Author response: The genetic basis for ecological adaptation of the Atlantic herring revealed by genome sequencing

Full text Figures and data Side by side Abstract eLife digest Main text Data availability References Decision letter Author response Article and author information Metrics Abstract Ecological adaptation is of major relevance to speciation and sustainable population management, but the underlying genetic factors are typically hard to study in natural populations due to genetic differentiation caused by natural selection being confounded with genetic drift in subdivided populations. Here, we use whole genome population sequencing of Atlantic and Baltic herring to reveal the underlying genetic architecture at an unprecedented detailed resolution for both adaptation to a new niche environment and timing of reproduction. We identify almost 500 independent loci associated with a recent niche expansion from marine (Atlantic Ocean) to brackish waters (Baltic Sea), and more than 100 independent loci showing genetic differentiation between spring- and autumn-spawning populations irrespective of geographic origin. Our results show that both coding and non-coding changes contribute to adaptation. Haplotype blocks, often spanning multiple genes and maintained by selection, are associated with genetic differentiation. https://doi.org/10.7554/eLife.12081.001 eLife digest The Atlantic herring is one of the most common fish in the world and has been a crucial food resource in northern Europe. One school of herring may comprise billions of fish, but previous studies had only revealed very few genetic differences in herring from different geographic regions. This was unexpected since Atlantic herring is one of the few marine species that can reproduce throughout the brackish Baltic Sea, which can be about a tenth as salty as the Atlantic Ocean. This unexpected finding could be explained in at least two different ways. Firstly, perhaps Atlantic herring are flexible enough to adapt to very different environments (i.e. high or low salinity) without much genetic change. Secondly, the previous studies only looked at a handful of sites in the Atlantic herring's genome and so it is possible that genetic differences at other genes control this fish's adaptation instead. Now, Martinez Barrio, Lamichhaney, Fan, Rafati et al. have sequenced entire genomes from groups of Atlantic herring and revealed hundreds of sites that are associated with adaptation to the Baltic Sea. The analysis also identified a number of genes that control when these fish reproduce by comparing herring that spawn in the autumn with those that spawn in spring. This is important because natural populations must carefully time when they reproduce to maximize the survival of their young. These new findings provide compelling evidence that changes in protein-coding genes and stretches of DNA that regulate the expression of other genes both contribute to adaptation in herrings. The analysis also clearly shows that variants of genes that contribute to adaptation were likely to evolve over time by accumulating multiple sequence changes affecting the same gene. Furthermore, these gene variants essentially form a rich "tool-box" that underlies the Atlantic herring's adaptation to its environment, and different subpopulations of herring were found to have their own optimal sets of gene variants. For instance, autumn-spawning herring and spring-spawning herring from the Baltic Sea both have gene variants that favor adaptation to low salinity. However, autumn-spawning Baltic herring also share gene variants that favor spawning in the autumn with autumn-spawning herring from the North Sea, but not with spring-spawning Baltic herring. The next step will be to study how the 500 or so genes identified affect adaptation at the molecular level. This will likely involve experiments with other model fish such as zebrafish and sticklebacks. Finally, these new findings can be directly applied to monitor stocks of herring to make herring fisheries more sustainable. https://doi.org/10.7554/eLife.12081.002 Main text The Atlantic herring (Clupea harengus) is a pelagic fish that occurs in huge schools, up to billions of individuals. The herring fishery has been crucial for food security and economic development in Northern