Summary
Intramuscular fat (IMF) is one of the main meat quality traits for breeding programmes in livestock species. The main objective of this study was to identify genomic regions associated with IMF content comparing two rabbit populations divergently selected for this trait, and to generate a list of putative candidate genes. Animals were genotyped using the Affymetrix Axiom OrcunSNP Array (200k). After quality control, the data involved 477 animals and 93 540 SNPs. Two methods were used in this research: single marker regressions with the data adjusted by genomic relatedness, and a Bayesian multiple marker regression. Associated genomic regions were located on the rabbit chromosomes (OCU) OCU1, OCU8 and OCU13. The highest value for the percentage of the genomic variance explained by a genomic region was found in two consecutive genomic windows on OCU8 (7.34%). Genes in the associated regions of OCU1 and OCU8 presented biological functions related to the control of adipose cell function, lipid binding, transportation and localisation (APOLD1, PLBD1, PDE6H, GPRC5D and GPRC5A) and lipid metabolic processes (MTMR2). The EWSR1 gene, underlying the OCU13 region, is linked to the development of brown adipocytes. The findings suggest that there is a large component of polygenic effect behind the differences in IMF content in these two lines, as the variance explained by most of the windows was low. The genomic regions of OCU1, OCU8 and OCU13 revealed novel candidate genes. Further studies would be needed to validate the associations and explore their possible application in selection programmes.
Keywords: divergent selection, genome‐wide association study, intramuscular fat, meat quality, rabbits
Introduction
Intramuscular fat (IMF) contributes to improve organoleptic properties and sensory attributes of the meat, as demanded by consumers (Hocquette et al. 2010). Hence, a large number of studies have investigated the genetic factors controlling IMF content in meat and their implications for several species, e.g. in beef cattle (Sapp et al. 2002; Garrick 2011; Ochsner et al. 2017), swine (McLaren & Schultz 1992; Gao et al. 2007), sheep (Hopkins et al. 2011; Mortimer et al. 2014) and goats (Peña et al. 2011). Following these studies, IMF has emerged as one of the most important meat quality parameters and in a few cases it has been included in breeding programmes (Gotoh et al. 2018; Pannier et al. 2018).
Moderate‐to‐high heritability and large variability have been reported for livestock IMF traits, which argue for a good potential for improving meat quality through genetic selection. IMF heritability is around 0.53 in swine (Ros‐Freixedes et al. 2016), 0.38 in cattle (Mateescu et al. 2015), 0.48 in sheep (Mortimer et al. 2014) and 0.54 in rabbit (Martínez‐Álvaro et al. 2016). Important limitations to IMF selection are the IMF being recorded mainly at slaughter and the phenotyping process being costly. In this context, genetic marker selection based on quantitative trait locus (QTL) with high or moderate effect size could overcome some of these limitations.
At genomic level, studies carried out in beef cattle suggest that IMF could be influenced by a large number of genes (Strucken et al. 2017). Nevertheless, studies in Japanese Black cattle have reported genomic markers with large effects on IMF or marbling score around the SCD, FASN, AKIRIN2, EDG1 and RPL27A genes (Gotoh et al. 2014; Sukegawa et al. 2014). Genomic markers on the genes SCD and FASN have been incorporated into a breeding programme for this breed to select elite sires (Gotoh et al. 2018). In swine, similarly to beef cattle, the results of experiments associating genetic markers with IMF are hardly conclusive with regard to the magnitude and importance of discovered associations (Pena et al. 2016). However, traits correlated to IMF such as fatty acid profiles have shown a noteworthy QTL on chromosome 14 in a Duroc commercial line (Uemoto et al. 2012; Ros‐Freixedes et al. 2016). So far, IMF appears as a troublesome trait for mapping studies in livestock species, owing to either the lack of validation in the results or insufficient power to detect genetic causal variants. Thus, genomic studies to understand the genetic control of IMF are still needed.
The rabbit has been shown to be an excellent animal model for other livestock species (Miller et al. 2014). Further, the recent availability of a high‐density SNP array has facilitated the performance of genomic studies. At the Universitat Politècnica de València, a successful divergent selection experiment for IMF has been carried out (Martínez‐Álvaro et al. 2016). The developed rabbit lines were kept in the same environment and selection criteria only differ for the IMF selection objective. Selection could have modified SNP frequencies in opposite directions, leading to intermediate allelic frequencies when both lines are jointly considered. This could increase the detection power of associated loci in a genome‐wide association study (GWAS) based on this experimental design.
The aim of this study was to carry out GWASs using these divergently selected rabbit lines to identify genomic regions associated with IMF and generate a list of putative candidate genes affecting this trait. Two different methods (single marker regression, SMR, and Bayesian multiple marker regression, BMMR) were applied to confirm the identified relevant genomic regions.
Materials and methods
Ethical statement
All experimental procedures were approved by the Ethical Committee of the Universitat Politècnica de València, according to Council Directives 98/58/EC (European Economic Community, 1998).
Animals and phenotypes
The animals of this study came from two rabbit lines divergently selected for IMF during nine generations at the Universitat Politècnica de València. The base population was composed of 83 does and 13 males from a synthetic rabbit line (Zomeño et al. 2013). The selection criterion was IMF content collected in two full siblings of the first parity. The selection of the males was within the sire family, avoiding mating between cousins to control inbreeding. At the ninth generation, the high‐IMF line consisted of 55 does and 10 males, and the low‐IMF line consisted of 61 does and 10 males. Over all animals, the mean was 1.09 g of IMF per 100 g of Longissimus thoracis et lumborum (LTH) muscle, after adjusting data for systematic effects (parity order, line, month‐season and sex) and a common litter random effect. The high‐IMF line had a mean of 1.27 g/100 g of LTH with 0.21 standard deviations, and the low‐IMF line had a mean of 0.83 g/100 g of LTH with 0.07 standard deviations. Details about the IMF divergent selection experiment can be found in Martínez‐Álvaro et al. (2016). The selection response was around 3.1 standard deviations at the ninth generation, calculated as the difference between lines. The phenotypic difference between lines was 41% of the mean of the base population.
The rabbits were brought up jointly from 33 days at weaning until slaughter under the same handling and feeding conditions. At 9 weeks from birth, the rabbits were slaughtered following a fasting period of 4 h. Carcasses were chilled 24 h at 2.5 °C after slaughter and dissected to obtain a sample of the left LTH muscle for each animal. These samples were minced, frozen, lyophilised and milled. The IMF data were obtained using near‐infrared spectroscopy (model 5000; FOSS NIRSystems Inc., Hilleroed, Denmark; Zomeño et al. 2013; Martínez‐Álvaro et al. 2016). In the last generation, 729 samples of the left LTH muscle of each animal were collected and IMF measured to compute the IMF selection response, and 480 rabbits were chosen from groups of an average size of four siblings per doe (dam) for the GWAS.
