Insights into adaption and growth evolution: a comparative genomics study on two distinct cattle breeds from Northern and Southern China

Abstract

Mongolian cattle (MG, Bos taurus) and Minnan cattle (MN, Bos indicus) are two different breeds of Chinese indigenous cattle, representing North type and South type, respectively. However, their value and potential have not yet been discovered at the genomic level. In this study, 26 individuals of MN and MG were sequenced for the first time at an average of 13.9- and 12.8-fold, respectively. Large numbers of different variations were identified. In addition, the analyses of phylogenetic and population structure showed that these two cattle breeds are distinct from each other, and results of linkage disequilibrium analysis revealed that these two cattle breeds have undergone various degrees of intense natural or artificial selection. Subsequently, 496 and 306 potential selected genes (PSRs) were obtained in MN and MG, containing 1,096 and 529 potential selected genes (PSGs), respectively. These PSGs, together with the analyzed copy number variation (CNV)-related genes, showed potential relations with their phenotypic characteristics, including environmental adaptability (e.g., DVL2, HSPA4, CDHR4), feed efficiency (e.g., R3HDM1, PLAG1, XKR4), and meat/milk production (e.g., PDHB, LEMD3, APOF). The results of this study help to gain new insights into the genetic characteristics of two distinct cattle breeds and will contribute to future cattle breeding.

Graphical Abstract

Mongolian cattle (Bos taurus) and Minnan cattle (Bos indicus) are two different breeds of Chinese indigenous cattle, representing North type and South type, respectively. Authors carried out comprehensive genome-wide analysis of them, and results revealed that these two cattle breeds have undergone various degrees of intense natural or artificial selection.

Introduction

More than 1,000 cattle breeds are living in the world,¹ making up a vital part of economic livestock by offering major sources (e.g., milk, meat, leather, and power) to humans. Modern cattle are stratified into two types according to the common usage, Bos taurus and Bos indicus, which have been differentiated from each other for >250,000 years2, 3, 4, 5, 6 and independently domesticated in the Fertile Crescent ∼8,000–10,000 years ago and the Indus Valley ∼6,000–8,000 years ago.^7,8 China has broad and diverse bovine genetic resources, with 53 indigenous cattle breeds. These Chinese breeds have various intrinsic characteristics, considered as important genetic resources for cattle around the world. It is generally known that for thousands of years, Chinese indigenous cattle breeds served as a major labor force in agricultural production and are well known for their endurance and adaption.⁹ Chinese cattle have long been used as draft animals and valued for their resistance to parasites, roughage-based diets, and the great tolerance to environmental challenges.^10,11 Among the mentioned breeds, Mongolian cattle (MG) and Minnan cattle (MN) are two distinct types representing Bos taurus and Bos indicus. The former, living in Northern China, is herded primarily in the Inner Mongolian region and seems well adapted to the cold environment and grazing. The latter originates from South China and is known for its tolerance to muggy weather (Figure S1). These two cattle breeds are varying in their intrinsic characteristics, whereas their value and potential have not yet been discovered. In the course of scientific and technological innovations, two projects, namely HapMap and bovine genome, have been completed.^11,12 In the meantime, the project of multiple bovine species has made rapid progress, and whole-genome resequencing includes Bos taurus,13, 14, 15 Bos indicus,^15,16 Bos mutus, and Gayal.^17,18 These projects explored the depth of evolution science of these large ungulates and better clarified the complex process of domestication and adaptation in these species, which suggested that the whole-genome sequencing is an efficient and effective method to explore genetic information of multiple species. In the present study, the whole genomes of 26 MG and MN cattle were sequenced to determine their genetic diversity. In addition, the genes with distinct characteristics that are positively selected through subtle combinations of human and natural selection were explored. The results of this study provided novel insights into their genetic difference in the whole genome and further identified genomic loci that might be highly important for cattle breeding programs.

Results

Whole-genome sequencing and mapping

Whole-genome sequencing of the genomic DNA extracted from MG (n = 13) and MN (n = 13) was performed on an Illumina HiSeq X Ten (Table 1). A total of 907.97 gigabases of high-quality paired-end reads (150 bp) was produced, and an average mapping depth of 13.9 (MN) and 12.8 (MG) fold were presented, respectively. 97.62% (MN) and 97.98% (MG) of the acquired reads were plotted to distinctive positions on the reference genome (Bos_taurus_UMD_3.1).¹²

Table 1.

