Whole-genome resequencing to identify candidate genes for the QTL for oleic acid percentage in Japanese Black cattle

2.1 Sample selection

In our previous study, we used Japanese Black cattle population comprising 1,836 animals that had been bred in the Hyogo Prefecture (Kawaguchi et al., 2018). After correcting for the C18:1 phenotype of each animal using an analytical model, we selected a total of 200 animals for conducting pool-based GWAS: 100 of these animals had higher C18:1 and the other 100 had lower C18:1 in the previous study.

In this study, eight animals were selected from the 200 animals used in the GWAS for whole-genome resequencing based on their sires and genotype of the most significant SNP in the GWAS (Hapmap43702-BTA-84086) (Table 1). We selected four animals with the TT genotype from the 100 “high” animals and four animals with the CC genotype from the 100 “low” animals, which were the progenies of different sires among the four animals in the high and low groups. Genomic DNA was extracted from each 50-mg longissimus cervicis muscle sample following the standard phenol–chloroform method.

2.2 Whole-genome resequencing

DNA degradation was monitored based on its concentration by spectrometry, fluorometry, and 1% agarose gel electrophoresis. A paired-end library was constructed using high-quality DNA for each individual, and its read length was 150 bp. Sequencing was performed using a HiSeq X Five Sequencing System (Illumina Inc., San Diego, CA, USA). Sequencing data were normalized by Genedata Expressionist 9.1.4a. We mapped the reads to the cattle reference genome assembly (UCSC bosTau8) downloaded from the UCSC Genome Browser assembly (https://genome-asia.ucsc.edu/cgi-bin/hgGateway) using BWA-MEM 0.7.12. and excluded PCR duplicates using Picard 2.2.4. GATK 3.6 (2016-12-08-g1c2527f) was used to call polymorphisms by comparing the genome sequences, including the reference sequence. The polymorphisms were annotated to the gene reference (NCBI RefSeq) based on their location (intron, exon, untranslated region, upstream, downstream, splice site, and intergenic region) and characteristics (synonymous/non-synonymous amino acid replacement, gain/loss of start/stop site, and frameshift mutations) using SnpEff v4.2 with the reference sequence (bosTau8).

2.3 Narrowing the polymorphisms and genes

In our previous study, we determined the region 5 Mbp upstream and downstream of Hapmap43702-BTA-84086 (64.9 –74.9 Mbp) to be the candidate region (Kawaguchi et al., 2018). In this study, we selected polymorphisms in the candidate region from among all the polymorphisms detected by whole-genome resequencing.

We first excluded intergenic polymorphisms and subsequently focused on the linkage disequilibrium (LD) between the polymorphisms and Hapmap43702-BTA-84086 because the responsible polymorphism is in LD with Hapmap43702-BTA-84086. We compared genotypes of four animals in high group with those of four animals in low group as an indicator of LD. The Hapmap43702-BTA-84086 genotypes completely differed between the high and low groups; it was defined that there were a total of eight allele differences between the high and low groups. Therefore, polymorphisms with a large number of allele differences were expected to be in LD with the SNP. In this study, we focused on the polymorphisms with over three allele differences.

We investigated the function of genes containing these polymorphisms in the NCBI database and previous reports and determined candidate genes based on their function in fatty acid metabolism, such as synthesis, transport, desaturation, and oxidation.

2.4 Genotyping the candidate SNPs

Polymorphisms were prioritized as candidates based on their location and characteristics. We selected candidate polymorphisms that were the most likely to affect the function of each candidate gene.

To test the effect of candidate polymorphisms on C18:1, we used a Japanese Black cattle population (n = 899) that had been randomly selected from 1,836 individuals bred in the Hyogo Prefecture, Japan, and graded them from 2010 to 2012. This population contained at least 10 offspring per each sire. We genotyped CYB5R4 c.*349G > T and MED23 c.3700G > A by PCR-RFLP. The primer sets for PCR amplification were designed based on the GenBank sequence (AC_000166.1) using OLIGO 7.41. The PCR reaction was performed at the annealing temperature and extension period shown in Table 2. The PCR products for CYB5R4 c.*349G > T and MED23 c.3700G > A were digested using MslI and ScaI, respectively.

2.5 Statistical analysis

We performed statistical analysis using the same samples and methods described in our previous study (Kawaguchi et al., 2018). To correct the phenotype, a linear mixed model was applied to C18:1 as follows:

where y is a vector of observations (C18:1); b is a vector of fixed effects including overall mean, slaughter year, slaughter month, sex of animals, linear covariate for inbreeding coefficient, and linear and quadratic covariates for age at slaughter; u₁ and u₂ are vectors of random farm and animal genetic effects, respectively; e is a vector of random residual effects; and X, Z₁, and Z₂ are known incident matrices. Restricted maximum likelihood by the expectation-maximization algorithm (EM-REML) and best linear unbiased prediction (BLUP) were employed to estimate variance components and all random effects in the model, respectively. A total of 8661 animals including ancestors were evaluated and the heritability was estimated to be 0.423 ± 0.119.

