Genetic diversity indices and demographic inferences
After applying the filtering criteria, the final dataset for the Swedish native breeds consisted of 143 samples and 111,914 SNPs. The genetic diversity summary statistics are in Table 1. A majority of the Swedish native cattle breeds displayed a high proportion of polymorphic loci (Pm), ranging from 0.7981 in Väne to 0.9736 in Swedish Holstein-Friesian. This indicates that the SNPs selected for the GGD HD150K array are highly informative for Swedish cattle breeds. Similarly, the expected heterozygosity (He) is relatively high with values ranging from 0.3041 in the Väne breed to 0.3883 in the Swedish Holstein-Friesian breed.
Demographic inferences in Swedish cattle breeds were made based on the analyses of ROH and LD decay. The ROH profile (ROH count and cumulative ROH length) varied a lot both between and within the Swedish cattle breeds (Fig. 2a and see Additional file 1: Table S3). In Fjällnära, the ROH count and cumulative ROH length ranged from 8 to 78, and from ~ 6 Mb to ~ 1.3 Gb, respectively. In Swedish Holstein-Friesian, ROH count ranged from 2 to 31, and cumulative ROH length ranged from ~ 6 Mb to ~ 295 Mb, while the corresponding numbers in Fjäll were 1 to 45, and ~ 4.5 Mb to ~ 310 Mb, respectively. These results indicate relatively large/diverse ancestral populations for Swedish Holstein-Friesian and Fjäll breeds. Swedish Red Polled individuals also displayed a large variation in ROH profile (ROH count ranged from 1 to 74, and cumulative ROH length ranged from ~ 4.3 Mb to ~ 1.3 Gb). Moreover, the cumulative ROH size of some Swedish Red Polled individuals (such as Individual SRP5 and SRP 8) was similar to that of other individuals (such as VAC7 and RMC13), but with lower ROH counts (see Additional file 1: Table S3), which indicates that some mating has occurred between closely related individuals. Conversely, Väne and Ringamåla cattle display relatively large ROH counts and cumulative ROH size, which indicates genetic isolation and a relatively small founder population. As expected, inferences that are drawn based on the pattern of pairwise LD decay (Fig. 2b) are consistent with those from ROH profiles. For example, for the Ringamåla and Väne breeds, we inferred a low level of haplotype diversity since they have the highest r2 values at all pairwise SNP distances and an overall slow LD decay. For the Fjäll, Swedish Holstein-Friesian and Swedish Red breeds, we inferred a wide haplotype diversity since they have the lowest r2 value at all pairwise SNP distances and a rapid LD decay.
Genetic structure and relationships among Swedish native cattle breeds
To assess the genetic relationships among the Swedish native cattle breeds, first we performed a PCA. The first principal component (EV1), which explains 5.76% of the total variance, clearly separated the Fjäll, Swedish Polled, Fjällnära and Bohus Polled breeds from the Swedish Red, Väne and Ringamåla breeds. The second principal component (EV2), which explains 3.03% of the total variance, separated the Swedish Holstein-Friesian breed from the cluster of native Swedish cattle breeds. Furthermore, among the native Swedish cattle breeds, Swedish Red Polled (which occupies the central position in the plot) and Väne clearly form separate clusters (Fig. 3a), whereas the Fjällnära samples are dispersed across the plot, which indicates either sub-structure or large variability within the population. Moreover, the genetic differentiation between Fjäll, Swedish Polled and Bohus Polled cattle, and that between Ringamåla and Swedish Red cattle are low.
The NJ-tree constructed based on Fst (Fig. 3b and see Additional file 2: Figure S1), and the ML-based phylogenetic tree (Fig. 3c) are in agreement with the patterns identified by using PCA (Fig. 3a). In general, these phylogenetic trees divide the Swedish native cattle breeds into two clusters: one cluster that comprises the Swedish Mountain cattle breeds, i.e. Fjäll, Swedish Polled, Fjällnära and Bohus Polled, and one cluster that includes the southern Swedish cattle breeds with horns—Swedish Red, Väne and Ringamåla. In particular, these trees also suggest close relationships between the Bohus Polled and Fjäll breeds, and between the Ringamåla and Swedish Red breeds. Moreover, the long branch observed for Väne cattle indicates considerable genetic drift/divergence. Interestingly, similar to the PCA pattern, the Swedish Red Polled breed displays an ambiguous clustering pattern, i.e. in the ML-based phylogenetic tree (Fig. 3c), it does not cluster with any of the Swedish cattle breeds, whereas in the Fst-based phylogenetic tree, it forms a sister clade with the cluster of Swedish mountain cattle breeds.
The clustering pattern (Fig. 4) based on shared ancestry that was inferred from the ADMIXTURE analysis also indicates a first split (at K = 2) between the Swedish mountain breeds (Fjäll, Swedish Polled, Fjällnära, and Bohus Polled) and the southern Swedish breeds (Swedish Red, Väne and Ringamåla). Conditioning the ancestral population to K = 3 separate the Swedish Holstein-Friesian breed from the clusters of native Swedish cattle breeds, and at K = 4, the Väne breed form a separate cluster. Starting from a K value of 5, we observe sub-structures within the Fjällnära breed. Furthermore, at all K values, the Ringamåla and Swedish Red breeds share a remarkably similar distribution of genetic variation. Likewise, Fjäll, Swedish Polled and Bohus Polled cattle also show a similar distribution of genetic variation, which indicates ancestral relatedness. It is worth noting that Swedish Red Polled cattle display a mosaic of ancestries between K = 2 to 5 but forms a separate cluster at K = 6.
Haplotype sharing analysis by using ChromoPainter and fineStructure reveals a pattern (Fig. 5) that not only reinforces the findings of all our previous analyses but also refines the clustering pattern of the Swedish native cattle breeds. In particular, these results helped to characterize in more detail the low differentiation between the Swedish mountain cattle breeds. In particular, we observed a low differentiation between Fjäll and Swedish Polled cattle. Interestingly, refined investigation of the Fjällnära individuals revealed four sub-clusters, which correspond to the farms from which the sampled individuals had their ancestries. Furthermore, similar to our previous analyses, the Ringamåla, Swedish Red, and Väne breeds are loosely included in the same cluster. However, Ringamåla and Swedish Red seem to be genetically closer to each other than to Väne. The Swedish Red Polled and Swedish Holstein-Friesian breeds seem to cluster together, but based on the Fst and PCA analyses, we hypothesize that this is due to the clustering of two other groups (Swedish mountain and southern Swedish breeds).
Genetic relationships between Swedish cattle breeds and other European cattle breeds
We investigated the relationships between Swedish cattle breeds and other European cattle breeds through analyses based on PCA, ADMIXTURE, and phylogenetic trees. The other European cattle breeds mainly originated from north-western Europe, and Norway (see Additional file 1: Table S2). In PCA, the first principal component, which explains 6.25% of the total variance, separated the Jersey breed from all the others (Fig. 6a). This can be attributed to the fact that the Jersey breed has remained genetically isolated for many generations. The second principal component (PC2) separated “red-coloured” cattle breeds (Swedish Red, Ringamåla, Väne and Norwegian Red) from the remaining breeds. As expected, the Swedish Holstein-Friesian individuals cluster with the Dutch cattle breeds. The clustering of “red-coloured” cattle breeds and of the Holstein-Friesian-derived breeds is also observed in the phylogenetic trees (Figs. 3c and 6b). Interestingly, we did not identify any historical relatedness between the Swedish mountain cattle breeds and the European cattle breeds studied here, which probably means that their contribution was not significant.