Submitted to: Rice Technical Working Group Meeting Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 1/19/2018
Publication Date: 10/16/2018
Citation: Eizenga, G.C., Edwards, J., Jackson, A.K., Jia, M.H., Boykin, D.L. 2018. Developing molecular markers for use in marker-assisted selection from rice genome-wide association studies. Rice Technical Working Group Meeting Proceedings. February 19-22, 2018, San Diego, California. p. 71-72.
Technical Abstract: Rice association mapping panels are collections of rice (Oryza sativa L.) accessions developed for genome-wide association (GWA) studies. One of these panels, the Rice Diversity Panel 1 (RDP1) was phenotyped by various research groups for several traits of interest, and more recently, genotyped with 700,000 SNP (single nucleotide polymorphism) markers using the high density rice array (HDRA). GWA analyses were conducted to identify GWA-QTL based on SNP markers associated with the particular trait of interest. To test the validity of these marker-trait associations and expedite the rice breeding process, markers are being developed from these associated SNPs and tested in biparental mapping populations and/or other association mapping panels. The RDP1 is composed of 423 diverse accessions originating from 79 countries and representing the two rice subspecies, Indica and Japonica, and five major subpopulation groups, indica, aus, aromatic, temperate japonica and tropical japonica. The panel was evaluated for 34 agronomic traits, most of which were yield components. The agronomic data collected from 2006 through 2010 was analyzed using a generalized linear mixed model with the GLIMMIX procedure in SAS, with year and replication nested within year as fixed effects, and accessions considered a random effect to calculate best linear unbiased predictors (BLUPs) and fixed effects for least squares means (LSmeans). Both BLUPs and LSmeans were evaluated as adjusted means for subsequent GWA analyses, with BLUPs selected for the quantitative traits and LSmeans for qualitative traits, in most cases. GWA analyses were performed using a mixed linear model in the Tassel 5.0 Pipeline (www.maizegenetics.net /tassel). The HDRA genotypic data (www.ricediversity.org) was further filtered for a minor allele frequency set at > 0.05. SNPs with heterozygosity >10%, > 40% missing calls, or not matching the Nipponbare reference, were excluded, resulting in 275,655 SNPs for GWA analyses. Tassel 5 was used to calculate population structure via principle component analysis (PC) and kinship. Manhattan and Q-Q plots were generated using the R package qqman. To group adjacent SNPs into significant regions, custom Perl scripts were developed with a threshold value to declare a SNP significant and find a maximum distance between significant SNPs to determine the start and stop of pseudomolecule positions for the significant region. To identify candidate genes, lists of annotated genes within 50Kb of the SNP with the highest significance in each region were obtained from the MSU7 and RAP1 rice gene annotations using a custom Perl script. Lists of annotated genes in the overlapping region were then manually examined to identify likely candidate genes based on annotated gene functions affecting the trait of interest found in either the Oryzabase or QTARO database based on pseudomolecule coordinates. To validate the GWA mapping, a biparental recombinant inbred line (RIL) population was developed from the RDP1 tropical japonica accessions, Estrela and NSFTV199. This population was genotyped with 132 SSR markers and phenotyped for 15 traits previously evaluated in RDP1 that were related to plant morphology, panicle architecture and seed shape. Across all traits, 42 of the 70 Estrela/NSFTV199 RIL-QTL overlapped with RDP1 GWA-QTL and were found in at least one of the six GWA analyses, which included all RDP1 accessions, the two subspecies (Indica including both indica and aus, and Japonica including temperate and tropical japonica), and three subpopulation groups (indica, tropical japonica, temperate japonica). A total of 40 known genes were identified in these overlapping regions. As proof of concept, we focused on two panicle architecture traits, number of primary branches per panicle and number of florets per panicle, to develop SNP markers based on the GWA-QTL. Th