Summary of sequence read alignments to the reference genome

Summary	MN (n = 13)	MG (n = 13)
Mean depth	13.9×	12.8×
GC content rate	44.84%	44.78%
Coverage rate
Coverage rate (≥1×)	97.62%	97.98%
Coverage rate (≥4×)	92.41%	92.08%
Coverage rate (≥10×)	41.62%	34.59%
Total reads (bp)	36,297,349,362 (100%)	33,546,860,746 (100%)
Clean reads (bp)	35,533,395,809 (97.85%)	32,492,362,240 (96.83%)
Q30	92.45%	91.43%
Mapped reads	98.37%	99.21%

SNP/insertion or deletion (indel) detection and annotation

In total of 25,501,400 single-nucleotide polymorphisms (SNPs; MN: 23,644,213, MG: 15,309,216) were identified, of which 9.78% (MG: 9.73%, MN: 9.76%) were novel in comparison with the latest cattle SNP database (dbSNP Build 140; ftp://ftp.ncbi.nlm.nih.gov/snp/organisms/archive/cow_9913/chr_rpts/) (Table 2). Transition-to-transversion (TS/TV) ratios were calculated as indicators of potential random sequence errors.¹³ In this study, the TS/TV ratios (MN: 2.40, MG: 2.38) were close to the empirical human TS/TV ratio > 2.1, suggesting the high quality of the identified SNPs in an oblique manner.¹⁹ In addition, the heterozygous/homozygous ratios of MN and MG were 1.47 and 2.54, respectively. The ratio of heterozygous/homozygous SNPs in MN was found to be lower than that of MG, which is noteworthy, since MN have been considered to be indigenous. In addition, over the past thousands of years, the population size has decreased significantly. This could be caused by natural or artificial selection of the genotype. With the use of the RefSeq and Ensembl gene sets, variations were annotated. 8,521,597 and 5,477,158 of the total SNPs in MN and MG were in introns, 492 and 382 in splice sites, and 191,076 and 129,293 in exonic regions, which includes 1,002,360 and 660,656 in untranslated regions (UTRs), 66,551 and 47,117 nonsynonymous, 114,421 and 75,073 synonymous, 573 and 399 stop-gain, and 102 and 76 stop-loss, respectively. Moreover, 2,800,408 and 1,785,757 indels were found in MN and MG genomes, respectively, 9.94% of which were newfound (Table S1). The insertion/deletion ratio of the total indels reached 0.73 (Table S1). The indels were mostly 1 bp in length (Figure S2). Interestingly, concerning the detected variations, 10,193,184 and 1,858,187 SNPs and 1,248,546 and 233,895 indels were breed specific in MN and MG, respectively (Figures S3 and S4). The higher ratios of total variations and breed-specific variations in MN than those in MG are consistent with the fact that cattle in South China have higher genetic diversity. Moreover, it is believed that MG have experienced more robust artificial selection than MN.

Table 2.

Functional classification of the detected single-nucleotide polymorphisms (SNPs)

SNP	MN	MG	Total
Total numbers	23,644,213	15,309,216	25,502,400
Heterozygous/homozygous	1.47	2.54	2.00
Transition/transversion ratio	2.40	2.38	2.39
Novel rate (%)	9.73	9.76	9.78
Intergenic	13,605,091	8,827,342	14,641,248
Up/downstream	323,597	214,385	399,595
Gene	9,715,525	6,267,489	10,461,557
Intronic	8,521,597	5,477,158	9,211,436
Splicing	492	382	557
Exonic	191,076	129,293	210,716
UTR	1,002,360	660,656	1,038,848
Non-synonymous	66,551	47,117	74,679
Synonymous	114,421	75,073	124,859
Stop-gain	573	399	669
Stop-loss	102	76	107
Others	9,429	6,628	10,402

