All experimental procedures were approved by the Institutional Animal Care and Use Committee of the Hubei Academy of Agriculture Sciences, and all methods involved pigs were in accordance with the agreement of Institutional Animal Care and Use Committee of the Hubei Academy of Agriculture Sciences (Permit number: 36/2016).

The animal genomics dataset used in the current study were gotten from pigs raised in a composite swine breeding farm located in Enshi, Hubei, China. In order to make the sample representative, we chose individuals from the core group to avoid full siblings. At last, there were 262 individuals consisting of approximately three generations. The 262 animals were genotyped using the Porcine Single nucleotide polymorphism SNP50 BeadChip (Illumina, United States), consisting of 51,315 Single nucleotide polymorphism (SNPs) evenly distributed along the pig genome. Genotype quality control was carried out using PLINK v1.90 (Chang et al., 2015) software based on the following filtering criteria: (1) the call rate of SNPs and individuals were higher than 0.9; (2) the minor allele frequency (MAF) was greater than 0.01; and (3) only SNPs mapped to autosomes were included. The latest version of the pig genome, Sus scrofa 11.1 was used in this study.

We identified ROH in individuals using PLINK v1.90 software, which uses a sliding window approach to detect autozygous segments. The algorithm is as follows: take a window of X SNPs and slide them across the genome. Determine at each window position whether the window looks sufficiently “homogeneous” (yes/no). Then, for each SNP, calculate the proportion of the “homozygous” window that overlaps that position. Call segmentation based on this metric, such as a threshold based on the average value. To define a ROH, the criterion and thresholds were as follows: (1) a minimum ROH length of 1Mb; (2) at least 50 homozygous SNPs included in a ROH; (3) a minimum density of a SNP in 100Kb; (4) a sliding window of 50 SNPs across the genome that moves one SNP at a time; and (5) up to one heterozygous SNP and five missing SNPs were allowed in a sliding window. Detected ROHs were later classified into three different classes based on their length: 1–5, 5–10, and >10Mb. The total number and length of ROHs were counted for all individuals.

We identified the genomic regions that were mostly associated with ROHs, by calculating the proportion of SNPs in ROH. This was done by counting the number of times the SNP was detected in those ROH across individuals. Afterward, we selected the top 1% of SNPs that were commonly observed in ROHs. Adjacent SNPs that were above this threshold were finally merged into genomic regions which are called ROH islands, which is characterized by being shared by a majority of individuals in the population (Dixit et al., 2020). We used the database provided by NCBI to annotate the genes in the ROH island. Through a large number of accurate literature searches, the biological function of each annotated gene in the ROH island was inferred.

In this study, two types of genomic inbreeding coefficients were calculated, one based on ROH (F_ROH) and the other based on excess of homozygosity (F_HOM). Genomic inbreeding coefficients (F_ROH) were computed for all individuals by the following formula, as proposed by McQuillan et al. (2008): FROH=∑LROHLauto,where ΣL_ROH is the length of ROHs, and L_auto is the total length of the genome covered by the SNPs included in this chip. F_ROH was also calculated based on three length classes: 1–5, 5–10, >10, and total (>1) Mb. Genomic inbreeding coefficients (F_HOM) were calculated as FHOM=O−EL−E, where O is the number of observed homozygous genotypes, E is the number of expected homozygous genotypes by chance, and L is the total number of genotyped autosomal SNPs. Pearson’s correlation was used to compare the inbreeding coefficients estimated by these methods using R.

The LD was measured using the r², which was calculated for each pair of SNPs per chromosome according to Hill and Robertson (1968). The pairwise LD (r²) were calculated using the parameters “--ld-window 99,999 --ld-window-kb 1,000 --ld-window-r2 0” in PLINK v1.90. To visualize the decline of LD, the physical distances between SNPs were divided into 100-Kb intervals, and the average of r² in each group was then estimated.

In this study, a total of 7,248 ROHs were identified in the 262 animals with an average of 27.66 ROH per animal. The average ROH length was 6.32Mb, and the longest fragment in Sus scrofa chromosome 8 (SSC8) was 57.84Mb (1,000 SNPs). Table 1 summarizes the descriptive statistics of the ROH number and length by classes. The total ROH number of the composite Xidu Black pigs was mainly composed of shorter segments (1–5Mb), which accounted for about 56.16% of all the detected ROH. The genome coverage of long segments (>10Mb) accounted for 37.11% of the total ROH. Short ROH reflects ancestral inbreeding history, while long ROH segments are usually formed by recent inbreeding. This indicated that both ancient and recent inbreeding events might have affected this population, but recent inbreeding or selection pressures have mainly influenced the genome of the Xidu Black pig population.

