This study was aimed at identifying genomic regions controlling feeding behavior in Danish Duroc boars and its potential implications for eating behavior in humans. Data regarding individual daily feed intake (DFI), total daily time spent in feeder (TPD), number of daily visits to feeder (NVD), average duration of each visit (TPV), mean feed intake per visit (FPV) and mean feed intake rate (FR) were available for 1130 boars. All boars were genotyped using the Illumina Porcine SNP60 BeadChip. The association analyses were performed using the GenABEL package in the R program. Sixteen SNPs were found to have moderate genome-wide significance (p<5E-05) and 76 SNPs had suggestive (p<5E-04) association with feeding behavior traits. MSI2 gene on chromosome (SSC) 14 was very strongly associated with NVD. Thirty-six SNPs were located in genome regions where QTLs have previously been reported for behavior and/or feed intake traits in pigs. The regions: 64–65 Mb on SSC 1, 124–130 Mb on SSC 8, 63–68 Mb on SSC 11, 32–39 Mb and 59–60 Mb on SSC 12 harbored several signifcant SNPs. Synapse genes (GABRR2, PPP1R9B, SYT1, GABRR1, CADPS2, DLGAP2 and GOPC), dephosphorylation genes (PPM1E, DAPP1, PTPN18, PTPRZ1, PTPN4, MTMR4 and RNGTT) and positive regulation of peptide secretion genes (GHRH, NNAT and TCF7L2) were highly significantly associated with feeding behavior traits. This is the first GWAS to identify genetic variants and biological mechanisms for eating behavior in pigs and these results are important for genetic improvement of pig feed efficiency. We have also conducted pig-human comparative gene mapping to reveal key genomic regions and/or genes on the human genome that may influence eating behavior in human beings and consequently affect the development of obesity and metabolic syndrome. This is the first translational genomics study of its kind to report potential candidate genes for eating behavior in humans.
Citation: Do DN, Strathe AB, Ostersen T, Jensen J, Mark T, Kadarmideen HN (2013) Genome-Wide Association Study Reveals Genetic Architecture of Eating Behavior in Pigs and Its Implications for Humans Obesity by Comparative Mapping. PLoS ONE 8(8): e71509. https://doi.org/10.1371/journal.pone.0071509
Editor: Peristera Paschou, Democritus University of Thrace, Greece
Received: April 5, 2013; Accepted: July 1, 2013; Published: August 19, 2013
Copyright: © 2013 Do et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.
Funding: Duy Ngoc Do is a PhD student funded by the Faculty of Health and Medical Sciences, University of Copenhagen. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
Competing interests: The authors have declared that no competing interests exist.
Feed represents a large proportion of the variable costs of breeding. Therefore, selection for reducing feed intake is a very important goal in breeding programs, at least in Danish pig breeds. Genetic improvement in feed efficiency was historically achieved as a correlated genetic change resulting from selection for growth rate and carcass lean content for animals tested in groups, where individual feed intake was too expensive to be measured on a large number of pigs. In recent years, the study of feed intake and behavior in pigs has been greatly facilitated by development of computerized systems that record the feed intake and related measures of individual animals within a group each time they enter the feeder. Several studies have shown low to moderate and positive genetic correlation between feeding behavior traits and daily feed intake. For instance, DFI had a positive genetic correlation with NVD (r = 0.27) . Labroue et al.  found FPV had positive genetic correlation to average daily gain, meaning that animals that eat more per visit tend to grow faster. These genetic associations underline the fact that genetic improvement of feed efficiency is also dependant upon genetic changes (improvement) in eating behavior of pigs. Furthermore, genomic control and gene pathways involved in eating or feeding behavior and its association to weight gain in pigs may translate to human eating behavior and obesity, because the pig is an excellent animal model genetically and physiologically very similar to humans . Feeding behavior has been reported to be highly related to social interaction of pigs and the number of pigs competing for access to the same feeder. Nielsen et al.  found that pigs with more frequent visits to the feeder were found to be positively correlated with less competition. Knowledge of molecular mechanisms of feeding behavior might help to improve our understanding of behavioral problems that are common in many fields of animal production (e.g. aggression, stress, pain). Quantitative trait loci (QTL) mapping is the first step to detect chromosomal regions affecting complex traits. Approximately 70 QTLs have previously been detected for feeding, drinking and socializing behaviors on 15 different pig chromosomes to date (http://www.animalgenome.org/cgi-bin/QTLdb/SS/index). However, QTLs are often mapped by linkage analysis to a large interval of 20 centimorgans (cM) or more that may contain several hundreds of genetic variants, not ideal for accurate mapping of potential causal variants –. Genome-wide association studies (GWAS) that survey most of the genome using dense genomic markers have been developed and applied widely in the analysis of complex traits in animals  and humans . GWAS take advantage of a large numbers of SNP markers in population-wide linkage disequilibrium with very small (QTL) regions potentially harboring candidate loci for the complex traits. Although some studies have identified QTLs for pig feeding behavior traits, this is is the first GWAS conducted to identify genetic variants and biological mechanisms for eating behavior in pigs.
The obesity epidemic has become one of the most important public health problems  and many of the common genetic variants for the risk of obesity, metabolic syndrome and related complications are associated with specific eating behaviors in human . A number of studies have shown that pigs are an excellent model for human obesity and metalolic syndrome . Eating behavior in humans (e.g. compulsive or comfort eating) can also be studied using the pig model, because eating behavior is closely related to development of obesity and metabolic syndrome. One of the objectives of this study was to conduct comparative pig-human genome mapping to identify potential candidate genes that may affect the way humans eat and develop obesity and related metabolic syndrome.
Materials and Methods
Recording of feeding behavior traits
A total of 7388 Duroc pigs had phenotypic records from the period of 2008–2011 and 1909 boars had 60 k SNP genotype records. The selection of boars to be genotyped and sent to the test station (i.e. phenotyping) was primarily based on their aggregate breeding value, but feeding behavior is not directly part of the breeding goal of Danish pig breeds and some genotyped boars have no recorded phenotype. For GWAS, anmials had to have both phenotypic and genotypic information; 1130 boars that had both genotypic and phenotypic records for feeding behavior traits were used in the study. Summary statistics of phenotypes are shown in Table 1. Data were recorded at the central Danish pig test station (Bøgildgård) during a period of four years (2008–2011) and the data were supplied by the Pig Research Centre of the Danish Agriculture and Food Council. The details of management and data records were described in Duy et al. . In summary, boars were put into pens of approx. 11 boars. Each pen had one ACEMO automatic dry feeding station and the boars were fed ad libitum from 30 kg to approximately 100 kg live weight with the same feed composition. The time, duration and feed consumption was recorded for each individual visit. Average daily feed intake (DFI) was derived from the total amount of recorded feed intake divided by the number of corresponding days at the feeder. The following feeding behavior traits were defined and calculated for each boar: DFI: total daily feed intake (kg/d), TPD: total time spent at feeder per day (minute), NVD: number of visits to the feeder per day, TPV: average duration of each visit ( = TPD/NVD), FPV: mean feed intake per visit (kg) and FR: mean feed intake rate (g/minute) ( = DFI/TPD) .
