Domesticated organisms have experienced strong selective pressures directed at genes or genomic regions controlling traits of biological, agricultural or medical importance. The genome of native and domesticated pigs provide a unique opportunity for tracing the history of domestication and identifying signatures of artificial selection. Here we used whole-genome sequencing to explore the genetic relationships among the European native pig Berkshire and breeds that are distributed worldwide, and to identify genomic footprints left by selection during the domestication of Berkshire. Numerous nonsynonymous SNPs-containing genes fall into olfactory-related categories, which are part of a rapidly evolving superfamily in the mammalian genome. Phylogenetic analyses revealed a deep phylogenetic split between European and Asian pigs rather than between domestic and wild pigs. Admixture analysis exhibited higher portion of Chinese genetic material for the Berkshire pigs, which is consistent with the historical record regarding its origin. Selective sweep analyses revealed strong signatures of selection affecting genomic regions that harbor genes underlying economic traits such as disease resistance, pork yield, fertility, tameness and body length. These discoveries confirmed the history of origin of Berkshire pig by genome-wide analysis and illustrate how domestication has shaped the patterns of genetic variation.
The over 730 pig (Sus scrofa) breeds or lines worldwide have undergone natural and artificial selection in various environments and produced high levels of phenotypic diversity, constituting a valuable resource for investigating how selection affects the genome1,2. Genes or genomic regions under selection in domesticated organisms, such as the ‘domestication genes’ found in silkworm3, chicken4, dog5, cattle6, yak7 and pig8,9, can be directly implicated in genetic breeding programs and greatly increases the efficiency of producing novel and desirable phenotypes10. The economic and biomedical importance of the domestic pig has led to significant efforts to decode the pig genome, such as that of the domestic Duroc pig1 and Tibetan wild boars9.
As part of a continuous effort to comprehensively document the genetic basis of pig phenotypic diversity, here we present genomic analyses among the European native Berkshire pig (three individuals) and other 38 pigs and wild boars distributed worldwide. The Berkshire pig is a typical traditional European breed that has been under intensive artificial selection since the early 18th century in England for rapid and efficient accumulation of muscle and desirable pork qualities such as juiciness, flavor, tenderness, pink-hued and heavily marbled. Using the whole genome sequencing approach, we explored the genetic relationships among Berkshire and other pigs, and identified genetic components under selection that are likely the consequence of domestication of Berkshire pig.
Sequencing, mapping, SNP and InDel calling
Sequencing of three female Berkshire pigs generated a total of 36.65 Gb of paired-end DNA sequence, of which 36.29 Gb (99.02%) high quality paired-end reads were mapped to the pig reference genome assembly (Sscrofa10.2) (Supplementary Table S1). Consequently, for each individual, ~78.75% of reads mapped to 79.39% of the reference genome assembly with 3.64-fold average depth (Supplementary Table S1). In addition, we also downloaded the genome data of 38 individuals from across the world from the EMBL-EBI database1,11, including 14 European domestic pigs from five breeds, 6 Asian domestic pigs from three breeds in China, 7 Asian wild boars from four locations, 6 European wild boars from four locations, 4 other species in the genus Sus, and an African warthog. The average depth for the compiled dataset is 6.46-fold, with average mapping rate of 95.17% and ~75.25% coverage of the reference genome assembly (Supplementary Table S2).
We performed single-nucleotide polymorphism (SNP) calling and identified 18.68 million (M) SNPs from 41 individuals (Supplementary Table S3). We then pooled the SNPs into three groups, including 5.25 M from the 23 domestic pigs, 5.47 M from the 13 wild boars, and 15.73 M from the four wild genus Sus and an African warthog (Supplementary Table S3). A small portion of (2.48 M of 18.68 M, or 13.28%) SNPs was shared among the three groups, indicative of substantial genomic differences among them.