Europe and currently ranks among the five largest fisheries in the world with nearly 2 million tons fish landed annually (FAO, 2014). The herring is one of few marine fishes that reproduce throughout the Baltic Sea where the salinity drops to 2–3‰ in the Bothnian Bay, compared with 35‰ in the Atlantic Ocean (Figure 1A). This ecological adaptation must be recent because the brackish Baltic Sea has only existed for 10,000 years following the last glaciation (Andrén et al., 2011). Fishery biologists have for more than a century recognized stocks of herring defined by spawning location, spawning time, morphological characters and life history parameters (Iles and Sinclair, 1982; McQuinn, 1997). Several decades of genetic studies based on limited numbers of genetic markers (allozymes, microsatellites or SNPs) have not been able to verify this divergence; extremely low levels of differentiation even between geographically distant populations as well as between spring- and autumn-spawning herring have been observed (Andersson et al., 1981; Ryman et al., 1984; Larsson et al., 2007; 2010, Limborg et al., 2012). It has been proposed that lack of precision in homing behaviour of herring causes sufficient gene flow between stocks to counteract genetic differentiation (McQuinn, 1997). However, in a recent study we constructed an exome assembly and used this in combination with whole genome sequencing of eight population samples and found more than 400,000 SNPs (Lamichhaney et al., 2012). We confirmed lack of differentiation at most loci, whereas a small percentage showed highly significant differentiation. Simulations demonstrated that the distribution of fixation index (FST)-values among herring populations deviated significantly from expectation for selectively neutral loci. Figure 1 with 1 supplement see all Download asset Open asset Demographic history and phylogeny. (A) Geographic location of samples. The salinity of the surface water in different areas is indicated schematically. Autumn spawners are marked with an asterisk. (B) Demographic history. Black circles indicate effective population size over time estimated by diCal (Sheehan et al., 2013); estimates are averages from four arbitrarily chosen genomic regions. The grey field is confidence interval ( ± 2 sd), while light grey lines show the underlying estimates from each genomic region. (C) Neighbor-joining phylogenetic tree. The evolutionary distance between Atlantic and Pacific herring was calculated using mtDNA cytochrome B sequences; right panel, zoom-in on the cluster of Atlantic and Baltic herring populations. Colour codes for sampling locations are the same as in Figure 1A. (D) Global distribution of FST –values based on 19 populations of Atlantic and Baltic herring. The inset illustrates the tail of the distribution. The mean and median of this distribution are indicated. To reduce the FST sampling variance, we only used SNPs with ≥30x coverage in each population. https://doi.org/10.7554/eLife.12081.003 Genetic studies of ecological adaptation in natural populations is challenging because genetic differentiation caused by natural selection is often confounded with genetic differences due to genetic drift caused by restricted effective population sizes. An ideal species for studying the genetic basis of ecological adaptation should comprise subpopulations of infinite size and exposed to different ecological conditions. In such a species there is minute genetic drift and genetic differentiation is caused by selection resulting in local adaptation. The herring is close to being such an ideal subject for studies of ecological adaptation due to the extremely low levels of genetic differentiation at most loci as documented in previous studies (Andersson et al., 1981; Ryman et al., 1984; Larsson et al., 2007; 2010; Limborg et al., 2012; Lamichhaney et al., 2012). This unique opportunity together with herring being such a valuable natural resource prompted us to generate a genome assembly and perform genome sequencing of populations adapted to different ecological conditions. Here we present a high-quality genome assembly for the Atlantic herring, and results of whole genome sequencing of 20 population samples using pooled DNA. The results were verified by individual genotyping using a custom-made 70k SNP