Genotyping and quality control
Obliquus abdominis muscle specimens (~50 g), obtained after slaughter of the animals, were used for DNA extraction using a standard protocol (Green et al. 2012). A total of 480 individuals were genotyped using the Affymetrix Axiom OrcunSNP Array (Affymetrix Inc., Santa Clara, CA, USA) at the ‘Centro Nacional de Genotipado’ (CeGen), Universidad de Santiago de Compostela. The SNP array contains 199 692 genetic molecular markers. The quality control was performed using axiom analysis suite version 3.0.1.4 and zanardi (Marras et al. 2017). SNPs with a call rate of at least 0.95, MAF of at least 0.03 and a known autosomal chromosome position according to OryCun2.0 assembly (Carneiro et al. 2014) were used in the analyses. Furthermore, animals missing more than 3% of marker genotypes, or failing a Mendelian inheritance test, were excluded. The remaining missing genotypes were imputed by the software beagle version 4.0 (Browning & Browning 2016). The SNPs with an imputation quality score R 2> 0.75 were included. After filtering, the data included 477 animals (240 from the high‐IMF line and 237 from the low‐IMF line) and 93 540 SNPs. In addition, the SNP density was described in this research because the rabbit SNP array is new (Blasco & Pena 2018).
Genome‐wide association study
Prior to performing the GWAS, we performed a multidimensional scaling analysis to evaluate the population structure in our genomic data. The method treats the distances as Euclidean distances and preserves the original distance metric, between points, as well as possible (Borg & Groenen 2005). The command cmdscale() from the R package stats was used to implement this analysis (R Core Team 2013).
Two methods were employed in this study: a frequentist and a Bayesian. Both methods included the mean and the systematic effects in the model: month‐season (five levels), sex (two levels), order‐parity (three levels) and line (two levels). The inclusion of a common litter random effect in the model was evaluated owing to the importance of this effect in previous studies of IMF in rabbits (Martínez‐Álvaro et al. 2016). Inclusion of this effect did not affect GWAS results (not shown), hence for simplicity we excluded this effect in the GWAS.
Single marker regression (SMR) with the data adjusted by genomic relatedness. The analysis was implemented using a family‐based score test for association (FASTA). The SNP effects were evaluated with FASTA based on a polygenic‐lineal mixed model that included the genomic kinship matrix to explain relatedness in the sampled population (Chen & Abecasis 2007). The model equation was:
where is the vector of IMF phenotypes, is a vector of ones, is the trait mean, is the design matrix for the systematic effects, is the vector of systematic effects, is the substitution effect for a particular SNP, is the vector of genotypes for each SNP denoted as the number of reference alleles for a particular SNP (0, 1 or 2), is the design matrix for random polygenetic effects, is the vector of random polygenic effects with a normal distribution and is the vector of random residual effects with a normal distribution; is the genomic variance and is the genomic kinship matrix computed using the genomic data by the method of Astle & Balding (2009). The identity matrix was denoted as and is the residual variance. The implementation of the association analysis was performed using R software package genABEL (Aulchenko et al. 2007). Furthermore, we utilised a genomic control method to avoid inflation in the statistic test. We calculated the lambda parameter that indicates the excess of false positives in the results. When its application is needed, the regression factor λ corrects the observed P‐values leading to new P‐values for every assessed SNP (Aulchenko et al. 2007). In this research, we used two thresholds: an LD‐adjusted Bonferroni (8.12 × 10−6) calculated for 10 Mb LD blocks according to LD analysis implemented in plink (Purcell et al. 2007), and also, a suggestive threshold of 1 × 10−4 owing to the high relatedness of the samples (Lander & Kruglyak 1995; Sahana et al. 2011; Do et al. 2018). As Bonferroni is a conservative method, we also implemented the suggestive threshold because it is less stringent as the samples from animals with high relatedness would have genomic segments of LD larger than those in humans (Wang et al. 2016c; Schmid & Bennewitz 2017). Therefore, the number of independent sites could be overestimated causing false‐negative results if SNP density is not large enough to adjust Bonferroni by LD (Spencer et al. 2009; Do et al. 2014).
Bayesian multiple marker regression (BMMR). This method is more robust to population structure than SMR approaches (Toosi et al. 2018). However, the line effect would correct for potential biases that might be derived by the family‐data structures in the investigated rabbit populations. Thus, the line effect remained in the BMMR model. The parameters were estimated with the following Bayes B model (Cesar et al. 2014; Ros‐Freixedes et al. 2016):
where , , and are the same as in the frequentist method shown above, is the vector including the genotypic covariate for each SNP or locus (0, 1 or 2), is the random substitution effect for and is the random 0/1 variable that represents the presence ( = 1 with probability 1 − π) or absence ( = 0 with probability π) of SNPs in the model for a given iteration. The value of π is defined as the proportion of SNPs with zero effects in the model. The value of π in our study was 0.9988, which means that between 100 and 200 SNP markers have non‐zero effects for every iteration. The parameters of the model were estimated with marginal posterior distributions using Markov chain Monte Carlo. After some exploratory analysis, a total of 825 000 iterations were performed, with a burn‐in period of 225 000 iterations. Only one sample every 60 iterations was saved to avoid the high correlation between consecutive samples. Gensel ® version 4.90 software (Garrick & Fernando 2013) was used for the GWAS analysis. The relevance of the association was assessed using two criteria, the Bayes factor (Stephens & Balding 2009; Ros‐Freixedes et al. 2016) and the percentage of the genomic variance explained for non‐overlapping genomic windows of 1 Mb, calculated by marginal posterior density. The genomic windows were defined for each chromosome and according to the OryCun2.0 rabbit genome assembly (Carneiro et al. 2014). In our study, 1999 genomic windows were defined. Those windows accounting for at least 1.0% of the total genomic variance were considerate as important to continue with the subsequent analysis (Cesar et al. 2014). This threshold was 20 times greater than the average genomic variance explained by a window (0.05%). We also considered the consecutive windows that explained at least 0.5% of genomic variance having a strong LD between them (Ros‐Freixedes et al. 2016) as SNPs associated with a causal variant can be located between consecutive windows and the estimated effect of association could be divided among these windows, hindering the detection of a genomic region (Beissinger et al. 2015).
In this study, we integrated the results from both frequentist and Bayesian methods to define the relevance of associations. This was established by the following procedure: first, we drew all genomic windows that overcame the condition expressed in the above paragraph. Then, the genomic windows harbouring SNPs above or around a Bayes factor of 20 (Kass & Raftery 1995) were extracted and considered as relevant genomic windows. These SNPs reaching at least one of thresholds, either suggestive or Bayes factor thresholds, were denoted as relevant polymorphisms. Finally, the genomic regions having relevant associations were chosen for functional gene analysis.