Generating dependent variable for GWAS
The estimated breeding values (EBVs) for DFI and feeding behaviors were calculated by single-trait animal model with fixed effect of herd-year-season, random effect of pen and a random additive genetic effect, as in Duy et al. . The phenotype used for association analysis was deregressed estimated breeding values (EBVs). The details of the estimation of deregressed EBVs are given by Ostersen et al.  following the deregression procedure of Garrick et al. . Briefly, the deregression adjusts for ancestral information, such that the deregressed EBV only contains their own and the descendant's information on each animal to avoid regressing information in both the generation of the dependent variable and the subsequent GWAS.
Genotyping and data validation
The details of the genotyping method have been described previously , . In summary, genomic DNA was isolated from all specimens by treatment with proteinase K followed by sodium chloride precipitation and SNPs were genotyped on the PorcineSNP60 Illumina iSelect BeadChip. The inclusion criteria for genomic data was a call rate per animal of 0.95. The inclusion criteria for SNP markers were a call rate of 0.95, Hardy Weinberg equilibrium test with p<0.0001 and minor allele frequency >0.05.
Statistical models for GWAS
The relationship matrix used by the “polygenic” linear mixed effects model was generated by the ibs() function of GenABEL which uses identity by state (IBS) genotype sharing to determine the realised pairwise kinship coefficient. Then a genome-wide association analysis was performed using a score test, a family-based association test, implemented in the mmscore() function of R/GenABEL . The full model: y = Xb + Wp + Za + e (1) is implemented in two steps in GenABEL. In the equation (1), y is the vector deregressed EBVs for a given trait, X is an incidence matrix for fixed non-genetic effects b (herd–week section and pen), W is a vector with genotypic indicators (−1, 0, or 1) associating records to the marker effect, p is a scalar of the associated additive effect of the SNP, Z is an incidence matrix relating phenotypes to the corresponding random polygenic effect, a is a vector of the random polygenic effect with the normal distribution a ∼ N(0, ), where A is the additive relationship matrix and is the polygenic variance, and e is a vector of random environmental deviates with the normal distribution , where is the error variance and R is the diagonal matrix containing weights of the deregressed estimated breeding values. Instead of fitting this full mixed model everytime a single SNP is fitted, the reduced model without the term Wp (SNP effect) is fitted only once and all fixed, polygenic and residual components are estimated using the REML approach. In the second step, with the estimated heritability estimate and kinship coefficients for each pair of relatives, the correlation between phenotypic records of relatives are adjusted and approximate IID (identical and independantly distributed) phenotypes with normality are obtained.
This mmscore test for family-based association is then conducted on the adjusted phenotype from the second step which takes into account pedigree structure and allows unbiased estimations of SNP allelic effect when relatedness is present between examinees . Multidimensional scaling plot of kindship distance based on IBS was used to check outliers and possible population stratification. The influence of population stratification after genomic control was also assessed in a quantile-quantile (q-q) plot by examining the distribution of test statistics generated from association tests and the deviation from the null hypothesis of no SNP association with the trait was assessed . The inflation factors before and after genomic control were 1.88 and 1.01, 2.16 and 1.04, 1.69 and 1.05, 2.14 and 1, 1.86 and 1.01 and 1.87 and 1.03 for DFI, TPD, NVD, TPV, FPV and FR, respectively. The genome-wide significance association at 5% significance level after Bonferroni multiple testing correction was p = 1.56E-06. However, the Bonferroni correction may result in a too stringent or very conservative threshold  and hence result in many false negative results, as this method assumes markers are independent. This is not the case in reality due to linkage disequilibrium (LD) between markers. Therefore, to avoid many false negative results caused by Bonferroni correction, the loci with p<5E-05 were considered as moderately genome-wide significant and loci with p<5E-04 were considered to be suggestively genome-wide significant. Both types of signifcant SNPs were included in downstream bioinformatics analysis. Linkage disequilibrium (LD) between SNPs in the chromosomal regions where multiple candidate SNPs were located was quantified as D' on all the animals of the GWAS using Haploview V4.2  and the LD block was defined by the criteria in . Frequency of defined haplotypes and their contribution to phenotypic variances of related traits was calculated using the PLINK software .
SNP positions were updated according to the newest release from Ensembl (Sscrofa10.2 genome version). Comparative mapping was performed by annotating significant SNP position to previously mapped QTL in pigs using the pig QTL database: http://animalgenome.org/cgi-bin/QTLdb/index  (assessed on 3rd, Feb, 2013). We also attempted to perform comparative mapping of chromosomal regions containing high numbers of tag (significant and suggestive) SNPs with human genomic map using RH map and comparative maps provided by Mayer et al.,  in the QTL database . Identification of the closest genes to tag SNPs was obtained using Ensembl annotation of Sscrofa10.2 genome version (http:// ensembl.org/Sus_scrofa/Info/Index). The positional candidate genes within 1 Mb bin size on either side of top SNPs peak were scanned using the function GetNeighGenes() in the NCBI2R package at http://cran.r-project.org/web/packages/NCBI2R/index.html using the R program . Investigation of functional categories in nearby genes was performed using the Database for Annotation, Visualization and Integrated Discovery (DAVID) at http://david.abcc.ncifcrf.gov/ . Human genes were used as background in annotation analysis, because many nearby genes have not been characterized in pigs and because translational gene aspects are of high interest.
Quality control, populations stratification assessments and phenotypic variation explained by markers
Following quality control of SNP data, 23795 markers were excluded as having a low (<5%) minor allele frequency, 1836 markers were excluded because of low (<95%) call rate and 3463 markers were excluded because they were not in HWE (p<0.001). A final set of 33945 SNPs and 1130 pigs was retained for GWAS. The number of markers on each chromosome and average distances between two markers after quality control are given in Table S1. Multidimensional scaling plot of IBS distances showed no outliers in populations (Figure S1). Total variance of all SNP markers explained 33, 42, 25, 38, 36 and 37% of the phenotypic variance (of the dependent variable, dEBVs) for DFI, TPD, NVD, TPV, FPV and FR, respectively.
