A genome-wide association study of five meat quality traits in Yorkshire pigs

Meat quality is an important trait in the pig industry. To identify genomic regions and haplotype blocks responsible for meat quality traits in pigs, a genome-wide association study was conducted for five traits including intramuscular fat content, pH at 45 min and 24 h, drip loss within 24 h and water-holding capacity in 231 Yorkshire barrows using illumina porcine 60k SNP chips. The results showed that a total of 344 single nucleotide polymorphisms (SNP) were significantly associated with five meat quality traits (P< 1 10). Moreover, 323 SNPs were within the reported QTL regions, of which 21 were novel. Also, 158 SNPs fell into the proximal region of meat quality related genes. In addition, 25 haplotype blocks based on 116 SNPs were revealed with SNP combination patterns for five traits. Our study added new SNP information for identification of meat quality traits in pigs and will help elucidate the mechanisms of meat quality in pigs.


Introduction
In the past, pig breeders have focused on increasing lean growth and decreasing backfat thickness. Unfortunately, meat qualities that influence meat flavor, tenderness and juiciness were not improved at the same time [1]. However, in modern pig breeding, improvements in these meat quality traits have been given higher priority [2,3]. In addition to intramuscular fat content (IMF), pH at 45 min and 24 h (pH 45m and pH 24h ), water-holding capacity (WHC) or drip loss within 24 h (DL) and color are now considered important traits for meat quality [4][5][6]. However, the genetic improvement of meat quality has proven to be difficult using established breeding methods because it relies on slaughtered measurements and carcass characterizations. To identify the quantitative trait loci (QTL) affecting meat quality traits, crossed populations of pigs have been widely used [7,8]. Many candidate genes, such as PRKAG3 gene, responsible for meat qualities have been identified in different populations [9,10]. Genomewide association study (GWAS) is an effective strategy to identify the candidate genes or specific genome region involved in quantitative traits such as meat qualities. Taking advantage of accuracy and unbiased prediction, GWAS has become widely used in candidate gene association and marker-assisted selection (MAS). Furthermore, the application of GWAS has been extended in porcine studies based on the draft of the pig genome and high-density single nucleotide polymorphism (SNP) chips [11]. In the study reported here, 116 significant SNPs, comprising 25 haplotypes blocks related to meat qualities in a Yorkshire pig population were identified. This work will potentially benefit genome selection for meat quality and shed light on the improvement of carcass quality in pigs.

Measurement of phenotypes
A total of 231 castrated Yorkshire pigs were fed ad libitum and slaughtered at 90 kg bodyweight. The longissimus dorsi muscle between the third and fourth ribs from the left carcass were excised, immediately frozen in liquid nitrogen and then stored at -80°C. The trait, pH 45m , was determined for one of the chops after the carcass was butchered and pH 24h taken directly on the longissimus muscle. WHC was determined for the longissimus muscle from the second to third lumbar by the filter paper press method [12]. Two-cm-thick muscle disks were taken between the third to fourth lumbar for determination of DL. At 2 h postmortem, the weight (initial weight) of another two 2-cm-thick chops was recorded and DL determined after 24 h. These chops were then stored in a sealed plastic bag at 4°C for 24 h, then blotted and reweighed (final weight). The DL at 24 h was calculated as follows: DL ¼ ðinitial weight-f inal weightÞ=initial weight Â 100, the IMF content in thoracic-lumbar longissimus muscle was determined by the petroleum ether extraction method [13]. These phenotypic data were determined independently three times, mean and standard errors were also calculated.

Whole genome SNP genotyping
Genomic DNA was isolated from ear tissue using a TIANamp Genomic DNA Kit (TIANGEN, Beijing) following the manufacturer's instructions. SNP genotyping was performed using PorcineSNP60 beadchip (Illumina, USA) that contains 61177 SNP sites across the whole pig genome.