In addition, the three main important polymorphisms within relevant genomic regions were tested according to genotypes using contrasts by frequentist statistic. This test was carried out within the IMF line in order to evaluate the statistical differences amongst genotypes of SNPs. To do that, a general linear model was implemented using R software (R Core Team 2013).
Linkage disequilibrium and functional gene analysis
To evaluate the number of independent sites across the rabbit genome, a computation of LD for blocks was performed. The plink software was utilised to identify LD blocks (Purcell et al. 2007). The number of independent sites was calculated every 0.5, 1, 5, 10 and 20 Mb (genomic physical distance) across the whole rabbit genome. The LD‐adjusted Bonferroni threshold used in this study was calculated using the number of independent sites for 10 Mb as the number of independent sites barely changed between 10 and 20 Mb. LD blocks were examined in the associated genomic regions through the Haploview software (Barrett et al. 2005). In order to visualise the genes into the relevant genomic regions (±500 kb of associated SNP), we initially used the programme UCSC Genome Browser (https://genome.ucsc.edu/cgi-bin/hgGateway). The gene annotations were determined using Ensembl Genes 96 Database in biomart (Aken et al. 2016). The functional enrichment and metabolic pathways analysis were finally performed using the Database for Annotation, Visualization and Integrated Discovery (david) version 6.8 (Jiao et al. 2012) and enrichr (Kuleshov et al. 2016). The computation for the functional analyses was carried out using the parameters recommended by the authors. In addition, the search for annotated functions for each gene was performed individually using the database of all annotated functions from Ensembl and david.
Results
Genomic data
A total of 93 540 autosomal SNPs with known chromosomal positions were retained after filtering for MAF and call rate (see details in Materials and Methods). The number of retained SNPs on each of the 21 rabbit autosomes is shown in Table 1. The average physical distance between these SNPs was 22.61 kb. The average SNP number within 1 Mb windows was 46. One extended genomic region on OCU14 (54–65 Mb) did not contain any SNPs.
Table 1.
OCU | Number of SNPs | Percentage of SNPs in OCU1 | Average distance (kb) | Chromosome size (Mb) |
---|---|---|---|---|
1 | 9288 | 63 | 20.98 | 194.85 |
2 | 7856 | 58 | 22.19 | 174.33 |
3 | 7006 | 59 | 22.22 | 155.69 |
4 | 3895 | 58 | 23.47 | 91.39 |
5 | 1721 | 67 | 21.84 | 37.99 |
6 | 1222 | 63 | 22.48 | 27.50 |
7 | 7626 | 57 | 22.78 | 176.68 |
8 | 5075 | 57 | 22.03 | 111.80 |
9 | 5136 | 57 | 22.58 | 116.25 |
10 | 2318 | 61 | 19.38 | 48.00 |
11 | 3827 | 56 | 22.81 | 87.55 |
12 | 7116 | 60 | 21.83 | 155.35 |
13 | 5945 | 56 | 24.11 | 143.36 |
14 | 5687 | 45 | 28.81 | 163.90 |
15 | 4657 | 55 | 22.71 | 109.05 |
16 | 3962 | 62 | 21.32 | 84.48 |
17 | 3836 | 59 | 21.94 | 85.01 |
18 | 3102 | 64 | 21.45 | 69.80 |
19 | 2574 | 64 | 21.00 | 57.28 |
20 | 1224 | 51 | 24.66 | 33.19 |
21 | 467 | 55 | 26.56 | 15.58 |
Total | 93 540 | 47 |
The proportion of SNPs after quality control divided by number total of SNPs into OCU (rabbit chromosome) from the rabbit SNP array.
GWAS for IMF
Figure 1 reports a multidimensional scaling plot obtained using the genotyped SNPs on the rabbits of the two divergent IMF lines. A strong structure separating the high‐ and low‐IMF lines is evident. Therefore, a line effect was included in the models. In addition, a polygenic effect was also included in the SMR to adjust this model owing to the plausible effects derived from family‐data structures, considering a genomic kinship matrix. After this correction, the calculated lambda parameter was 1.065, indicating that the correction of bias derived from the population structure was not enough. Hence, we also implemented the correction by the lambda parameter in the SMR analysis. Note that the first and second components of multidimensional scaling accounted for 29.26% and 3.26% of genomic variance, respectively (Fig. 1).
Two methods were used in this research: SMR with the data adjusted by genomic relatedness and a BMMR (Bayes B method). We employed the term of "relevant" in order to denote those SNPs and genomic windows that we considered as true positive associations. In this research, we understand the GWAS as an exploratory analysis, which works as a mechanism for deriving promising genomic regions associated with IMF, and retrieving annotated rabbit genes. Table 2 shows the SNPs and genomic windows associated with IMF according to the procedure for defining the relevant associations (see details in Materials and Methods). For both methods, the associated SNPs and genomic windows were located on OCU8 and OCU13. The two genomic windows on OCU13 (2 Mb), containing 10 relevant SNPs for both methods, accounted together for 1.30% of the total genomic variance. On OCU8, 10 relevant polymorphisms showed the lowest P‐values for the SMR method, and had high Bayes factors for the BMMR method (Fig. 2). The two genomic windows containing these relevant polymorphisms accounted for 7.34% of the genomic variance. In addition, a genomic window on OCU1 was found to be associated with IMF by BMMR, explaining 2.03% of the genomic variance. The associated SNPs in this latter genomic window presented values close to the Bayes factor threshold, but these SNPs were distant from the P‐value (suggestive) threshold for SMR method.
Table 2.