Genome-wide association analysis and functional categories of nearby genes
Among 92 significant SNPs, 16 were found to have moderate genome-wide significance (Table 2) and 76 were found to have suggestive (Table S2) associations with feeding behavior traits. Number of significant and suggestive loci associated with DFI, TPD, NVD, TPV, FPV and FR were 1 and 10, 6 and 11, 6 and 16, 1 and 10, 1 and 19 and 1 and 10, respectively. While associated SNP with DFI, TPD and NVD were located on SSC 1, 11 and 12, the associated SNP with other traits were distributed around different chromosomes. Eleven SNPs were in unassembled scaffolds of the Sscrofa10.2 genome version. The locus DRGA00169471 on SSC 18 was found associated with both TPF and FPV. Nineteen of 92 loci were found in the intronic regions of known genes. The chromosomes and exact positions based on Sus scrofa Genebuild 10.2 (SSC10.2 build) as well as their nearest genes for SNPs were listed in Table 2. Quantitile-quantitle plots of observed and expected p values and Manhattan plots of GWAS of all traits after genomic control are shown in Figure S2 and Figure 1, respectively. Three haplotype blocks were detected in genomic regions affecting DFI on SSC1 (Figure 2). The major haplotypes with occurrence frequency is shown in Table 3. Two haplotype blocks were detected in genomic regions influencing NVD on SSC 12 (Figure 3) and their frequency and contribution to variances of the trait are shown in Table S4.
The horizontal line indicates genome-wide significant threshold. On vertical, Manhattan plot for total daily feed intake (DFI), total time spent at feeder per day (TPD), number of visits to the feeder per day (NVD), time spent to eat per visit (TPV), mean feed intake per visit (FPV), and mean feed intake rate (FR), respectively. Chromosome 19 stands for X chromosome. Chromosome 0 stands for unmapped SNPs.
LD blocks are marked with triangles. Values in boxes are LD (r2) between SNP pairs and the boxes are colored according to the standard Haploview color scheme: LOD >2 and D' = 1, red; LOD >2 and D'<1, shades of pink/red; LOD<2 and D' = 1, blue; LOD<2 and D'<1, white (LOD is the log of the likelihood odds ratio, a measure of confidence in the value of D').
LD blocks are marked with triangles. Values in boxes are LD (r2) between SNP pairs and the boxes are colored according to the standard Haploview color scheme: LOD >2 and D' = 1, red; LOD >2 and D'<1, shades of pink/red; LOD<2 and D' = 1, blue; LOD<2 and D'<1, white (LOD is the log of the likelihood odds ratio, a measure of confidence in the value of D').
A total of 652 gene identities (Entrez ID) was located in 1Mb window size from SNP positions (Table S3). However, 283 genes were reported as repetitions, since they were located in overlapping regions between two or more windows. The final list of 369 genes with unique identity was used for functional annotation. The functional categories based on protein resource information (SP_PIR_KEYWORDS) and biological processes (GOTERM_BP_FAT) of nearby genes involved in feeding behavior are shown in Table 4.
Comparisons with previously mapped QTL in pigs
A total of 36 SNPs were located in the genomic region where QTLs have previously been mapped for behavior and/or feed intake traits in pigs (Table 5). Eight loci on SSC 1 and a locus on SSC 2 associated with DFI were located on previous QTL regions for feed intake/daily feed intake in other pig populations. Several significant SNPs for FR, FPV and TPD were found in QTL regions for drinking and socializing from previous studies. Moreover, we also detected five SNPs located in the genome regions where QTL/SNPs have been previously detected by GWAS for backfat traits in pigs.
Comparative mapping of significant QTL with human genome
We indentified the five most significant QTL (contained more than 5 significant SNPs) for eating behavior traits including regions of 64–65 Mb on SSC 1 influencing DFI, 124–130 Mb on SSC 8 influencing both FR and TPD, 63–68 Mb on SSC 11 influencing TPD, 32–39 Mb and 59–60 Mb influencing NVD and TPV on SSC 12, respectively. The QTL region for DFI on SSC 1 located on p2.1 cytogenetic band (Figure 4a) which is homologous with the 136–157 Mb region on the human chromosome (HSA) 6 (Figure 4b and c). We also found that pleotropic QTL for FR and TPD on SSC 8 (124–130 Mb) was homologous with 90–101 Mb region on HSA 4 (HSA 4q22–24) (Figure S3). The QTL for TPD on SSC 11 was homologous with the 84–99Mb region on HSA13 (HSA 13q31–32) (Figure S4). Two QTL regions for NVD and TPV on SSC 12 located in q1.1–1.2 and q1.5 cytogenetic band (Figure 5a) were homologous with 36–48 Mbp (17q21 cytogentic band) and 4–8 Mb (17p13 cytogentic band) on HSA 17 (Figure 5b), respectively. Thus, our pig-human comparative mapping approaches revealed key genomic regions and/or genes on the human genome that may influence eating behavior in human beings and consequently obesity.
(a) Cytogenetic band, approximate positions of QTL shown in both cM and Mb, (b) linkage map, radiation hybrid mapping and human map of selected regions based on QTL database (release19), (c) human cytogentic band and physical map. The red band indicates QTL presence.
Comparison with previously mapped QTL in pigs
Since no GWAS study for feeding behavior in pigs has been previously published, we have made an attempt to overlap our association signals with those of previously reported QTLs. However, direct comparison between data obtained in this study and those from previous QTL studies is hindered by the fact that locations given in centimorgan on different genome assemblies do not necessarily reflect the same physical location on the genome . Therefore, the physical locations on the QTL (in Mb) as given in the SSC10.2 build in the pig QTLdb were used to compare to results from previous studies.
