Genome sequencing reveals loci under artificial selection that underlie disease phenotypes in the laboratory rat

Abstract

Large numbers of inbred laboratory rat strains have been developed for a range of complex disease phenotypes. To gain insights into the evolutionary pressures underlying selection for these phenotypes, we sequenced the genomes of 27 rat strains, including 11 models of hypertension, diabetes, and insulin resistance, along with their respective control strains. Altogether, we identified more than 13 million single-nucleotide variants, indels, and structural variants across these rat strains. Analysis of strain-specific selective sweeps and gene clusters implicated genes and pathways involved in cation transport, angiotensin production, and regulators of oxidative stress in the development of cardiovascular disease phenotypes in rats. Many of the rat loci that we identified overlap with previously mapped loci for related traits in humans, indicating the presence of shared pathways underlying these phenotypes in rats and humans. These data represent a step change in resources available for evolutionary analysis of complex traits in disease models.

Graphical abstract

Figure 1

Phylogenetic Tree of 28 Rat Strains The phylogenetic tree was constructed using 9.6 million SNVs across 28 laboratory rat strains, including the Brown Norway reference strain (BN/Mcwi). The scale represents genetic distance; the distance matrix was calculated by dividing the number of SNVs between a given pair of strains by the length of the BN reference genome. The phylogenetic tree was constructed using the Fitch-Margoliash method with 1,000 bootstraps.

Figure 2

Illustration of the Mirror Tree Approach In this illustration, hypothetical genes 1, 4, and 6 evolved together, as they show identical phylogenetic history, whereas genes 2 and 5 also show an identical evolutionary history, though these two groups of genes evolved at a different evolutionary rate. These evolutionary patterns of individual genes were converted into phylogenetic vectors. Phylogenetic vectors of coevolving genes such as genes 1, 4, and 6 are more highly correlated than those that evolved at different rates such as genes 2 and 6. Networks of genes that coevolved were generated from significantly correlated genes. Phylogenetic vectors were derived by taking into account population structure. See also Figure S2 and Tables S3, S4, and S8.

Figure 3

Examples of Putative Artificial Selective Sweep Regions Identified in Laboratory Rat Strains (A) Relative SNV density (RSD) in 10 kb windows for four rat strains plotted along rat chromosomes separated by colors. Black arrows indicate putative artificial selective sweep (PASS) regions identified in respective rat strains. Closely adjacent but distinct PASS regions on the same chromosome are represented by a single arrow. The PASS region on chromosome 7 in SS is adjacent to but not overlapping with the chromosome 7 PASS region in SR. (For detail, see Table S6). (B) Average linkage disequilibrium (LD) between pairs of SNVs within a distance ranging from 1 bp to 0.5 Mb. The blue line represents average LD at whole-genome level; the red line represents LD in PASS regions. (C) Distribution of Tajima’s D in 10 kb windows in the entire genome (blue line) and in PASS regions (red line). (D) Ratio of nonsynonymous coding (NSC) variants to synonymous coding (SC) variants at genome level (blue bar) and in PASS regions (red bar). See also Figure S4 and Tables S6 and S7.

Figure 4

FHH PASS Regions on Chromosome 15 (A) Blue dots represent relative SNV density (RSD) in 10 kb windows for the FHH rat strain on chromosome 15. The genomic region unique to the FHH strain is highlighted by the gray dotted lines. Heatmap (bottom) showing RSD in the genomic region unique to FHH and flanking regions across all laboratory rat strains sequenced. Only the FHH rat strain showed an RSD value equal to 100. Remaining strains showed an RSD value of zero across almost all of the PASS regions, indicating that only the FHH strain shows SNVs against the BN reference genome in this region. Linkage disequilibrium (LD) structure of SNVs in the genomic region unique to FHH show that ∼98% of the SNVs in this region were in significant LD. (B)Tajima’s D in genomic region unique to FHH rat strain. The PASS region unique to FHH is highlighted, showing highly negative Tajima’s D value. (C) The genomic region unique to FHH shows a markedly increased ratio of NSC to SC variants.

Figure S1

Pairwise Comparison of SNVs, Stop Gain or Stop Loss Variants, Indels, and Frameshift Coding Variants, Related to Table 1 Heat maps showing (A) Number of SNVs between any pair of 28 rat strains including the BN reference genome, calculated by counting the number of alleles differing between any pair of strains; (B) Number of stop gain or stop loss variants between any pair of 28 rat strains. (C) Number of Indels between any pair of 28 rat strains including the BN reference genome, calculated by counting the number of alleles differing between any pair of strains; (D) Number of frameshift coding variants between any pair of 28 rat strains.

Figure S2

Principal Component Analysis of Rat Strains Sequenced, Related to Figure 2 (A) Principal component analysis shows that the first two principal components contribute to 70% of the sequence variability across the rat strains. (B) The first two principal components separate Japanese Wistar-derived strains, Sprague-Dawley-derived strains and remaining rat strains.

Figure S3

Linkage Disequilibrium in Coevolutionary Gene Clusters, Related to Table 2 Red bars show linkage disequilibrium (LD) between the SNVs in the genes within the cluster while the green bar shows LD between the SNVs after random shuffling. A) LD structure in all the clusters; B) LD structure in the clusters where at least 40% of the SNVs in the cluster show strong LD (r2 > = 0.8).

Figure S4

Linkage Disequilibrium in Putative Artificial Selective Sweep Regions, Related to Figure 3 Red bars show LD between the SNVs within the putative artificial selective sweeps: in the majority of the putative artificial selective sweep (PASS) regions almost all SNVs show strong LD. After random shuffling of all SNVs in the genomes of the 27 rat strains, the strong LD between SNVs in PASS regions is completely lost (represented as green bars).