Complexities of cell differentiation

In his Reference Frame (Physics Today, March 2009, page 8 ), Leo Kadanoff discussed how the function of biological systems could be represented in terms of computer models of their control networks. In that discussion he invoked a 1969 paper in which Stuart Kauffman speculated that different limit cycles of a nonlinear biological cellular network might represent different cell types—a liver cell, the skin on your nose, and so on—so that even though the sequence of nucleotides in the DNA of each cell is identical, different cell types can develop.

However, it’s important for readers of Physics Today to understand that developmental biology has seen tremendous progress in figuring out what machinery drives cellular differentiation. It turns out that Nature developed a fascinating mechanism for differentiating cell types which involves attaching a variety of covalent epigenetic markings such as acetylation of the histone proteins that support the DNA in chromosomes, or methylation of the nucleotides themselves, at specific sites on the cellular DNA of eukaryotes. By that mechanism, the DNA of different cell types becomes modified—without its sequence being changed—in ways that control the expression of the 20 000—30 000 or so genes encoded in human DNA.

An equally significant discovery, which arose out of the human genome project, is that only about 2% of the human genome codes for proteins. Accumulating evidence shows that 50%—80% or maybe more of human DNA, the so-called noncoding type, or ncDNA, is transcribed into RNA molecules of various sizes, many of them quite short (as few as 22 base pairs). And it appears that the whole business of epigenetic marking may be controlled by those various RNAs. ¹ Thus the expression of specific genes, which needs to be highly controlled as to timing and position in cells, is probably under the control of the “playbook” represented by the vast amount of genetic information embedded in the ncDNA. ²

An even more fascinating possibility is that human memory may also be controlled by epigenetic labeling of DNA. ³ Memory-inducing events are known to stimulate protein synthesis in the cell body of neurons, a process associated with synapse formation. It is also conceivable that neuronal gene expression is associated with RNA control of epigenetic markings. ⁴ Such a hypothesis would help explain the fact that consolidated memories can last a lifetime in humans, and maybe in tortoises and elephants as well.

Epigenetic labeling could also help account for the huge multiplicity of memory states. If one thinks of marking each of just 25 sites of neuronal DNA with one of three epigenetic markers, one could define a different state of that neuron corresponding to every millisecond of a human lifetime.

So it appears that Nature has come up with a much more robust set of tools for eukaryotes than the simplified nonlinear dynamical networks beloved of physicists. However, bacteria are much simpler systems, having only about 1% of their DNA non-coding. So bacterial metabolism may well be simple enough to be a jumping-off point for understanding complex networks.