On SSC 1, we found that eight SNPs associated with DFI are in previously mapped QTL which spanned 49–73 (cM) for feed intake in a Pietrain/Meishan F2 family . Moreover, we also found other SNPs associated with DFI very close to the QTL region mapped for DFI in full-sibs families based on cross-bred Pietrain, Large White, Landrace, and Leicoma . This may imply that the same gene affected the traits across different pig breeds. On SSC 6, a QTL for TPV in pigs were also found on regions for time spent per day in a Pietrain x Meishan cross . Other SNPs associated with feeding rate also found in QTL mapped for time spent feeding and socializing , drinking  and daily feed intake . For instance, SNPs associated with FPV and TPD on SSC 8 were also found in the regions affecting DFI in Duroc×Petrain populations . Because Lui et al.  did not find QTL for FPV and TPD, it is difficult to make any conclusions about pleiotropic effects of these QTL. Several SNPs associated with TPD on SSC 11 were also found in the QTL for time spent socializing in a Pietrain x Meishan cross . Because the QTLs for fat deposition traits can be found over all pig chromosomes , we only compared our GWAS results with previous studies for backfat and obesity-related traits. Two SNPs associated with NVD on SSC18 in our study were found very close to a SNP detected for backfat thickness in an Italian breed . Fontanesi et al,  found the neuronal genes play important roles in controlling fat deposition in this chromosome. These results suggested possible pleiotropic QTL/genes in the nervous system controlling both fat metabolism and feeding behavior. Some other QTLs and SNPs overlapping with previous studies might also be interesting for further investigation. Nevertheless, comparative mapping is useful for narrowing down QTL regions and targeting candidate genes for complex traits such as eating behavior.
Haplotype block and haplotype frequency
Understanding linkage disequilibrium profiles and haplotype diversity in genomic regions of interest helps to better understand the genetic basis of these traits. The average LD observed in a Danish Duroc pig population was quite high (r2 = 0.56 between adjacent markers) . High LD limits fine-mapping the QTL because of SNPs quite far from the actual QTL position, but it does not have much influence on an association test. In the candidate region (64–65 Mb) for DFI on SSC1, we found three haplotypes blocks with high LD between adjacent markers. An interesting haplotype block is 2222122 of seven markers including ASGA0003043, MARC0003007, ASGA0003045, ASGA0003049, ASGA0003051, ALGA0003690 and DRGA0000958 which contributed most (0.48 %) to phenotypic variance of DFI. Moreover, two SNPs in the haplotype were located in the intron region of two different genes (GABRR2 and SRSF12); hence, it could be interesting to further investigate the functional involvement of these two genes in relation to DFI. Adjacent to GABRR2 is the GABRR1 gene which encodes the GABA receptor γ1 subunit (Figure 2). In humans, GABRR1 and GABRR2 are highly linked and located in the GABA receptor cluster on SSC 6. Details of molecular functions and possible roles of GABRR1 and GABRR1 in relation to daily feed intake are discussed below. Furthermore, we also found that the haplotype 21222 for block 1 had the highest contribution to variances of NVD on SSC 12. All these SNPs were located in ankyrin-repeat and fibronectin type III domain containing the ANKFN1 gene (Figure 3). ANKFN1 was previously identified as a candidate gene in a genomic study of general vulnerability to substance use disorders in humans . No functional investigations of the genes in pigs has been reported so far.
Potential candidate genes
Potential candidate genes for average daily feed intake.
Daily feed intake is an important trait for animal production and of general biological interest. Therefore, many studies have been conducted to investigate the genetic background underlying this trait. Only locus ALGA0003690 (G/A) was found to be significantly associated with DFI in the current study and it is located in the intron region of the Gamma-aminobutyric acid receptor subunit rho-2 (GABRR2) gene. GABRR2 encodes for a receptor of Gamma-aminobutyric acid (GABA) which is the most important inhibitory neurotransmitter in the vertebrate central nervous system (CNS) and is involved in manifold physiological and pathological processes . Moreover, a suggestive SNP associated with DFI was identified close to the GABRR1 gene, which is in the same transcriptional orientation, suggesting a similar expression and regulatory pattern as GABRR2. The GABA and these receptors have a known function in controlling feed intake, as shown in different species such as rats , chickens , and ruminants . Expression of GABRR2 was significantly changed after fasting and refeeding in the hypothalamus in mice . Baldwin et al.  showed that GABA and the GABA agonist stimulate feeding in satiated pigs by an action on central GABA receptors. However, the mechanism of GABA and these receptors in controlling feed intake and feed behavior is not well understood. Some other interesting genes in adjacent regions such as SRSF12, ANKRD6, RRAGD, PM20D2, RNGGT, MDN1, and UBE2J1 might be interesting to investigate, since these functions are related to regulation of gene expression or signaling pathway (Table S3).
Potential candidate genes for time spent to eat per day.
The significant loci MARC0085057 was closest to ALX1 gene (Table 2), whose function has not been extensively studied even in humans. However, it is interesting to note that in a 1 Mb window around the SNP position we found the NTS gene which encodes a common precursor for two peptides, neuromedin N and neurotensin (Table S3). Neurotensin is a secreted tridecapeptide, which is widely distributed throughout the central nervous system and may function in controlling feeding behavior . Intranigral microinjection of neurotensin suppressed feeding in food-deprived rats . Nearby DDIT4L gene regulates the TOR signaling pathway and in turn mammalian target of rapamycin (mTOR) as a key fuel sensor in hypothalamic neurons . Nutritional regulation of the mTOR-signaling pathway is mediated by their corresponding plasma membrane transporters ; therefore, DDIT4L may involve feeding behavior via nutritional impacts.
Potential candidate genes for number of visit to feeder per day.
Two loci ASGA0054177 and M1GA0016584 had the strongest association with NVD (p = 9.65E-7 and 2.19E-6, respectively; Table 2). The musashi homolog 2 (MIS2) is the known gene located closest to them. This gene encodes an RNA-binding protein and play central roles in posttranscriptional gene regulation in mammals . The gene also plays a role in the proliferation and maintenance of stem cells in the central nervous system in mice . During neurogenesis, the MSI2 expression persisted in a subset of neuronal lineage cells, such as parvalbumin-containing GABA neurons in the neocortex . As mentioned earlier, the GABA receptors also play a role in controlling feed intake and feeding behavior. It could be interesting to investigate how MSI2 gene and GABA receptor genes are connected in controlling feeding behavior. Moreover, we found two mutations in the TCF7L2 gene suggestively associated with NVD (Table S2). TCF7L2 encodes for a transcriptional factor involved in Wnt signaling that can regulate the tumor necrosis factor-α induced antiadipogenesis, pancreatic β-cell survival and function  as well as primary immune response . TCF7L2 mutations were associated with backfat  and with meat color traits  and residual feed intake traits  in pigs.
Potential candidate gene for time spent to eat per visit.