We identified 3.65 M SNPs from three Berkshire pigs, of which 21,905 coding SNPs leading to 7,773 nonsynonymous nucleotide substitutions (7,713 missense, 44 stop gain and 16 stop loss) were detected in 3,978 genes (Table 1 and Supplementary Data S1). Top 1,000 genes containing the highest number of nonsynonymous SNPs (nsSNPs) were mainly over-represented in olfactory-related categories, such as ‘olfactory transduction (52 genes, P = 1.68 × 10−10)’, ‘sensory perception of smell (53 genes, P = 8.70 × 10−9)’, ‘olfactory receptor activity (52 genes, P = 1.19 × 10−8)’ and ‘sensory perception (77 genes, P = 2.35 × 10−7)’ (Supplementary Data S2). The olfactory receptors, known to be involved in sensing of the extracellular environment, are encoded by the largest gene superfamily in the mammalian genome12,13. Pigs have one of the largest repertoire of functional olfactory receptor genes14, reflecting the strong reliance of pigs on their sense of smell while scavenging for food1 and other odor-driven behavior (particularly mate recognition and sexual receptive behavior)15,16.
We also identified 2.93 M small insertion or deletion polymorphisms (InDels) ranging from 1–30 bp in length (Supplementary Table S4), which tend to be detected with greater frequency than long InDels. Only 2,991 (0.10%) InDels were located in coding sequences, of which 29.35% were multiples of 3 bp (Supplementary Fig. S1 and Data S3). The enrichment of in-frame InDels that are expected to preserve reading frame, can be explained by previous findings that in-frame InDels were under weaker negative selection than frame-shift InDels with lengths that are not evenly divided by three17,18. These InDels affected genes enriched mainly in terms related to basic cellular functions, such as the ‘binding of adenyl nucleotide, purine nucleoside, ATP, cation, ion, metal ion, and nucleoside’ and ‘protein kinase activity’ (Supplementary Data S4), which is similar to previous reports in mammals about the effect of InDels on the functions of genes1,5,18,19.
Phylogenetic and admixture analysis
To explore relatedness among the Berkshire pig and other pigs distributed worldwide, we conducted principle component analysis (PCA) using genomic SNPs. The first eigenvector geographically distinguishes 23 individuals in Europe from 17 individuals in Asian and a warthog in Africa, whereas the second eigenvector captures the biological differentiation between pigs (including domestic and wild boar) and other outgroup (i.e., wild genus Sus and warthog) (Fig. 1a and Supplementary Table S5). The neighbor-joining (NJ) tree confirmed these findings and further revealed genetically distinct clusters that relate to geographic locations rather than by domestic versus wild (Fig. 1b). This result is consistent with a deep phylogenetic split between European and Asian pigs since domestication about 10,000 years ago in multiple locations across Eurasia1,20.
It is well documented that a clear signal for admixture between domestic pigs in Asia and Europe1,9,21 is likely due to the importation of Chinese breeds into Europe (especially UK) at the onset of the agricultural revolution in the late 18th and 19th century22. To investigate the amount genetic material of Chinese origin in the Berkshire pigs relative to that shared by other five reprehensive European domestic pigs, we performed an admixture analysis (D-statistics) using ‘ABBA/BABA’ single nucleotide sites, which was originally developed to test for admixture between Neanderthals and modern humans23,24. We divided the genome into N blocks and computed the variance of the statistics over the genome N times, leaving each block aside and derived a standard error using the theory of the Jackknife24. Given the standard error of the D-statistics of different block sizes were very similar, we used 2 Mb as the block size for further analyses (Supplementary Table S6).
The excess of ABBA sites (0 < D < 1) indicates that the Berkshire pig has a stronger signal of introgression from Chinese domestic pig than other 5 European domestic pigs (Fig. 1c). Especially, when compared with Duroc pig, the Berkshire pig exhibits an highest excess of ABBA sites across 18 autosomes, giving a significantly positive D of 0.337 ± 0.010 (two-tailed Z-test for D = 0, P = 1.680 × 10−243) (Fig. 1c). The higher amount of Chinese genetic material in the Berkshire pig is consistent its history of origin: in the county of Berkshire in England, a reddish or sandy colored pig strain (sometimes spotted) was latterly refined with a cross of Siamese and Chinese blood (~300 years ago), bringing the color pattern we see today along with more efficient meat production22. Currently, the purebred Berkshire is recorded as a ‘transboundary’ (occurring in more than one country) breed.
Genome-wide selective sweep signals
To accurately detect the genomic footprints left by selection, we measured the genome-wide variations between six European wild boars and three Berkshire pigs, which are geographically close and genetically indistinguishable.