CNV discovery and annotation

Copy number variations (CNVs) refer to DNA segments no less than 1 kb with varying copy numbers compared to a reference genome. In the genome, CNVs have wide distribution. In comparison with the SNPs, CNVs are capable of affecting most genomes and leading to effects similar to changing gene structure and dosage, as well as regulating and exposing recessive alleles. CNVs also have a correlation with many diseases and probably promote several missing heritabilities. After quality control, a total of 4,760 and 3,946 CNVs were found in MN and MG genomes (Tables S2 and S3), respectively, covering 32.19 Mb (1.13%) and 26.41 Mb (0.93%) of the reference and containing 442 and 385 genes (Tables S4 and S5). As previous reports suggested, some of these CNV-related genes were correlated with coat color, dairy traits, meat production/quality traits, and reproduction traits, as well as immunity and adaption. KIT, ASIP, and AP3B1 are likely to be involved in coat color, not only in cattle but also in other domesticated mammals.^15,20,21 In particular, because of the expression of the ASIP gene in adipocytes and its significance in obese yellow mice, this transcript was also believed to be related to the milk composition character of the dairy line and the intramuscular fat content of other varieties.²² It has been shown that the SPP1 gene plays a critical part in regulating milk protein gene expression. The allelic variants of this gene have also been shown to have a relation to the variation in milk composition.^23,24 It has been suggested that the expression of PPARGC1A is required by the beginning and developing stages of lactation in dairy cattle, and it is also associated with milk composition, reproduction, growth, carcass traits, and meat quality.²⁵ SLC27A6 and ITFG1 were considered as candidate genes of milk fat mammary glands and daily production.^26,27 PRKG1 is involved in feed efficiency and gap junction, acting as a candidate gene for intramuscular fat in pigs.¹⁶ It is well known that TG affects meat quality traits and carcass in beef cattle.^28,29 CRH can significantly suppress glucose uptake and stimulate fat breakdown. This gene is critical for lipid metabolism, highly associated with ribeye area and marbling as well as the OMEGA 6:3 ratio.³⁰ In addition, it was identified that GPC1 correlated with meat production and quality, which is also an economically significant trait, having been considered broadly during artificial selection.³¹ It has been recently found that NEB is associated with both lipid and organoleptic traits.³² ATP8A1 is also correlated with weight gain, residual feed intake, feed conversion ratio, and feed intake.¹⁶ In addition, it was reported that some other genes are involved in body size, fertility, production, and milk fatty acid profile (e.g., FANCC,³³ SRY,¹³ SLC5A4,³⁴ and CTNNBL1³⁵). Moreover, several immune system activation genes respond to environmental stress, including CD59,³⁶ CDH9,³⁷ and PROCR.³⁸ RLBP1,³⁹ BOLA,⁴⁰ and STOM¹⁵ have close relevance to parasitic diseases, thereby clarifying the function of these genes and unraveling the underlying mechanisms of the innate immunity against several important tropical environments. Variants in SOX5 have effects on phenotype in colder climates.²¹ Furthermore, 1,164 and 596 breed-specific CNVs were included in MN and MG, covering 124 and 67 breed-specific CNV-related genes, respectively (Figures S5 and S6). According to enrichment analysis, genes in MN were significantly upregulated in adaption-related terms (e.g., GO:0042612∼MHC class I protein complex, GO:0002474∼antigen processing and presentation of peptide antigen via MHC class I, bta04360:Axon guidance, and GO:0006955∼immune response, which might result from thousand-year natural selection in a particular environment) (Table S6). In addition to this, the MG-specific CNV-related genes were upregulated (e.g., GO:0003383∼apical constriction, GO:0032525∼somite rostral/caudal axis specification, GO:0006654∼phosphatidic acid biosynthetic process, bta04726:Serotonergic synapse, etc.) (Table S7). This is consistent with the fact that MN (indicine cattle) are also more resistant to ticks, gastrointestinal parasites, and rinderpest than are MG (taurine cattle).

Population structure

The neighbor-joining tree and population structure indicated that MN and MG can be stratified into two groups (Figures 1A, 1B, and S7). Also, according to principal-component analysis (PCA), MN and MG are significantly separated along the first principal component, accounting for the largest proportion of variability (Figures 1C and S8).

Figure 1.

Population analysis of the two sequenced cattle breeds

(A) Neighbor-joining phylogenetic. (B) Genetic structure. The length of each colored segment represents the proportion of the individual’s genome from K = 2 ancestral populations. (C) Principal-component analysis (PCA). (D) Linkage disequilibrium (LD) analysis.

Moreover, it was observed that linkage disequilibrium decays more slowly in MG than that in MN (Figure 1D). This may result from the population bottleneck event during domestication. Furthermore, the results of this study showed that the genetic diversity of MN cattle was higher than that of MG (Figure 2), which agrees with the different living environments and artificial selection directions of these two cattle types.