The ElaC homolog 2 (ELAC2) was close to significant SNPs associated with TPV (Table 2). This gene encodes for a protein which has a C-terminal domain with tRNA, processing endoribonuclease activity which catalyzes the removal of the 3′ trailer from precursor tRNAs. Mutations in this gene result in an increased risk of prostate cancer in humans . No functional characterization of the gene in pigs are available so far. A mutation in intron regions of GATA binding protein 3 was suggested to be linked to TPV (Table S2). GATA3 is a transcription factor of the Gata Zn-finger family which performs important functions during organogenesis . In mice, the expression of gene was changed in obesity induced by different diet .
Potential candidate genes for feed intake per visit.
The ACO1 gene was close to significant SNPs associated with FPV (Table 2). The gene encodes for soluble aconitase, a bifunctional protein involved in the control of iron metabolism or as the cytoplasmic isoform of aconitase . However, the gene has not been extensively studied in pigs. Disks large-associated protein 2 is a protein encoded by the DLGAP2 gene (Table S2). The DLGAP2 protein is one of the membrane-associated guanylate kinases localized at postsynaptic density in neuronal cells  and may play a role in the molecular organization of synapses and in neuronal cell signaling. The DLGAP2 variants were found significantly associated with autism spectrum disorders . Ceroid-lipofuscinosis neuronal 8 (CLN8) plays a role in cell proliferation during neuronal differentiation . Both DLGAP2 and CLN8 were located on SSC15 (Table S2) and may be of interest for feeding behavior traits, because it functions in the neuronal center controlling feed intake.
Potential candidate genes for rate of feed intake.
The PPA2 gene may be an interesting candidate gene for rate of feed intake, since two variants of the gene were found suggestively associated with the trait. The protein encoded by this gene is localized to the mitochondrion and contains the signature sequence essential for the catalytic activity of PPase . PPA2 may have a function in feeding behavior via controlling the phosphate level of the cell. Neuromedin U Receptor 2 (NMU2) is the most interesting gene for FR (Table S2). Neuromedin U is a known neuropeptide with potent activity on smooth muscle which is widely distributed in the gut and central nervous system . The NMUR2 gene is expressed in the ventromedial hypothalamus in the rat brain and its level is significantly reduced following fasting . Neuromedin U receptor 2-deficient mice display differential responses in sensory perception, stress, and feeding .
Functional categories of potential candidate genes
The results of functional annotation of nearby genes showed many genes involved in synapses that are essential to neuronal functions. The GABRR2, PPP1R9B, SYT1, GABRR1, CADPS2, DLGAP2 and GOPC genes were involved in activities for synapses based on protein resource information (Table 4). Functions of DLGAP2, GABRR1 and GABRR2 in feeding behavior have been discussed above. In humans, SYT1 encodes for Synaptotagmin-1 protein SYT1 which is the master switch responsible for allowing the human brain to release neurotransmitters . Protein encoded by protein phosphatase 1, regulatory subunit 9B (PPP1R9B) plays an important role in linking the actin cytoskeleton to the plasma membrane at the synaptic junction . The CADPS2 gene encodes a member of the calcium-dependent activator of secretion (CAPS) protein family, which are calcium-binding proteins that regulate the exocytosis of synaptic and dense-core vesicles in neurons . Dephosphorylation is the essential process of removing phosphate groups from an organic compound as adenosine triphosphates (ATP) by hydrolysis. Feeding behavior has been linked to ATP concentration in the liver with satiety occurring as fuels are oxidized and ATP is produced, and hunger occurring as oxidation decreases and ATP is depleted . Seven nearby genes have been classified in dephosphorylation based on their functions and may play significant roles in this mechanism (Table 4). The PTPN4, PTPN18 and PTPRZ1 genes are members of the protein tyrosine phosphatase (PTP) family. A recent review described PTPs as central regulators of metabolism, specifically highlighting their interactions with the neuronal leptin and insulin signaling pathways . On the other hand, PPM1E was located in the nucleus of the cell and it encodes a member of the PPM family of serine/threonine-protein phosphatases. The encoded protein dephosphorylates and inactivates multiple substrates such as 5′-AMP-activated protein kinase (AMPK) which is well documented to play key roles in controlling energy balance . AMPK appears to play a role in hypothalamic glucose and nutrient sensing . Therefore, the function of the PPM1E gene on feeding behavior may be mediated by AMPK. Another significant biological process involves the nearby genes (GHRH, NNAT, and TCF7L2) having a positive regulation of peptide secretion (Table 4). Growth hormone-releasing hormone (GHRH) is well known to stimulate food intake  and will therefore not be discussed further. The TCF7L2 has been proven as candidate gene for residual feed intake, as discussed above.
Implications for humans by comparative QTL/genomic mapping
Our pig-human comparative mapping approaches revealed key genomic regions and/or genes on the human genome that may influence eating behavior in human beings and consequently lead to obesity and metabolic syndrome. For instance, the QTL for DFI on SSC 1 was homologous with HSA 6q23–24 region (Figure 4a and b) which has been found to significantly affect obesity-related traits in humans such as waist circumference, body mass index or fasting glucose and insulin levels in different studies (reviewed in ). The region also contains several genes associted with obesity or metabolic syndrome such as ENPP1 with obesity and risk of glucose intolerance and type 2 diabetes , SGK1 with insulin secretion in type 2 diabeties . Frequency of eating and meal time are important indicators for eating behavior in humans. QTL for NVD was homologous with HSA 17q21 regions which contained many obesity candidate genes including PPY, PON1 and 2, GAST, PNMT, STAT3 and HCRT (reviewed in ). Moreover, some of the genes have been found to play very important roles in controlling feed intake in both human and animal models. For instance, the HCRT gene encodes a hypothalamic neuropeptide precursor protein that gives rise to two mature neuropeptides, orexin A and orexin B, which stimulate feed intake in rats . Peptide YY (PYY) also plays a very important role in energy homeostasis by balancing food intake  by acting as an “ileal brake” leading to a sensation of fullness and satiety . Other homologous regions including HSA 4q22–24, HSA 13q31–32 and HSA 17p13 also contain a number of candidate genes for obesity/ metabolic syndorome and eating behavior in both human and animals. For instance, microsomal triglyceride transfer protein (MMTP) gene located in HSA q24 were found as a candidate gene for obesity  in humans. The inhibitation of this gene by JTT-30 was found to suppress also the food intake in rats . The function of the MTTP gene in feed intake may be due to its involvement in the gut leptin-melanocortin pathway . Although pigs and humans have similar genetic structure, comparative genomic mapping between these species has a limitation on accuracy of homolgous regions. This limitation can be overcome by fine mapping or meta-analysis of QTL in each species and by taking systems biology approaches that links genomic regions with phenotypes through transcriptomics to detect potential causal genes (– and ). Nevertheless, the results of comparative QTL mapping from this study are useful for understanding the genetic background of eating behavior in humans (more QTL for traits) as well as in pigs (more candidate genes with functional validations).