Compared with the wild boars, the domestic Berkshire pigs have lower levels of linkage disequilibrium (LD) across the range of distances separating loci (P < 10−16, Mann-Whitney U test) (Fig. 2a), reflecting relatively higher inbreeding under artificial breeding programs and thus a lower genomic diversity in Berkshire pig.
Out of 272,292 windows of 100 kb in length sliding in 10 kb steps across the pig genome, 210,266 windows contain ≥ 50 SNP and cover 77.22% of the genome (Supplementary Fig. S2), which were used to detect signatures of selective sweeps. We used an empirical procedure and selected windows simultaneously with significantly high log2 (θπ ratio (θπ, wild boar/θπ, Berkshire)) (10% right tail, where log2 (θπ ratio) is 3.14) and significantly high FST values (10% right tail, where FST is 0.71) of the empirical distribution as regions with strong selective sweep signals along the genome, which should harbor genes that underwent selective sweep. Consequently, we identified a total of 11.95 Mb genomic regions (4.75% of the genome, containing 482 genes) with strong selective sweep signals in Berkshire pigs (Fig. 2b), which also exhibited significant differences (P < 10−16, Mann-Whitney U test) in log2 θπ ratio and FST values when compared to genomic background (Fig. 2c). SNPs from these regions formed two distinct clusters (i.e. Berkshire pigs and European wild boars) (Supplementary Fig. S3).
In total, 482 genes embedded in selected regions were predominantly related to immune (such as ‘defense response to virus (4 genes, P = 0.001)’ and ‘Immunoglobulin’ (11 genes, P = 0.001)), growth (such as ‘regulation of growth (11 genes, P = 0.003)’), reproduction (such as ‘oocyte meiosis (6 genes, P = 0.004)’ and ‘reproductive developmental process (9 genes, P = 0.005)’) (Table 2). This result coincides with previous reports of pig domestic genes1,8,11,21,25 and may be responsible for dramatic phenotypic changes in domestic pigs that are of economic values, such as disease resistance, pork yield and fertility. In addition, we also identified genes related to neuron functions (such as ‘neurotrophin signaling pathway (8 genes, P = 0.003)’) that experienced selective sweep (Table 2), which support the hypothesis that selection for altered behavior (such as tameness or aggression towards humans) was important during pig domestication and that mutations affecting developmental genes may underlie these changes10,26,27. For example, one of the genes under selective sweep in Berkshire pig is transcription factor SOX6 (SRY (sex determining region Y)-box 6) (Supplementary Data S5), a modulator of cell fate during neocortex development28, which plays roles in brain development and related to the differences in the development or maturation of the frontal cortex in domesticated animals29.
Body length in domestic pigs
Notably, we detected numerous well-characterized genes related to body length embedded in selected regions (Fig. 3a), which is the most characteristic morphological change between the wild boar and domestic pig. Wild boars, which are ancestors of domestic pigs, have 19 vertebrae. In comparison, European commercial breeds have 21–23 vertebrae, probably owing to selective breeding for enlargement of body size30.
Eight genes exhibiting strong selective sweep signals are significantly over-represented in ‘OMIM-disease term: Many sequence variants affecting diversity of adult human height’ (P = 0.002)’ (Supplementary Data S5), which has been documented to associate significantly with adult human height31. For example, ADAMTSL3 (a disintegrin-like and metalloprotease domain with thrombospondin type I motifs-like 3), a glycoprotein in extracellular matrix, is associated with the chondrogenesis, morphogenesis and growth of the skeleton in human31,32,33 and other mammals (cattle)34, which is also an attractive candidate genetic marker to identify animal body size or type. GPR126 (G-protein coupled receptor 126), an orphan receptor of the adhesion-G-protein coupled receptor family, is essential for mammalian embryonic viability35, myelination36, osteoclast function and regulation of bone mineral density37. Association between variation at GPR126 with height in childhood38 and adult31,33 as well as the skeletal frame size39 has been shown. PRKG2 (cGMP-dependant type II protein kinase) is involved in preovulatory follicles as a response to luteinizing hormone and progesterone, which is highly expressed in brain and in cartilage, and contributed to the determination of dwarfism in mammals. The knockout mouse40 and naturally occurring rat41 and cattle42 PRKG2 mutants resulted in unorganized growth plate with abnormal stacking of chondrocytes and dwarfism. In addition, the RASGEF1B (RasGEF domain family, member 1B) is a highly conserved guanine nucleotide exchange factor for Ras family proteins43, which is neighboring with the PRKG2. Ras superfamily proteins function as molecular switches in fundamental events such as signal transduction, cytoskeleton dynamics and intracellular trafficking. In human, the microdeletion (1.37 Mb) at chromosome 4q21 that encompass PRKG2 and RASGEF1B resulted in growth restriction, mental retardation and absent or severely delayed speech43.