Figure 2.

Violin plots showing the genetic diversity value of MG and MN cattle calculated by different parameters

(A) Hp. (B) π (pi).

Selective sweep analysis

As a result, 496 and 306 PSRs were obtained in MN and MG, a total of 21.79 and 7.16 Mb, covering 1,096 and 529 potential selected genes (PSGs) (Figure 3A; Tables S8 and S9). Under domestication, extensive selection created different phenotypic features and biological characteristics for a variety of cattle breeds.⁹ Natural selection usually results in directional changes for adaptation traits, particularly acclimatization to harsh environments and resistance to parasites and other diseases for survival in a certain environment. Likewise, artificial selection exerts its impacts on the development of specific traits of economic importance (e.g., milk, meat, and fertility) through genetic improvement.⁹ According to further investigation of this study, a considerable number of the PSGs were correlated with shaping characteristics of the populations, including morphological and production traits (e.g., adaption, coat color, meat traits, etc.) (Table 3).

Figure 3.

Overview of selective sweeps in the MG and MN genomes

(A) Distribution of θπ ratios ln(θ_π,MN/θ_π,MG) and F_ST values, calculated in 50 kb windows sliding in 10 kb steps. Data points respectively located to the left and right of the vertical dashed lines (representing 5% left and right tails of the empirical θπ ratio distribution) and above the horizontal dashed line (the 5% right tail of the empirical F_ST distribution) were recognized as the regions selected for MG (blue points) and MN (green points), respectively. (B) Example of genes with strong selective sweep signals. F_ST values and Tajima’s D values are shown on the top and middle. Haplotypes inferred from genotyping the SNPs across the DVL2 gene are shown on the bottom (blue, red, and deep red colors represent non-mutation, heterozygous mutations, and heterozygous mutations in comparison with the reference, respectively).

Table 3.

Candidate genes putatively selected affecting phenotypes in cattle

Gene	Description	Chr.	FST	Traits	Reference
Minnan cattle
VPS13A	vacuolar protein sorting 13 homolog A	8	0.6781	adaptability	41
GNA14	G protein subunit alpha 14	8	0.6932	adaptability	41
KLHL3	kelch like family member 3	7	0.7136	adaptability	41
DVL2	dishevelled segment polarity protein 2	19	0.8091	adaptability	42
HSPA4	heat shock 70kDa protein 4	7	0.6965	adaptability	15
GPR50	G protein-coupled receptor 50	X	0.7483	adaptability	43
FGF9	fibroblast growth factor 9	12	0.6802	adaptability	44
MON1A	MON1 homolog A, secretory trafficking associated	22	0.8486	parasite resistance	39
MST1R	macrophage stimulating 1 receptor	22	0.7560	parasite resistance	39
DNAH2	dynein axonemal heavy chain 2	19	0.7540	parasite resistance	38,45
LCT	lactase	2	0.6614	parasite resistance	33
THPO	thrombopoietin	1	0.7626	parasite resistance	11
SAR1B	secretion associated Ras related GTPase 1B	7	0.7129	immune response	26
IFNAR2	interferon alpha and beta receptor subunit 2	1	0.6516	immune response	11
HSPA9	heat shock protein family A member 9	7	0.6390	immune response	46
CD5	CD5 molecule	29	0.6496	immune response	32
VAMP7	vesicle-associated membrane protein 7	X	0.8344	immune response	47
IL15	interleukin 15	17	0.6597	immune response	11
FOXC2	forkhead box C2	18	0.7322	disease resistance	32
TPM2	tropomyosin 2	8	0.6497	disease resistance	32
CNGB1	cyclic nucleotide gated channel beta 1	18	0.6554	feed efficiency	43
R3HDM1	R3H domain containing 1	2	0.6497	feed efficiency	22
PIK3CB	phosphatidylinositol-4,5-bisphosphate 3-kinase catalytic subunit beta	1	0.7196	feed efficiency	16
FGF1	fibroblast growth factor 1	7	0.6860	reproduction	46
ADRB2	adrenoceptor beta 2	7	0.7676	meat traits	14
PDHB	pyruvate dehydrogenase beta	22	0.7017	meat traits	32,38
SLC2A4	solute carrier family 2 member 4	19	0.7437	meat traits	42
PCSK4	proprotein convertase subtilisin/kexin type 4	7	0.7427	fertility	46
SPOCK1	SPARC/osteonectin, cwcv and kazal like domains proteoglycan 1	7	0.7334	fertility	46
CABIN1	calcineurin binding protein 1	17	0.6795	fertility	34
SLC38A3	solute carrier family 38 member 3	22	0.8155	meat traits	48
EXOC3	exocyst complex component 3	20	0.8345	meat traits	48
STAT5B	signal transducer and activator of transcription 5B	19	0.6768	meat traits	48
EMD	emerin	X	0.9695	meat traits	48
Mongolian cattle
MC1R	melanocortin 1 receptor	18	0.6532	coat color	15
RXFP2	relaxin/insulin like family peptide receptor 2	12	0.7337	horn development	42,46
CDHR4	cadherin related family member 4	22	0.6483	parasite resistance	39
IP6K1	inositol hexakisphosphate kinase 1	22	0.7258	parasite resistance	39
RNF123	ring finger protein 123	22	0.7129	parasite resistance	39
LEMD3	LEM domain containing 3	5	0.6871	growth, meat traits	42
NCOA1	nuclear receptor coactivator 1	11	0.6699	muscular development, reproduction	42,43
APOF	apolipoprotein F	5	0.7752	meat traits	43
HEYL	hes related family bHLH transcription factor with YRPW motif-like	3	0.6418	meat traits	48
SLC29A1	solute carrier family 29 member 1	23	0.7452	meat traits	48
TFIP11	tuftelin interacting protein 11	17	0.6634	meat traits	48
ASNS	asparagine synthetase	4	0.6548	meat traits	48
RPL15	ribosomal protein L15	27	0.7344	meat traits	48
RELA	RELA proto-oncogene, NF-kB subunit	29	0.6763	dairy traits	48
RDH5	retinol dehydrogenase 5	5	0.7020	meat traits	48
ITPR3	inositol 1,4,5-trisphosphate receptor type 3	23	0.6633	meat traits	48