Feeding or eating behavior are important traits in pig production, as they are directly related to feed efficiency and hence cost of pig production, but their genetic mechanisms have not been extensively studied. This is the first GWAS study pinpointing a number of significant SNPs associated with feeding or eating behavior in pigs. This study presented a comprehensive approach by combining GWAS and post-GWAS bioinformatics as well as comparative mapping approaches to elucidate genomic regions and candidate genes associated with eating behavioral traits in pigs. Post-GWAS analyses highlighted potential candidate genes for feeding behavior. Several nearby genes have been mentioned directly or indirectly as being involved in the genetic control of eating or feeding behavior traits in either pigs or other species. Pigs are a well-known animal model for studying human obesity. We have conducted pig-human comparative gene mapping to reveal key genomic regions and/or genes on the human genome that may influence eating behavior in human beings and consequently affect the development of obesity and metabolic syndrome, both of which are key societal and public health problems. This is the first study to report results on genes that may affect human eating behavior via such translational genomics approaches.
Multidimensional scaling plot of identity by state distances. The principal component analysis fitted the genetic distances along the two components. The results showed that no population stratification in the data. Each point on the plot corresponds to a pig, and the 2D distances between points were fitted to be as close as possible to those presented in the original identity by state matrix. You can see that study subjects clearly cluster in a group.
A quantile-quantile plot of observed and expected p-values for feeding behavior traits. The inset shows a quantile-quantile (qq) plot with the observed plotted against the expected p-values for total daily feed intake (DFI), total time spent at feeder per day (TPD), number of visits to the feeder per day (NVD), time spent to eat per visit (TPV), mean feed intake per visit (FPV), and mean feed intake rate (FR) from top to bottom, respectively.
Comparative mapping between QTL on pig chromosome 8 and human chromosome 4. (a) Cytogenetic band, approximate positions of QTL for mean of feed intake rate FR) and total time spent at feeder per day (TPD) shown in both cM and Mb, (b) linkage map, radiation hybrid mapping and human map of selected regions based on QTL database (release19). The read band indicated QTL presence.
Comparative mapping between QTL on pig chromosome 11 and human chromosome 13. (a) Cytogenetic band, approximate positions of QTL for total time spent at feeder per day (TPD) shown in both cM and Mb, (b) linkage map, radiation hybrid mapping and human map of selected regions based on QTL database (release19). The red band indicates QTL presence.
Distribution of SNPs after quality control and average distances on each chromosome.
Suggestive SNPs associated to studied eating behavioral traits, their positions and nearest genes for feeding behavior traits.
List of nearby genes in 1 Mb region flanking the associated SNPs.
We would like to thank the Department of Breeding and Genetics of the Danish Pig Research Centre for providing all data for the research reported in this study and useful comments on this manuscript. We would also like to thank Gilda Kischinovsky for proofreading the manuscript. Duy Ngoc Do is a PhD student funded by the Faculty of Health and Medical Sciences, University of Copenhagen.
Conceived and designed the experiments: HNK. Performed the experiments: DND TO. Analyzed the data: DND TO ABS HNK. Contributed reagents/materials/analysis tools: DND TO HNK TM JJ. Wrote the paper: DND ABS TO JJ TM HNK. Bioinformatic and systems genetic analyses and biological interpretations: DND HNK.
- 1. Do DN, Strathe AB, Jensen J, Mark T, Kadarmideen HN (2013) Genetic parameters for different measures of feed efficiency and related traits in boars of three pig breeds. Journal of Animal Science Accepted.
- 2. Labroue F, Gueblez R, Sellier P (1997) Genetic parameters of feeding behaviour and performance traits in group-housed Large White and French Landrace growing pigs. Genetics Selection Evolution 29: 451–468.
- 3. Kogelman LJA, Kadarmideen HN, Mark T, Karlskov-Mortensen P, Bruun CS, et al.. (2013) An F2 pig resource population as a model for genetic studies of obesity and obesity-related diseases in humans: Design and genetic parameters. Frontiers in Genetics 4.
- 4. Nielsen BL, Lawrence AB, Whittemore CT (1995) Effect of Group-Size on Feeding-Behavior, Social-Behavior, and Performance of Growing Pigs Using Single-Space Feeders. Livestock Production Science 44: 73–85.
- 5. Kadarmideen H, Rohr P, Janss LG (2006) From genetical genomics to systems genetics: potential applications in quantitative genomics and animal breeding. Mammalian Genome 17: 548–564.
- 6. Kadarmideen HN, Reverter A (2007) Combined genetic, genomic and transcriptomic methods in the analysis of animal traits. CABI review: perspectives in agriculture, veterinary science, nutrition and natural resources 2: 16.
- 7. Pearson TA, Manolio TA (2008) How to interpret a genome-wide association study. Jama-Journal of the American Medical Association 299: 1335–1344.
- 8. Goddard ME, Hayes BJ (2009) Mapping genes for complex traits in domestic animals and their use in breeding programmes. Nature Reviews Genetics 10: 381–391.
- 9. McCarthy MI, Abecasis GR, Cardon LR, Goldstein DB, Little J, et al. (2008) Genome-wide association studies for complex traits: consensus, uncertainty and challenges. Nature Reviews Genetics 9: 356–369.
- 10. Chan RSM, Woo J (2010) Prevention of Overweight and Obesity: How Effective is the Current Public Health Approach. International Journal of Environmental Research and Public Health 7: 765–783.
- 11. Grimm ER, Steinle NI (2011) Genetics of eating behavior: established and emerging concepts. Nutrition Reviews 69: 52–60.
- 12. Ostersen T, Christensen OF, Henryon M, Nielsen B, Su GS, et al.. (2011) Deregressed EBV as the response variable yield more reliable genomic predictions than traditional EBV in pure-bred pigs. Genetics Selection Evolution 43.
- 13. Garrick DJ, Taylor JF, Fernando RL (2009) Deregressing estimated breeding values and weighting information for genomic regression analyses. Genetics Selection Evolution 41.
