In 1994, as cited in the same issue of Science, Robin Holliday voiced a commonly apprehended drift in meaning, and redefined epigenetic as "Nuclear inheritance which is not based on differences in DNA sequence." These two memes are freely circulating and can cause muddle or mischief mainly when they recombine, namely when epigenetic-H is automatically applied to epigenetic-W.
For example, epigenetic-H might be taken to imply that the developmental modulation of cell phenotype is always nuclear but not sequence-related (what I will suggest be called epinucleic). But there are many epigenetic-W (read developmental) processes that do depend on sequence-related changes, such as cellular senescence following upon telomere-shortening, or immunocyte diversification from promiscuous recombination and mutagenesis. Cytoplasmic heredity is still contentious, but it has strong foundations in studies of protozoa. If nuclear in epigenetic-H means chromosomal, then we have innumerable exceptions in the role of mitochondria and other endocellular plasmids and symbionts, which indubitably play roles in phenotypic development, sometimes even to the point of its extinction, not to mention sex determination.
Conversely, genetic transmission through the germline may be subject to imprinting and other epinucleic phenomena. So there are many traps for the unwary, especially the presumption that epigenetic-W is never to be traced back to nucleic sequence changes, the dogma of the invariance of somatic cells. That dogma greatly delayed the formulation and acceptance of the clonal selection theory of immunity; and who knows what else may be impeded by it?
In 1958, epigenetic was already a semantic morass. At a conference on somatic cell variation,2 I attempted to "define a category of genetic information as being nucleic; that is, depending on the sequence of nucleotides in a nucleic acid. By contrast, epinucleic information is expressed in another form; that is, as an aspect of nucleic acid configuration other than nucleotide sequence [e.g.,] polypeptide or polyamine adjuncts to the polynucleotide. We also have extranucleic information in molecules or reaction cycles not directly connected with nucleic acid. In accord with Ephrussi's suggestion, we might propose that nucleic information has the pervasiveness and static precision connoted by genetic, whereas the epinucleic information regulates the manifestation of nucleic potentialities in the dynamic, temporally responsive functioning of actual development." (I knew nothing of methylation in those days; it would be prototypically epinucleic.)
This neology of nucleic, epinucleic, extranucleic, has attracted few followers, I think largely because so few people had really thought through the distinctions. There is much merit in Ben Johnson's caution about unbridled proliferation of terms: "A man coins not a new word without some peril, and less fruit; for if it happen to be received, the praise is but moderate; if refus'd, the scorn is assur'd." But is a polysemy to be preferred, with thought-muddling as a further peril?
1. C.H. Waddington, Endeavor, 1:18, 1942.
2. J. Lederberg, "Genetic approaches to somatic cell variation: summary comment," Journal of Cellular and Comparative Physiology, Suppl. 1, 52:383-402, 1958.