SNP name | OCU | Position (bp) | P‐Value | Bayes factor | Window | MAF | |
---|---|---|---|---|---|---|---|
Name | Percentage of variance | ||||||
Affx‐151793092 | 1 | 121151928 | 1.10 × 10−3 | 15.95 | 118 | 2.03 | 0.24 |
Affx‐151803947 | 1 | 121280205 | 1.10 × 10−3 | 19.59 | 0.24 | ||
Affx‐151888965 | 1 | 121308004 | 1.10 × 10−3 | 16.03 | 0.25 | ||
Affx‐151956200 | 8 | 14893810 | 3.51 × 10−4 | 19.51 | 831 | 1.21 | 0.31 |
Affx‐151962168 | 8 | 14913105 | 3.51 × 10−4 | 24.86 | 0.32 | ||
Affx‐151945237 | 8 | 14939285 | 3.51 × 10−4 | 28.58 | 0.31 | ||
Affx‐151973204 | 8 | 14972879 | 1.83 × 10−4 | 18.38 | 0.31 | ||
Affx‐151800097 | 8 | 25087426 | 2.13 × 10−6 | 21.78 | 841 | 6.20 | 0.16 |
Affx‐151900210 | 8 | 25227502 | 3.33 × 10−6 | 44.73 | 0.16 | ||
Affx‐151917268 | 8 | 25262821 | 2.13 × 10−6 | 20.64 | 0.16 | ||
Affx‐151813008 | 8 | 25268392 | 2.13 × 10−6 | 22.57 | 0.16 | ||
Affx‐151795704 | 8 | 25467177 | 3.12 × 10−6 | 20.99 | 0.16 | ||
Affx‐151972842 | 8 | 25643667 | 2.06 × 10−6 | 24.15 | 0.16 | ||
Affx‐151964185 | 8 | 25732369 | 2.06 × 10−6 | 21.78 | 0.16 | ||
Affx‐152000638 | 8 | 25751303 | 2.06 × 10−6 | 21.17 | 0.16 | ||
Affx‐151808634 | 8 | 25863739 | 2.06 × 10−6 | 23.27 | 0.16 | ||
Affx‐151853378 | 8 | 25874631 | 2.12 × 10−6 | 21.25 | 0.16 | ||
Affx‐151824236 | 8 | 26115758 | 2.66 × 10−3 | 21.87 | 842 | 1.14 | 0.16 |
Affx‐151867012 | 13 | 84307591 | 7.14 × 10−5 | 11.73 | 1380 | 0.79 | 0.09 |
Affx‐151824373 | 13 | 84431723 | 7.14 × 10−5 | 10.62 | 0.09 | ||
Affx‐151874466 | 13 | 84447172 | 8.45 × 10−5 | 11.90 | 0.09 | ||
Affx‐151883028 | 13 | 84453332 | 7.14 × 10−5 | 11.73 | 0.09 | ||
Affx‐151801561 | 13 | 84537466 | 7.14 × 10−5 | 25.39 | 0.09 | ||
Affx‐151841215 | 13 | 84723427 | 2.20 × 10−5 | 25.39 | 0.09 | ||
Affx‐151846540 | 13 | 84738337 | 2.20 × 10−5 | 26.98 | 0.09 | ||
Affx‐151790364 | 13 | 84751504 | 2.23 × 10−5 | 25.30 | 0.09 | ||
Affx‐151939801 | 13 | 85316544 | 3.40 × 10−4 | 43.81 | 1381 | 0.51 | 0.08 |
Affx‐151937959 | 13 | 85333053 | 6.31 × 10−6 | 15.69 | 0.09 |
Percentage of variance: percentage of genomic variance explained by window. OCU, rabbit chromosome; bp, base pair.
Regarding the LD analysis, we found that in our data the rabbit genome could be divided into 2338 LD blocks and 6158 independent sites, with the longest LD blocks having a maximum length of 10 Mb. The associated SNPs on OCU13 and on OCU8 displayed a high LD within the chromosomal region (Fig. 3). The associated genomic region on OCU13 (window 1380 and 1381) holds two LD blocks. The second LD block (of 1506 kb) included almost all of the two associated windows (Fig. S1). The associated genomic region on OCU8 (window 841 and 842) presented just one block of 1945 kb, containing both windows (Fig. S2).
After the previous analysis (GWAS and LD), four relevant genomic regions were used to continue searching for putative candidate genes based on the functional annotation analysis (Table 3). In these regions, we also tested the IMF differences between genotypes within lines. Most of the SNPs tested presented statistical differences between one of the homozygous genotypes and the other genotypes within the high‐IMF line. In the low‐IMF line, except in region located 14.01–15.47 Mb in OCU8, these SNPs were not segregating (Fig. S3).
Table 3.
Cluster | OCU | Position (bp) | Number of genes | Annotated rabbit gene | |
---|---|---|---|---|---|
Start | End | ||||
1 | 1 | 120,651,928 | 121,986,803 | 9 | MAML2, MTMR2, CEP57, FAM76B, ENSOCUG00000025632 1, SESN3, ENDOD1, KDM4D, CWC15 |
2 | 8 | 14,014,437 | 15,472,879 | 9 | RASSF8, LMNTD1, RF00001, KRAS, ETFRF1, CASC1, LRMP, BCAT1, ENSOCUG00000021067 1 |
3 | 8 | 24,587,426 | 26,948,204 | 25 | PDE6H, ARHGDIB, ERP27, MGP, ART4, SMCO3, ENSOCUG00000017177 1, H2AFJ, HIST4H4, GUCY2C, PLBD1, ATF7IP, ENSOCUG00000017095 1, ENSOCUG00000021765 1, GRIN2B, RF00411, ENSOCUG00000021882 1, EMP1, GSG1, FAM234B, HEBP1, GPRC5D, GPRC5A, DDX47, APOLD1 |
4 | 13 | 83,807,591 | 85,998,108 | 3 | RF00026, ENSOCUG00000027270 1, RF00001 |
CLUSTER, denotes the genomic region; OCU, rabbit chromosome; bp, base pair.
Novel genes are named according to their Ensembl gene ID.
Functional annotation analysis and putative candidate genes
The final objective of our study was to generate a list of putative candidate genes, in order to guide further research for investigating the genetic determination of IMF content. Overall, 46 genes were annotated to the four relevant genomic regions (Table S1).
Only three genes (two non‐coding‐protein genes and one protein‐coding gene) mapped to the genomic region on OCU13 (Table 3). Among them stands out a novel annotated gene with Ensembl gene ID: ENSOCUG00000027270 (84.56 Mb), which is linked to metal ion binding in rabbits. The genes located on the genomic region on OCU8 were those showing a clearer relationship to lipid metabolism pathways. The ‘apolipoprotein L domain containing 1’ gene (APOLD1) shows functions related to lipid binding, transportation and localisation. The ‘phospholipase B domain containing 1’ (PLBD1) and ‘phosphodiesterase 6H’ (PDE6H) genes show functions linked to hydrolase activity (phospholipases) and lipid metabolic processes. In humans, several functional annotations, including the sphingolipid signalling pathway, have been found for the ‘K‐RAS proto‐oncogene, GTPase’ (KRAS) gene. Moreover, two members of the retinol‐induced G protein‐coupled protein receptors also stand out in OCU8: ‘G protein‐coupled receptor class C group 5 member D’ (GPRC5D) and ‘G protein‐coupled receptor class C group 5 member A’ (GPRC5A; Table 3). On OCU1, the ‘myotubularin‐related protein 2’ (MTMR2) gene displays biological functions linked to lipid metabolic processes. In addition to the biological and molecular functional annotations, a list of pathways that include these genes was generated from david, the KEGG and Wiki pathways databases (Table S2).
Discussion
Knowledge and understanding of control mechanisms of IMF content would be useful in the meat industry. Thus, a GWAS was performed in order to identify genomic regions associated with IMF content in rabbits owing to the increasing importance of meat quality in livestock for consumers (Hocquette et al. 2010; Pena et al. 2016; Strucken et al. 2017).
