Tracking signatures of response over 20 generations of selection for long leg length in mice Layla Hiramatsu Postdoc with Frank Chan Friedrich Miescher Laboratory Max-Planck Campus, Tübingen, Germany
How do small populations respond to selection? • Selection vs. genetic drift Quantitative Population • Hard vs. soft sweeps genetics genetics • Allelic interactions • How many genes • Effect sizes • How repeatable is the response Replicated, controlled, pedigreed selection experiments 1
Longshanks selection experiment for long tibiae Three lines of CD-1 outbred mice, 16 pairs per generation Ctrl LS1 13.1% LS2 12.7% 2.5 mm 2 bioRxiv link! Marta Marchini, Campbell Rolian (U. of Calgary)
Longshanks selection experiment for long tibiae Three lines of CD-1 outbred mice, 16 pairs per generation Ctrl LS1 LS2 N=1331 N=3060 N=3098 ℎ 2 = 0.51 𝑇 = 0.02* N e = 46 3 M. Marchini, C. Rolian (U. of Calgary) & QG estimates N. Barton (IST Austria)
All breeders whole-genome sequenced N = 1,550 Ctrl Ctrl LS1 LS1 LS2 LS2 N=1331 N=3060 N=3098 N WGS = 25 (F0) +33 (F17) + 316 N WGS = 26 + 32+ 546 N WGS = 25 + 32 + 514 Founders (F0) and F17 at 100x coverage (~5x / individual) All other generations 30x coverage (~0.5x / individual) 4 F0F17: João Castro, Mish Yancoskie et al., eLIFE, in revision
Change in allele frequencies in Ctrl line Genetic drift ∆z 2/ π 2 Ctrl F0 vs. F17 chr 5 Castro, Yancoskie et al., eLIFE, in revision
Broad polygenic response, parallel and line-specific ∆z 2/ π 2 F0 vs. F17 chr 6 Castro, Yancoskie et al., eLIFE, in revision
8 loci of major effect, top 2 are in parallel Significant threshold from Hitchhiking filtered pedigree-calibrated infinitesimal model with LD LS2 LS1 Nick Barton, Stefanie Belohlavy (IST Austria) 7 Castro, Yancoskie et al., eLIFE, in revision
8 loci of major effect, top 2 are in parallel Significant threshold from pedigree-calibrated infinitesimal model with LD >5Mb region in chr10 Nkx3-2 in chr5 LS2 LS1 8 Castro, Yancoskie et al., eLIFE, in revision
8 loci of major effect, top 2 are in parallel Significant threshold from pedigree-calibrated infinitesimal model with LD >5Mb region in chr10 Nkx3-2 in chr5 No coding changes Functional test of enhancers LS2 F0 allele vs. F17 allele (3 SNPs) Loss of bone growth F0 F17 repression LS1 9 Castro, Yancoskie et al., eLIFE, in revision
How did the Longshanks respond to selection? • Selection vs. genetic drift Drift strong, selection stronger Quantitative Population Near fixation by F17 genetics genetics • How many genes Very polygenic Nkx3-2 region 8 major loci DevBio No coding changes • Effect sizes Cis- regulatory Loss of function Nkx3-2 : 10% • Can we model the response Yes! Linkage important 10 bioRxiv link! Castro, Yancoskie et al., eLIFE, in review
Reconstruction of selection response generation-to-generation Ctrl LS1 LS2 generations 11
Reconstruction of selection response generation-to-generation Ctrl LS1 LS2 generations 30x coverage (~0.5x / individual) 12
Reconstruction of selection response generation-to-generation Ctrl LS1 LS2 Allele trajectories • Allelic interactions (e.g., dominance) • Timing and relative importance of genes Haplotype reconstruction with pedigree • Compare against models of linkage 13
LS1 Average each 100kb window across all chromosomes chr 1 || 2133 SNPs Frequency of minor allele at each of 2133 SNPs Average frequency 14
LS1 Average each 100kb window across all chromosomes * Reduces noise due to low coverage but haplotypes are likely larger than 100kb chr 1 || 2133 SNPs Frequency of minor allele at each of 2133 SNPs Average frequency 15
LS1 Trajectories of windows which started at 20-25% in founding population 16
LS1 Trajectories of windows which started at 20-25% in founding population Selected windows in chr 10 17
LS1 Trajectories of windows which started at 20-25% in founding population Selected windows in chr 10 LS2 18
LS1 Side story. Late alleles: recessive or broke linkage with deleterious allele? multi-locus adaptation 19
LS1 Side story. Chr 6 region broke from deleterious allele at generation 11? Or seeing many background haplotypes? 20