PLINK-based GWAS
GWAS using IMF, pH 45m , pH 24h , DL and WHC phenotypes was implemented by PLINK software [14] with SNP call rate > 90%, minor allele frequency (MAF) > 0.03 and P value of χ 2 test for Hardy-Weinberg Equilibrium (HWE) > 1Â10 -6 . The animals with sample call rates < 90% were excluded. The phenotypic data were subjected to logarithmic transformation for fitting normal distribution before GWAS. The PLINK-based GWAS were performed with maximum one million permutations and the threshold empirical P value at 1Â10 -4 for each phenotype.

SNP annotation based on proximal QTLs and genes
The significant SNPs (P < 1Â10 -4 ) were locational annotated by their proximal QTLs and genes. Previously reported QTLs for meat quality were obtained from Animal QTLdb (http://www.animalgenome.org) [15], and the SNPs that fell into the region of QTLs or genes based on linkage analysis were scanned on porcine chromosomes using the Variant Effect Predictor tool in the Ensembl website [16].

Linkage disequilibrium analysis
Linkage disequilibrium (LD) analysis and haplotype block detection of all significant SNPs were carried out on the chromosomal region, which contained all significant SNPs associated with each phenotype. The LD blocks were first generated within 500 kb, then haplotype blocks were estimated by an accelerated EM algorithm [17] using Haploview software [18].

Phenotypic variation and trait correlations
The descriptive statistics for IMF, pH 45m , pH 24h , DL and WHC measures of meat quality were normalized according to the corresponding sample size (Table 1 and AppendixA-Table S1). The range of IMF content varied from 1.18% to 3.90% across the population, and pH 24h was significantly lower than pH 45m as a consequence of the meat aging process. Pearson's correlation coefficients for pair-wise comparisons between the traits were calculated (Table 1). This analysis showed that pH 24h was moderately correlated with the DL (r = 0.22) and pH 45m (r = 0.24) parameters.

Genome-wide association study
The GWAS for the five traits identified 46320 SNPs over 19 porcine chromosomes (Chr.) after applying the quality control thresholds. Of these, 344 were identified as significant (P < 1Â10 -4 ) candidate SNPs associated with the different traits (Table 2 and Appendix B). Manhattan plots for five traits based on empirical P values also showed that GWAS of the population had more power to detect the SNPs associated with the IMF, pH 24h and pH 45m than the DL and WHC traits, because more significant SNPs were identified for the first three traits. Additionally, the SNPs associated with pH 45m and pH 24h traits were dispersed over the porcine genome, which indicates that the meat pH measures were extensively influenced by multiple gene interactions. In contrast, 12 out of 45 SNPs for IMF trait tended to be located in porcine Chr. 14 ( Fig.  1).

SNPs and functional annotation
There were four SNPs on Chr. 16   0.32, 0.16, 0.10 and 0.07, respectively. The other block included three haplotypes, GAAGG, ACGAA and AAGAG, with frequencies of 0.26, 0.72 and 0.02. The coefficient of connection between two blocks was 0.66 (Fig. 2a). In addition, five blocks on Chr. 1 were composed of 22 SNPs (Fig. 2b), two blocks on Chr. 11 composed of 13 SNPs (Fig. 2c) and two blocks on Chr. 4 composed of 11 SNPs (data not shown) were also determined by the same analysis. Moreover, two blocks, 391 and 308 kb, on Chr. 6 contained nine SNPs associated with the pH 45m trait. The r 2 of SNPs in the blocks ranged from 0.93 to 1.00. There were four haplotypes in 391 kb haplotype, TACA, CGTC, CACA and CGTA, with frequencies of 0.50, 0.43, 0.04 and 0.03 and two haplotypes in 308 kb haplotype, ACCCC and GTTTA with frequencies 0.28 and 0.72. The coefficient of connection between two blocks was 0.46 (Fig. 2d). A further five blocks composed of 23 SNPs were identified for pH 45m trait, one block on each of Chr. 2, 8, 18 and two on Chr. 17.