Following GWAS detection power studies (Spencer et al. 2009; Visscher et al. 2017), the distribution of SNPs (after quality control) across the rabbit genome in our data was suitable for GWAS analysis in livestock, given the LD and SNP density (Fan et al. 2010; Zhang et al. 2012). For instance, LD blocks having distance of 98 kb show r 2 = 0.5 as a measure of LD within rabbit breeds (Carneiro et al. 2011). This would indicate that the 93 540 SNP having an average distance of 22.61 kb between SNPs can be useful for discovering true associations amongst SNPs and the causal variants of IMF.
A challenge in GWAS analysis is the impact of confounding factors in the results. To avoid problems owing to population structure, we fit the genomic kinship matrix (Sul et al. 2018). The obtained λ value of 1.065 shows that this was almost enough to correct the population stratification effect. The purpose of implementing two methods was to corroborate the presence of associations between genomic windows or SNPs with IMF. The causal variants of moderate to high effect size can be detected by both methods in GWAS analyses when polymorphisms present high LD with these causal variants (López de Maturana et al. 2014). SNPs on OCU13 and OCU8 were found to be associated with IMF for both frequentist and Bayesian methods. However, the two associated windows on OCU13 (window 1380 and 1381) explained the low percentage of genomic variance (<1%). In addition, the LD block containing the most important SNPs on OCU13 covered a short physical distance and was uneven with regard to LD within this block (Fig. S1). This indicates that in this area of the genome a selective sweep process might not have been produced by divergent selection, since short‐term selection increases LD and the expected length of the LD block that contains an important causal variant (Vitti et al. 2013). In addition, the reference alleles of these associated SNPs presented low allelic frequencies (close to zero) for the low‐IMF line. The MAF value of the reference SNPs was also low (<0.09) in both low‐ and high‐IMF lines (Table 2). All SNPs were fixed or near fixation in the low‐IMF line, therefore the associations of these SNPs with IMF were uncovered given their segregation in the high‐IMF line. This could affect the association detection power even when the sample size is large (López de Maturana et al. 2014). For instance, if SNPs associated with the causal variants present a low MAF, the effects and association can be underestimated, generating false‐negative results.
In contrast, the associated region on OCU8 in 24.59–26.95 Mb explained a larger percentage of genomic variance between both associated windows (7.34%). Moreover, this region presented a strong and long LD block between windows 841 and 842, which could imply a selective sweep process owing to divergent selection (Fig. S2). The MAF values of the SNPs in this region were higher than on OCU13, reaching a maximum value of 0.16 (Table 2). Most SNPs in OCU8 were fixed or near fixation in the low‐IMF line. It seem that the causative variants and their surrounding SNPs would be at low frequency in the base population. This might explain the fixation of SNPs in the low‐IMF line and their segregation in the high‐IMF line of the ninth generation. Therefore, this genomic region showed more evidence than the region on OCU13 for considering it as an important association driving the control mechanism for IMF. Finally, another potentially interesting genomic region was identified on OCU1. This region explained 2.03% of the IMF genomic variance, although the SNPs show − log (P‐values) or Bayes factors below thresholds (Fig. 2). This suggests that the association of these SNPs could be better captured by a method that considers the percentage of variance explained by the windows instead of evaluating each SNP individually. In addition, these SNPs present MAF values around 0.24 (0.48 for the high‐IMF line and close to zero for the low‐IMF line), which might suggest that the differences might be a consequence of the divergent selection process.
This is the first GWAS study for IMF in rabbits. Therefore, comparisons within rabbits are limited to previous candidate gene studies. In this sense, as in Migdał et al. (2018), we did not find an association between the FABP4 (OCU3) candidate gene and IMF. Our results are not in agreement with the studies for FTO (OCU5) (Zhang et al. 2013), CAST (OCU11) (Wang et al. 2016b) and MYPN (OCU18) (Wang et al. 2017), which found associations in two, one and one SNP within genes, respectively (P‐values between 0.032 and 0.044). However, these associations should be taken with caution as the significance threshold was more liberal (P‐value < 0.05, without applying correction for multiple testing) than in our GWAS (P‐value < 1 × 10−4). In agreement with GWAS studies for IMF in swine, our results suggest that there is a large polygenic component influencing the trait (Pena et al. 2016; Ros‐Freixedes et al. 2016; Won et al. 2017). However, our results also showed important genomic regions associated with IMF. Especially in OCU8, a region of 2 Mb explains a notable percentage of the genomic variance (7.34%) in comparison with other GWAS studies for IMF (Cesar et al. 2014; Pena et al. 2016).
Several genes related to lipid metabolism (on OCU1, OCU8 and OCU13) were found in the associated regions. In OCU13, orthologues of a novel gene (Ensembl gene ID: ENSOCUG00000027270) have been reported in other species. In rabbits, there are no functional annotations related to lipid metabolism or IMF linked to this gene. However, in humans and mice this gene is known as EWS or EWSR1, and regulates the genetic expression of the transcription factor ‘Y‐Box Binding Protein 1’ gene (YBX1). This transcription factor activates the expression of the gene BMP7 (‘Bone Morphogenetic protein 7’), which in turn promotes the development of brown adipocytes (Wang & Seale 2016).
The genomic regions on OCU8 contained the genes with the most important biological functions. Hence, the genes on this region can be considered as candidates for further research, given that this window explains a large percentage of the IMF genomic variance (7.34%). In particular, APOLD1, PLBD1, PDE6H and GPRC5A were involved in functions of lipid transport, localisation and binding or in the control of adipose cell function. Two of these genes (PLBD1 and PDE6H) participated in the catabolism of phospholipids, which are the major components of cell membranes and have important implications in adipocyte hypertrophy (Chaves et al. 2011; Aloulou et al. 2012). As a result, PLBD1 has been related to lipid catabolic processes, skeletal muscle weight and body mass index in mice (Lionikas et al. 2012; Nyima et al. 2016) and humans (Wahl et al. 2017). In addition, KRAS (OCU8) was associated with the control of fat deposition in chickens (Claire D’Andre et al. 2013) and was involved in the sphingolipid signalling pathway. In humans, this gene was related to abnormal lipid metabolism in therapy for pancreatic cancer (Swierczynski et al. 2014). Another promising gene is GPRC5A, also known as RAI3, which is a key factor in repressing the differentiation of adipocytes in humans (Jin et al. 2017). This gene encodes for a member of the G‐coupled proteins, a large family including over 800 receptors, amongst them the olfactory receptors. GRPC5A belongs to a small subfamily of four members that are activated by retinol, the bioactive version of vitamin A. Although the role of GPRC5A is not well characterised at present, initial investigation reports a link with lung cancer, and also as a negative regulator or with adipogenesis (Song et al. 2019). Given the dual role of retinol during the adipogenesis (a positive regulator of pre‐adipocyte hyperplasia but a negative regulator of final maturation; see Wang et al. 2016a), GRPC5A rises as an interesting gene to mediate the inhibitory effect of retinoids in adipogenesis (Amisten et al. 2017).
