Trends in genome-wide and region-specific genetic diversity in the Dutch-Flemish Holstein-Friesian breeding program from 1986 to 2015

Abstract

Background: In recent decades, Holstein-Friesian (HF) selection schemes have undergone profound changes, including the introduction of optimal contribution selection (OCS; around 2000), a major shift in breeding goal composition (around 2000) and the implementation of genomic selection (GS; around 2010). These changes are expected to have influenced genetic diversity trends. Our aim was to evaluate genome-wide and region-specific diversity in HF artificial insemination (AI) bulls in the Dutch-Flemish breeding program from 1986 to 2015.

Methods: Pedigree and genotype data (~ 75.5 k) of 6280 AI-bulls were used to estimate rates of genome-wide inbreeding and kinship and corresponding effective population sizes. Region-specific inbreeding trends were evaluated using regions of homozygosity (ROH). Changes in observed allele frequencies were compared to those expected under pure drift to identify putative regions under selection. We also investigated the direction of changes in allele frequency over time.

Results: Effective population size estimates for the 1986-2015 period ranged from 69 to 102. Two major breakpoints were observed in genome-wide inbreeding and kinship trends. Around 2000, inbreeding and kinship levels temporarily dropped. From 2010 onwards, they steeply increased, with pedigree-based, ROH-based and marker-based inbreeding rates as high as 1.8, 2.1 and 2.8% per generation, respectively. Accumulation of inbreeding varied substantially across the genome. A considerable fraction of markers showed changes in allele frequency that were greater than expected under pure drift. Putative selected regions harboured many quantitative trait loci (QTL) associated to a wide range of traits. In consecutive 5-year periods, allele frequencies changed more often in the same direction than in opposite directions, except when comparing the 1996-2000 and 2001-2005 periods.

Conclusions: Genome-wide and region-specific diversity trends reflect major changes in the Dutch-Flemish HF breeding program. Introduction of OCS and the shift in breeding goal were followed by a drop in inbreeding and kinship and a shift in the direction of changes in allele frequency. After introduction of GS, rates of inbreeding and kinship increased substantially while allele frequencies continued to change in the same direction as before GS. These results provide insight in the effect of breeding practices on genomic diversity and emphasize the need for efficient management of genetic diversity in GS schemes.

Fig. 1

Number of genotyped bulls by year of birth

Fig. 2

Generation interval for bull sires, bull dams and bull parents by year of birth

Fig. 3

Correlations between different genome-wide estimates of inbreeding (left) and kinship (right) by year of birth. Note the different scales for the y-axes for inbreeding and kinship. Self-kinships were excluded from the computation to remove the influence of the number of bulls per year on the correlations. Error bars represent ± 2 standard errors. $F_{\mathit{PED}}$ and $f_{\mathit{PED}}$: genealogical inbreeding and kinship; $F_{\mathit{ROH}}$ and $f_{\mathit{SEG}}$: segment-based inbreeding and kinship;$HO{M}_{\mathit{SNP}}$ and $SI{M}_{\mathit{SNP}}$: marker-by-marker homozygosity and similarity

Fig. 4

Average genome-wide inbreeding (left) and kinship (right) by year of birth. Coefficients of IBD ($F_{\mathit{PED}}$, $F_{\mathit{ROH}}$, $f_{\mathit{PED}}$, $f_{\mathit{SEG}}$) and IBS ($HO{M}_{\mathit{SNP}}$, $SI{M}_{\mathit{SNP}}$) are shown on the primary and secondary y-axis, respectively. $F_{\mathit{PED}}$ and $f_{\mathit{PED}}$: genealogical inbreeding and kinship; $F_{\mathit{ROH}}$ and $f_{\mathit{SEG}}$: segment-based inbreeding and kinship;$HO{M}_{\mathit{SNP}}$ and $SI{M}_{\mathit{SNP}}$: marker-by-marker homozygosity and similarity

Fig. 5

Rate of change per year (top) and generation (bottom) for genome-wide parameters within 5-year periods. $F_{\mathit{PED}}$ and $f_{\mathit{PED}}$: genealogical inbreeding and kinship; $F_{\mathit{ROH}}$ and $f_{\mathit{SEG}}$: segment-based inbreeding and kinship;$HO{M}_{\mathit{SNP}}$ and $SI{M}_{\mathit{SNP}}$: marker-by-marker homozygosity and similarity

Fig. 6

Positional inbreeding coefficients ($F_{\mathit{ROH}}$) per 5-year period between 1986 and 2015. Grey bars cover gaps between consecutive markers of > 500 kb (with an additional 3.75 Mb on both sides of the gap). BTA: Bos taurus autosome. Note that the scale of the x-axis differs between chromosomes

Fig. 7

Absolute allele frequency changes from 1986–1990 to 2011–2015 ($\left|p_{2011-2015}-p_{1986-1990}\right|$) observed in data and gene drop. Changes are shown for different minor allele frequencies (MAF) in the 1986–1990 period, using MAF-classes of 0.5% (e.g. 0.0–0.5%). The red line represents the 99.9%-threshold of the gene drop distribution per MAF class

Fig. 8

Moving average of absolute changes in allele frequency from the 1986–1990 to the 2011–2015 period ($\left|p_{2011-2015}-p_{1986-1990}\right|$). Moving average is based on 31 SNPs. The SNPs in red (N = 490) have an allele frequency change above the 99.9%-threshold of the gene drop distribution (see Fig. 7). BTA Bos taurus autosome