All experimental processes, including breeding, feeding, and phenotype measurements, were conducted under identical conditions in the windowless house at the National Institute of Animal Science (NIAS, RDA) in Jeju Island. A three-generation resource population was produced and maintained as described by Cho et al [19]. Briefly, 19 purebred KNPs were crossed with 17 purebred Landrace. From these parental crosses, 91 F₁ progeny and 1,105 F₂ progeny (568 males and 537 females) were generated. None of the F₂ males were castrated. The Animal Care Committee at the Jeju National University approved all procedures of experiments. From the F₂ progeny, we measured BFT between the 4th and 5th ribs (BF_tho4_5), the 11th and 12th ribs (BF_tho11_12), and the last rib and first lumbar vertebrae (BF_tho_lum). The fat thickness was measured in millimeters.

Pools of genomic DNA samples from the parental lines (KNP [n = 5] and Landrace [n = 5]) of the F₂ intercross were re-sequenced to detect SNP markers across the whole genome, including the GRIP1 gene (ENSSSCG00000000478), using the HiSeq2000 massively parallel sequencing technology (Illumina Inc., San Diego, CA, USA). We identified a total of 8 exonic SNP markers in the porcine GRIP1 gene. The 8 SNP markers were confirmed through pyrosequencing, using the primer sequences presented in Table 1. Among these, a non-synonymous SNP marker (c.3316 C>G [R1106G]) and synonymous SNP marker (c.2442 T>C) were found to be polymorphic in the study population. Thus, these SNP markers were genotyped in all of the F₂ offspring (n = 1,105), using pyrosequencing (Biotage AB, Uppsala, Uppland, Sweden).

For the pyrosequencing analysis of missense and silent mutations in the GRIP1 gene, we made a primer set and minisequencing primer to be used in polymerase chain reaction (PCR) (Supplementary Table S1). Reactions were performed using 50 ng of genomic DNA, 0.4 μM of each primer, 1.5 mM of MgCl₂, reaction buffer (Genetbio, Daejeon, Korea), 0.2 mM of dNTPs, and Taq DNA polymerase (Genetbio, Daejeon, Korea) in a 25 μL reaction volume. The PCR conditions were as follows: 35 cycles each of 30 s at 94°C (denaturation), 45 s at 63°C or 66°C (annealing), and 45 s at 72°C (elongation), 2 minutes at 94°C (denaturation). The genotype of the PCR products was analyzed using a pyrosequencer.

The descriptive statistical analysis and ascertainment of normality of BFT traits were conducted using MINITAB Release 14 (MINITAB Inc., State College, PA, USA). We applied a mixed-effects linear model to perform SNP association analyses between the genetic markers and traits with simultaneous consideration of familial structure in an F2 intercross between Landrace and Korean native pigs:

Where, Y_ijklmn is the measured phenotype; μ is the population mean, S_i is the fixed effect of ith sex, B_j is the fixed effect of jth batch, P_k is the fixed effect of kth parity, G_l is the fixed effect of lth genotype of GRIP1, b₁ is a regression coefficient, CW_ijklm is the covariate for the carcass weight, A_ijklm is the random additive polygenic animal effect and ɛ_ijklm is the random residual effect. The mean and variance for random additive polygenic animal effects can be defined as: A~N(0, Aσ_a²), where A is based on the relationship matrix computed from the F₂ intercross in this study and σ_a² is the additive polygenic variance. The mean and variance for the residual random effect of individuals can be defined as: ɛ~ N(0, Iσ_e²), where I is the identity matrix and σ_e² is the residual variance [20].

The significance level of the fixed effects in MODEL 1 was computed by the Wald procedure implemented in ASReml-R. The p-values of less than 0.05 were considered as statistically significant and p-values of less than 0.001 were considered as statistically highly significant. To estimate genotypic values and their standard error of each SNP genotype, the predict command in ASReml-R was used [21]. In order to estimate the additive and dominance coefficients of each SNP found in GRIP1, G_l in the Model 1 was replaced by additive variable X_additive (coded as 1 for rare allele homozygote, −1 for common allele homozygote, and 0 for heterozygote), and dominance variable X_dominance (coded as 0 for two homozygotes, 1 for heterozygote). Then, trait values were regressed onto the additive and dominance coefficients at each SNP marker. The PedPhased program based on the pedigree structure was used to infer the haplotypes of GRIP1 in the F₂ population [22].

