New World cattle show ancestry from multiple independent domestication events

Results

Our samples of New World cattle included Texas Longhorn cattle (n = 114), Mexican Corriente cattle (n = 5), and Colombian Romosinuano cattle (n = 8). To place these individuals in a global phylogeographic context, we also included previously published data from individuals of 55 other breeds (n = 1,332; Table 1) (6). These cattle were genotyped for nuclear SNP loci across all 29 autosomal chromosomes using the Illumina BovineSNP50 BeadChip, the Illumina 3K chip, or 6K chip. We analyzed two datasets: one (termed the 1.8k dataset) included 1,814 SNP loci present on all three chips, and the other (termed the 50k dataset) included 47,506 SNP loci from the Bovine SNP50 chip. The 1.8k dataset included more extensive sampling of Texas Longhorn cattle (n = 114) compared with the 50k dataset (n = 40), but a less thorough sampling of the genome.

Average heterozygosity within breeds ranged from 15% (SD, 1%) in the indicine breed Gir, to 30% (SD, 1%) in the taurine Belgian Blue cattle (Table S1). The highest heterozygosity was 31% (SD, 1%) in the recent hybrid Beefmaster. Generally, as expected from the ascertainment panel for the SNP chip (39), taurine breeds had higher heterozygosity. Breeds of taurine origin averaged heterozygosity of 27%, whereas breeds of indicine origin averaged heterozygosity of 16%. Across New World cattle, average heterozygosity was 28% (Texas Longhorns: 29%, SD, 2%; Corriente; 27%, SD, 2%; Romosinuano: 27%, SD, 1%).

Discussion

Simulations have demonstrated that inference of complex historical migration models using PCA is difficult because multiple processes can result in the same patterns (9, 41). Indeed, even under relatively simple scenarios, such as the admixture between two ancestral groups, admixed individuals can be incorrectly assigned to a third group that appears to be geographically intermediate (42). However, when ancestral groups are known, coalescent estimates of admixture between distinct populations are mathematically straightforward. McVean (43) showed that, although PCA is a nonparametric analysis method, coordinates can be predicted from pairwise coalescence times between individuals. This prediction allows a genealogical interpretation to principal component scores. The first principal component can be interpreted as the deepest coalescent event in a tree, and the projection of admixed individuals onto this axis can be used to estimate the proportion of mixture between two parental groups (43). As a test case, we were able to correctly reconstruct the known ancestry of recent taurine–indicine hybrid breeds created for agricultural purposes: Santa Gertrudis (group 45), a Brahman-Shorthorn cross developed in 1918, and Beefmaster (group 46), a cross between Hereford, Shorthorn, and Brahman cattle developed in the 1930s (the “Hybrid” groups shown in Figs. 1 and 2). In addition, we were able to recover the taurine–indicine hybridization cline across Africa along the first principal component (PC1) shown in Fig. 1.

The second principal component (PC2) shown in Fig. 1 separates Eurasian from African cattle, indicating a distinctive genomic component in African breeds. Our samples of African cattle breeds all appear to have admixed taurine–indicine ancestry, based on the intermediate position of African cattle on PC1 (Fig. 1) and the STRUCTURE analyses when K = 2 (Fig. 2). However, the distinctiveness of northern African breeds on PC2 (Fig. 1), as well as in the STRUCTURE analyses when K = 3, indicates additional genomic differentiation in northern African cattle. If African breeds are derived entirely from a mixture of European and Asian cattle, this differentiation must have occurred after the importation of domestic cattle to Africa. Alternatively, this unique African component may be derived from additional domestication events involving north-African aurochsen, as has been suggested previously (16, 26, 27).

