This study was carried out in the province of Poni in the South-western administrative region of Burkina Faso. We selected three study sites with different management and breeding systems of Baoulé cattle and Baoulé x shorthorn Zebu crossbreds, including 27 villages from the Bouroum-Bouroum, Kampti and Loropéni departments (Figure 1). In the Bouroum-Bouroum department, we worked with 55 sedentary farmers of the ethnic group of Lobi, who keep purebred Baoulé and are the owners of these animals. In Kampti, we included 18 farms with mostly Baoulé x Zebu crossbreds and some pure Baoulé owned by the Mossi ethnic group, but herded by the transhumant Fulani people. Finally, in Loropéni 15 farms were included mainly with crossbred animals kept by Lobi and Djan breeders. As all samples were collected within close geographic distances (< 50 km) in the tsetse infested province of Poni, it is reasonable to assume that all individuals were exposed to the same trypanosome infection challenge.

A total of 737 blood samples, including 387 Baoulé and Baoulé x Zebu crossbreds from Bouroum-Bouroum, and 350 crossbreeds from Kampti and Loropéni were collected in EDTA tubes during the health monitoring activities of the “Characterization and Sustainable Utilization of Local Cattle Breeds” (LoCaBreed) project. DNA extraction from EDTA blood samples was performed with the MasterPureTM DNA Purification Kit for Blood Version II (Biozym Scientific, Oldendorf, Germany) following the manufacturer’s protocol. The trypanosomosis status was recorded by indirect ELISA test to diagnose positive or negative trypanosome infection in the blood samples (Desquesnes et al., 2003) resulting in a total of 377 positive and 360 negative animals (Table 1).

The genotyping of the 737 DNA samples with the Illumina Bovine SNP50 BeadChip was performed at Neogen (Lincoln, United States). Additional genotypes from 30 purebred Zebu and 35 crossbreds (Pérez et al., 2014) were included in the study to represent the two ancestral populations (Baoulé and Zebu) and to increase the number of crossbreds, summing to a total of 802 animals. Stringent quality filtering of the data was performed with PLINK 1.9 (Purcell et al., 2007; Chang et al., 2015). Specifically, the dataset was controlled to exclude non-autosomal SNPs, and SNPs with a minor allele frequency (MAF) lower than 0.05, a call rate less than 90% and those that deviated from Hardy Weinberg equilibrium with Fisher’s exact test with P-value 1 × 10 − 6. After quality control, 28,034 SNPs and 776 animals were available for subsequent analyses.

Unsupervised global ancestry estimation was performed with the full set of quality controlled SNPs using ADMIXTURE software (Alexander et al., 2009) with the number of ancestral populations (Baoulé and Zebu) fixed at two (K = 2). The admixture bar plots for ancestry proportions were created in R with the barplot function (The R Development Core Team, 2020). We calculated the frequencies of the admixture levels for all animals in Excel and plotted them in categories of 0.1 steps. We assigned 30 purebred Baoulé (global admixture levels ≥ 0.999 Baoulé) and 30 purebred Zebu (global admixture levels ≥ 0.987) as reference populations to investigate local admixture levels in 716 animals that were considered as potential crossbreds based on the sampling information. Animals found to be purebred (global admixture levels ≥ 0.995 Baoulé) were removed from the pool of crossbreds.

Local Ancestry in Mixed Populations (LAMP) is a program for estimating locus-specific ancestries in admixed individuals, using allele frequencies of the reference populations (Sankararaman et al., 2008). We applied the LAMP-ANC mode implemented in LAMP and provided the estimated allele frequencies files for Baoulé and Zebu as the purebred ancestral populations. LAMP-ANC is a modification of the LAMP mode and shows higher accuracy allowing triple mixing to be estimated, while LAMP cannot determine frequencies for more than two ancestral populations (Sankararaman et al., 2008). The following parameters were set: admixture proportions (alpha) = 0.8 for Baoulé and ≥ 0.2 for Zebu based on the global ancestry estimation using ADMIXTURE program, number of generations since admixture (g) = 2 and recombination rate (r) = 10^–8. We estimated the local ancestry proportion, as well as the “delta ancestry” with R in trypanosome positive and negative trypanosomosis animals following Khayatzadeh et al. (2016) using a custom script (see section “Data Availability”). The “delta ancestry” reflects the extreme fluctuations in ancestry differences across the genome, which are calculated by subtracting the genome-wide ancestry from locus-specific ancestry for each ancestry component. Such extreme fluctuations in ancestry differences are unlikely to have occurred by random genetic drift and potentially exhibit a selection signature in the admixed individuals (Tang et al., 2007). To identify significant deviations from the genome-wide average ancestry, we performed permutation tests (Doerge and Churchill, 1996) of the local ancestry proportions over the whole genome of admixed animals as proposed and carried out by Tang et al. (2007) in an admixed human population (Puerto Ricans) and replicated by Gautier and Naves (2011), Flori et al. (2014) in African Taurine, and Khayatzadeh et al. (2016) in composite cattle breed (Swiss Fleckvieh) to find significant thresholds for the deviations of local genetic ancestries from global ancestries. Separating animals with positive and negative trypanosomosis status, for each animal we concatenated the local ancestry estimations of all 29 autosomes and then permuted the circularized genome by cutting at a random location and rearranging the two resulting pieces of the genome for each individual independently. This type of permutation largely preserves the extent of Linkage Disequilibrium (LD), assuming that it is homogeneously distributed over the whole genome. We implemented 1,000 permutations. The distributions of maximum and minimum over all permutations were then used to define the 5 and 10% genome-wide thresholds levels that indicated significant deviation of the observed local ancestries from the genome-wide average ancestry (Tang et al., 2007; Gautier and Naves, 2011; Khayatzadeh et al., 2016).

