SciAm Dec. 2003 - DNA, Epigenetics, and Complexity

["Tim Gwinn" <***>]

The December 2003 issue of Scientific American contains an article entitled "The Unseen Genome: Beyond DNA" by W. Wayt Gibbs. The article begins:

    "DNA was once considered the sole repository of heritable information. But biologists are starting to decipher a separate, much more malleable layer of information encoded within the chromosomes. Genetics, make way for epigenetics." [p. 106]

The "epigenetic" factors described are "chemicals and proteins that surround and stick to DNA" [p. 108]. Some of these factors are apparently known to even be passed from parent to child. These factors are not all stable like the sequences in DNA. Instead, some of them "are in constant flux", according to the article. As these factors change, they appear to (at the least) alter expressibility of various regions of DNA. In effect, they alter what constitutes active sites of the genome at any given moment.

 

The article attempts to implicitly maintain the mechanistic view, and in particular, to maintain the computer metaphor, where genome = software. The article ends:

    "The new view of the genomic machine is energizing, because it opens avenues to genomic engineering. Those 30,000-odd protein-coding genes, so important yet so immutable, are not the only instruction set to which cells refer. Noncoding DNA matters. Chemical attachments to DNA and to the histones matter. The shape of chromatin matters. And all these are subject to manipulation. "There is a whole new universe out there that we have been blind to," Bestor says. "It is very exciting."" [p. 113, bold added]

The assertion being that manipulation of this more inclusive "instruction set" which ostensibly comprises genome will then be  dutifully implemented via the cell's "transcription machinery" [p.113]. But can this computer metaphor hold given the evidence they present?

 

One of the key questions raised by Rosen in Essays on Life Itself, Chapter 1 is: can a fixed, inert substance (such as DNA) be equated with genome, since genotype is what forces phenotype? Rosen's answer is, briefly, that:

"...the question is much more context dependent that that; its answer involves not just inherent properties of a "molecule" in itself (e.g., "aperiodicity"), but also the properties of what system is being forced, and the preceding levels of forcing (of which "genome" is to be the last). 

    "Thus, very little remains of Schrodinger's simple cyptographic picture of "order from order," in which rigid molecular structures get transduced somehow into nonrigid phenotypic ones. Rather, the initial "order" appears as a pattern, or graph, of interpenetrating constraints, which determines what happens, and how fast it happens, and in what order, in an underlying open system.  The arrows in such graphs, which I suggest constitute the real "aperiodic solid," are operators; they express "gravitational" effects on the underlying system. In terms of inertia, it is much more appropriate to speak of active sites rather than of molecules. The two are not the same." [EL 24-25]

The SciAm article essentially describes that what will constitute the "active sites" are constantly in flux, due to "many epigenetic marks being constantly in flux"[p. 113]. Clearly, this "flux" is not some random occurrences, but rather constitutes a definite behavior: a phenotype of the system. Therefore, being a phenotype, it is forced - at least in part - by genotype. Yet, at the same time, this "flux" of active sites is also identified as being part of the genome. So, we have an impredicativity where phenotype and genotype are intertwined.

 

In other words, as Rosen asserted long ago, genotype is not fractionable from phenotype in an organism. Co-extensive with that result is the result that a computer software metaphor of genome is inapplicable. As Rosen also notes, this situation is highly context-dependent: it is (at least) the dynamic, changing relationships occurring between DNA and the various epigenetic factors which constitute genome. Fractionating these epigenetic factors from each other and/or from DNA breaks these relationships, removing the information (e.g., the "interpenetrating constraints" Rosen mentions) contained in them. As the last paragraph of the article indicates, these relationships "matter".

 

It is entirely unclear to me what might result by tinkering with these factors. Simple, mechanistic, models will certainly not provide the necessary predictive answers to these questions about this Rosennean complex system. Conservatively, such tinkering might 'only' alter what constitutes the active sites of the genome. At first glance, that would limit the dangers (if one considers such a danger as 'limited'!) of such engineering to the maximal range of permutations of active sites allowable by a given set of (DNA + epigenetic factors). 

 

I suspect, however, that such permutations would likely not result in single, fixed changes in the organism. Unlike the tinkering with the rigid holonomic constraints of DNA, tinkering with some of these epigenetic factors will likely alter nonholonomic constraints in the genome. If we take into account Rosen's view of genome as a pattern of "interpenetrating constraints", then alteration of some nonholonomc constraint(s) of the genome would seem to me to lead to not only marked alterations what can (or will be) active site(s); but, rather, given the nature of nonholonomic constraints, of the sequences and rates of what can (or will) become active site(s) and would likely lead, therefore, to possibly very marked alterations in the organism as a whole. In other words, the variations are not limited to a maximal range of permutations of active sites allowable by a given set of (DNA + epigenetic factors); but rather, to a maximal range of permutations of both sequences and rates of change of a maximal range of permutations of active sites allowable by a given set of (DNA + epigenetic factors).

 

To me, this not only raises questions of unpredictability of dangers, but also questions about tractability and utility of such approaches for therapeutic purposes: I wonder about reversibility (i.e., therapy) with regard to these nonholonomic constraints.

 

Perhaps, though, the effects of such tinkering will not be limited to 'only' altering what constitutes active sites in the genome. For example, the article refers to "transposons" which are genes that, when active, can alter the DNA sequence itself by copying itself into distant regions of DNA. So, perhaps the effects of altering epigenetic factors will be even more pervasive, in which case it is very difficult, to say the least, to predict the consequences and possible dangers (as well as possible benefits) of such tinkering or engineering.

 

Regards,

Tim