Both the principal components analysis (Fig. 1) and the model-based STRUCTURE analyses (Fig. 2; Fig. S1) support a hybrid ancestry for New World cattle, although the patterns of hybridization are distinct from the recently constructed hybrid breeds. New World cattle are largely of taurine descent, but they exhibit an average of 11% indicine ancestry (as estimated from the STRUCTURE analyses of the 1.8k data, with K = 2). In this regard, New World cattle are much like some modern breeds from southern Europe. However, when K = 3 in the STRUCTURE analyses (Fig. 2), much of this “indicine” component in southern European and New World cattle appears to be more specifically associated with cattle from northern Africa. The PCA is also consistent with the hypothesis that New World cattle (as well as modern breeds from southern Europe) are influenced by ancestral gene flow from northern Africa, based on the placement of these breeds at intermediate positions along PC1 and PC2 in Fig. 1.

The pattern of African admixture in southern Europe is consistent with movement of cattle across the Straits of Gibraltar during the Moorish invasion and occupation of the Iberian peninsula in the 8th to 13th centuries CE (6, 18, 44). However, sequencing of Bronze Age cattle mtDNA from Spain suggests that earlier African introgression into Iberia may also have occurred (45). The elevated disease resistance of Texas Longhorn cattle (compared with northern European cattle breeds that have been imported to southwestern North America) may be partially related to the portions of their genomes that stem from this African ancestry. African N’Dama cattle also exhibit some substantial types of disease resistance (46), consistent with this hypothesis.

Using model-based clustering analyses, we found Spanish-derived New World cattle breeds—Texas Longhorns (group 39), Corriente (group 41), and Romosinuano (group 42)—did not differ significantly in levels of indicine introgression (Fig. 3 and Table S1). The Brazilian breeds Nelore (group 54) and Guzerat (group 56) are recently developed breeds from indicine stock, which is reflected in our estimates of their ancestry. All sampled New World cattle that are descended from old Spanish imports (114 Texas Longhorns, 5 Corriente, and 8 Romosinuano) show indicine ancestry (estimated by STRUCTURE, with K = 2). As well, these breeds group together in our PCAs in a position consistent with African introgression. This result suggests that introgression from African cattle occurred before the introduction of these cattle to the New World. This conclusion is supported by the STRUCTURE analysis of the breeds sampled from southern Europe, particularly Italy, which also show indicine and African ancestry. In fact, among the breeds that we sampled, and using the coarse geographical resolution of “country,” we found a significant correlation between latitude in Europe and degree of indicine introgression as estimated from STRUCTURE at K = 2.

The signal of “indicine” introgression in southern Europe may be somewhat misleading, however, depending on the complexity of domestication history in African cattle. In our STRUCTURE analyses at K = 3, the variation captured by the African-like group was not a subset of either of the groups distinguished at K = 2, as would be expected from a strictly bifurcating evolutionary process. The African subdivision at K =3 is at least partly composed of hybrid taurine–indicine genotypes. However, if African cattle are partly derived from a third domestication event involving aurochsen from northern Africa, this deep divergence may be a more important driver of the differentiation between European and African cattle than is indicine introgression. In that case, the indicine component of African and European lineages at K = 2 may reflect African diversity rather than true indicine ancestry.

Additional analyses, including a more thorough sampling of African and Iberian cattle, are needed for a conclusive determination of the number of independent domestication events in cattle. Although our results cannot exclude the possibility of an independent domestication of aurochsen in northern Africa, the relatively low level of variation captured by the second principal component (2–3%) is more consistent with European and African taurine cattle both being derived primarily from a single domestication in the Middle East, with the likely continued but occasional incorporation of genetic material from wild aurochsen in both areas. However, our results do suggest that if there was a third distinct domestication event, it took place in Africa.

Achilli et al. (21) found a novel haplogroup in Italian cattle (Cabannina, not sampled here), for which the timing of divergence was consistent with introgression from European aurochsen. Although continued introgression of aurochs derived genetic material after the original domestication events probably led to greater diversity in European taurine cattle populations, this diversity is not expected to have been indicine-like and therefore is not the likely explanation for the indicine genetic component observed in southern European cattle.