We also found IGF1 (Insulin-like growth factor 1), a hormone similar in molecular structure to insulin, which is a primary mediator of the effects of growth hormone, could stimulate systemic body growth, especially skeletal muscle, cartilage and bone, and has been recognized as a major determinant of body size in mammals44,45. In particular, NR6A1 (nuclear receptor subfamily 6, group A, member 1), which is involved in neurogenesis and germ cell development46, to be embedded in the most significantly selected regions (simultaneously with high log2 (θπ ratio) (1% right tail) and FST values (1% right tail) (Fig. 3b). It has been well documented that the NR6A1 is a strong candidate for being a causal gene underlying the elongation of the back and an increased number of vertebrae in pigs varies8,47,48.
Analogously with the human height31, the porcine trunk length is a highly heritable (the number of vertebrae, h2 = 0.62) and classic polygenic trait49. The strong selective sweeps of these genes related to ‘body length’ reveal the specific evolutionary scenarios triggered by artificial selection for agricultural production.
This study presented the genetic relationships between the Berkshire and other pigs, and uncovered genetic footprints of domestication that provide an important resource for further improvements of this important livestock species. The work performed here will serve as a typical demonstration for future deciphering the genomic differences shaped by the artificial selections.
All research involving animals were conducted according to the Regulations for the Administration of Affairs Concerning Experimental Animals (Ministry of Science and Technology, China, revised in June 2004) and approved by the Institutional Animal Care and Use Committee in College of Animal Science and Technology, Sichuan Agricultural University, Sichuan, China under permit No. DKY- S20123130.
Sequencing of Berkshire pigs
Genomic DNA was extracted from the ear tissues of each of three female Berkshire pigs. There is no direct and collateral blood relationship within the last three generations among them. Sequencing was performed on the Illumina HiSeq 2000 platform. In addition, we also downloaded the genome data of 38 Sus scrofa individuals across the world from the EMBL-EBI database (ftp.sra.ebi.ac.uk/vol1/fastq/ERR173/).
Sequence quality checking and filtering
First, to avoid reads with artificial bias (i.e. low quality paired reads, which mainly result from base-calling duplicates and adapter contamination), we removed the following types of reads: (a) reads with ≥ 10% unidentified nucleotides (N); (b) reads with > 10 nt aligned to the adapter, allowing ≤ 10% mismatches; and (c) reads with > 50% bases having phred quality < 5; and (d) putative PCR duplicates generated by PCR amplification in the library construction process (i.e. read 1 and read 2 of two paired-end reads that were completely identical). Second, high quality paired-end reads were mapped to the pig reference genome sequence (Sscrofa10.2) using the BWA software50. The reference was indexed and the command ‘aln -o 1 -e 10 -t 4 -l 32 -i 15 -q 10’ was used to find the suffix array coordinates of good matches for each read. The best alignments were generated in the SAM format given paired-end reads with command ‘sampe’. We further improved the alignment results with the following three steps: (a) filter the alignment read with mismatches ≤ 5 and mapping quality = 0; (b) the alignment results were corrected using the package Picard (http://sourceforge.net/projects/picard/) with two core commands. The ‘AddOrReplaceReadGroups’ command was used to replace all read groups in the INPUT file with a new read group and assigns all reads to this read group in the OUTPUT BAM. ‘FixMateInformation’ command was used to ensure that all mate-pair information was in sync between each read and its mate pair; and (c) remove potential PCR duplication. If multiple read pairs have identical external coordinates, only the pair with the highest mapping quality was retained.
