bioRxiv preprint doi: https://doi.org/10.1101/2020.07.15.204214; this version posted July 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 1 The genetic architecture underlying ecological adaptation in 2 Atlantic herring is not consistent with the infinitesimal model 3 4 Fan Han1, Minal Jamsandekar2, Mats E. Pettersson1, Leyi Su1, Angela Fuentes-Pardo1, Brian 5 Davis2, Dorte Bekkevold3, Florian Berg4,5, Michele Cassini6,7, Geir Dahle5, Edward D. 6 Farell8, Arild Folkvord4,5, Leif Andersson1,2,9,* 7 8 1Department of Medical Biochemistry and Microbiology, Uppsala University, Uppsala, 9 Sweden. 10 2Department of Veterinary Integrative Biosciences, Texas A&M University, College Station, 11 USA. 12 3National Institute of Aquatic Resources, Technical University of Denmark. 13 4Department of Biological Sciences, University of Bergen, Bergen, Norway. 14 5Institute of Marine Research, Bergen, Norway. 15 6Department of Aquatic Resources, Institute of Marine Research, Swedish University of 16 Agricultural Sciences, Lysekil, Sweden. 17 7Department of Biological, Geological and Environmental Sciences, University of Bologna, 18 Bologna, Italy.8School of Biology and Environmental Science, Science Centre West, 19 University College Dublin, Belfield, Dublin 4, Ireland.9Department of Animal Breeding and 20 Genetics, Swedish University of Agricultural Sciences, Uppsala, Sweden. 21 *Corresponding author: Email:[email protected] 22 1 bioRxiv preprint doi: https://doi.org/10.1101/2020.07.15.204214; this version posted July 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 23 Abstract 24 Atlantic herring is widespread in North Atlantic and adjacent waters and is one of the 25 most abundant vertebrates on earth. This species is well suited to explore genetic 26 adaptation due to minute genetic differentiation at selectively neutral loci. Here we 27 report hundreds of loci underlying ecological adaptation to different geographic areas 28 and spawning conditions. Four of these represent megabase inversions confirmed by 29 long read sequencing. The genetic architecture underlying ecological adaptation in the 30 herring is in conflict with the infinitesimal model for complex traits because of the large 31 shifts in allele frequencies at hundreds of loci under selection. 32 33 Main 34 Atlantic herring (Clupea harengus) constitutes the basis for one of the world’s most important 35 commercial fisheries1 and has been a valuable food resource throughout human history in 36 North Europe. It is one of the most abundant vertebrates on earth with a total estimated 37 breeding stock of one trillion individuals2. Its short-term (<10,000 years) effective population 38 size is enormous whereas the long-term effective population size is much more restricted 39 most likely due to the impact of periods of glaciation3. As a consequence, the nucleotide 40 diversity is moderate (~0.3%)3, only three-fold higher than in human although the census 41 population size is at least two orders of magnitude higher in herring and until recently (>1,000 42 years ago) many orders of magnitude higher. 43 Detailed population genomic studies of Atlantic herring are justified for two reasons. 44 Firstly, a better understanding of the population structure and development of diagnostic 45 markers have the potential to revolutionize the procedures of stock assessments of herring that 46 has been in operation for more than hundred years, which will add key scientific knowledge 47 to safeguard a sustainable fishery. Secondly, the Atlantic herring can be used as a model to 2 bioRxiv preprint doi: https://doi.org/10.1101/2020.07.15.204214; this version posted July 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 48 study the genetic architecture underlying phenotypic diversity and ecological adaptation. The 49 herring has adapted to different marine environments including the brackish Baltic Sea, where 50 salinity drops to 2-3 PSU and water temperature shows a much larger variation among 51 seasons than in the Atlantic Ocean with typically 34-35 PSU (Extended Data Fig. 1). 52 Furthermore, herring spawn during different periods of the year, which involves 53 photoperiodic regulation of reproduction. There is minute genetic differentiation at selectively 54 neutral loci among herring populations from different geographic regions3,4 and the 55 distribution of per locus FST values deviates significantly from the one expected for 56 selectively neutral loci with an excessive number of outlier loci demonstrating signatures of 57 selection3,5. These properties make the Atlantic herring a powerful model to explore how 58 natural selection shapes the genome in a species where genetic drift plays a minor role in 59 genome-wide population divergence. 