基于核磁共振的玉米不同籽粒类型单粒质量和含油率分析

针对现有玉米单倍体核磁共振分选系统基于一个含油率阈值，无法对胚败育籽粒和单倍体籽粒正确分选的问题，分别对玉米生物诱导产生的二倍体、单倍体和胚败育3种不同籽粒类型的单粒质量和含油率进行分析，提出了利用籽粒含油率双阈值提高单倍体正确识别率的分选方法。该研究以2个普通玉米杂交种和3个自交系为母本，以高油型诱导系为父本，进行生物诱导产生的3种不同类型籽粒为研究对象，利用核磁共振分选系统分别对不同类型籽粒的单粒质量和含油率进行测定，结果表明：单粒质量整体表现为单倍体>二倍体>胚败育，除二倍体籽粒与胚败育籽粒间存在极显著差异外，其他籽粒类型间差异不显著；不同类型籽粒的单粒质量平均变异系数为16.62%，并且每个材料的3种籽粒类型间出现较大的重叠区域。而不同类型籽粒含油率整体表现为二倍体>单倍体>胚败育，变异性以二倍体最小，平均变异系数仅为12.52%，其次是单倍体，而胚败育籽粒最高（34.14%），但其含油率最低且均≤2%；每个材料各自的3种类型籽粒间含油率呈现梯度分布，存在较明显的界限。由此可见，利用籽粒含油率能够区分玉米生物诱导的3种不同籽粒类型，而单粒质量则不能；通过设置二倍体籽粒的最小含油率为上限，胚败育籽粒的最大含油率为下限，利用含油率的双阈值可提高单倍体的正确识别率，为玉米生物诱导单倍体高效自动化分选提供依据。

Analysis on single kernel weight and oil content of different grain types in maize based on NMR

Maize haploid breeding as an important technology of modern breeding programs can shorten breeding cycles and enhance the breeding efficiency. The method of haploid produced by in vivo has become a routine process and has been adopted widely in maize breeding. The rapid and accurate recognition of haploid kernels is a prerequisite for in vivo haploid breeding technology. Therefore, how to efficiently select haploid seeds from mass hybridization kernels has been an issue during the use of parthenogenetic haploid. At present, spectral information and image recognition based on the gene R-nj expressing Navajo marker in kernels were usually used to automatically sort haploids in maize, but the expression of gene R-nj were significantly depended on the maternal genetic backgrounds, it is unavailable for some flint germplasms with the presence of a supressor. With the development of inducers with high oil and high inducing rate, the automatic sorting of maize haploid kernels become a reality based on nuclear magnetic resonance (NMR). The embryo of haploid only developed from female gamete without the fusion of the sperm cell, so the oil content of haploid kernels is usually lower than that of diploid kernels because of xenia effect. Presently, most studies only set one threshold of oil content to identify haploid and diploid kernels based on NMR. However, there are some embryo abortion kernels with very lower oil content during the process of haploid induction, it is difficult to make distinction among diploid, haploid and embryo abortion kernels based on one threshold of oil content using NMR sorting system. In this paper, the single kernel weight and oil content of diploid, haploid and embryo abortion kernels by in vivo induction were analyzed, respectively, a new concept of increasing the correct discrimination rate of haploid based on two thresholds of oil content was developed. Three different types of kernels were produced by in vivo induction using two common maize hybrids and three inbred lines as females and inducers with high oil content as the males. The measurement of single kernel weight and oil content for three types of kernels were carried out by NMR sorting system. The result showed that the rank of the single kernel weight of different grain types was haploid > diploid> embryo abortion kernels, the highly significant difference existed between diploid and embryo abortion kernels, but there was no difference between diploid and haploid kernels, haploid and embryo abortion kernels. The variation coefficient of the average performance of single kernel weight was 16.62%, lager overlaps appeared among three different types of kernels. The oil content showed that the diploid > haploid > embryo abortion kernels. The coefficient of variation of oil content for diploid kernels was the lowest with only 12.52%, followed by haploid kernels, the highest was embryo abortion kernels with 34.14%, but the oil content was not more than 2% for all the embryo abortion kernels. The step distribution with obvious boundaries for oil content among the three different grain types was found. It indicated that the oil content can be used to recognize among three different types of kernels, but the single kernel weight cannot be as a sorting standard. According to the oil content, minimum of diploid and maximum of embryo abortion kernels were set respectively as the upper and lower limit value to form double thresholds, which could improve the rate of correct discrimination for haploid and provide a reference during the process of automatically sorting haploid kernels with high efficiency.