LS1 “Drift zone” by simple Wright -Fisher simulation, 95% of 1000 simulations 21
LS1 Estimate selection coefficient for a window 22
LS1 Estimate selection coefficient for a window 23
Change in allele frequency with selection AA Aa aa Genotypes p 2 q 2 2pq Frequency before selection dominance coefficient h = 0 : recessive A 1+s 1+hs 1 Fitness h = 0.5 : additive h = 1 : dominant A p 2 (1+s) 2pq(1+hs) q 2 Frequency after selection p 2 (1 + s) + pq (1 + hs ) iterate for 20 p t+1 = generations 1 + sp 2 + 2hspq 24
Change in allele frequency with selection AA Aa aa Genotypes p 2 q 2 2pq Frequency before selection dominance coefficient h = 0 : recessive A 1+s 1+hs 1 Fitness h = 0.5 : additive h = 1 : dominant A p 2 (1+s) 2pq(1+hs) q 2 Frequency after selection p 2 (1 + s) + pq (1 + hs ) iterate for 20 p t+1 = generations 1 + sp 2 + 2hspq Try many h and s combinations and calculate the likelihood that they would produce the observed trajectory 25 with help from Felicity Jones
LS1 This allele in chr10 acts partially dominantly and selection is strong - loglik Dominance coefficient h best likelihood at s, h = Selection coefficient s 26
LS1 Calculated coefficients fit the observed data well p 2 (1 + s) + pq (1 + hs ) p t+1 = 1 + sp 2 + 2hspq s = 0.43 h = 0.73 27
LS1 Estimate selection coefficient for Nkx3-2 selected alleles Loss of bone growth F0 F17 repression 28
LS1 Overdominance in Nkx3-2 allele: reached an equilibrium s = 0.35, h = 1.60 , freq eq = 0.73 29
LS1 Overdominance in Nkx3-2 allele: reached an equilibrium s = 0.35, h = 1.60 , freq eq = 0.73 Check allelic interactions (e.g., dominance) with genotype frequencies 30
LS1 Genome-wide: Most windows are not under selection Windows under selection tend to act in dominance (?) 31
LS1 Genome-wide: Most windows are not under selection Windows under selection tend to act in dominance (?) Windows outside “drift zone” for at least 5 generations 1.6% 32
LS1 Genome-wide: Most windows are not under selection Windows under selection tend to act in dominance (?) Windows outside “drift zone” for at least 5 generations 1.6% Nkx3-2 Selected windows in chr 10 Side story Recessive(?) alleles 33
Genome-wide: Most windows are not under selection Windows under selection tend to act in dominance Similar pattern genome-wide for replicate lines LS1 LS2 34
Single-locus simulation with within-family selection 35
Reconstruction of selection response generation-to-generation Founding Allele trajectories haplotypes • Allelic interactions (i.e., dominance) time • Timing and relative importance of genes Haplotype reconstruction with pedigree • Compare against models of linkage Selected haplotypes Can we fully map the timing and ? location of the selection response? How does strong selection distort the recombination landscape? 36
bioRxiv link! How did the Longshanks respond to selection? Quantitative Population • Selection vs. genetic drift genetics genetics Drift strong, selection stronger Near fixation by F17 DevBio • How many genes Very polygenic 8 major loci • Effect sizes Nkx3-2 : 10% • Can we model the response Yes, and linkage important • How do the alleles increase tibia length cis - regulatory “breaking enhancers” • Allelic interactions Most selected alleles act dominantly (?) • Haplotype segregation Coming soon! 37 Castro, Yancoskie et al., eLIFE, in review
bioRxiv link! Thank you! Quantitative Population • Selection vs. genetic drift genetics genetics Drift strong, selection stronger Acknowledgements Near fixation by F17 Campbell Rolian DevBio Frank Chan • How many genes Marta Marchini João Castro Very polygenic Mish Yancoskie Isabella Skuplik Marek Kuçka John Cobb 8 major loci Nick Barton Felicity Jones • Effect sizes Stefanie Belohlavy Bill Beluch Nkx3-2 : 10% WE ’ RE Chan Lab Ronald Naumann HIRING! • Can we model the response Jones Lab Yes, and linkage important • How do the alleles increase tibia length cis - regulatory “breaking enhancers” • Allelic interactions Most selected alleles act dominantly (?) • Haplotype segregation Coming soon! 38
Recommend
More recommend