We identified 8 putative nucleotide substitutions in GRIP1 in the exon regions using HiSeq2000 massively parallel sequencing technology (Supplementary Table S1). Among these eight putative substitutions, we detected one missense substitution [c.3316 C>G (R1106G)] and seven synonymous exonic nucleotide substitutions. To conduct the association analysis, we selected c.3316 C>G (R1106G) and c.2442 T>C because the other six substitutions were invariant, as shown in (Supplementary Table S2).

The genotyping results of c.3316 C>G (R1106G) and c.2442 T>C obtained using pyrosequencing are shown in (Supplementary Figure S1). Data on the identified genotypes and allele frequencies of c.3316 C>G (R1106G) and c.2442 T>C in GRIP1 are presented in (Table 1). The Haploview results indicated that the alleles of the two SNPs were highly associated, so it was expected that the significant results would be very similar each other (Figure 1).

Liu and Cull-Candy [23] reported that the dynamic regulation of AMPA receptors by GRIP and protein interacting with C-kinase-1 (PICK) provide a mechanism for controlling the Ca²⁺ permeability of the synaptic receptors. Wu et al [18] reported that glutamatergic signaling is implicated in a neuronal circuit mediating regulation of feeding behavior and body weight in mice [18]. Hence, we investigated the association between SNP markers within the exons of GRIP1 and BFT-related traits in pigs.

The BFT traits were measured in more than 1,010 F₂ animals. A linear mixed-effects model accounting for familial relatedness was used to evaluate the associations between each SNP marker and the BFT traits in the F₂ intercross population. These two SNP markers were significantly associated with BFT between the 4th and 5th thoracic (p value = 0.00096 for c.3316 C>G [R1106G]; p value = 0.00022 for c.2442 T>C), BFT between the 11th and 12th thoracic (p value = 6.84×10⁻⁵ for c.3316 C>G [R1106G]; p value = 9.62×10⁻⁶ for c.2442 T>C), and BFT between the last thoracic and first lumbar vertebrae (p value = 7.92×10⁻¹² for c.3316 C>G [R1106G]; p value = 1.99×10⁻¹⁴ for c.2442 T>C) (Table 2).

The highly significant two SNPs showed additive effects on BF_tho4_5 and BF_tho11_12, as no significant dominant effect was observed. In case of BF_tho_lum, however, the c.3316 C>G and c.2442 T>C showed both additive and dominance effects. Regarding the SNP c.3316 C>G, C3316 allele was associated with higher values of BFT values. For the SNP c.2442 T>C, T2442 allele showed higher value than C2442 allele.

Following the genome-wide linkage analysis to identify QTLs for BFT traits, a positional candidate gene analysis was conducted using an intercross between KNP and Landrace pigs. We screened the GRIP1 full-length exons for polymorphisms and detected a novel association between the missense mutation in the GRIP1 gene (c.3316 C>G [R1106G]) and BFT traits (Table 2). The change from glycine (GGA) to arginine (CGA) in exon 25 likely affects the function of GRIP1. To assess the importance of this SNP, we conducted multiple protein sequence alignment analysis [24] using human (ENSP00000381098), horse (ENSECAP00000006920), dog (ENSCAFP00000000554), cattle (XP_002687693.1), and pig (ENSSSCP00000000508, reference sequence) proteins, as well as our KNP sequences (Figure 2). The result showed that the nucleotide involved in the c.3316 C>G (R1106G) missense mutation is highly conserved among these species, providing further evidence for the putative functional importance of this SNP. However, any effect of the missense mutation (c.3316 C>G [R1106G]) on BFT traits must be validated in other independent populations prior to commercial application.

Haplo-genotype effects of the two SNPs on BFT traits are shown in Table 3. A total of seven haplo-genotypes were identified, and these haplo-genotypes exhibit different effects. The association results showed that the haplo-genotypes significantly affected BFT traits (BF_tho4_5, BF_tho11_12, and BF_tho_lum).

In this study, porcine GRIP1 was evaluated as positional candidate gene for BFT traits. We screened the GRIP1 exon regions to identify polymorphic DNA markers, and two SNPs (c.2442 T>C, c.3316 C>G [R1106G]) were found to significantly affect BFT traits in an F₂ intercross population between Landrace and KNP. GRIP1 plays an important role in transmission in GABAergic and glutamatergic synapses [16,17]. After validating its functional implications in obesity related traits and this positive associations in other independent populations, the markers in GRIP1 can be used for the selection of individuals with desirable BFT traits and can contribute the understanding genetic structure of BFT traits.