Revealing new candidate genes for reproductive traits in pigs: combining Bayesian GWAS and functional pathways Lucas L. Verardo, Fabyano F. Silva, Marcos S. Lopes, Ole Madsen, John W. M. Bastiaansen, Egbert F. Knol, Mathew Kelly, Luis Varona, Paulo S. Lopes, Simone E. F. Guimarães To cite this version: Lucas L. Verardo, Fabyano F. Silva, Marcos S. Lopes, Ole Madsen, John W. M. Bastiaansen, et al.. Revealing new candidate genes for reproductive traits in pigs: combining Bayesian GWAS and functional pathways. Genetics Selection Evolution, 2016, 48 (1), pp.9. 10.1186/s12711-016-0189-x. hal-01341343 HAL Id: hal-01341343 https://hal.archives-ouvertes.fr/hal-01341343 Submitted on 4 Jul 2016 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Verardo et al. Genet Sel Evol (2016) 48:9 Genetics DOI 10.1186/s12711-016-0189-x Selection Evolution RESEARCH ARTICLE Open Access Revealing new candidate genes for reproductive traits in pigs: combining Bayesian GWAS and functional pathways Lucas L. Verardo1,2*, Fabyano F. Silva1, Marcos S. Lopes2,3, Ole Madsen2, John W. M. Bastiaansen2, Egbert F. Knol3, Mathew Kelly4, Luis Varona5, Paulo S. Lopes1 and Simone E. F. Guimarães1 Abstract Background: Reproductive traits such as number of stillborn piglets (SB) and number of teats (NT) have been evalu- ated in many genome-wide association studies (GWAS). Most of these GWAS were performed under the assumption that these traits were normally distributed. However, both SB and NT are discrete (e.g. count) variables. Therefore, it is necessary to test for better fit of other appropriate statistical models based on discrete distributions. In addition, although many GWAS have been performed, the biological meaning of the identified candidate genes, as well as their functional relationships still need to be better understood. Here, we performed and tested a Bayesian treatment of a GWAS model assuming a Poisson distribution for SB and NT in a commercial pig line. To explore the biological role of the genes that underlie SB and NT and identify the most likely candidate genes, we used the most significant single nucleotide polymorphisms (SNPs), to collect related genes and generated gene-transcription factor (TF) networks. Results: Comparisons of the Poisson and Gaussian distributions showed that the Poisson model was appropriate for SB, while the Gaussian was appropriate for NT. The fitted GWAS models indicated 18 and 65 significant SNPs with one and nine quantitative trait locus (QTL) regions within which 18 and 57 related genes were identified for SB and NT, respectively. Based on the related TF, we selected the most representative TF for each trait and constructed a gene-TF network of gene-gene interactions and identified new candidate genes. Conclusions: Our comparative analyses showed that the Poisson model presented the best fit for SB. Thus, to increase the accuracy of GWAS, counting models should be considered for this kind of trait. We identified multiple candidate genes (e.g. PTP4A2, NPHP1, and CYP24A1 for SB and YLPM1, SYNDIG1L, TGFB3, and VRTN for NT) and TF (e.g. NF-κB and KLF4 for SB and SOX9 and ELF5 for NT), which were consistent with known newborn survival traits (e.g. congenital heart disease in fetuses and kidney diseases and diabetes in the mother) and mammary gland biology (e.g. mammary gland development and body length). Background different parities [2]. In humans, it has been shown that Reproductive traits, such as number of stillborn piglets kidney diseases and diabetes in the mother and congeni- (SB) and number of teats (NT), are widely included in the tal heart disease in the fetus are some of the main causes selection indices of pig breeding programs due to their of the occurrence of stillbirths [3–5]. In pigs, however, importance to the pig industry. The number of SB is a a better biological understanding of these traits is still complex trait that is directly affected by the total num- needed to improve selection against SB. Number of teats ber of piglets born [1] and by temporal gene effects in is a trait with a large influence on the mothering ability of sows [6], since it is a limiting factor for increasing the number of weaned piglets. Biologically, the development *Correspondence: [email protected] 1 Department of Animal Science, Universidade Federal de Viçosa, of embryonic mammary glands requires the coordination Viçosa 36570000, Brazil of many signaling pathways to direct cell shape changes, Full list of author information is available at the end of the article © 2016 Verardo et