The relationship between the total ROH number and the total length of the genome covered by ROH in each individual is shown in Figure 1, and is greatly different among animals. In this population, the most extreme animal with long ROHs had a length of 777.06Mb (34.41% of the pig genome). The variability of the total number and length of ROH among individuals was high. Similar distributions were also observed in other pigs (Xu et al., 2019; Shi et al., 2020) and livestock species, such as sheep (Mastrangelo et al., 2017) and cattle (Peripolli et al., 2018).

For chromosomes, the distribution of total number of ROHs in each chromosome and percentage coverage per chromosome are presented in Figure 2. The number of ROHs per chromosome was greater on SSC1 (710 segments), while the smallest number of ROHs was on SSC17 (118 segments). Previous studies on pigs (Xu et al., 2019; Zhan et al., 2020) have also reported the highest number of ROH on SSC1, possibly because SSC1 is the largest chromosome in the pig genome, and has more markers than other chromosomes. The highest ROH coverage was observed on SSC8 (3.26%), whereas the lowest was on SSC10 (1.19%). Our result suggests that the chromosomes with high ROH coverage might have been influenced by positive selection, which consequently increases the accumulation of advantageous alleles on the chromosome. According to our results, some genomic regions, with the highest ROH coverage, on the SSC4 require more consideration.

Here, we calculated the genomic inbreeding coefficients (F_ROH and F_HOM) using the genotype data of 262 individuals. The mean value of the inbreeding coefficient based on total observed ROHs (F_{ROH_total}) was 0.077 and ranged from 0.002 to 0.344. The estimated F_HOM inbreeding coefficients was −0.054 with a range from −0.199 to 0.251 in this population (negative values correspond to individuals with lower-than-average homozygosity). Traditionally, the inbreeding coefficient was estimated based on pedigree data. However, when it comes to cross-bred individuals, things get complicated because the genealogical relationships between parental breeds cannot be established. Moreover, in reality, pedigree information might be incorrect and incomplete, and does not usually take into account the various stochastic events of recombination, which might have occurred during meiosis (Marras et al., 2015). Thus, the inbreeding coefficient value estimated based on pedigree data could not totally show the actual relatedness among individuals within a population. In this study, we used genomic data to estimate inbreeding coefficient.

To further investigate the inbreeding coefficients which were obtained by different estimation methods, we conducted a pairwise comparisons between F_HOM and F_ROH. The pairwise correlations among five types of inbreeding coefficients were shown in Figure 3. Among all pairwise correlations, the highest correlation was 0.94 between F_{ROH_total} and F_ROH>10Mb. This result showed that long ROH segments (>10Mb) were the main source of F_{ROH_total}. The inbreeding coefficients obtained by different categories of ROHs with F_HOM ranged from 0.63 to 0.83, with the highest correlation found between F_{ROH_total} and F_HOM. These results are in line with previous research in other pig populations (Xu et al., 2019; Shi et al., 2020) and cattle (Mastrangelo et al., 2016; Biscarini et al., 2020). Furthermore, we found that many individuals had a negative F_HOM value, which might be because F_HOM was sensitive to allele frequency for populations with a higher level of heterozygosity compared to F_ROH estimators (Zhang et al., 2015). Zhang et al. (2015) found a negative F_HOM value for Danish Red Cattle (RDC), a composite breed, which is likely due to the admixture present in RDC. A similar result was obtained for a crossbred cattle in Vrindavani (Chhotaray et al., 2021). Therefore, the results of this study suggest that the inbreeding level based on total ROH may be a useful method, especially for crossbreed or composed populations.

Furthermore, we plotted the percentage of SNPs in ROHs against their respective positions along the chromosomes (Figure 4). The result shows a non-uniformity in the frequency of different SNPs within the ROH regions across the genome. The most frequent SNP in ROH (121 occurrences, 46.18%) was mapped at ∼35Mb on SSC11, and the closest gene to this SNP was the U6 gene. Regions of the genome with high homozygosity around the ROH islands may contain positively selected targets and might be under strong selection pressure (Pemberton et al., 2012). To identify the genomic regions that were mostly associated with ROH in all individuals, we considered the top 1% of SNPs with the highest occurrences (over 30.15% of the samples) in the ROH as candidate SNPs (Figure 4). We identified a total of eight ROH island regions (Table 2), with length ranging from 0.804Mb on SSC2 to 3.188Mb on SSC11.