array. Our study addresses two fundamentally different types of adaptations; one example of niche expansion (adaptation to low salinity), and one example of sympatric balancing selection (variation in the timing of reproduction). The results provide a comprehensive list of hundreds of independent loci underlying ecological adaptation and shed light on the relative importance of coding and non-coding variation. The results have important implications for sustainable fishery management, and provide a road map for cost effective high-resolution characterization of genetic diversity in natural populations. Results Genome assembly and annotation Clupeiformes represents an early diverging clade of the otomorpha (Near et al., 2012) (Figure 2A). The genome size for herring has been estimated at ~850 Mb (Hinegardner and Rosen, 1972; Ida et al., 1991; Ohno et al., 1969) with no recent whole genome duplications reported. We performed whole genome assembly based on short read sequencing of libraries ranging from 170 bp to 20 kb insert sizes (Supplementary file 1A). The 808 Mb assembly had a scaffold N50 of 1.84 Mb with 23,336 predicted coding gene models. It showed a high degree of completeness based on RNAseq alignments, core gene analyses and comparisons to other fish gene sets (Table 1, Supplementary files 2, 3A–D, Figure 2B, Figure 2—figure supplements 1–2). The GC content was 44%, and repetitive elements made up 31% of the assembly (Table 1). Alignments of synthetic long reads (SLRs; Illumina) failed to significantly improve the assembly due to coincidental gaps between the assembly and the SLRs, but proved useful in phasing parental alleles (Materials and methods; Figure 2—figure supplements 3–4) and dramatically improved the discovery of indels larger than 30 bp compared to short Illumina reads (Supplementary file 1F). We identified 150 endogenous retroviruses (ERVs) constituting ~0.14% of the genomic sequence but none included open reading frames in all gag, pol and env genes (Supplementary file 1, Figure 2—figure supplement 5). Figure 2 with 5 supplements see all Download asset Open asset Genome assembly and annotation. (A) Phylogeny of ray-finned fishes (Actinopterygii) from the Devonian to the present, time-calibrated to the geological time scale based on Near et al. (2012). Geological abbreviations: C (Carboniferous), CZ (Cenozoic), D (Devonian), J (Jurassic), K (Cretaceous), Ng (Neogene), P (Permian), Pg (paleogene) and Tr (Triassic). Dating of the specific rounds of whole genome duplication is based on Glasauer and Neuhauss (2014). Abbreviations: Ts3R (teleost-specific third round) and Ss4R (salmonid-specific fourth round) of duplication. The number of species with a genome assembly available is marked within parentheses after their group's name. Atlantic herring belongs to Clupeiformes, the order indicated in red letters. (B) Orthologous gene families across four fish genomes (C. harengus, D. rerio, L. chalumnae and G. morhua). https://doi.org/10.7554/eLife.12081.005 Table 1 Summary of the herring assembly compared to other sequenced fish genomes. https://doi.org/10.7554/eLife.12081.011 SpeciesHerring (Clupea harengus)Zebrafish (Danio rerio)Cod (Gadus morhua)Coelacanth (Latimeria chalumnae)Stickleback (Gasteosteus aculeatus)Estimated genome size (Mb)8501,454a830b3,530c530dAssembly size (Mb)8081,412753b2,861e463fContig N50 (kb)21.325.02.812.783.2Scaffold N50 (Mb)1.841.550.690.9210.8Sequencing technologygIS+IR+IISRepeat content30.952.225.427.725.2%GC content44.136.745.443.044.6Heterozygosity1/309n.a.1/5001/4351/700Protein-coding gene count23,33626,45922,15419,03320,787 a(Freeman et al., 2007; Vinogradov, 1998; Howe et al., 2013) b(Star et al., 2011) cGenome size calculated as pg x 0.978 × 109 bp/pg; picogram values taken from Cimino and Bahr (1974) d(Vinogradov, 1998; Jones et al., 2012) e(Amemiya et al., 2013) f(Jones et al., 2012) gI=Illumina sequencing; S=Sanger sequencing; R=Roche 454 n.a.