al. This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/ publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Verardo et al. Genet Sel Evol (2016) 48:9 Page 2 of 13 cell movements, and cell–cell interactions that are neces- SNPs. The genes that are linked to significant SNPs can sary for proper morphogenesis of mammary glands [7]. be used to examine the sharing of pathways and func- In addition, number of vertebrae, which determines the tions, as well as the enrichment of significantly related TF body length of the sow, may also have a direct relation in the selected genes. Some studies have shown that TF with the final NT that is observed in pigs [8]. genes can be associated with important traits in pigs, e.g. Since these traits are directly involved with higher PIT1 in carcass traits [23] and SREBF1 in the regulation production and welfare of piglets, several genome-wide of muscle fat deposition [24]. association studies (GWAS) have been performed for SB Thus, providing evidence for an interaction between a and NT [2, 9, 10]. However, in these GWAS, these traits TF gene that is known to be related with a given trait and were assumed to be normally distributed, which may not its predicted target genes via the analysis of regulatory be true. Both SB and NT are measured as count varia- sequences and the construction of gene-TF networks bles, and therefore they follow discrete distributions such serves as an in silico validation for the gene-gene interac- as the Poisson distribution. Although the Poisson distri- tions. A gene-TF network facilitates the identification of bution has already been implemented in animal breeding the most probable group of candidate genes for the stud- in the context of traditional mixed models [11, 12] and ied trait. In addition, these genes can be used in further quantitative trait locus (QTL) mapping [13, 14], there are functional analyses of in vitro and in vivo validation. Sim- no reports of GWAS for SB and NT using such models. ilar approaches have been performed for puberty-related Poisson models can be fitted using a Bayesian Markov traits in cattle [25, 26] to identify related candidate genes chain Monte Carlo (MCMC) approach [15]. By applying and TF. In pigs, however, this approach has just begun to Poisson models to GWAS with a traditional mixed ani- be exploited [27]. mal linear model, it is possible to fit all markers simul- In this study, we performed a Bayesian treatment of a taneously by using the genomic relationship matrix [16], GWAS model assuming Poisson and Gaussian distribu- as performed in genomic best linear unbiased predic- tions for SB and NT in a commercial pig line. Then, we tion (GBLUP). GBLUP has been widely used in genome- used the significant SNPs to obtain related genes and wide selection (GWS), in which the vector of genome generate gene-TF networks, in order to explore the bio- enhanced breeding values (GEBV), from each MCMC logical roles that underlie the considered traits and iden- iteration, can be directly converted into a vector of sin- tify the most probable candidate genes. gle nucleotide polymorphism (SNP) allele substitution effects [17, 18]. Based on this, samples of the posterior Methods distribution for each SNP effect are generated at each Phenotypic and genotypic data cycle and significance tests based on highest posterior Stillbirth records from 1390 Large White (LW) sows with density (HPD) intervals can be performed [19, 20] in an average of 3.9 parities were evaluated. The average SB order to identify the most relevant SNPs. In addition, the number in this population was 1.2, ranging from 0 to 16 posterior probability (PPN0) of the estimated effect being SB per litter. NT was counted at birth for 1795 LW ani- smaller than 0 (for negative effects) or greater than 0 (for mals. The average NT in this population was 15.3, rang- positive effects), as proposed by Ramírez et al. [20] and ing from 14 to 20 teats. All animals were genotyped using Cecchinato et al. [21], can also be used to report the sig- the Illumina 60 K + SNP Porcine Beadchip [28]. As a part nificance of SNP effects under a Bayesian approach. Fur- of quality control procedures, SNPs with a GenCall score thermore, when considering all SNPs simultaneously in less than 0.15 [29], a minor allele frequency less than the model (in this case, the GBLUP assuming a general 0.01, a frequency of missing genotypes greater than 0.05, shrinkage parameter), some issues such as the influence unmapped SNPs, and SNPs located on the Y chromo- of gene length (i.e.