During the development of the brain, young neurones need to form connections with other cells. But equally important, they need to avoid making connections with themselves.
Unfortunately, the chance of this happening is rather high. As a neurone grows and branches out in all directions, many of the branches will inevitably come into contact with others from the same cell. They're right next to each other.
So, how do brains achieve self-avoidance? The answer, according to a new Nature paper building on previous work, is a clever mechanism involving a single protein, Dscam1. The DNA code which produces it contains three sections (exons), which can each vary in several ways. There are 12 variants of exon 4, 48 of exon 6, and 33 of exon 9.
That means that Dscam1 can end up in 12 x 48 x 33 = 19,008 different configurations (isoforms). It's as if whenever the protein is formed, it rolls a 12 sided dice, a 48 sided dice, and a 33 sided dice, and then ends up signalling the result. The diagram illustrates this nicely. The clever part is that each developing neurone expresses only a few isoforms, entirely at random.
If a growing neuronal branch encounters another branch with the same Dscam1 isoform, the two identical proteins interact and the branches repel each other. Because every part of any given cell expresses the same "fingerprint", this produces self-avoidance. But the chance that another neighbouring cell will have the same protein is very small. There are billions of neurones in the brain, so many will share the same protein, but the chance of a cell encountering another nearby with the identical fingerprint is tiny.
In this paper, the authors genetically engineered fruit flies (Drosophila) so that they had fewer than the normal 19,008 Dscam1 variants. (Previous work suggests that the system is similar in mammals.) Flies with 4,752 variants developed normally, but with only 1,152, problems arose: neurones got repelled from other nearby neurones because they shared the same protein. With 576, 24, or 12 isoforms, the problem became progressively worse, as the chance of two cells having the same isoform rose.
So, in order to avoid tying themselves in knots, brains need somewhere between about one thousand and five thousand Dscam1 variants. It's an elegant solution to the problem of neurite self-avoidance, and a lovely example of evolution at work.
Hattori D, Chen Y, Matthews BJ, Salwinski L, Sabatti C, Grueber WB, & Zipursky SL (2009). Robust discrimination between self and non-self neurites requires thousands of Dscam1 isoforms. Nature, 461 (7264), 644-8 PMID: 19794492