=not available Population genetics and demographic history Whole genome pooled sequencing was done using 20 population samples of herring from the Baltic Sea, Skagerrak, Kattegat, North Sea, Atlantic Ocean and Pacific Ocean (Figure 1A; Table 2); the latter sample represents the closely related Pacific herring (Clupea pallasii). Each pool comprised 47–100 fish and was sequenced to ~30x coverage. Furthermore, 16 fish, eight Baltic and eight Atlantic herring (Table 2), were sequenced individually to ~10x coverage. All data were aligned to the reference assembly and SNPs were called after rigorous quality filtering. We found 8.83 million SNPs when Pacific herring was included and 6.04 million among Atlantic and Baltic herring. Table 2 Samples of herring used for whole genome resequencing. https://doi.org/10.7554/eLife.12081.012 LocalityaSamplenPositionSalinity (‰)Date (yy/mm/dd)Spawning seasonBaltic SeaGulf of Bothnia (Kalix)bBK47N 65°52'E 22°43'3800629springBothnian Sea (Hudiksvall)BU100N 61°45'E 17°30'6120419springBothnian Sea (Gävle)BÄV100N 60°43'E 17°18'6120507springBothnian Sea (Gävle)BÄS100N 60°43'E 17°18'6120718summerBothnian Sea (Gävle)BÄH100N 60°44'E 17°35'6120904autumnBothnian Sea (Hästskär)cBH50N 60°35'E 17°48'6130522springCentral Baltic Sea (Vaxholm)bBV50N 59°26'E 18°18'6790827springCentral Baltic Sea (Gamleby)bBG49N 57°50'E 16°27'7790820springCentral Baltic Sea (Kalmar)BR100N 57°39'E 17°07'7120509springCentral Baltic Sea (Karlskrona)BA100N 56°10'E 15°33'7120530springCentral Baltic SeaBC100N 55°24'E 15°51'8111018unknownSouthern Baltic Sea (Fehmarn)bBF50N 54°50'E 11°30'12790923autumnKattegat, Skagerrak, North Sea, Atlantic OceanKattegat (Träslövsläge)bKT50N 57°03'E 12°11'20781023unknownKattegat (Björköfjorden)KB100N 57°43'E 11°42'23120312springSkagerrak (Brofjorden)SB100N 58°19'E 11°21'25120320springSkagerrak (Hamburgsund)bSH49N 58°30'E 11°13'25790319springNorth SeabNS49N 58°06'E 06°10'35790805autumnAtlantic Ocean (Bergen)bAB149N 64°52'E 10°15'35800207springAtlantic Ocean (Bergen)cAB28N 60°35'E 05°00'33130522springAtlantic Ocean (Höfn)AI100N 65°49'W 12°58'35110915springPacific OceanStrait of Georgia (Vancouver)PH50--35121124- aPlaces where the sample was landed (if known) are given in parenthesis bSamples from previous study (Lamichhaney et al., 2012) cEight Baltic herring from the BH sample and eight Atlantic herring from the AB2 sample were used for individual sequencing n=number of fish Average nucleotide diversity was estimated by counting the frequency of heterozygous sites in the reference individual after stringent filtering for sequence quality and coverage (within one standard deviation of mean coverage). The estimate was one heterozygous site per 309 bp, giving a nucleotide diversity of 0.32%; no estimate based on the 16 herring sequenced individually deviated significantly from this value and there was no significant difference between Atlantic and Baltic herring. The average decay of linkage disequilibrium between loci was very steep, with average r2 falling to 0.1 at a distance of 100 base pairs (Figure 1—figure supplement 1A). The allele frequency distribution deviated significantly from the one expected for selectively neutral alleles at genetic equilibrium (p<2x10-16, Kolmogorov-Smirnov test), due to an excess of rare alleles (Figure 1—figure supplement 1B) consistent with population expansion. The result is supported by the genome-wide distribution of Tajima's D, which shows a global shift towards negative values (mean=−0.57 ± 0.01; Figure 1—figure supplement 1C). A demographic analysis using the diCal software (Sheehan et al., 2013) confirmed that herring have experienced an expansion in effective population size, roughly five- to ten-fold, and that the current Ne is on the order of 106 individuals (Figure 1B); the results for Baltic and Atlantic herring were essentially identical. The result indicates that the effective population size minimum occurred at around one to two MYA, after the onset of the Quaternary ice age. Phylogeny The neighbor-joining phylogenetic tree including Atlantic, Baltic and Pacific herring shows a large phylogenetic distance between Pacific and Atlantic herring, as compared with the tiny genetic divergence among samples of Atlantic and Baltic herring (Figure 1C). We estimated the split between Atlantic and Pacific herring to ~2.2 million years ago based on mtDNA cytochrome B sequence divergence. The phylogenetic tree is consistent with minute differentiation at selectively neutral loci in Atlantic herring (Ryman et al., 1984; Lamichhaney et al., 2012); all subpopulations in the Eastern North Atlantic may have expanded from a common ancestral population after the