Genome-wide effects of long-term divergent selection

Abstract

To understand the genetic mechanisms leading to phenotypic differentiation, it is important to identify genomic regions under selection. We scanned the genome of two chicken lines from a single trait selection experiment, where 50 generations of selection have resulted in a 9-fold difference in body weight. Analyses of nearly 60,000 SNP markers showed that the effects of selection on the genome are dramatic. The lines were fixed for alternative alleles in more than 50 regions as a result of selection. Another 10 regions displayed strong evidence for ongoing differentiation during the last 10 generations. Many more regions across the genome showed large differences in allele frequency between the lines, indicating that the phenotypic evolution in the lines in 50 generations is the result of an exploitation of standing genetic variation at 100s of loci across the genome.

Figure 1. Selection response in the Virginia chicken lines.

The red and blue lines show the average male body weight in the high and low body weight lines, respectively. The high line is still responding to selection, whereas the low line appears to have phenotypically plateaued. The photos are from generation 37.

Figure 2. Schematic view of the groups of genotyped individuals and outline of the comparisons made in this study.

The number of genotyped birds in each group is shown as n. The arrows represent the different comparisons made; A is across time but within the same line, B is between lines within generation.

Figure 3. Allele frequency changes between generation 40 and 50 across chromosome 1 as measured by association analysis.

There are several regions on chromosome 1 with significant allele frequency differences between generation 40 and 50 for in the high line (red) but not in the low line (blue). The solid grey line indicates the Bonferroni corrected significance level p<0.001 and the dashed grey line the corrected p<0.05.

Figure 4. Allele frequency differences between the high and low body weight selected lines.

An association test for comparison B in Figure 2 shows a big difference in how quickly allele frequencies change between generation 40 and 50 in the high- and low lines. In the region between 60 and 80 Mb on chromosome 4 there are large differences between the results from generations 40 and 50. The sliding window mean of 20 markers is shown as a red line for generation 40 and as a purple line for generation 50. The scale is normalized for easier comparison between the time points because their maximum χ² is different (80 at generation 40 and 118 at generation 50).

Figure 5. Significance for allele frequency changes on chromosome 1.

Average allele frequency changes (comparison A in Figure 2) over blocks of 5 markers were calculated and the blue and red lines indicate the number of loci with unexpectedly high changes in a window of 20 blocks. The light grey line corresponds to a 95% confidence genome wide (p = 10⁻⁵).

Figure 6. Clusters of SNPs fixed for different alleles in the two lines.

Comparisons of clusters consisting of at least 2 SNPs in generation 40 (reddish pink) and at least 5 SNPs in generation 50 (blue). The overlap is considerable, as some clusters already contained more than 5 SNPs at generation 40 and others that with 2 SNPs in generation 40 were extended over those last 10 generations to include at least 5 SNPs in generation 50. Interestingly, several novel sets have appeared, indicating possible sites for ongoing selection. The sets that have not increased are unlikely to be under selection at this time in the selection process.

Figure 7. Clusters with at least 5 SNPs fixed for different alleles in the two lines.

The grey lines represent the regions on the chromosomes covered by markers. Chromosome Z is numbered 31 and the linkage groups LGE22C19W28_E50C23 and LGE64 are numbered 29 and 30, respectively. Clusters at generation 40 are shown in red with the additional regions covered by clusters at generation 50 shown in purple. The maximum distance between subsequent fixed SNPs is 1 Mb. The significant QTL regions reported in are shown in blue. As can be seen, all QTL regions contain one or several clusters.