The Swedish short-tailed sheep populations represented in this study were: Dalapäls, Fjällnäs, Gotland, Gute, and Klövsjö sheep breeds. These five breeds represented different geographic origins within Sweden and different population sizes. Locations of the origins of breeds have been included in Supplementary Figure 1 and population information on registered individuals in Sweden have been included in Supplementary Table1.
Dalapäls (also known in English as Dala fur or Dala pelt) sheep are from Dalarna (a county located in the middle of Sweden that shares a border with Norway) and have been bred for their skin which was used in traditional costumes. Their wool is curly and finer than that of most other local Swedish breeds. These sheep are usually white and males are horned while females are usually polled.
Accepted officially as a breed in 2011, Fjällnäs sheep are currently found in Norrbotten, a county bordering with Finland. Individuals are light fawn in colour and males have small brittle horns while females are polled. The breed has only 10 adults in 2011 and all of these individuals were genotyped and included in this study.
Gotland and Gute sheep originate from the same population of sheep on the island/county of Gotland in the Baltic Sea. Polled sheep became popular and eventually became the Gotland breed and in order to preserve the primitive sheep, the remaining horned individuals were gathered in the 1920s to become Gute sheep breed. Today Gute sheep are bred for conservation purposes while the Gotland sheep are bred for improved pelt quality (colour and curl of the fibre). Many Gotland and Gute sheep are some shade of grey with the Gotland sheep fleece more uniform in colour. Gotland males and females are polled while Gute males and females are horned.
Klövsjö sheep are from the surrounding area of a village, Klövsjö, in Jämtland (a county north of Dalarna, also sharing a border with Norway). These sheep are usually black or white and both sexes are usually polled.
Blood samples were collected from native Swedish sheep. Ethical permission was obtained prior to the collection of blood samples. Genomic DNA used in this study was extracted from fresh blood samples using a QIAsymphony® DSP DNA extraction robot and DSP DNA Midi Kit (QIAGEN). For this study 21 Dalapäls, 10 Fjällnäs, 19 Gotland, 22 Gute and 21 Klövsjö sheep were genotyped for 653,305 autosomal SNPs (Ovine Infinium® HD SNP BeadChip) at Labogena, INRA, Paris, France. Of the 22 Gute sheep genotyped: 19 sheep were genotyped for seven microsatellite markers and included in a study by Rochus and Johansson ; and eight sheep had retroviruses and microsatellites genotyped and were included in both studies by Mukiibi et al.  and Rochus and Johansson . Of the 21 Klövsjö sheep included in this study, three sheep had retroviruses genotyped and were included in Mukiibi et al. .
Quality control and filtering of SNPs included removing SNPs with a minor allele frequency of zero, removing SNPs with a missing rate greater than 0.01 and removing sheep with a missing genotype rate greater than 0.15 using PLINK 1.9 [18, 19]. All of the 93 sheep, and a total of 502,144 SNPs passed quality control and filtering steps.
Genetic diversity measures were estimated using PLINK 1.9 [18, 19] including the number and size of runs of homozygosity (ROH) (using the following settings: max inverse density (kb/SNP) = 1000, minimum length = 100, minimum SNP count = 20, scanning window size = 20, maximum heterozygotes in a scanning window = 1), and three inbreeding estimates: Fhat1 (variance-standardized relationship minus 1), Fhat2 (heterozygosity), and Fhat3 (based on the correlation between uniting gametes).
Three methods were used to study population structure in the five Swedish breeds: principal components analysis using the software PLINK 1.9 [18, 19], a mixture model to estimate individual ancestry coefficients estimated using the software sNMF (R version 2.0)  (using default settings) and a population tree model which also estimates mixture events determined using the software Treemix 1.13  (grouping SNPs by 100 and using the Asian Mouflon as the tree root). For the population tree model we included SNP frequencies from Asian Mouflon sequences, which was data generated by the NEXTGEN consortium (http://nextgen.epfl.ch/).
We then combined our dataset of Swedish breeds with a French sheep dataset (also genotyped with Ovine Infinium® HD SNP BeadChip, first presented in a study by Rochus et al.  and available from https://zenodo.org/record/237116) and European and southwestern Asian sheep (genotyped for 49,034 SNP first presented in  available from http://www.sheephapmap.org). Only the SNPs shared between all datasets were used to study population structure using principal components analysis [18, 19], estimating individual ancestry coefficients  and estimating a population tree . We included 38,589 SNPs and 2938 sheep from 87 populations. For the population tree model, we again included the SNP frequencies from Asian Mouflon sequences to root the tree.
We estimated the historic effective population sizes of the five native Swedish breeds and two French populations (Texel and Romanov). We included the two French populations for comparative purposes. Texel sheep are a common breed originating from the Netherlands but is bred all around the world today for commercial meat production. Romanov sheep are a north European short-tailed sheep breed that originated in Russia. This breed is also used more widely. We estimated historic effective population size using the default settings (MAF ≥0.05, minimum distance between SNPs: 50 kb, maximum distance between SNPs: 4000 kb, and ≤ 100,000 SNPs for each chromosome) using the software SNeP 1.1 .