In addition, MTMR2 (OCU1) was linked to the metabolic process of lipids. This gene has been proposed as a functional candidate gene for IMF in GWAS and signatures of selection studies in a Duroc pig population selected for IMF (Kim et al. 2015).
Conclusions and implications
This is the first GWAS study for IMF in rabbits and hence provides a benchmark for continuing research in the field. Our findings support the hypothesis that four genomic regions (on OCU1, OCU8 and OCU13) influence IMF content. The genomic variance explained by these associated regions is important although no major causal variants seem to segregate in the analysed rabbit populations. Therefore, according to what we observed in these divergently selected lines, it seems that IMF content is mainly driven by a polygenetic effect. In addition, we identified some candidate genes on the associated genomic regions of OCU13 (EWSR1), OCU8 (APOLD1, PLBD1, PDE6H, GPRC5A and KRAS) and OCU1 (MTMR2) related to IMF. Nevertheless, further research would be necessary in order to corroborate these results; for instance, a genotype refinement or sequencing of promoter and exonic regions of the candidate genes and its validation in independent populations of rabbits. Our results could be important for further studies to discover polymorphisms that can assist in IMF genetic improvement.
Conflict of interests
The authors declare that there is no conflict of interests.
Funding
The work was funded by project AGL2014‐55921‐C2‐1‐P from National Programme for Fostering Excellence in Scientific and Technical Research – Project I+D. BSS was supported by a FPI grant from the Ministry of Economy and Competitiveness of Spain+ (BES‐2015‐074194). NIB was supported with a “Ramon y Cajal” grant provided by Ministerio de Ciencia e Innovación (RYC‐2016‐19764). CSH and PN were supported by the Medical Research Council (United kingdom, grants MC_PC_U127592696 and MC_PC_U127561128). CSH was supported by Biotechnology and Biological Sciences Research Council (United Kingdom, Grant/Award Number: BBS/E/D/30002276).
Author’s contributions
BSS carried out the statistical analyses and drafted the manuscript. AB, PH, LF and MAS conceived of the study and secured substantial funding. PH, BSS and RP performed the phenotypic data recording and collected DNA samples. PN, CSH and NIE supervised analyses and helped draft the manuscript. All authors read and approved the final manuscript.
Supporting information
Acknowledgements
The authors thank Federico Pardo, PhD. Marina Morini, Antonnella Della Badia, Ilaria Giora, and Rosalia Rodríguez Esteban for technical assistance; and PhD. Yanni Zeng and PhD. Ricardo Pong‐Wong for their collaboration with bioinformatics tools.
Data Availability Statement
The datasets used and analysed in the current study are available from the Figshare Repository (https://doi.org/10.6084/m9.figshare.9934058.v1).
References
- Aken B.L., Ayling S., Barrell D. et al (2016) The Ensembl gene annotation system. Database 2016, baw093. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Aloulou A., Ali Y.B., Bezzine S., Gargouri Y. & Gelb M.H.(2012) Phospholipases: an overview In: Lipases and Phospholipases: Methods and Protocols(Ed. by Sandoval G.), pp. 63–85. Humana Press, Totowa, NJ. [DOI] [PubMed] [Google Scholar]
- Amisten S., Mohammad Al‐Amily et al (2017) Anti‐diabetic action of all‐trans retinoic acid and the orphan G protein coupled receptor GPRC5C in pancreatic beta‐cells. Endocrine Journal 64, 325–38. [DOI] [PubMed] [Google Scholar]
- Astle W. & Balding D.J. (2009) Population structure and cryptic relatedness in genetic association studies. Statistical Science 24, 451–71. [Google Scholar]
- Aulchenko Y.S., Ripke S., Isaacs A. & van Duijn C.M. (2007) GenABEL: an R library for genome‐wide association analysis. Bioinformatics 23, 1294–6. [DOI] [PubMed] [Google Scholar]
- Barrett J.C., Fry B., Maller J. & Daly M.J. (2005) Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 21, 263–5. [DOI] [PubMed] [Google Scholar]
- Beissinger T.M., Rosa G.J.M., Kaeppler S.M., Gianola D. & de Leon N. (2015) Defining window‐boundaries for genomic analyses using smoothing spline techniques. Genetics Selection Evolution 47, 30. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Blasco A. & Pena R.N. (2018) Current status of genomic maps: genomic selection/GBV in livestock In: Animal Biotechnology 2 (Ed. by H. Niemann & C. Wrenzycki), pp. 61–80. Springer International Publishing, Cham. [Google Scholar]
- Borg I. & Groenen P.J.F. (2005) Modern Multidimensional Scaling: Theory and Applications, 2nd edn Soviet journal of communications technology & electronics Springer International Publishing, New York, NY. [Google Scholar]
- Browning B.L. & Browning S.R. (2016) Genotype imputation with millions of reference samples. The American Journal of Human Genetics 98, 116–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carneiro M., Afonso S., Geraldes A., Garreau H., Bolet G., Boucher S., Tircazes A., Queney G., Nachman M.W. & Ferrand N. (2011) The genetic structure of domestic rabbits. Molecular Biology and Evolution 28, 1801–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Carneiro M., Rubin et al. (2014) Rabbit genome analysis reveals a polygenic basis for phenotypic change during domestication. Science 345, 1074–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Cesar A.S., Regitano L.C.A., Mourão G.B. et al. (2014) Genome‐wide association study for intramuscular fat deposition and composition in Nellore cattle. BMC Genetics 15, 39. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Chaves V.E., Frasson D. & Kawashita N.H. (2011) Several agents and pathways regulate lipolysis in adipocytes. Biochimie 93, 1631–40. [DOI] [PubMed] [Google Scholar]