Discussion
GWAS has been considered as a useful technique for candidate gene association and QTLs identification for complex traits, which are mostly focused on important production traits in domestic animals such as backfat and milk yield. In this study, a GWAS was conducted to detect potential sites responsible for the IMF, pH, DL and WHC traits of meat in pigs. As a consequence of the close relationship between accuracy of GWAS and heritability (h 2 ) of trait, many studies have reported that the heritabilities of meat quality related traits varied from 0.13 to 0.32 [19]. However, the higher h 2 of IMF trait (0.44) and pH 24h trait (0.32) found in the current work indicates that the genome-wide selection for IMF content and meat pH might be possible in pig population [20,21]. Based on the heritabilities and trait correlations, meat quality traits can be used in selection programs, if they are related to improved meat quality but not strongly opposed to other important economic traits. Based on the comparisons of QTLs in the PigQTLdb database, a total of 358 SNPs found in the current study also validated the previous reported QTL regions related to meat qualities. ASGA0097315 on Chr.13 for pH 24h trait falls in the intron of the FHL2 gene, which is considered as a key regulator of fatty acids metabolism and important for meat quality. In addition, this SNP and its 375 kb haplotype blocks were located within the reported QTL region for pH 24h , pH 45m and DL traits [22,23]. MARC0110724 and MARC007 5674 related to the pH 24h trait in the current study were also detected in another pig population [24]. These two SNP and ALGA0000021, ALGA0000046 were located in one haplotype block of 453 kb on Chr. 1. Moreover, they fell into the QTLs for DL and WHC traits. In addition, MARC0110724 was mapped in the intron of the WDR27 gene that was associated with type I diabetes in previous GWAS. DNAJC3 was another candidate gene that was associated with the pH 24h trait. The significant SNP identified in the intron of DNAJC3 was ASGA0051428.
Studies have demonstrated that DNAJC3 null mice showed pancreatic-cell failure and diabetes, and the expression of DNAJC3 at the protein level was increased in human T2D islets. The pH of pork depends mostly on the glycogen content of and glycolysis in postmortem muscle. Furthermore, this SNP was not only located within the reported QTL for DL and meat color score but also in one block of 247 kb on Chr. 11. ASGA0048299, as one of the significant SNPs for pH 24h , was located in the intron of PIP4K2A. The same locus and candidate gene for meat quality traits has been identified in the Duroc pigs. PIP4K2B -/mice result in increasing insulin sensitivity of skeletal muscle and reducing adiposity. Insulin stimulates glucose transport into muscle and fat, where it is metabolized through glycolysis, or converted to glycogen or triglycerides, resulting in adipose tissue accumulation. Therefore, the mutation of PIP4K2B leads indirectly to a decrease in transport of glucose into white adipose tissues, because the increased insulin stimulates glucose transport in skeletal muscle causing a lower basal insulin-secretion requirement. Also, PIP4K2A might compensate for the loss of PIP4K2B according to the high ratio of PIP4K2B expression to PIP4K2A expression in skeletal muscle. The H3GA0052676 SNP in the intron of the RAB31 gene and a 308 kb haplotype block on Chr. 6 for the pH 45m trait confirmed that the enhanced insulinstimulated Glut4 translocation and glucose uptake appeared in the knockdown of Rab31 in 3T3-L1 adipocytes. The other Chr. 6 ASGA0103250 SNP in the intron of the ARHGEF19 gene for pH 45m trait was related to adipogenesis through DNA methylation change [25]. SNP H3GA0056439 in the upstream of MFAP2 gene and a 391 KB haplotype block on Chr. 6 was highly related to adipocyte hypertrophy and predisposition to metabolic dysfunction, which indicated it is correlated with IMF in pigs.

Conclusions
This study of 231 Yorkshire pigs identified 344 significant SNPs (P < 1Â10 -4 ) associated with five meat quality traits, in which 323 SNPs were located in the reported QTL regions and 21 were novel. Twenty-five haplotype blocks originated from 116 SNPs provided the combination patterns of SNPs for five traits. Also, the potential genes associated with SNPs supplied candidate targets for improving meat quality in pigs. These GWAS results may provide extra landmarks to elucidate the mechanism of meat quality in pigs, as well as shedding light on the efficiency of genome wide selections.