Large scale sequence-based screen for recessive variants allows for identification and monitoring of rare deleterious variants in pigs

Abstract

Most deleterious variants are recessive and segregate at relatively low frequency. Therefore, high sample sizes are required to identify these variants. In this study we report a large-scale sequence based genome-wide association study (GWAS) in pigs, with a total of 120,000 Large White and 80,000 Synthetic breed animals imputed to sequence using a reference population of approximately 1,100 whole genome sequenced pigs. We imputed over 20 million variants with high accuracies (R2>0.9) even for low frequency variants (1-5% minor allele frequency). This sequence-based analysis revealed a total of 14 additive and 9 non-additive significant quantitative trait loci (QTLs) for growth rate and backfat thickness. With the non-additive (recessive) model, we identified a deleterious missense SNP in the CDHR2 gene reducing growth rate and backfat in homozygous Large White animals. For the Synthetic breed, we revealed a QTL on chromosome 15 with a frameshift variant in the OBSL1 gene. This QTL has a major impact on both growth rate and backfat, resembling human 3M-syndrome 2 which is related to the same gene. With the additive model, we confirmed known QTLs on chromosomes 1 and 5 for both breeds, including variants in the MC4R and CCND2 genes. On chromosome 1, we disentangled a complex QTL region with multiple variants affecting both traits, harboring 4 independent QTLs in the span of 5 Mb. Together we present a large scale sequence-based association study that provides a key resource to scan for novel variants at high resolution for breeding and to further reduce the frequency of deleterious alleles at an early stage in the breeding program.

Fig 1. Imputation accuracies (R2) of all SNPs grouped by Minor Allele Frequency (MAF).

Fig 2. Additive (red) and recessive (blue) Manhattan plots for growth rate (GR) and backfat (BF).

For peaks with a p-value significance below 1E-10 (green line) the genomic location is shown. QTLs only supported by a single significant SNP are neglected. A) Large White growth rate B) Large White backfat C) Synthetic growth rate D) Synthetic backfat. *The 121MB and 10MB peaks have top SNPs with p-values of 6,91E-154 and 1,63E-135 respectively.

Fig 3. Number of significant GWAS QTLs overlapping between breeds and traits.

Fig 4. Phenotypic effect sizes of top SNPs from the significant QTLs.

QTLs with phenotypic effects < 0.1 standard deviations are excluded. Square size indicates significance (bigger square is lower p-value, ranging from 1E-10 to 1E-50+).

Fig 5. Genomic overview of the OBSL1 frameshift variant.

A) OBSL1 gene model adapted from Ensembl 110 showing the genes three transcripts with exons (boxes) and introns (lines) The exon where the frameshift is located (exon 5) is highlighted in blue. B) DNA sequence alignment showing the 5bp deletion causing a frameshift in affected animals. C) Protein alignment showing the frameshift inducing a premature stop-codon on exon 6.

Fig 6. Manhattan plots of region on chromosome 1 showing several QTLs for growth rate (GR) and backfat (BF).