We applied BayeScan 2.1 (Foll and Gaggiotti, 2008) to identify F_ST outlier loci putatively under selection between the trypanosome positive (n = 244) and negative (n = 266) crossbred animals using a cut-off at p < 0.05 corrected for a false discovery rate [FDR; (Benjamini and Hochberg, 1995)]. BayeScan uses a Reversible Jump Markov Chain Monte Carlo (RJ-MCMC) algorithm to obtain posterior distributions, with 100,000 iterations and a Burn-in length of 50,000 iterations. The regions within ± 0.5 Mb of the most significant SNPs were searched for any potential associated genes based on the ARS UCD1.2 Bos Taurus Genome Assembly on the NCBI database.

We aimed to identify SNPs with the highest F_ST differentiation between the 30 pure Baoulé and 30 pure Zebu. We re-filtered the original dataset for MAF < 10%, individual and genotype missingness < 10%, respectively. The F_ST values were calculated in PLINK following Weir and Cockerham (1984). With these we were able to provide a set of top 200 F_ST markers, which were then used as a starting point to manually remove markers less than 5 Mb to each other—preference given to higher F_ST markers. Based on this, we selected the top 15, 25, 50, and 100 SNPs, and extracted these for the crossbred animals and repeated the global admixture analysis (K = 2). We used the cor function in R to calculate the Pearson correlation coefficient (Pearson’s r) for pairwise determining the linear association between admixture levels (ancestry proportion) estimated based on the different sets of ancestry informative SNPs [all SNPs (35,952 SNPs) versus the top 100, 50, 25, 15 SNPs]. Significance of the Pearson’s r was assessed with the P-value from the Pearson Correlation Coefficient Calculator (Social Science Statistics, 2021).

The individual admixture proportions using the full set of SNPs were estimated for all pure and admixed animals and are presented in Figure 2. The distribution of the global admixture proportions for the 802 animals is presented in Figure 3. Notably, we detected 91 cattle with a Baoulé ancestry > 0.995 among the presumed crossbreds, which we excluded from the subsequent analysis of local ancestry in admixed animals.

The average ancestry estimation for every single SNP was performed across 29 autosomes for trypanosome positive and negative Baoulé x Zebu crossbreds, respectively. The permutation tests over all chromosomes indicated significant local ancestry deviation from the average (above the 5 and 10% genome-wide thresholds) in chromosomes 8 and 19 for trypanosome positive crossbreds (Figure 4), and in chromosomes 6, 19, 21, and 22 for trypanosome negative animals, respectively (Figure 5). The local admixture proportions for these chromosomes are presented in Figure 6 and for all other chromosomes in Supplementary Figures 1, 2. We further visualized the deviations from the average ancestry in the respective chromosomes and identified regions of higher delta ancestry (wide peaks) on chromosome 8 between 35–50 Mb and in chromosome 21 between 20–35 Mb and 40–50 Mb, respectively (Figure 7). These genomic regions might harbor candidate genes associated to tolerance or susceptibility of trypanosomosis.

We screened the genomes of the Baoule and Zebu crossbred animals for outlier SNPs with high F_ST values and disregarded the pure-bred Baulé and Zebu. Among these crossbred animals we grouped them in trypanosome positive and negative animals to avoid detection of breed differences unrelated to trypanosome tolerance. We detected seven variants with a FDR corrected threshold of p < 0.05 (Figure 8). The seven outlier SNPs with the highest levels of F_ST values were found in chromosomes 2, 3, 5, 20, 21, and 23, and are presented in Table 2, together with their neighboring genes. The positions of the SNPs were not located in regions with higher delta ancestry.

