Our results establish that diversity is low for Y-linked genes, rather than being unusually high for X-linked ones, since the mean X diversity is slightly lower than the autosomal loci, and close to the value expected with a 1 : 1 sex ratio and the same variance in male and female reproductive success. Hitch-hiking processes on the non-recombining Y chromosome are predicted to lower Y diversity, but a possible alternative reason for low diversity of Y-linked genes is a high variance in male reproductive success. The autosomes should then also have lower diversity than expected, relative to the X, i.e. the X/A ratio should be increased above 0.75 (Laporte & Charlesworth 2002
; Ellegren 2009
). We therefore conclude that there is no evidence that sexual selection is important in populations of this plant.
Another possibility, however, is a recent change in the demographic history of the population. Changes in population size affect the X and Y chromosomes differently from the autosomes, and can increase or reduce X/A and Y/A diversity ratios for many generations (Fay & Wu 1999
; Pool & Nielsen 2007
). Rapid growth in population size leads to increased X/A ratios, lasting for considerable times after the size change, whereas a reduction in population size decreases the ratio (the decrease can be severe, but not for long, see Pool & Nielsen 2007
). A temporary bottleneck generally lowers X/A ratios, but a very severe bottleneck, followed by growth as the population recovers, can have the opposite effect. The Y/A diversity ratio (and the mitochondrial/autosomal one) are affected in similar ways, but the low effective size of Y-linked genes causes their diversity to recover much faster than for X-linked genes, so that prolonged effects of a bottleneck are less likely to be found.
We therefore compared diversity estimates from two geographical regions, since the size change models predict different ratios in populations whose sizes differ. Although the regions are not strongly isolated (see above), the samples from northern populations (as defined in §2), when analysed separately, tend to have slightly higher diversity than southern ones (14 of 17 genes for which data are currently available), with a south/north diversity ratio for silent sites of 0.78 for autosomal genes (but no difference for the Y-linked sequences). The difference is more pronounced for the X-linked genes: southern populations' silent site diversity is only 56 per cent of that in the northern sample, so that their X/A ratio is much lower (0.35) than the overall value, consistent with a greatly reduced southern population in the recent past. However, Tajima's D-values are not more negative in the southern sample.
A low population size in the recent past, from which the population has not yet recovered, could temporarily greatly reduce diversity for Y-linked genes, relative to X-linked ones, and is thus a possible alternative to hitch-hiking to explain the observed low Y sequence diversity. illustrates this. A larger decrease in size could give an even larger X/Y ratio. On the other hand, any such difference will be lessened if Y genes have a higher mutation rate than X ones (as seems to be the case in S. latifolia, see above). It therefore seems unlikely that a population size change can explain the ratios observed in S. latifolia.
Figure 2. Plot X/Y ratios predicted using equation (4) of Pool and Nielsen (Pool & Nielsen 2007). The calculations assumed the same mutation rate for all chromosomes, and a fivefold decrease (black symbols and line) or increase (grey symbols and line) in (more ...)
Our observed higher KST
values for Y-linked genes than for other loci, regardless of the spatial scale of the samples used, suggest that the reduction in Y diversity is mainly caused by within-population processes, such as elimination of deleterious mutations, or local adaptations involving Y-linked genes, and is not principally an effect of a species-wide selective sweep owing to an advantageous mutation fixing in a Y-linked gene (which should lead to lower KST
for the Y, until equilibrium is re-established). The highest within-population X/Y diversity ratios average 31 for silent sites (electronic supplementary material, table S4). Overall, we therefore estimate that S. latifolia
Y-linked sequences have a 20–30-fold reduction in effective population size, consistent with the previous estimates using slightly larger within-population sample sizes, but fewer localities and loci (Laporte et al. 2005
). Taking into account the higher mutation rate of Y-linked genes, the low diversity of Y-linked genes is quantitatively consistent with strong within-population hitch-hiking processes associated with the ongoing Y chromosome genetic degeneration (Kaiser & Charlesworth 2009
). However, the different kinds of processes are not exclusive, and a species-wide selective sweep could also have contributed to reducing diversity.