Discussion

Candidate genes correlated with environmental adaptability

This study investigated whether the domestication and artificial selection could have shaped the genomes of MG and MN and how to adapt to local environmental challenges (e.g., parasite and viral challenges). It has been found that some of the PSGs identified in MN are associated with environmental adaptability (e.g., DVL2, VPS13A, GNA14, KLHL3, HSPA4, GPR50, and FGF9). Previous study suggested the regulation of the downstream of DVL2 in an endogenous Wnt pathway that can be operated in outer root sheath cells. DVL2 is a cell type that was previously shown to play a role in limiting hair growth and the candidate for wool production because of its link to the hair follicle cycle.⁴² In this study, the frequencies of DVL2 mutations are obviously different between the MG and MN, especially the three mutations in CDS including Chr19:27574636A>G, Chr19:27574810G>A, and Chr19:27578136G>A (Figure 3B). These variations might affect the normal transcription and expression of the DVL2 gene and further have impacts on hair growth, which makes it adapt to the warm climate in the south and the cold climate in the north. Moreover, for the other PSGs in MN, VPS13A and GNA14 genes were located in significant sweep regions of the pig, which were correlated with blood coagulation and circulation and enabled temperature adaptation.⁴¹ KLHL3 was over-represented in biological processes correlated with kidney development, which can affect water reabsorption in extreme temperature conditions in several species.⁴¹ HSPA4 belongs to the family of heat shock 70 kDa protein, known for improving cell protection against heat damage and preventing protein denaturation. Genome-wide analysis of African cattle also showed selective sweeps for heat tolerance in the gene region.¹⁵ GPR50 plays a role in thermogenesis and might be directly correlated with wisent adaptation to colder climatic conditions.⁴³ It was suggested that FGF9, the functions of which are correlated with the developments of the respiratory system, respiratory tube, and lung, is the adaptive gene in a test of genetic differentiation between domestic and argali sheep.⁴⁴ Altogether, the selection on those genes agrees with a previous report that zebu breeds have a better ability to regulate body temperature to respond to heat stress. In addition, we further detected genes (e.g., CDHR4, IP6K1, and RNF123 under selection in the MG genome and MON1A, MST1R, DNAH2, LCT, and THPO under selection in the MN genome) that were reported as involved in internal and external parasite tolerance in different cattle breeds.^33,38,39,43