The e-values of 899 animals were analyzed by one-way analysis of variance (ANOVA) with genotype as a source of variation and Tukey-Kramer's honestly significant difference (HSD) test implemented in JMP13 (SAS institute Inc., Cary, NC, USA).

The proportion of additive genetic variance was calculated following the method of Matsuhashi et al. (2011). LD coefficients (r²) between the polymorphisms were calculated using HAPLOVIEW 4.0.

3.1 Detection of candidate polymorphisms by genome resequencing

We conducted whole-genome resequencing in eight Japanese Black cattle (Supporting Information Table S1). A total of 39,658 polymorphisms were identified in the candidate region (Chr 9: 64,956,436–74,956,436) by comparing genome sequences between nine animals, including the reference (bosTau8). Of these, 10,045 polymorphisms were located within the genes. We further narrowed the polymorphisms to 1993 with over three allele differences between the high and low groups, which were located within a total of 23 genes. These SNPs were counted based on genes and annotations to select the candidate polymorphisms (Supporting Information Table S2).

3.2 Candidate genes in terms of gene function

We investigated the functions of the 23 genes based on information in the NCBI database and previous reports (Supporting Information Table S3). Three genes, that is, CYB5R4, MED23, and VNN1, were determined to be candidate genes in terms of their function in fatty acid metabolism.

CYB5R4 is an electron donor for fatty acid desaturation by stearoyl-CoA desaturase (SCD) (Deng et al., 2010; Zhu et al., 2004). Larade et al. (2008) reported that the C18 desaturation index (C18:1/C18:0) was markedly lower in CYB5R4-knockout mouse than in a wild-type mouse. Polymorphisms within the CYB5R4 gene may affect SCD activity, which would alter the ratio between C18:1 and C18:0.

MED23 is a subunit of the Mediator complex and transmits information between RNA polymerase II and transcription factors, such as FOXO1 (Chu et al., 2014; Knuesel & Taatjes, 2011). A previous study reported that FOXO1 suppresses the expression of acetyl-CoA carboxylase (ACC), which catalyzes the carboxylation of acetyl-CoA to malonyl-CoA, that is, the first step of long-chain fatty acid synthesis (Bastie et al., 2005). MED23 may control fatty acid synthesis by regulating the expression of ACC.

VNN1 is an enzyme that controls the amount of pantethine generated by pantetheinase activity (Kavian et al., 2015; Pitari et al., 2000). Acetyl-CoA and malonyl-CoA are converted into acetyl-ACP and malonyl-ACP, respectively, during one step of fatty acid synthesis. As pantethine bypasses this reaction as a metabolically active substrate for CoA and ACP (Kelly, 1997), VNN1 appears to be involved in fatty acid synthesis.

3.3 The effect of the candidate SNPs on C18:1

A total of 77, 125, and 28 polymorphisms were detected within the CYB5R4, MED23, and VNN1 genes, respectively (Supporting Information Table S2). We selected one candidate polymorphism from each candidate gene to test its effect on C18:1. We primarily focused on amino acid substitutions and secondarily focused on SNPs related to gene expression (5′UTR, 3′UTR, the promoter region, and splice site). The amino acid substitutions, V1234I and T66M, were selected within MED23 and VNN1 genes, respectively, while we selected the SNP on 3′UTR in CYB5R4 gene as amino acid substitution was not detected within the gene (Supporting Information Table S2). In short, CYB5R4 c.*349G > T, MED23 c.3700G > A (V1234I), and VNN1 c.197C > T (T66M) were selected as candidate SNPs for the QTL. VNN1 c.197C > T was selected as a candidate SNP in our previous study (Kawaguchi et al., 2018).

We previously genotyped the most significant SNP in GWAS (Hapmap43702-BTA-84086) and VNN1 c.197C > T for a Japanese Black cattle population (n = 899) (Kawaguchi et al., 2018). In addition, we genotyped CYB5R4 c.*349G>T and MED23 c.3700G>A for the same population in this study. The minor allele frequencies of Hapmap43702-BTA-84086, CYB5R4 c.*349G>T, MED23 c.3700G > A, and VNN1 c.197C > T were 0.261, 0.272, 0.280, and 0.201, respectively (Table 3). These SNPs were in relatively high LD with Hapmap43702-BTA-84086 (r² = 0.53–0.83) (Fig. 1). In our previous study, statistical analysis revealed that Hapmap43702-BTA-84086 and VNN1 c.197C > T were significantly associated with C18:1 (p = 0.0080 and 0.0162, respectively) (Kawaguchi et al., 2018). Therefore, we also conducted statistical analysis for CYB5R4 c.*349G > T and MED23 c.3700G > A to investigate and compare their effects on C18:1. ANOVA revealed that both SNPs were significantly associated with C18:1 (p = 0.0075 and 0.0295) (Table 4). Significant differences between genotypes were observed by the Tukey–Kramer's HSD test. The proportions of additive genetic variance for Hapmap43702-BTA-84086, CYB5R4 c.*349G > T, MED23 c.3700G > A, and VNN1 c.197C > T were 3.70, 4.20, 3.30, and 2.24, respectively.