- 14. Gregersen VR, Conley LN, Sorensen KK, Guldbrandtsen B, Velander IH, et al.. (2012) Genome-wide association scan and phased haplotype construction for quantitative trait loci affecting boar taint in three pig breeds. Bmc Genomics 13.
- 15. Aulchenko YS, Ripke S, Isaacs A, Van Duijn CM (2007) GenABEL: an R library for genorne-wide association analysis. Bioinformatics 23: 1294–1296.
- 16. Chen WM, Abecasis GR (2007) Family-based association tests for genomewide association scans. American Journal of Human Genetics 81: 913–926.
- 17. Price AL, Zaitlen NA, Reich D, Patterson N (2010) New approaches to population stratification in genome-wide association studies. Nature Reviews Genetics 11: 459–463.
- 18. Sahana G, Kadlecová V, Hornshøj H, Nielsen B, Christensen OF (2013) A genome-wide association scan in pig identifies novel regions associated with feed efficiency trait. Journal of Animal Science.
- 19. Barrett JC, Fry B, Maller J, Daly MJ (2005) Haploview: analysis and visualization of LD and haplotype maps. Bioinformatics 21: 263–265.
- 20. Gabriel SB, Schaffner SF, Nguyen H, Moore JM, Roy J, et al. (2002) The structure of haplotype blocks in the human genome. Science 296: 2225–2229.
- 21. Purcell S, Neale B, Todd-Brown K, Thomas L, Ferreira MAR, et al. (2007) PLINK: A Tool Set for Whole-Genome Association and Population-Based Linkage Analyses. The American Journal of Human Genetics 81: 559–575.
- 22. Hu Z-L, Park CA, Wu X-L, Reecy JM (2013) Animal QTLdb: an improved database tool for livestock animal QTL/association data dissemination in the post-genome era. Nucleic Acids Research 41 (D1): D871–D879.
- 23. Meyers SN, Rogatcheva MB, Larkin DM, Yerle M, Milan D, et al.. (2005) Piggy-BACing the human genome – II. A high-resolution, physically anchored, comparative map of the porcine autosomes. Genomics 86: 739−+.
- 24. Team RD (2008) R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria.
- 25. Huang DW, Sherman BT, Lempicki RA (2009) Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources. Nature Protocols 4: 44–57.
- 26. Hoglund JK, Guldbrandtsen B, Lund MS, Sahana G (2012) Analyzes of genome-wide association follow-up study for calving traits in dairy cattle. Bmc Genetics 13.
- 27. Reiner G, Kohler F, Berge T, Fischer R, Hubner-Weitz K, et al. (2009) Mapping of quantitative trait loci affecting behaviour in swine. Animal Genetics 40: 366–376.
- 28. Duthie C, Simm G, Doeschl-Wilson A, Kalm E, Knap PW, et al. (2008) Quantitative trait loci for chemical body composition traits in pigs and their positional associations with body tissues, growth and feed intake. Animal Genetics 39: 130–140.
- 29. Liu G, Jennen DG, Tholen E, Juengst H, Kleinwachter T, et al. (2007) A genome scan reveals QTL for growth, fatness, leanness and meat quality in a Duroc-Pietrain resource population. Animal Genetics 38: 241–252.
- 30. Dufour CR, Levasseur M-P, Pham NHH, Eichner LJ, Wilson BJ, et al. (2011) Genomic convergence among ERRα, PROX1, and BMAL1 in the control of metabolic clock outputs. PLoS Genetics 7: e1002143.
- 31. Johnson C, Drgon T, McMahon FJ, Uhl GR (2009) Convergent Genome Wide Association Results for Bipolar Disorder and Substance Dependence. American Journal of Medical Genetics Part B-Neuropsychiatric Genetics 150B: 182–190.
- 32. Martinez-Delgado G, Estrada-Mondragon A, Miledi R, Martinez-Torres A (2010) An Update on GABArho Receptors. Curr Neuropharmacol 8: 422–433.
- 33. Weedon MN (2007) The importance of TCF7L2. Diabetic Medicine 24: 1062–1066.
- 34. Lyssenko V, Lupi R, Marchetti P, Del Guerra S, Orho-Melander M, et al. (2007) Mechanisms by which common variants in the TCF7L2 gene increase risk of type 2 diabetes. Journal of Clinical Investigation 117: 2155–2163.
- 35. Du ZQ, Fan B, Zhao X, Amoako R, Rothschild MF (2009) Association Analyses Between Type 2 Diabetes Genes and Obesity Traits in Pigs. Obesity 17: 323–329.
- 36. Poplawski MM, Mastaitis JW, Yang XJ, Mobbs CV (2010) Hypothalamic responses to fasting indicate metabolic reprogramming away from glycolysis toward lipid oxidation. Endocrinology 151: 5206–5217.
- 37. Patient RK, McGhee JD (2002) The GATA family (vertebrates and invertebrates). Current Opinion in Genetics & Development 12: 416–422.
- 38. Hawkins MF (1986) Central-Nervous-System Neurotensin and Feeding. Physiology & Behavior 36: 1–8.
- 39. Vaughn AW, Baumeister AA, Hawkins MF, Anticich TG (1990) Intranigral Microinjection of Neurotensin Suppresses Feeding in Food Deprived Rats. Neuropharmacology 29: 957–960.
- 40. Barendse W, Reverter A, Bunch RJ, Harrison BE, Barris W, et al. (2007) A validated whole-genome association study of efficient food conversion in cattle. Genetics 176: 1893–1905.
- 41. Sakakibara S, Nakamura Y, Satoh H, Okano H (2001) RNA-binding protein Musashi2: developmentally regulated expression in neural precursor cells and subpopulations of neurons in mammalian CNS. Journal of Neuroscience 21: 8091–8107.
- 42. Sakakibara S, Nakamura Y, Yoshida T, Shibata S, Koike M, et al. (2002) RNA-binding protein Musashi family: Roles for CNS stem cells and a subpopulation of ependymal cells revealed by targeted disruption and antisense ablation. Proceedings of the National Academy of Sciences of the United States of America 99: 15194–15199.
- 43. Kadarmideen HN, Watson-Haigh NS, Andronicos NM (2011) Systems biology of ovine intestinal parasite resistance: disease gene modules and biomarkers. Mol Biosyst 7: 235–246.
- 44. Gilbert H, Bidanel JP, Gruand J, Caritez JC, Billon Y, et al. (2007) Genetic parameters for residual feed intake in growing pigs, with emphasis on genetic relationships with carcass and meat quality traits. Journal of Animal Science 85: 3182–3188.