In addition, some PSGs found in MN (e.g., SAR1B, IFNAR2, HSPA9, CD5, VAMP7, IL15, FOXC2, and TPM2) were briefly reported to be associated with immune response and disease resistance in cattle.^11,32,46,47 It is noteworthy that PSGs in MG also cover the RXFP2 and MC1R gene. RXFP2 serves as a receptor for the relaxin and insulin-like factor 3 proteins, of which the impacts on horn status and size are dependent on its biochemical interaction with testosterone. It has been suggested to be critical for developing the horns of goats and cattle and to determine the size and presence of the horns of wild and domestic sheep.^9,42,46 Previous reports suggested that mutations in MC1R generate red (or chestnut) coat colors in many different species (e.g., dogs, mice, horses, and cattle)¹⁵ Selection on these genes associated with horn or coat color may be responses to adapt the free-range environment of most MG cattle. Some PSGs in MN (e.g., CNGB1 and R3HDM1) were olfaction and food conversion efficiency,^22,43 respectively. Moreover, PIK3CB was reported to be involved in the Akt/PI3K and MAPK signaling pathways, which is important for high feed efficiency.¹⁶ Selection of these genes in MN may facilitate the adaption to available food resources and vegetation diversity (forest/steppe habitat). These genes are recognized the promising candidates for further investigation.

Candidate genes correlated to production and quality traits

As agriculture mechanization has been developed, these traits are gradually depreciated, and the demand for beef is increasing with the rapid improvement of people’s living standards. Hence, genetic improvement through selection on genes that are correlated with economic characters (e.g., meat production and quality) has been largely considered during artificial selection.

In MN, because of their role in fertilization, growth, and embryonic development, some of the PSGs (e.g., SLC38A3, EXOC3, STAT5B, and EMD) were also recommended as candidate genes for meat traits by whole-genome resequencing of the Japanese native cattle Kuchinoshima-Ushi.⁴⁸ FGF1 was implicated in reproduction traits.⁹ It was found that polymorphisms in ADRB2 have impacts on the selected carcass traits of Qinchuan cattle.⁴⁹ PDHB, located in the regions of homozygosity of Hanwoo (Korean cattle), has been involved in the tricarboxylic acid cycle.^32,38 High-resolution analysis between fine-wool Merino and coarse-wool Churra sheep breeds found the SLC2A4 gene was harbored in the selection sweeps and linked to meat production/quality traits.⁴² In addition, PCSK4, SPOCK1, and CABIN1 have vital roles in fertility and are recognized to be vital candidates for reproduction traits in animals as well.^9,34 Likewise, in MG, some genes were found to have relationships with animal production traits. LEMD3 showed extreme allele frequencies among different sheep breeds in previous research.⁴² Intronic variation in this gene has direct correlations with a pleiotropic quantitative trait locus (QTL) reported in cattle for calving ease direct, birth weight, ribeye muscle area, and marbling.⁴² According to Gautier et al.,⁴³ NCOA1 could be considered as a candidate gene for skeletal and muscular development, reproduction, and embryonic and organ development in cattle. APOF has key functions in lipid metabolism.⁴³ Furthermore, it was reported that some genes, including HEYL, SLC29A1, TFIP11, ASNS, RPL15, RELA, RDH5, and ITPR3, have underlying correlations with economically significant traits (e.g., domestication, growth rate, and meat/milk production) in other cattle breeds.⁴⁸ These results above might be responses to diet adaptation and artificial selection on meat and dairy traits.

Materials and methods

Ethics approval

All the experimental procedures with cattle used in the present study were approved by the Experimental Animal Manage Committee (EAMC) of Northwest A&F University (2011-31,101,684). All the operations and experimental procedures complied with the National Standard of Laboratory Animals Guidelines for Ethical Review of Animal Welfare (GB/T 35892-2018) and Guide for the Care and Use of Laboratory Animals: Eighth Edition.

Sample preparation and sequencing

Blood samples of MG (n = 13, female) and MN (n = 13, female) were randomly collected from Inner Mongolia and Fujian, China, respectively. The 5 μg of blood extracted genomic DNA underwent shearing into small fragments of 200–800 bp using the Covaries system (Life Technologies). The DNA library was built from these fragments through ligation of paired-end adopters and insertion with 500 bp PCR amplification. The amplicons underwent sequencing via the Illumina HiSeq X Ten platform, and 150 bp paired-end reads were produced.