- Chen W.‐M. & Abecasis G.R. (2007) Family‐based association tests for genomewide association scans. The American Journal of Human Genetics 81, 913–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Claire D’Andre H., Paul W., Shen X., Jia X., Zhang R., Sun L. & Zhang X. (2013) Identification and characterization of genes that control fat deposition in chickens. Journal of Animal Science and Biotechnology 4, 1–16. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Do D.N., Strathe A.B., Ostersen T., Pant S.D. & Kadarmideen H.N. (2014) Genome‐wide association and pathway analysis of feed efficiency in pigs reveal candidate genes and pathways for residual feed intake. Frontiers in Genetics 5, 1–10. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Do D.N., Schenkel F.S., Miglior F., Zhao X. & Ibeagha‐Awemu E.M. (2018) Genome wide association study identifies novel potential candidate genes for bovine milk cholesterol content. Scientific Reports 8, 13239. [DOI] [PMC free article] [PubMed] [Google Scholar]
- European Economic Community (1998) Council Directive 98/58/EC of 20 July 1998 concerning the protection of animals kept for farming purposes. Official Journal L 8, 221. [Google Scholar]
- Fan B., Du Z.‐Q., Gorbach D.M. & Rothschild M.F. (2010) Development and application of high‐density SNP arrays in genomic studies of domestic animals. Asian‐Australasian Journal of Animal Sciences 23, 833–47. [Google Scholar]
- Gao Y., Zhang R., Hu X. & Li N. (2007) Application of genomic technologies to the improvement of meat quality of farm animals. Meat Science 77, 36–45. [DOI] [PubMed] [Google Scholar]
- Garrick D.J. (2011) The nature, scope and impact of genomic prediction in beef cattle in the United States. Genetics Selection Evolution 43, 1–11. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Garrick D.J. & Fernando R.L. (2013) Implementing a QTL detection study (GWAS) using genomic prediction methodology In: Genome‐Wide Association Studies and Genomic Prediction(Ed. by C. Gondro, van der Werf J. & B. Hayes), pp. 275–98. Humana Press, Totowa, NJ. [DOI] [PubMed] [Google Scholar]
- Gotoh T., Takahashi H., Nishimura T., Kuchida K. & Mannen H. (2014) Meat produced by Japanese Black cattle and Wagyu. Animal Frontiers 4, 46–54. [Google Scholar]
- Gotoh T., Nishimura T., Kuchida K. & Mannen H. (2018) The Japanese Wagyu beef industry: current situation and future prospects – a review. Asian‐Australasian Journal of Animal Sciences 31, 933–50. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Green M.R., Sambrook J. & Sambrook J. (2012) Molecular Cloning: A Laboratory Manual. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY. [Google Scholar]
- Hocquette J.F., Gondret F., Baéza E., Médale F., Jurie C. & Pethick D.W. (2010) Intramuscular fat content in meat‐producing animals: development, genetic and nutritional control, and identification of putative markers. Animal 4, 303–19. [DOI] [PubMed] [Google Scholar]
- Hopkins D.L., Fogarty N.M. & Mortimer S.I. (2011) Genetic related effects on sheep meat quality. Small Ruminant Research 101, 160–72. [Google Scholar]
- Jiao X., Sherman B.T., Huang D.W., Stephens R., Baseler M.W., Lane H.C. & Lempicki R.A. (2012) DAVID‐WS: a stateful web service to facilitate gene/protein list analysis. Bioinformatics 28, 1805–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Jin C., Wang W., Liu Y. & Zhou Y. (2017) RAI3 knockdown promotes adipogenic differentiation of human adipose‐derived stem cells by decreasing β‐catenin levels. Biochemical and Biophysical Research Communications 493, 618–24. [DOI] [PubMed] [Google Scholar]
- Kass R.E. & Raftery A.E. (1995) Bayes factors. Journal of the American Statistical Association 90, 773–5. [Google Scholar]
- Kim E.‐S., Ros‐Freixedes R., et al (2015) Identification of signatures of selection for intramuscular fat and backfat thickness in two Duroc populations. Journal of Animal Science 93, 3292–302. [DOI] [PubMed] [Google Scholar]
- Kuleshov M.V., Jones M.R., Rouillard A.D. et al. (2016) Enrichr: a comprehensive gene set enrichment analysis web server 2016 update. Nucleic Acids Research 44, W90–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Lander E. & Kruglyak L. (1995) Genetic dissection of complex traits: guidelines for interpreting and reporting linkage results. Nature Genetics 11, 241–7. [DOI] [PubMed] [Google Scholar]
- Lionikas A., Meharg C., Derry J.M.J., Ratkevicius A., Carroll A.M., Vandenbergh D.J. & Blizard D.A. (2012) Resolving candidate genes of mouse skeletal muscle QTL via RNA‐Seq and expression network analyses. BMC Genomics 13, 592. [DOI] [PMC free article] [PubMed] [Google Scholar]
- López de Maturana E., Ibáñez‐Escriche N., González‐Recio Ó., Marenne G., Mehrban H., Chanock S.J., Goddard M.E. & Malats N. (2014) Next generation modeling in GWAS: comparing different genetic architectures. Human Genetics 133, 1235–53. [DOI] [PubMed] [Google Scholar]
- Marras G., Rossoni A., et al (2017) zanardi: an open‐source pipeline for multiple‐species genomic analysis of SNP array data. Animal Genetics 48, 121. [DOI] [PubMed] [Google Scholar]
- Martínez‐Álvaro M., Hernández P. & Blasco A. (2016) Divergent selection on intramuscular fat in rabbits: responses to selection and genetic parameters. Journal of Animal Science 94, 4993–5003. [DOI] [PubMed] [Google Scholar]
- Mateescu R.G., Garrick D.J., Garmyn A.J., VanOverbeke D.L., Mafi G.G., & Reecy J.M. (2015) Genetic parameters for sensory traits in longissimus muscle and their associations with tenderness, marbling score, and intramuscular fat in Angus cattle. Journal of Animal Science 93, 21–7. [DOI] [PubMed] [Google Scholar]
- McLaren D.G. & Schultz C.M. (1992) Genetic Selection to Improve the Quality and Composition of Pigs. In 45th Reciprocal Meat Conferences Proceedings Colorado State University, pp. 115–21.