- 45. Fan B, Lkhagvadorj S, Cai W, Young J, Smith RM, et al. (2010) Identification of genetic markers associated with residual feed intake and meat quality traits in the pig. Meat Science 84: 645–650.
- 46. Noda D, Itoh S, Watanabe Y, Inamitsu M, Dennler S, et al. (2006) ELAC2, a putative prostate cancer susceptibility gene product, potentiates TGF-[beta]//Smad-induced growth arrest of prostate cells. Oncogene 25: 5591–5600.
- 47. Koza RA, Nikonova L, Hogan J, Rim JS, Mendoza T, et al. (2006) Changes in gene expression foreshadow diet-induced obesity in genetically identical mice. Plos Genetics 2: 769–780.
- 48. Eisenstein RS (2000) Iron regulatory proteins and the molecular control of mammalian iron metabolism. Annual Review of Nutrition 20: 627–662.
- 49. Ranta S, Zhang YH, Ross B, Takkunen E, Hirvasniemi A, et al. (2000) Positional cloning and characterisation of the human DLGAP2 gene and its exclusion in progressive epilepsy with mental retardation. European Journal of Human Genetics 8: 381–384.
- 50. Pinto D, Pagnamenta AT, Klei L, Anney R, Merico D, et al. (2010) Functional impact of global rare copy number variation in autism spectrum disorders. Nature 466: 368–372.
- 51. Cannelli N, Cassandrini D, Bertini E, Striano P, Fusco L, et al. (2006) Novel mutations in CLN8 in Italian variant late infantile neuronal ceroid lipofuscinosis: another genetic hit in the Mediterranean. Neurogenetics 7: 111–117.
- 52. Curbo S, Lagier-Tourenne C, Carrozzo R, Palenzuela L, Lucioli S, et al. (2006) Human mitochondrial pyrophosphatase: cDNA cloning and analysis of the gene in patients with mtDNA depletion syndromes. Genomics 87: 410–416.
- 53. Brighton PJ, Szekeres PG, Willars GB (2004) Neuromedin U and its receptors: Structure, function, and physiological roles. Pharmacological Reviews 56: 231–248.
- 54. Gartlon J, Szekeres P, Pullen M, Sarau HM, Aiyar N, et al. (2004) Localisation of NMU1R and NMU2R in human and rat central nervous system and effects of neuromedin-U following central administration in rats. Psychopharmacology 177: 1–14.
- 55. Zeng HK, Gragerov A, Hohmann JG, Pavlova MN, Schimpf BA, et al. (2006) Neuromedin U receptor 2-deficient mice display differential responses in sensory perception, stress, and feeding. Molecular and Cellular Biology 26: 9352–9363.
- 56. Perin MS, Johnston PA, Ozcelik T, Jahn R, Francke U, et al. (1991) Structural and Functional Conservation of Synaptotagmin (P65) in Drosophila and Humans. Journal of Biological Chemistry 266: 615–622.
- 57. Meng X, Kanwar N, Du Q, Goping IS, Bleackley RC, et al. (2009) PPP1R9B (Neurabin 2): Involvement and dynamics in the NK immunological synapse. European Journal of Immunology 39: 552–560.
- 58. Brunk I, Blex C, Speidel D, Brose N, Ahnert-Hilger G (2009) Ca2+−dependent Activator Proteins of Secretion Promote Vesicular Monoamine Uptake. Journal of Biological Chemistry 284: 1050–1056.
- 59. Allen MS (2000) Effects of diet on short-term regulation of feed intake by lactating dairy cattle. Journal of Dairy Science 83: 1598–1624.
- 60. Tsou R, Bence K (2013) Central regulation of metabolism by protein tyrosine phosphatases. Frontiers in Neuroscience 6.
- 61. Kola B (2008) Role of AMP-activated protein kinase in the control of appetite. Journal of Neuroendocrinology 20: 942–951.
- 62. Vaccarino FJ, Sovran P, Baird JP, Ralph MR (1995) Growth hormone-releasing hormone mediates feeding-specific feedback to the suprachiasmatic circadian clock. Peptides 16: 595–598.
- 63. Rankinen T, Zuberi A, Chagnon YC, Weisnagel SJ, Argyropoulos G, et al. (2006) The human obesity gene map: The 2005 update. Obesity 14: 529–644.
- 64. Meyre D, Bouatia-Naji N, Tounian A, Samson C, Lecoeur C, et al. (2005) Variants of ENPP1 are associated with childhood and adult obesity and increase the risk of glucose intolerance and type 2 diabetes. Nature Genetics 37: 863–867.
- 65. Friedrich B, Weyrich P, Stancakova A, Wang J, Kuusisto J, et al.. (2008) Variance of the SGK1 Gene Is Associated with Insulin Secretion in Different European Populations: Results from the TUEF, EUGENE2, and METSIM Studies. Plos One 3.
- 66. Choi DL, Davis JF, Fitzgerald ME, Benoit SC (2010) The Role of Orexin-a in Food Motivation, Reward-Based Feeding Behavior and Food-Induced Neuronal Activation in Rats. Neuroscience 167: 11–20.
- 67. Murphy KG, Bloom SR (2006) Gut hormones and the regulation of energy homeostasis. Nature 444: 854–859.
- 68. Druce MR, Small CJ, Bloom SR (2004) Minireview: Gut peptides regulating satiety. Endocrinology 145: 2660–2665.
- 69. Hata T, Mera Y, Ishii Y, Tadaki H, Tomimoto D, et al. (2011) JTT-130, a Novel Intestine-Specific Inhibitor of Microsomal Triglyceride Transfer Protein, Suppresses Food Intake and Gastric Emptying with the Elevation of Plasma Peptide YY and Glucagon-Like Peptide-1 in a Dietary Fat-Dependent Manner. Journal of Pharmacology and Experimental Therapeutics 336: 850–856.
- 70. Iqbal J, Li XS, Chang BHJ, Chan L, Schwartz GJ, et al. (2010) An intrinsic gut leptin-melanocortin pathway modulates intestinal microsomal triglyceride transfer protein and lipid absorption. Journal of Lipid Research 51: 1929–1942.
- 71. Fan B, Onteru SK, Du ZQ, Garrick DJ, Stalder KJ, et al. (2011) Genome-wide association study identifies Loci for body composition and structural soundness traits in pigs. PLoS ONE 6: e14726.