its epigenome.