SNP and InDel calling
After alignment, we performed SNP calling on a population-scale for three groups (23 domestic pigs, 13 wild boars, and four species of the wild genus Sus and an African warthog) using a Bayesian approach as implemented in the package SAMtools51. The genotype likelihoods from reads for each individual at each genomic location were calculated, and the allele frequencies were also estimated. The ‘mpileup’ command was used to identify SNPs with the parameters as ‘-q 1 -C 50 -S -D -m 2 -F 0.002 –u’. Then, to exclude SNP calling errors caused by incorrect mapping or InDels, only high quality SNPs (coverage depth ≥ 4 and ≤ 1,000, RMS mapping quality ≥ 20, the distance between adjacent SNPs ≥ 5 bp, no InDel present within a 3 bp window and the missing ratio of samples within each group < 50%) were kept for subsequent analysis. We also performed InDel calling using the ‘mpileup’ command with the parameters as ‘-m 2 -F 0.002 -d 1,000’ as implemented in the package SAMtools51.
Functional enrichment analysis
Functional enrichment analysis of Gene Ontology (GO), pathway and InterPro domains was performed using the DAVID web server52. Genes were mapped to their respective human orthologs, and the lists were submitted to DAVID for enrichment analysis of the significant overrepresentation of GO biological processes (GO-BP), molecular function (GO-MF) terminologies, and KEGG-pathway and InterPro categories. In all tests, the whole known genes were appointed as the background, and P values (i.e. EASE score), indicating significance of the overlap between various gene sets, were calculated using Benjamini-corrected modified Fisher's exact test. Only terms with a P value less than 0.05 were considered as significant and listed.
Phylogenetic genetic analyses
We performed the PCA with the population scale SNPs using the package EIGENSOFT4.253, and the eigenvectors were obtained from the covariance matrix using the R function reigen. The significance level of eigenvectors was determined using the Tracey-Widom test53. The phylogenetic tree was inferred using TreeBeST (http://treesoft.sourceforge.net/treebest.shtml) under the p-distances model using SNPs in a population scale.
Admixture analysis – D-statistics (ABBA-BABA tests)
To detect admixture between Chinese domestic pigs and Berkshire or other five European domestic pigs, we computed D-statistics based on ABBA and BABA SNP frequency differences using the expression23:where P1, P2, P3 and P4 are the four different populations under comparison, P1 (separately, each of 5 European domestic pigs) and P2 (Berkshire) are sister taxa, P3 is the Chinese domestic pigs and P4 (African warthog) is an outgroup, pij is the observed frequency of the derived “B” SNP i in taxon j, and n is the total number of SNPs.
It is possible to compute the number of derived alleles common between P1 and P3 (ABBA count) and between P2 and P3 (BABA count). Under the null hypothesis of solely incomplete lineage sorting and no gene flow between P3 and either P2 or P1, we expect a similar count of ABBA and BABA patterns. Under an alternative scenario of gene flow, the count of ABBA must be significantly higher than BABA counts (or vice versa)23,24,54. In addition to calculating D for the entire genome, to examine variation in D across the genome, separate D-statistics were evaluated for each of the 18 autosomes. A standard error (s.e.) of the D-statistics was computed using a Weighted Block Jackknife approach24.
Linkage-disequilibrium (LD) analysis
To estimate the LD patterns between Berkshire and European wild boars, we used 4.92 M SNPs of six European wild boars and merged them with SNPs of the Berkshire pigs, resulting in 8.57 M SNPs in total. To evaluate LD decay, the coefficient of determination (r2) between any two loci was calculated using Haploview55. Average r2 was calculated for pairwise markers in a 500 kb window and averaged across the whole genome.
Calculation of θπ and FST
A sliding window approach (100 kb windows sliding in 10 kb steps) was applied to quantify the polymorphism levels (θπ, pairwise nucleotide variation as a measure of variability) and genetic differentiation (FST) between the Berkshire and European wild boars.
Accessions codes The genome resequencing reads have been deposited into the NCBI sequence read archive (SRA) under the accession SRP030626.
This work was supported by grants from the National High Technology Research and Development Program of China (863 Program) (2013AA102502), the National Special Foundation for Transgenic Species of China (2014ZX0800950B and 2011ZX08006-003), the Fund of Fok Ying-Tung Education Foundation (141117), the Fund for Distinguished Young Scientists of Sichuan Province (2013JQ0013), the Specialized Research Fund of Ministry of Agriculture of China (NYCYTX-009), the Program for Changjiang Scholars and Innovative Research Team in University (IRT13083), the Postdoctoral Fellowship of Peking-Tsinghua Center for Life Sciences, and the China Postdoctoral Science Foundation (2012M520123).
Supplementary Data S1-S5