Significant CpG-rich regions in angiosperm genes

Abstract

CpG islands, which are hypomethylated regions with clustered CpG dinucleotides resulting from a high G+C content and a lack of the normal CpG deficiency, are associated with a large number of vertebrate nuclear genes. There is some evidence that CpG islands may also be a feature of angiosperm genes. In mammalian and avian genes, candidate CpG islands can be readily identified by a simple set of sequence criteria. However, identification of candidate angiosperm CpG islands is more difficult due to the much higher ratio of observed/expected (O/E) CpG in the average angiosperm gene relative to the average vertebrate gene (see accompanying paper). We have developed sets of objective criteria that readily detect angiosperm DNA regions with a consistently high O/E CpG compared to the rest of the genome. These regions, which we call significant CpG-rich regions, always showed a CpG frequency equivalent to or greater than that expected from the base composition, suggesting that they may undergo positive selection for the presence of CpG dinucleotides. The significant CpG-rich regions were similar to vertebrate CpG islands in that they tended to be associated with the 5′ ends of genes, particularly of housekeeping genes, and in that they varied in location in tissue-specific genes. However, unlike the situation in mammalian and avian genes, DNA with a high O/E CpG did not surround the transcription start site of all angiosperm housekeeping genes. Significant CpG-rich regions in monocotyledonous species were G+C rich, like CpG islands. Significant CpG-rich regions in dicotyledonous species, on the other hand, differed from CpG islands but were similar to some CpG-rich regions in cold-blooded vertebrates (Xenopus and fish) in that they were A+T rich. We have argued that both the A+T-rich and the G+C-rich significant CpG-rich regions may fulfill the same function as the G+C-rich CpG islands of vertebrates.