last glaciation as indicated by demographic analysis (Figure 1B). A closer examination of the tight cluster of Atlantic and Baltic herring populations reveals some structure consistent with geographic origin (Figure 1C). Samples from the Baltic Sea cluster on one half while samples from marine waters cluster on the other half of the tree. Only three populations are located at intermediate positions. Two of these are autumn-spawners from the Baltic Sea (BÄH and BF), indicating that autumn-spawning herring are genetically distinct from spring- and summer-spawning herring. The third sample (KT) at an intermediate position was sampled outside the spawning season and at the border between Kattegat and Baltic Sea and may represent a mixed sample of local Kattegat population and fish that spawn in the Baltic Sea but migrate into Kattegat for feeding. Genetic adaptation to a new niche environment The Atlantic (Clupea harengus harengus) and Baltic herring (Clupea harengus membras) were classified as subspecies by Linnaeus (1761) in the 18th century. They are adapted to strikingly different environments, in particular regarding salinity that ranges from 2–3‰ in the Gulf of Bothnia to 12‰ in Southern Baltic Sea, whereas salinity in Kattegat, Skagerrak, North Sea and Atlantic Ocean is in the range 20‰–35‰ (Figure 1A; Table 2). To reveal loci underlying genetic adaptation associated with the recent niche expansion into brackish waters after the last glaciation we compared allele frequencies, SNP by SNP, in two superpools: one Atlantic including all populations from Atlantic Ocean, Skagerrak and Kattegat and a pool comprising all samples collected in Baltic Sea; this is justified by low differentiation at neutral loci as documented by the low FST-values when comparing all samples of Atlantic and Baltic herring (Figure 1D). Samples of autumn-spawning herring, a possible confounding factor, were excluded from the analysis. We used a stringent significance threshold of p<1x10-10 (Bonferroni correction, p=8.2x10-9). We identified 46,045 SNPs that showed an allele frequency difference with p<1x10-10 in the χ2 test (Figure 3A; Supplementary file 3A). An important question is how many independent loci these represent. A conservative estimate of 472 independent loci was obtained (i) by only using SNPs with p<1x10-20, (ii) by taking into account gaps in the assembly and (iii) by using the Comb-P software (Pedersen et al., 2012) to combine strongly correlated SNPs from the same genomic region (see Materials and Figure illustrates one of the most For a large of scaffold there are no significant differences among Atlantic and Baltic samples whereas there are allele frequency differences over a kb this is a for indicating that genetic adaptation typically as large blocks, often including multiple A phylogenetic tree based on SNPs showing genetic differentiation between Atlantic and Baltic (Figure from the tree based on all SNPs (Figure 1C). the of the two autumn-spawning populations and from the Baltic Sea, the position of all other populations the in salinity with the population samples from the North Sea and Atlantic Ocean at one of the tree and samples from the brackish Baltic Sea at the other and with samples from Skagerrak and Kattegat at intermediate positions. The low genetic differentiation among Baltic the two autumn-spawning populations and that adaptation to brackish waters is a Figure with 2 supplements see all Download asset Open asset Genetic differentiation between Atlantic and Baltic herring. (A) of significance values for allele frequency differences between of herring from marine waters Skagerrak, Atlantic Ocean) the brackish Baltic Sea. panel, for scaffold both and FST-values are (B) Neighbor-joining phylogenetic tree based on all SNPs showing genetic differentiation in this (C) of allele in five strongly regions. The major allele in the sample (Atlantic Ocean) was used as reference at each panel, neighbor-joining tree based on by SNPs from scaffold (D) map showing number the gene. of is marked with an the position of SNPs significant in the χ2 test is indicated by Population samples and salinity at sampling locations are indicated to the are explained in Table genetic differentiation between Atlantic and Baltic herring in a region of significance based on the χ2 