Wednesday, 4 November 2009

rice genetics watch: structure in Chinese rices but not domestication genes?

  • In the Oct. 2009 issue of Theoretical and Applied Genetics, a large Chinese research group (Zhang et al.) looked Genetics structure among Chinese rice landraces, with over 3000 Chinese rice populations. They find clear population structure, not just between indica and japonica as expected, but also within each of these. Interesting they report that the structure in indica seems to relate to flowering time (early, middle or late flowering varieties), which suggests that early differentiation after indica originated may be focused on seasonality (and constraints of seasonal land and water availability). In the case of japonica (primarily temperate japonica one presumes), seasonality is pretty much always restricted to the warm wet summer, as China has dry, cool winters that are not conducive to rice. Instead structure seems to divide Chinese japonica landraces on the ground of soil and water adaptations, and whether they are best grown ion paddies or on upland rainfall. Indeed, as predicted from the archaeology the earliest ecological efforts in rice domestication in China are likely to have focused on water manipulation (see Fuller & Qin in World Archaeology), while early dispersal must have also seen diversification in rainfed and less labour intensive systems of cultivation. On the whole an interesting approach that one would like see extended beyond China.

  • Another paper also with Zhang et al. authorship (but a different Zhang), that came out at the end of the summer in New Phytologist looked in more detail at the phylogeny of sh4 and qsh1 non-shattering (domestication) genes, and provides a coallescent model of their origin in terms of fixation time. Their estimate this trait should have been fixed in ~100 years seems a throwback to the kind results that models produced a decade ago, now at odds with the archaeobotanical evidence on domestication rates. The authors are at odds to explain this by positing thhat the now universal(?) sh4 domestication gene evolved after initial domestication and then diffused throughout rice (and replaced some earlier domestication genes). Not a particular elegant, nor historically/archaeologically compelling model. I am forced to assume that something is amiss in the math or the assumptions of the model. Can an apparent rapid bottleneck be artefact of another process in the way the apparent monophyly can (as per Allaby et al 2008). I also note that the phylogeny that relate domesticated sh4 to wild populations the same or a close gene, on the surface suggests an origin of sh4 from a Lao rufipogon or an Indian nivara-- but surely these wild taxa, and the indica and japonica types deriving from them should not group together in a population phylogeny when they have different chloroplast genomes (with a common ancestor in excess of 70,000 years ago!). Of course a Neighbourjoining tree, however much bolster by bootstraps and Montecarlo methods is still just a cluster analysis that is not a particularly logical or robust way to look for phylogenetic relationships within a species that hybridizes. Thus the method employed here denies the reticulate evolution which is so clearly a part of evolutionary story of rice, as so elegantly argued in earlier papers by Sang & Ge or more robustly in the recent papers of Kovach et al or McNally et al. I am therefore provisionally not at all sure what this sh4 data is actually telling us.


maundering gadabout said...

My interpretation is that the data from the New Phytologist paper are pretty straightforward: monophyly of sh4 suggests a common origin of the domesticated allele, presumably via gene flow between species (under one of the two models proposed in Sang and Ge, for example), and the popgen analysis points to a rapid fixation of that allele. One important thing to note, however, is that the popgen analysis does not rule out weak selection and long fixation times. If you look at figure 2, for example, much of the probability density occurs in regions of the parameter space with fixation times >1000 years. While the best point estimate still points to strong selection and fast fixation, it cannot exclude many scenarios of slow fixation under weaker selection.

I am also not convinced that a fast fixation of sh4 necessarily conflicts with the archaeological record. This could be the case if other alleles or genes contributed to observed change in shattering in archaeological samples, and sh4 -- which happened to provide a better shattering phenotype -- arose later and spread quickly.

VI said...
This comment has been removed by a blog administrator.