There are at least two alternatives to our interpretation of introgression from Africa into Europe before the introduction of cattle to the New World: (i) relatively recent hybridization with indicine cattle in the New World (within the last 150 y); and (ii) direct importation of cattle from Africa to the Caribbean early in Spanish colonization. Although we cannot completely rule out the possibility of either of these alternatives, neither of these hypotheses explains the signal of shared ancestry between southern Europe and the Americas. Moreover, the first explanation is inconsistent with the clear African-like genomic component in the New World breeds. Finally, the admixture of genomes across chromosomes indicates ancient, rather than recent, introgression. Thus, the simplest explanation is that introgression of genetic material from African cattle occurred before the importation of cattle to the New World by Spanish colonists. The genetic diversity captured by this hybridization likely provided variation for selection when the ancestors of these animals were transported to North America in the late 1400s to early 1500s. However, there is individual variation among Texas Longhorn cattle, with some individuals showing elevated levels of indicine introgression (Fig. 2). This suggests that additional, more recent introgression with indicine cattle may also have occurred in some Texas Longhorn herds.

Our analyses made use of SNP data from across the genome. SNP-chip data have the advantage of being easily replicable, and data reuse across laboratories is straightforward, allowing results to be readily comparable. Furthermore, informative sequence data can inexpensively be generated, allowing investigators to sample many individual cattle. However, it is important to keep in mind the limits of these analyses. As the SNPs selected for the chip were chosen by resequencing individuals on an ascertainment panel, genetic diversity represented in that panel is expected to be overrepresented in future samples (47). In the case of the cattle 50k SNP chip, the ascertainment panel consisted mostly of taurine cattle. The SNPs were selected to be common polymorphisms in these animals (39); therefore, diversity estimates based on these data will overestimate diversity in taurine lineages and underestimate diversity in indicine lineages. Average heterozygosity within groups sampled in this study is consistent with this bias. Because of this bias, we did not attempt to estimate diversity metrics such as F_ST (48). In addition, the bias toward polymorphisms found in European taurine breeds and for alleles with high minor allele frequencies make these data inappropriate for identifying selective sweeps in New World cattle (39). Although mathematical methods have been developed to correct for ascertainment biases in some cases (49, 50), we did not have the appropriate data regarding the ascertainment process to do so in this case. Nonetheless, McVean (43) showed that, although ascertainment bias has an effect on principal component projections, it does not affect the relative placing of samples. Therefore, ancestry estimation by this method is robust to this source of bias.

Our results are complementary to previous work on the relationships and genetic diversity among cattle breeds (5, 6). Our conclusions match those of the Bovine HapMap Consortium (5) for the breeds that were sampled in both studies.

Our findings of introgression in New World Cattle breeds suggest that European–African admixture (which results in greater apparent divergence) may have driven the apparent sister group relationship between Texas Longhorns and all other European taurine cattle in some analyses presented by Decker et al. (6). Our results also suggest that finding may have resulted from imposing a tree-like structure on populations that arose through complex introgression events.

Although we have only a very sparse sampling of Asian cattle breeds from outside India, our results suggest that these animals are also of hybrid taurine–indicine origin. The possibility of introgression of genetic material from populations or species not sampled in our analysis limits our ability to make inferences about Asian cattle, but they promise to be an interesting area for future research. Although Kawahara-Miki et al. (51) suggested that Japanese cattle are sister to all other domesticated cattle, their omission of an indicine breed in their analyses makes this conclusion difficult to test. In addition, introgression among taurine and indicine lines would produce a similar result in a tree-based analysis.

The recent publication of the first Bos indicus genome sequence (52) will provide an opportunity to identify specific alleles of African or indicine origin that have contributed to the adaptation of New World cattle breeds. Such analyses will be of particular interest given the rapidly changing global climate. New World cattle in general, and Texas Longhorns in particular, are reported to exhibit resilience to drought and harsh climatic conditions (13, 53). Previous work has shown that New World cattle are an important reservoir of genetic diversity (37). As we show here, some of this diversity appears to derive from ancient introgression via African cattle.

Materials and Methods

Supplementary Material