There are two main varieties of domesticated rice (Oryza sativa). One variety, O. sativa indica can be found in India and Southeast Asia while the other, O. sativa japonica, is mostly cultivated in Southern China.
Extensive studies of the varieties has demonstrated that they were independently derived from the wild rice species Oryza rufipogon (left). The domesticated varieties show much less variation (polymorphism) than the wild species. This is not unexpected since they were presumably bred from a small number of plants when rice first began to be cultivated more than 10,000 year ago. This phenomenon of restricted variation after a speciation event is called the bottleneck effect because it represents a severe reduction in the number of different individuals that contributed to the new species. This bottleneck effect is thought to be a major factor in reducing variation in domesticated strains as well. In this case, the "speciation" event is man-made.
Bottleneck effects are similar to founder effects and both of them are forms of random genetic drift. The changes in the frequency of alleles within the new populations are due to chance and not to adaptation.
Caicedo et al. (2007) have recently explored the patterns of nucleotide polymorphisms in rice. They looked at SNP's, or single nucleotide polymorphisms, in 111 different regions of the genome. The idea was to see if certain polymorphisms tended to cluster together to form distinct patterns. If one genetic locus had a particular nucleotide "A" at site SNP-32, for example, it would be interesting to see whether the nearby regions of the genome were similar or different. If the same pattern, say SNP-32(A), SNP-33(G), and SNP-34(T), occurred at a high frequency then it indicates that there hasn't been enough time for recombination to separate the three distinctive variations.
When such patterns are found, there's a tendency to attribute the pattern to "selective sweeps." In a selective sweep the pattern becomes rapidly fixed in the genome due to selection for one of the markers. Assume, for example, that the presence of nucleotide "G" at site SNP-33 conferred some selective advantage on the plant. As this allele is rapidly selected, the nearby alleles (SNP-32(A) and SNP-34(T)) will be swept up in the adaptive event. Their frequency in the population will increase because they are hitchhiking on the SNP-33(G) allele. It's important to remember that the fixation of the flanking alleles are accidents—they are not being selected for their own phenotype.
But selective sweeps are not the only way that specific patterns of alleles can become widespread in a population. It can also happen if the population goes though a bottleneck where much of the variation was eliminated by chance. If the bottleneck occurred relatively recently then the pattern can look very much like a rapid fixation due to adaptation and hitchhiking.
The pattern of evolution in domestic rice varieties compared to Oryza rufipogon shows many examples of associated alleles, or haplotypes. The O. japonica variety has only 19% of the total polymorphism of the wild type genome and the other main variety, O. indica, has only 43%. The authors note that such patterns are often associated with selected sweeps but there are other possibilities.
An excess of high-frequency derived SNPs is often interpreted as a result of genetic hitchhiking during recent selective sweeps . Because the site-frequency spectrum in rice varieties is observed from randomly selected loci, and the loci contributing high frequency derived SNPs are distributed across the genome (Fig. S4), this pattern suggests that strong linkage to positively selected mutations occurred within most of the genome. However, demographic forces may have also played a role in shaping the rice genomes. We developed several demographic models and a multiple selective sweeps model to test which evolutionary processes may best explain the observed patterns of polymorphism in rice.In addition to a selective sweep model, the authors tested a neutral population bottleneck model defined as,
The most widely accepted demographic model for crop domestication is a neutral bottleneck model [27-29]. In this model, rice domestication is assumed to be a result of recent population divergence, with one of the two daughter populations experiencing a reduction in population size at divergence associated with the founder effect at the time of domestication, followed by population growth as cultivation of the crop increases.The other models were combinations of bottleneck and migration between populations, and bottleneck plus selection.
It isn't easy to test these models, even with an extensive database such as the one from rice genomes. The mathematics is complicated and many simplifying assumptions have to be made. Nevertheless, Caicedo et al. (2007) conclude from their analysis that bottlenecks alone are not sufficient to explain the SNP patterns they see in domesticated rice. They conclude that the patterns result from a combination of drift (bottlenecks) and adaptation (selective sweeps).
A more complex demographic scenario involving very strong bottlenecks that led to the fixation of alternate alleles during the two rice domestication events (with concurrent gene flow between variety groups) can explain the site-frequency spectrum of indica and O. rufipogon. However, this pure demography model requires a bottleneck four-fold stronger in indica and twice as strong in tropical japonica relative to the model that incorporates selection (Table 2; Figure 5), and a relatively high migration rate between domesticated rice and wild O. rufipogon populations. It is also important to note that the model is a poor fit to the observed frequency distribution of alleles in tropical japonica.The lesson here is that it is very difficult to distinguish selection from drift and one should be cautious in attributing results to only one of these mechanisms of evolution.
Domestication, however, is characterized by strong directional selection on a suite of traits that lead to the establishment of cultivated species as distinct entities from their wild progenitors within agricultural settings. We show that, in contrast to the complex demographic model, a simple bottleneck with sweeps model fits data from both tropical japonica and indica well without requiring an extremely strong domestication bottleneck. Since domesticated Asian rice has been subject to artificial selection, the selection plus demography model is a very plausible explanation for the observed strong excess of high frequency derived alleles in domesticated rice varieties, and is consistent with recent reports about domestication genes in rice [45,46].
Caicedo, A., Williamson, S., Hernandez, R.D., Boyko, A., Fledel-Alon, A., et al. (2007) Genome-Wide Patterns of Nucleotide Polymorphism in Domesticated Rice. PLoS Genet. In press. [doi:10.1371/journal.pgen.0030163.eor]
[The drawing of Oryza rufipogon is from Naples, M.L. (2005). The middle photograph is of Japanese short grained rice from the Wikipedia article on rice. The lower figure is Figure 3 from Caicedo et al. (2007)]