Alignment and variation identification

The paired-end reads of 100 bp length reads were aligned to the reference Bos_taurus_UMD_3.1¹² by Burrows-Wheeler-Alignment (BWA) software with the parameters as follows: “bwa aln-m200000-o1-e30-15-L-I-t4-n0.04-R20-f” and “bwa sampe-a 650-n30-N30.”⁵⁰ The output was aligned and converted to bam format from SAM format with the use of SAMtools.⁵¹ The resultant bam files underwent sorting, and the reads underwent filtering in duplicate using Picard package (http://broadinstitute.github.io/picard/, version 1.92). The indels and SNPs were identified using Genome Analysis Toolkit (GATK, version 2.4-9).⁵² Beagle⁵³ was employed to refine the genotype calls and infer the haplotypes using genotype likelihoods from the GATK result. CNVs were identified using CNVnator v0.2.7⁵⁴ with default parameters. CNVs with fewer than 3 reads supported were removed.

Variation annotation

Computer software ANNOVAR was applied for the identification of the variation resulting from these variants in protein coding regions and consequently on the amino acid sequences altered.⁵⁵ The word “upstream” was used for the variants overlapping the 1,000 bp upstream region relative to transcription start site, while the term “downstream” annotates the variants overlapping 1,000 bp downstream region relative to the end site of the gene. Likewise, “upstream/downstream” reveals that the variant exists in both upstream and downstream positions (probably two different genes). The term “stop gain” refers to the situation in which a stop codon was created by a nonsynonymous SNP in the variant sequence, while the term “stop loss” means loss of stop codon in the variant sequence site by the nonsynonymous SNP. The word “unknown” refers to the unknown function caused by certain errors within the genetic structure (defined in the database file). The term “splicing” refers to the variant’s placement position being inside the 2 bp splicing junction. The missense mutation functional impact was predicted by the SIFT algorithm.⁵⁶ During annotation through ANNOVAR software, the source databases used included NCBI, dbSNP build 140 (ftp://ftp.ncbi.nlm.nih.gov/snp/organisms/cow_9913/chr_rpts/) and NCBI RNA refseq, Ensembl release 78 provided by UCSC (ftp://hgdownload.cse.ucsc.edu/goldenPath/bosTau6/database).

Phylogenetic and population structure analyses

By using PHYLIP v3.695 program (https://evolution.genetics.washington.edu/phylip.html), this study built the neighbor-joining phylogenetic tree of the SNP data. By using ADMIXTURE,⁵⁷ this study further inferred the structure of the population. The smartPCA program of the EIGENSOFT package was applied for the analysis of principal components.⁵⁸

Linkage disequilibrium

For the estimation of the genome-wide linkage disequilibrium (LD) of each breed, the mean r² values for pairwise markers were calculated using Haploview⁵⁹ software. This study only employed SNPs with a minor allele frequency above 0.05.

Selective sweep analysis and enrichment analysis

Domestication and artificial selection led to a reduction in nucleotide diversity and changed allele frequency. To identify the genes that have undergone positive selection and investigate the differences between these two cattle breeds, the differentiation (F_ST) was measured, and their genetic diversity ratio (θπ) was compared with the use of a sliding window method (50 kb window and 10 kb step). The 5% of windows with the highest F_ST and θπ ratio were considered the potential selected windows (PSWs), and the adjacent windows were merged into a single region, thereby becoming the potential selected region. The population variation was estimated by pairwise FST using an unbiased estimator.⁶⁰ A sliding window approach was employed for measuring mean pairwise nucleotide diversity (θπ) and Hp value of each breed through VCFtools software with default parameters and 50 kb sliding windows in 10 kb steps.⁶¹ The θπ ratio between group MN and MG was calculated as ln(θπ,MN/θπ,MG), reflecting the loss of nucleotide diversity in MG relative to MN. Tajima’s D statistic was calculated with the use of VCFtools for each candidate gene. Selective sweep regions were labeled with cattle QTLdb release 29 from the Animal Quantitative Trait Loci Database.⁶² Functional enrichment analyses of PSGs were conducted using DAVID tools.⁶³