Chromosome position, the start and end position of ROH, ROH length, number of SNPs, and the number of genes within the ROH islands were reported in Table 2. To evaluate the potential functional importance of the detected ROH islands, we analyzed the gene content of the identified regions. In summary, we annotated a total of 199 genes that were detected within these ROH islands. The chromosome position, start and end, gene name and Ensembl Gene ID were provided in Supplementary Table S1.

In this study, we focus on the genes related to some specific livestock-traits which are also important in breeding. We identified numerous candidate genes associated with muscular development and fat deposition. Among these genes is the TRAF7 gene, a MyoD1 transcriptional target that can regulate NF-κB activity during myogenesis. Studies have shown that missense mutations in TRAF7 causes developmental delay or skeletal dysplasia (Tsikitis et al., 2010; Tokita et al., 2018). IGFBP7 can promote lipid accumulation and triglyceride production in mature adipocytes and plays an important regulatory role in the differentiation of preadipocyte cells that can affect fat deposition (Hu et al., 2021). PRSS33 was related to lipid transport and metabolism, and was also detected in reported selection signature regions in Enshi black pigs, one of the founder breeds. Since Xidu black pigs are characterized for meat quality, we also detected some important genes that are associated with specific meat quality traits: TRAP1 and CREBBP are associated with pork meat pH in Finnish Yorkshire pigs (Verardo et al., 2017); SLC9A3R2 gene has been shown to be differentially expressed in longissimus muscle tissues of Meishan and Large White pigs (Jiugang et al., 2011), and was considered to be the candidate genes for meat quality (Wu et al., 2020). OTX1 is a novel regulator of proliferation, migration, invasion, and apoptosis in lung adenocarcinoma (Yang et al., 2020), and were involved in the battle between foot-and-mouth disease virus and the host (Zhang et al., 2018a). We identified several candidate genes related to reproduction traits: XPO1, a nuclear transport receptor, plays an essential role in meiotic resumption in porcine full-grown and growing oocytes (Onuma et al., 2018); SLC26A8, also known as testis anion transporter 1, is required for sperm terminal differentiation and male fertility in the mouse (Toure et al., 2007).

OR1F1, an olfactory receptor, was demonstrated to function in odor perception activation (Li et al., 2013). CLDN9 played an essential role in maintaining barrier function in airway epithelial cells (Gon et al., 2017), and E4F1 is essential for skin homeostasis (Lacroix et al., 2010). Notably, the Xidu black pigs reside in a subtropical region, which is characterized by high temperature and humidity. Therefore, we considered that OR1F1, CLDN9, and E4F1 as key factors for the environmental adaptability of Xidu black pigs. The most interesting candidate gene in this population seems to be PPARD gene, which was shown to affect the shape of the external ear and fat deposition in pigs (Meidtner et al., 2009; Ren et al., 2011). Due to the hybrid parents (Hubei white pig, Meishan pig, and Enshi black pig) have different ear shapes, the ear shapes of the base population had erected, forward sloping and drooping phenotype. Therefore, during the breeding process of Xidu black pigs, the forward sloping and slightly drooping ear shape was constantly selected. This gene may play a key role in ear shape in this population and needs more attention.

A total of 38,275 SNPs were used to calculate the average LD between all adjacent SNPs, with a distance less than 100Kb. The average and SD of estimated r² was 0.289±0.316, which was similar to the result that (Joaquim et al., 2019) estimated in a crossbred Landrace pig population. It was observed that the LD value decreased with the increase of the distance between markers (Figure 5). When the distance was greater than 1,000Kb, the average r² was about 0.15. The LD extent (r²=0.3) in Xidu black pigs was about 25Kb. This value was less than 79.54Kb (r²=0.3) for Meishan pigs (one of the original parental breed) obtained by the same method (Liu et al., 2020). The effectiveness of GWAS and genomic selection (GS) relies on the LD between the markers. According to the literature, a mean r² value above 0.30 is considered as a strong LD sufficient for QTL mapping (Shifman et al., 2003). However, an average r² of 0.20 is considered enough to achieve an accuracy of 0.85 for genomic estimated breeding value (GEBV; Meuwissen et al., 2001). In this study, we found that moderate LD (r²≥0.2) extended up to 150–200Kb. Assuming that the total length of the pig genome is 2.5Gb, this suggests that at least 12,500–16,667SNPs would be required for effective GWAS in this breed. This result could be particularly useful when designing breed-specific SNP array panels in future genomic study and selection programs for this composite pig breed.