To reliably detect hybrids even with a small set of SNPs applicable for routine genetic monitoring, we selected the most ancestry informative markers resulting in the highest differentiation between Baoulé and Zebu cattle. The 100 SNPs with the highest divergence presented F_ST values ranging between 0.98 and 0.79 (Supplementary Table 1). We estimated admixture levels of the crossbred individuals using the top 15, 25, 50, and 100 SNPs (Supplementary Figure 5). The Pearson correlation coefficients r between the levels of admixture obtained with the full dataset of 35,952 SNPs and each of the sets of top SNPs were generally high and ranged between 0.949 (allSNPs/top15) and 0.990 (allSNPs/top100) (Table 3). All Pearson correlation coefficients were statistically significant (p < 0.001).

The high amounts of global admixture observed in the taurine cattle population in the three studied departments of Burkina Faso indicated mixed genetic backgrounds of the cattle in Bouroum-Bouroum, Kampti and Loropeni (Figures 2, 3). The observed admixture levels within the departments are likely due to unrestricted mating among cattle of different genetic backgrounds. Long-distance migrations within and across countries, utilization of communal pastures, exchange of breeding animals, and uncontrolled mating facilitate constant gene flow. Houessou et al. (2019) explained this situation by lack of selection and high levels of gene flow due to cyclical cross-border cattle herd movements known as “transhumance” and extensive commercial transactions of cattle in the West African region.

The uncontrolled mating in extensive production systems, which are typically practiced in West Africa, can lead to the introgression of Zebu genes in the small taurine cattle population, which represents a threat to their genetic integrity (Dossa and Vanvanhossou, 2016), and might lead to a potential dilution of their trypanotolerance (Traoré et al., 2015; Albert et al., 2019). The increasing importance of Zebu in the South-western region of Burkina Faso might endanger Baoulé cattle in the long term. As the North, which is the preferred area for Zebu cattle, is hit by drought, increasing numbers of Zebu cattle breeders looking for pasture are moving to the South-west where the climate is quite favorable and grass is still abundant. Thus, suitable management is required for the sustainable use of local breeds, and recently community-based breeding programs (CBBP) for Baoulé cattle and their crossbreds have been implemented (Ouédraogo et al., 2020). Within the CBBP, Zoma et al. (2020) identified four distinct types of cattle production systems sedentary Lobi farms, sedentary crossbreed farms, semi-transhumant Fulani Zebu farms, and transhumant Fulani Zebu farms. The admixture between Zebus and Baoulé cattle observed in this study could be due to differences in the production systems. Furthermore, notable size differences between purebred and crossbred Baoulé were confirmed (Yougbaré et al., 2020) and breeders prefer to have large animals like Zebu cattle. As shown in Figure 2, we identified several purebred Baoulé cattle that had been considered as admixed based on the sampling information. These animals originated from the populations of Loropeni and Bouroum-Bouroum where the farmers have a preference for breeding purebred Baoulé (Zoma et al., 2020).

In a recently admixed population, ancestral populations have been mixing for a relatively small number of generations, resulting in a new population with different proportions of their genome derived from the original parental populations (Khayatzadeh et al., 2016). Local ancestry analysis of admixed populations has been successfully used to detect recent selection in admixed Swiss Fleckvieh cattle (Khayatzadeh et al., 2016), as well as selection for Zebu introgressed regions in Colombian creole taurine cattle (Pitt et al., 2018). In our study, we applied this approach to identify significantly different local admixture levels and detected five chromosomes with higher deviation from the average ancestries, with an excess of Baoulé ancestry, which might account for a higher tolerance to trypanosomiasis. Similarly, Decker et al. (2014) investigated the population structure of domesticated cattle and calculated Asian indicine (B. t. indicus), Eurasian taurine, and African taurine (both B. t. taurus) ancestry proportions.

We applied an approach of significance testing and performed a permutation test of circularizing the genome by concatenating the SNPs of all autosomes in a single string, cutting this string once and rearranging the two resulting segments, as proposed by Tang et al. (2007). The permutation approach removes not only the effects of selection, but also the local effects of genetic drift; the threshold is considered non-conservative. Nevertheless, based on simulations (Tang et al., 2007) outliers are unlikely to be due to genetic drift. Therefore, this procedure is considered robust to find significant signals for selection while accounting for confounding effects of demographic history of the admixed cattle.

We found regions deviating from the average ancestry with a higher amount of Baoulé proportions on chromosomes 6, 8, and 19 in trypanosome negative individuals. A previous study (Noyes et al., 2011) identified VAV1, PIK3R5, RAC1, VAV2, GAB2, and INPP5D genes in chromosome 8 to be genes under selection in Muturu and N’Dama cattle breeds in response to trypanosomes infection. Surprisingly, we also found higher Baoulé ancestry in chromosome 8 (35–50 Mb) also in trypanosome positive cattle, which could indicate that these regions harbor beneficial Baoulé haplotypes, which are not connected to trypanosomosis tolerance. These regions might harbor genes of general importance for adaptation to the environment. Some canadidate genes on chromosome 6 at 71373513-71421283 (PDGFRA) and chromosome 19 at 41185975-41196948 (CDC6) for trypanotolerance in West African taurines have been found on these chromosomes (Tijjani, 2019) overlapping with the regions identified in our study. Furthermore, Smetko et al. (2015) identified chromosomes 7 and 22 as regions with the highest Baoulé ancestry proportion, similar to our results.

