MAKIN' BRAINS

Dick Pountain/Wed 16 March 2005/10:56 am/Idealog 128

The Massachusetts Institute of Technology isn't much given to academic gimmickry, so when MIT introduced its first new course for 29 years, in 'bioengineering', I sat to attention. Bioengineering differs from biotechnology by applying standard principles from other engineering disciplines (say structural or aeronautical) to living systems - first make a mathematical model of the desired product, test this model, and only then implement the system from standard prefabricated modules, which in this case will be custom lengths of protein or DNA. Early projects include using viruses to deposit metallic traces for microcircuits and inserting counting devices into living cells. While I wish the new discipline well, I'm very much afraid that, like nanotechnology before it, it will soon be disfigured by wild hubristic claims to have usurped the powers of nature.

What triggers this thought is that I've finally been catching up with the work of Gerald Edelman, something I should have done several years ago but put off. Edelman won the Nobel Prize in 1972 for his work in immunology, but then turned his attention to neuroscience and the workings of the brain. His key insight in immunology lay in uncovering the way new antibodies get generated. If you or I or a bioengineer were designing an immune system from scratch, we'd probably create one standard antibody type that somehow reads the shape of intruding molecules by 'taking an impression', like pressing a key into a bar of soap. Nature didn't do it that way: instead it creates vast numbers of randomly-shaped antibodies, and if one of them happens to fit a new intruder, that one gets massively duplicated to combat the infection. In other words the immune system works by selection, not by instruction.

Edelman proposes that brain structure and function can be understood in a similar way: we need to know how brain evolved in animals by natural selection (as discovered by Darwin); how the brain gets constructed within each individual from a single cell (called morphology); and how morphology itself has evolved from species to species. The answer in each case is by selection, but on different units and under different forces. A bioengineer designing a human would probably first build the skeleton, then all the organs and mount them in it, wire them up with nerves and finally cover them with skin. A fertilised human egg works in exactly the opposite way. The dividing cell mass first forms a disk that becomes the skin and nervous system, which then wraps itself into a tube in which the organs eventually form. Nature works on principles utterly different from those of engineering: it can never start from scratch but has to work from what's already there; its materials must always remain contiguous, so modules can only emerge, never be created separately; and there's no-one waiting outside to write the software.

As the brain develops in an embryo its fantastically complex structure is *not* fully encoded in its genes. On the contrary, DNA merely records the starting conditions (as in Conway's Game of Life) and some guiding 'rules'. The brain then constructs itself using three simple operations: moving sheets of nerve cells from place to place; pinning them in a certain place; and killing them off (that is, cutting away unwanted portions). These operations are triggered by chemicals called growth factors that *are* encoded in the genes, but get expressed in a highly context-related way: a particular nerve cell only makes a certain factor if surrounded by certain neighbours at a certain time. This is the second of Edelman's types of selection, carried out on nerve sheets by growth factors, and it results in every brain having a different structure, even in identical twins. 

After the embryo is born or hatched it starts to take in information from its environment through its eyes, ears, taste, smell, touch, and such inputs act on brain structure by strengthening certain synaptic connections at the expense of others, which in immature brains may cause neurons to be killed off or created. Even in adults new neurons are created in more limited numbers. Edelman's third kind of selection then operates at the level of groups of neurons, which get wired together on-the-fly into functional structures - filters, recognizers, memory structures - in a way that's dictated partly by the actual data that they work on. The metaphor of brain as a Turing Machine that runs a program could hardly be further from the truth, because it's impossible to distinguish hardware from software, and the software in part writes itself and in part is written by the data stream.

Those parts of the brain that evolved first historically (brainstem and limbic system) act like thermostats regulating all bodily processes. Newer regions like the cortex analyse sensory inputs in the light of these primitive regulatory needs, attaching a 'meaning' to all processed information in crude terms like edibility, screwability or avoidability (danger). The brain is therefore less like a machine than like a congealed blob of accumulated meanings. The glorious final twist to the tale is that brain structure controls behaviour, and it's behaviour on which Darwinian natural selection operates. Selection within selection within selection ad infinitum... I really hope those new bioengineers don't imagine that they can second-guess all this stuff and build a brain the way you build a PC, by stuffing mobo, power supply and disks into a box.