Trichromatic vision is one of the characteristics that set humans and higher primates apart from our ancestors. The evolution of our visual system from an earlier dichromatic system seems to have started with the duplication of a visual pigment gene, which subsequently gave rise to the genes that code for red and green pigments. Expression of the two pigments is usually mutually exclusive, but how does a cone cell decide which gene to express? The 'standard' model suggests that this is pre-determined by the differential expression of transcription factors between red and green cones, but Smallwood et al. now present evidence for a simpler 'stochastic' model, in which the pigment gene promoters compete to pair with a cis-acting regulatory sequence.

The red and green pigment genes are arranged in tandem on the X chromosome, and they share a locus control region (LCR), which is situated at the 5′ end, adjacent to the red pigment gene. Smallwood et al. generated a series of transgenic mouse lines, using the red and green pigment gene promoters to drive the expression of histochemical reporter genes (alkaline phosphatase for red and β-galactosidase (lacZ) for green). By arraying the genes in different configurations, they examined how proximity to the LCR and gene order influence the relative expression of the two genes.

First, they showed that if a 9-kb spacer was inserted between the genes, the proportion of lacZ-expressing cones was reduced, indicating that the distance of the gene from the LCR can bias promoter selection. On the other hand, if the gene order was swapped so that the green pigment gene was adjacent to the LCR, there was a large increase in the proportion of cones that expressed lacZ only. This implies that the green pigment promoter has a higher affinity for the LCR than the red pigment promoter, but presumably this difference is usually negated by the effect of distance.

The LCR clearly has a strong influence on pigment gene expression, but is competition for the LCR the main factor that drives promoter choice? To test this, the authors removed the element of competition by duplicating the LCR, and they found that a large proportion of cones now expressed both pigment genes simultaneously. This indicates that, contrary to the standard model, the cells are intrinsically able to express both pigment genes, but only one promoter can interact with the LCR at any given time.

So, the data from these experiments strongly support the stochastic model. Perhaps most remarkable was the fact that the transgenic mouse retinal cells were able to choose whether to express the red or the green pigment gene, a decision that they are never faced with in real life. The beauty of the stochastic model is that it does not require any pre-existing differences in transcription factor expression between red and green cones. Instead, all of the information that determines promoter choice is designed into the gene array. This also has important evolutionary implications, as it shows how the ability to express a new visual pigment in a distinct cone population could have resulted from a single gene-duplication event.