- Migdał Ł., Kozioł K., Pałka S., Migdał W., Otwinowska‐Mindur A., Kmiecik M., Migdał A., Maj D. & Bieniek J. (2018) Single nucleotide polymorphisms within rabbits (Oryctolagus cuniculus) fatty acids binding protein 4 ( FABP4) are associated with meat quality traits. Livestock Science 210, 21–4. [Google Scholar]
- Miller I., Rogel‐Gaillard C., Spina D., Fontanesi L., de Almeida A.M. (2014) The rabbit as an experimental and production animal: from genomics to proteomics. Current Protein and Peptide Science 15, 134–45. [DOI] [PubMed] [Google Scholar]
- Mortimer S.I., van der Werf J.H.J., Jacob R.H. et al. (2014) Genetic parameters for meat quality traits of Australian lamb meat. Meat Science 96, 1016–24. [DOI] [PubMed] [Google Scholar]
- Nyima T., Müller M., Hooiveld G.J.E.J., Morine M.J. & Scotti M. (2016) Nonlinear transcriptomic response to dietary fat intake in the small intestine of C57BL/6J mice. BMC Genomics 17, 1–13. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Ochsner K.P., MacNeil M.D., Lewis R.M. & Spangler M.L. (2017) Economic selection index development for Beefmaster cattle I: terminal breeding objective. Journal of Animal Science 95, 1063–70. [DOI] [PubMed] [Google Scholar]
- Pannier L., Gardner G.E., O'Reilly R.A. & Pethick D.W. (2018) Factors affecting lamb eating quality and the potential for their integration into an MSA sheepmeat grading model. Meat Science 144, 43–52. [DOI] [PubMed] [Google Scholar]
- Peña F., Juárez M., Bonvillani A., García P., Polvillo O. & Domenech V. (2011) Muscle and genotype effects on fatty acid composition of goat kid intramuscular fat. Italian Journal of Animal Science 10, e40. [Google Scholar]
- Pena R.N., Ros‐Freixedes R., Tor M. & Estany J. (2016) Genetic marker discovery in complex traits: A field example on fat content and composition in pigs. International Journal of Molecular Sciences 17, 2100. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Purcell S., Neale B., Todd‐Brown K. (2007) PLINK: a tool set for whole‐genome association and population‐based linkage analyses. The American Journal of Human Genetics 81, 559–75. [DOI] [PMC free article] [PubMed] [Google Scholar]
- R Core Team (2013) R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna. [Google Scholar]
- Ros‐Freixedes R., Gol S., Pena R.N., Tor M., Ibáñez‐Escriche N., Dekkers J.C.M. & Estany J. (2016) Genome‐wide association study singles out SCD and LEPR as the two main loci influencing intramuscular fat content and fatty acid composition in Duroc pigs. PLoS ONE 11, e0152496. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Sahana G., Guldbrandtsen B. & Lund M.S. (2011) Genome‐wide association study for calving traits in Danish and Swedish Holstein cattle. Journal of Dairy Science 94, 479–86. [DOI] [PubMed] [Google Scholar]
- Sapp R.L., Bertrand J.K., Pringle T.D. & Wilson D.E. (2002) Effects of selection for ultrasound intramuscular fat percentage in Angus bulls on carcass traits of progeny. Journal of Animal Science 80, 2017–22. [DOI] [PubMed] [Google Scholar]
- Schmid M. & Bennewitz J. (2017) Invited review: genome‐wide association analysis for quantitative traits in livestock – a selective review of statistical models and experimental designs. Archives Animal Breeding 60, 335–46. [Google Scholar]
- Song H., Sun B., Liao Y. et al (2019) GPRC5A deficiency leads to dysregulated MDM2 via activated EGFR signaling for lung tumor development. International Journal of Cancer 144, 777–87. [DOI] [PubMed] [Google Scholar]
- Spencer C.C.A., Su Z., Donnelly P. & Marchini J. (2009) Designing genome‐wide association studies: sample size, power, imputation, and the choice of genotyping chip. PLoS Genetics 5, e1000477. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Stephens M. & Balding D.J. (2009) Bayesian statistical methods for genetic association studies. Nature Reviews Genetics 10, 681–90. [DOI] [PubMed] [Google Scholar]
- Strucken E.M., Al‐Mamun et al (2017) Finding the marble – the polygenic architecture of intramuscular fat. Journal of Animal Breeding and Genomics 1, 69–76. [Google Scholar]
- Sukegawa S., Miyake et al (2014) Multiple marker effects of single nucleotide polymorphisms in three genes, AKIRIN2, EDG1 and RPL27A, for marbling development in Japanese Black cattle. Animal Science Journal 85, 193–7. [DOI] [PubMed] [Google Scholar]
- Sul J.H., Martin L.S. & Eskin E. (2018) Population structure in genetic studies: Confounding factors and mixed models. PLoS Genetics 14, e1007309. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Swierczynski J., Hebanowska A. & Sledzinski T. (2014) Role of abnormal lipid metabolism in development, progression, diagnosis and therapy of pancreatic cancer. World Journal of Gastroenterology 20, 2279–303. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Toosi A., Fernando R.L. & Dekkers J.C.M. (2018) Genome‐wide mapping of quantitative trait loci in admixed populations using mixed linear model and Bayesian multiple regression analysis. Genetics Selection Evolution 50, 32. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Uemoto Y., Nakano H., Kikuchi T., Sato S., Ishida M., Shibata T., Kadowaki H., Kobayashi E. & Suzuki K. (2012) Fine mapping of porcine SSC14 QTL and SCD gene effects on fatty acid composition and melting point of fat in a Duroc purebred population. Animal Genetics 43, 225–8. [DOI] [PubMed] [Google Scholar]
- Visscher P.M., Wray N.R., Zhang Q., Sklar P., McCarthy M.I., Brown M.A. & Yang J. (2017) 10 Years of GWAS discovery: biology, function, and translation. The American Journal of Human Genetics 101, 5–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Vitti J.J., Grossman S.R. & Sabeti P.C. (2013) Detecting natural selection in genomic data. Annual Review of Genetics 47, 97–120. [DOI] [PubMed] [Google Scholar]
- Wahl S., Drong et al (2017) Epigenome‐wide association study of body mass index and the adverse outcomes of adiposity. Nature 541, 81–6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wang W. & Seale P. (2016) Control of brown and beige fat development. Nature Reviews Molecular Cell Biology 17, 691–702. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wang B., Yang Q., Harris C.L., Nelson M.L., Busboom J.R., Zhu M.‐J. & Du M. (2016a) Nutrigenomic regulation of adipose tissue development – role of retinoic acid: a review. Meat Science 120, 100–06. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wang J., Elzo et al (2016b) A single nucleotide polymorphism in CAST gene is associated with meat quality traits in rabbits. Animal Science Papers and Reports 34, 269–78. [Google Scholar]
- Wang X., Tucker N.R. & Rizki G. (2016c) Discovery and validation of sub‐threshold genome‐wide association study loci using epigenomic signatures. eLife 5, 1–24. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Wang J., Shi Y., Elzo M.A., Su Y., Jia X., Chen S. & Lai S. (2017) Myopalladin gene polymorphism is associated with rabbit meat quality traits. Italian Journal of Animal Science 16, 400–04. [Google Scholar]
- Won S., Jung J., Park E. & Kim H. (2017) Identification of genes related to intramuscular fat content of pig using genome‐wide association study. Asian‐Australasian Journal of Animal Sciences 31, 157–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
- Zhang H., Wang Z., Wang S. & Li H. (2012) Progress of genome wide association study in domestic animals. Journal of Animal Science and Biotechnology 3, 26. [DOI] [PMC free article] [PubMed] [Google Scholar] [Retracted]
- Zhang G.‐W., Gao L., Chen S.‐Y., Zhao X.‐B., Tian Y.‐F., Wang X., Deng X.‐S. & Lai S.‐J. (2013) Single nucleotide polymorphisms in the FTO gene and their association with growth and meat quality traits in rabbits. Gene 527, 553–57. [DOI] [PubMed] [Google Scholar]
- Zomeño C., Hernandez P. & Blasco A. (2013) Divergent selection for intramuscular fat content in rabbits. I. Direct response to selection. Journal of Animal Science 91, 4526–31. [DOI] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Data Availability Statement
The datasets used and analysed in the current study are available from the Figshare Repository (https://doi.org/10.6084/m9.figshare.9934058.v1).