Lei Liu, Yanfang Du, Dongao Huo, Man Wang, Xiaomeng Shen, Bing Yue, Fazhan Qiu, Yonglian Zheng, Jianbing Yan and Zuxin Zhang

Theoretical and Applied Genetics, November 2015, Volume 128, Issue 11, pp 2243-2254

http://link.springer.com/article/10.1007/s00122-015-2581-

Abstract

Key message

Maize kernel row number might be dominated by a set of large additive or partially dominant loci and several small dominant loci and can be accurately predicted by fewer than 300 top KRN-associated SNPs.

Abstract

Kernel row number (KRN) is an important yield component in maize and directly affects grain yield. In this study, we combined linkage and association mapping to uncover the genetic architecture of maize KRN and to evaluate the phenotypic predictability using these detected loci. A genome-wide association study revealed 31 associated single nucleotide polymorphisms (SNPs) representing 17 genomic loci with an effect in at least one of five individual environments and the best linear unbiased prediction (BLUP) over all environments. Linkage mapping in three F_2:3 populations identified 33 KRN quantitative trait loci (QTLs) representing 21 QTLs common to several population/environments. The majority of these common QTLs that displayed a large effect were additive or partially dominant. We found 70 % KRN-associated genomic loci were mapped in KRN QTLs identified in this study, KRN-associated SNP hotspots detected in NAM population and/or previous identified KRN QTL hotspots. Furthermore, the KRN of inbred lines and hybrids could be predicted by the additive effect of the SNPs, which was estimated using inbred lines as a training set. The prediction accuracy using the top KRN-associated tag SNPs was obviously higher than that of the randomly selected SNPs, and approximately 300 top KRN-associated tag SNPs were sufficient for predicting the KRN of the inbred lines and hybrids. The results suggest that the KRN-associated loci and QTLs that were detected in this study show great potential for improving the KRN with genomic selection in maize breeding.

Fig. 3  Predictability using tagSNPs for the kernel row number. (a) Predictability of the top tagSNPs and randomly selected tagSNPs in the inbred lines. Continuous lines the prediction accuracies using 5–24 K top tagSNPs (strategy 1); Dotted lines the prediction accura- cies using 5–24 K randomly selected tagSNPs (strategy 2). (b) Prediction accuracies using 5–24 K top tagSNPs for 54 hybrids. (c) Predic- tion accuracies of 5–24 K top tagSNPs in different subpopulations using different training sets and validation populations. (d) Predictabil- ity of different sizes of training sets using 300 top tagSNPs in inbred lines and hybrids