60 The genetic basis for complex traits and disorders is of central importance in current 61 biology and human medicine. The aim of the present study was to test whether the genetic 62 architecture underlying ecological adaptation in herring is consistent with the expectations 63 from Fisher’s classical infinitesimal model for polygenic traits6. This has been accomplished 64 by comprehensive analysis of whole genome sequence data from 53 population samples 65 spread across the entire species distribution. 66 67 Results 68 Genome sequencing 69 We performed pooled whole-genome sequencing using 28 population samples of Atlantic 70 herring. These were analyzed together previously published data3,7, making a total number of 71 53 population samples. The current study has added populations around Ireland and Britain, 72 from Norwegian fjords and from the Southern Baltic Sea. Together the 53 population samples 3 bioRxiv preprint doi: https://doi.org/10.1101/2020.07.15.204214; this version posted July 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 73 cover the entire species distribution and include spring, autumn, winter and summer spawning 74 stocks (Fig. 1 and Table S1). A previously sequenced population of Pacific herring (Clupea 75 pallasii) was used as a closely related outgroup in the bioinformatic analysis. We pooled 76 genomic DNA from 35-110 individuals per population, and conducted whole-genome 77 resequencing to a minimum of 30 coverage per population. In order to generate individual 78 genotype data we sequenced 12 fish from four localities to ~10 coverage per individual. 79 Together with data from previous studies3,4,7, we used a total of 55 individually sequenced 80 herring in the analysis (Table S2). All sequence reads were aligned to the recently released 81 chromosome-level assembly for Atlantic herring8, and sequence variants were called and 82 filtered with a stringent pipeline (Methods). In total, we identified ~11.5 million polymorphic 83 biallelic sites among the 53 population samples and ~15.9 million sites when including 84 Pacific herring. 85 86 Loci under selection reveal population structure 87 A genetic distance tree, based on genome-wide SNPs, groups the 53 population samples into 88 seven primary clusters: (i) autumn- and (ii) spring-spawning herring from the brackish Baltic 89 Sea, (iii) populations from the transition zone, close to the entrance to the Baltic Sea, 90 spawning at lower salinity than in the Atlantic Ocean, (iv) Norwegian fjord populations, (v) 91 populations around Ireland and Britain, (vi) autumn- and (vii) spring-spawning herring from 92 the North Atlantic Ocean (Extended Data Fig. 2). This clustering, in general, fits with their 93 geographical origin, but the branch lengths separating some subpopulations are very short, 94 e.g. autumn- and spring-spawning populations from the North Atlantic Ocean. Pairwise FST 95 values among all populations are in the range 0.013 to 0.061 (Fig. S1). 96 To explore the contribution of loci under selection to the observed population 97 structure, we separated SNPs showing no significant genetic differentiation between 4 bioRxiv preprint doi: https://doi.org/10.1101/2020.07.15.204214; this version posted July 15, 2020. The copyright holder for this preprint (which was not certified by peer review) is the author/funder. All rights reserved. No reuse allowed without permission. 98 populations from those showing strong genetic differentiation based on their standard 99 deviation of allele frequencies across Atlantic and Baltic populations (Methods; Fig. S2). We 100 carried out Principal Component Analysis (PCA) based on these two sets of markers 101 independently. The PCA using 169,394 undifferentiated markers separated some of the major 102 groups but three distinct groups (Atlantic Ocean spring and autumn spawners, and 103 populations around Ireland and Britain) were merged into one tight cluster (Fig. 2a). The first 104 two principal components explained only ~11% of the variance. This is in line with the 105 conclusion that the majority of the genome shows minute genetic differentiation across the 106 entire species distribution of Atlantic herring3,4,7. In contrast, the PCA based on only 794 107 markers showing the most striking genetic differentiation separated the populations into the 108