Identifying recent positive selection signatures in domesticated animals can provide information on beneficial mutations and their underlying biological pathways for economically important traits. Global F_ST values are one useful method to detect selection signatures across breeds (Biswas and Akey, 2006). The seven outlier SNPs, which we identified between trypanosome positive and negative crossbreds, were on chromosome 2, 3, 5, 20, 21, and 23. The chromosomes BTA 2, 3, 5, and 23 have previously been identified harboring common candidate genes in Muturu and N’Dama breeds linked to trypanotolerance in West African taurine population as well as selected candidate genes in Muturu cattle only (Tijjani et al., 2019). Functional annotation and enrichment analyses based on Reactome pathways in PANTHER ver 13.1 (Thomas et al., 2003) confirmed their relevance in response to trypanosome infection pathways. In our study, we identified other genes (Table 2) such as LOC100138101, LMO4, LTRM2, ISL1, PTCHD4, and HIVEP1 as genes potentially responsible for trypanotolerance.

From previous studies genes such as TICAM1, ARHGAP15, SLC40A1, GFM1, and INHBA have been proposed as candidate genes for trypanotolerance on chromosomes 2, 3, and 5 (Dayo et al., 2009; Noyes et al., 2011). Bahbahani et al. (2018) identified the genes LTA4H, IL7, IL15, FCN, LTA4H, and NFAM1 as potential targets of natural selection related to immunity in Sheko cattle, which are a mixture of Asian zebu and African taurine ancestry and considered a trypanotolerant breed with high potential for milk production.

The indigenous cattle breeds are disappearing because of indiscriminate crossbreeding by individual farmers, and schemes for genetic improvement developed without concern for preserving locally adapted breeds (Belemsaga et al., 2005). Many breeding programs or genetic improvement strategies in developing countries failed due to the lack of involvement of farmers in the different steps of implementation (Wurzinger et al., 2011). In many developing countries, livestock crossbreeding has been implemented with poor or no pedigree recording. Thus, ancestry informative markers would provide a great opportunity to estimate the level of admixture in a cost-effective way. Sölkner et al. (2010) proposed that individual admixture levels were estimated more accurately based on the genomic data using panels of pure reference animals, compared to estimation based on pedigree. (Getachew et al., 2017) indicated that the Ovine 50KSNP array is a powerful tool to identify small sets of AIMs for admixture studies in crossbred sheep populations in Ethiopia.

The minimum set of the 25 highest differentiating SNPs (Supplementary Table 1) can be used to develop an efficient competitive allele-specific PCR (KASP^TM, LGC Group, United States) genotyping assay. Such an easy and fast genotyping array can be implemented at any laboratory equipped with Real-Time PCR machine and can be used for routine monitoring of hybridization in Baoulé cattle. Getachew et al. (2017) identified a total of 74 SNPs from the Ovine 50K SNP data as AIMs. The SNPs were selected based on their F_ST values showing the highest levels of allele frequency differentiation between the two parental breeds similar to our methodology. These AIMs provided close estimation with pedigree information. Correlation coefficient between breed level based on admixture estimates from 25 SNP data obtained in this study (r = 0.99; Table 3) was higher compared to the correlation value of 0.96 obtained from ∼500 AIMs suggested to predict breed composition in cattle (Frkonja et al., 2012) or the correlation values in the range of 0.89 to 0.96 reported for different human populations in prediction of admixture levels (Halder et al., 2008). Other studies (Judge et al., 2017) recommended at least 300 informative SNPs identified based on similar diversity parameters to be used for genomic-based breed composition prediction. However, as the purpose of our AIM set was to differentiate between only two ancestral breeds (Baoulé and Zebu) the number of 25 highest differentiating SNPs was sufficient to detect admixed individuals (Table 3 and Supplementary Figure 5). The existence of such a 25 SNP set allows their genotyping locally in Burkina Faso, providing a sustainable and low-cost solution to monitor admixture rates in these populations. We will further validate the 25 most AIMs in a larger group of confirmed crossbreds in Burkina Faso. Understanding the relationship between genetic admixture and performances is crucial for the success for local cattle breed conservation and crossbreeding programs. Ideally, a combination of pedigree and genomic information is used in breeding programs. Applying small sets of AIMs is a cost-effective option to estimate the levels of admixture in situations where pedigree recording is difficult like in Burkina Faso.