test is indicated. Figure shows estimated allele for highly SNPs from five genomic in population each region showing an underlying genetic architecture with large and defined The Atlantic Ocean and North Sea samples are both nearly for the reference allele at these In the samples of Baltic herring were close to fixation for the the sample collected in Skagerrak is most to the Atlantic Ocean and North Sea but shows a towards more intermediate allele at these loci. We a 70k SNP to study in more and to use data from individual fish to by pooled The included neutral SNPs across the genome and SNPs showing genetic differentiation between fish each from populations were used in the SNP was an between allele estimated with pooled sequencing and with the SNP (Figure supplement 1). We constructed a phylogenetic tree (Figure for of highly SNPs from scaffold present among individual fish from after phasing using and expected all fish from Atlantic Ocean and North Sea closely related Two major groups were present among Baltic herring and with few Baltic herring only from Skagerrak Atlantic but with a of Baltic for other are in Figure supplement are many and ecological differences between Atlantic Ocean and Baltic Sea of the Baltic Sea, and but the most difference salinity. We used the and 2013) software to reveal which of the 472 independent loci with the χ2 test showed the most consistent with salinity. This analysis identified SNPs from independent with highly significant to salinity (Supplementary file 3A). of the genes in these have been associated with in and of these genes showed expression in in or water (Supplementary file 3A). Here we present three loci with to salinity. Firstly, the kb region in scaffold (Figure a that is for but has a for in fish and in et al., Secondly, genetic differentiation was also observed at scaffold (Figure 3A; This high This was also identified as one of the most region in for changes (Supplementary file A kb region including of the coding sequence showed a number that had a negative with salinity (Figure The Pacific herring, showed an intermediate the Pacific herring in waters et al., often in and where salinity in to Atlantic herring. is a also that the of the and of this in to salinity has been et al., In herring, we found no coding changes In of the region is expected to gene of the is a challenging of development for a marine fish to brackish conditions. a kb region of shows with salinity (Figure Supplementary file 3A). which has an in is associated with in and shows expression in between in or water et al., 2014). Genetic basis underlying timing of spawn from early to to this study it was spawning time is due to by and or genetic factors contribute (McQuinn, 1997). For it has been that spawning time in the Baltic Sea is by of the affecting of fish to spawning To study this important question we collected spawning herring from the same geographic close to in and (Table 2). Our sampling included two other autumn-spawning populations collected in one from North Sea and the other from Southern Baltic Sea. We two including three autumn-spawning and spring-spawning population the and one population of herring in Table were excluded from the analysis. We identified SNPs with significant allele frequency differences between and with number (Figure the highly SNPs at least independent loci based on (see Materials and The result for the time that and spring-spawning herring are genetically distinct and indicates that genetic factors affect spawning In a phylogenetic tree based on these SNPs the autumn-spawning populations from the Baltic Sea and North Sea to cluster with spring-spawning herring from the Atlantic Ocean (Figure Figure with 2 supplements see all Download asset Open asset Genetic differentiation between spring- and autumn-spawning herring. (A) of significance values for allele frequency (B) Neighbor-joining phylogenetic tree based on all SNPs showing genetic differentiation in this (C) of allele in four strongly regions. The major allele in the sample (Atlantic Ocean) was as reference at each and have been in this since it was that they were in scaffold is present kb of and kb of (D) Neighbor-joining tree based on by SNPs from scaffold same populations as in Figure

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.