大豆育种的分子标记系统最新进展（英文版）

现代分子标记技术与大豆育种（英文版）系统地介绍了各种类型分子标记的发展概况和开发技术，QTL分析和分子辅助育种的基本原理，阐述了大豆产量性状、品质性状、抗病虫性状、动态发育性状分子标记和QTL的*研究进展及大豆分子设计育种的基本设想及展望。该书反映了近年分子标记在大豆遗传改良方面获得的成果。

Wenbin Li
Key Laboratory of Soybean Biology in Chinese Ministry of Education(Agriculture Research System of China and the Key Laboratory of Northeastern Soybean Biology and Breeding/Genetics of Chinese Agriculture Ministry),Northeast Agricultural University,Harbin,China.150030
David A.Lightfoot
Illinois Soybean Center for Excellence in Soybean Research,Southern Illinois University at Carbondale,Carbondale,IL 62901,USA

Acknowledgement
Preface
Chapter 1 Essential principles of molecular marker and quantitative trait loci associations
Chapter 2 Method development for identifying functional markers in the soybean genome
Chapter 3 Map development and marker-assisted selection for oligogenic and polygeneic traits
Chapter 4 QTL analysis of fungal disease resistance
Chapter 5 Identification of loci for resistance to insects and nematodes
Chapter 6 QTL analysis for seed quality traits
Chapter 7 Seed yield and component trait loci
Chapter 8 Loci underlying soil edaphic, physiological and developmental traits
Chapter 9 Dynamic QTL analysis of major quantitative traits
Chapter 10 Molecular mapping and breeding for biotic stress resistance
Colored Figures

Chapter 1
Essential principles of molecular marker and quantitative trait loci associations
David A. Lightfoot
Illinois Soybean Center of Excellence in Research, Teaching and Outreach, Department of Plant, Soil and Agricultural Systems, 176 Ag. Building, MC 4415, Southern Illinois University at Carbondale, Carbondale, Illinois 62901?4415, USA
Abstract
In soybean genetics databases across the world there are more than 2,000 loci that underlie quantitative traits, 500 that underlie oligogenic traits and 300 loci that underlie simple traits recognized by association mapping in populations. The number of loci will increase as new phenotypes are measured in more diverse genotypes and genetic maps based on saturating numbers of SNPs are developed. A period of locus re-evaluation will decrease the number of important loci as those underlying mega-environmental effects are recognized. A second wave of re-evaluation of loci will follow from developmental series analysis, especially for harvest traits like seed yield and composition. Breeding methods to properly use the accurate maps of QTL are being developed. New methods to map, fine map and isolate the genes underlying the loci will be critical to future advances in soybean biotechnology. The selective genotyping approaches, including bulked segregation event (segregant) analysis (BSA), will be increasingly important. Association mapping across panels of germplasm will become more widely used. The identification of eQTL based on transcript, protein or metabolite abundance shifts are another new and promising method. Mutational approaches like TILLING, RNAi and VIGS will also identify some genes underlying important loci. Proteomics and protein-protein-metabolite interactions are fourth area promising critical advances. Examples of the methods are presented for monogeneic, oligogeneic and polygeneic traits. Examples of successful mapping, fine mapping and gene isolation are given. When combined with high-throughput methods for genotyping and a genome sequence advanced population structures and by the use of association mapping in panels adequately reflecting the diversity in soybean across the world the new methods described herein will provide critical advances in the analysis of many economically important simple traits.
Keywords bulked segregation event (segregant) analysis (BSA); monogeneic; oligo- geneic; polygeneic traits; TILLING; quantitative trait loci; eQTL
Introduction
The analysis of QTL stands on a new threshold due to several recent advances. In genomics critical advances include the development of a genome sequence for soybean (Schmutz et al. 2010), soybean genome resequencing (Lam et al. 2010) and high throughput methods for marker scoring have been made (Wu et al. 2010). Worldwide advances in genetics like microarrays (Gupta et al. 2008; Tyler et al. 2009; Yuan et al. 2009) and RNA sequencing (Severin et al. 2010) based on inexpensive second and third generation sequencing capacity (Van Orsouw et al. 2007; Wu et al. 2010) have also impacted QTL identification. In parallel, new powerful methods for QTL identification have been developed including eQTL identification (Tyler et al. 2009) and developmental QTL identification (Li et al. 2007b, 2010). Simultaneously, methods to isolate the gene(s) underlying QTL have made exciting advances (Yuan et al. 2009; Liu et al. 2011) since the development of large scale positional cloning projects (Ashfield et al. 2003; Searle et al. 2006; Ruben et al. 2006). Large scale transformation projects using model plants are possible (Shultz et al. 2006; Ullah et al. 2009). Even soybean transformation has become routine and may be used at high throughput for gene inhibition (Chai et al. 2010). Combined these tools provide tremendous power to those scientist engaged in the identification of genes underlying soybeans economically and scientifically important traits. In the next decade many more traits will be associated with their underlying genes. This book will provide examples of these techniques applied to different groups of traits.
Trait Analysis
The distinction between quantitative, oligogeneic and simple traits lies with the number of polymorphic loci underlying a trait (Lightfoot 2010). Polygeneic traits are underlain by polymorphisms at 10 loci or more. Oligogeneic traits vary at the alleles of 3-9 loci. Simple traits are controlled by allelic variation at 1-2 loci. Trait classification is context dependent, the diversity of the population or germplasm collection used to analyze the trait will determine its genetic complexity. Therefore, traits exist within a continuum and may be moved from one class to another by population design. For
example polygenes underlying traits like partial disease resistances or seed yield and composition at harvest (Table 1-1; Yuan et al. 2002; Kassem et al. 2004) can be split and Mendelized by analysis over developmental time in a NIL population segregating for a single region (Figure 1-1; Njiti et al. 1998; Meksem et al. 1999; Triwitayakorn et al. 2005; Ruben et al. 2006).
Table 1-1 Examples of markers associated with seed yield at harvest among recombinant inbred lines in two soybean populations
Population, marker and map position Location P value R2/% LOD QTL var /% Yield mean ± SEM / (Mg ha.1) (alleles from) P1 P2
Essex × Forrest Essex Forrest
Satt337 LG K(Gm9) C96 0.0328 5.8 1.9 13.7 2.88 ± 0.06 2.71 ± 0.06
47.4 cM R96 0.0269 6.2 1.3 7.5 4.51 ± 0.05 4.38 ± 0.04
D97 0.0197 6.9 1.6 10 3.02 ± 0.03 2.91 ± 0.03
Mean. 0.0042 10.1 2.1 10.7 3.40 ± 0.03 3.29 ± 0.03
Satt167 LG K(Gm9) D97 0.0047 10 1.5 7.7 3.01 ± 0.03 2.87 ± 0.04
45.7 cM Mean 0.0023 11.6 1.9 9.6 3.39 ± 0.03 3.27 ± 0.03
Satt294 LG C1(Gm4) D97 0.0002 16.4 3.7 23.7 2.88 ± 0.03 3.06 ± 0.03
78.6 cM Mean 0.0056 9.5 2 12.1 3.29 ± 0.02 3.41 ± 0.03
Satt440 LG I(Gm20) D97 0.001 14.8 2.5 14.7 2.86 ± 0.03 3.02 ± 0.04
112.7 cM Mean 0.0071 10.2 1.6 10.3 3.28 ± 0.03 3.39 ± 0.03
TMD1 LG G(Gm18). Mean 0.2964 3.32 ± 0.03 3.36 ± 0.03
Flyer × Hartwig Flyer Hartwig
Satt337 LG K(Gm9) N98 0.0042 10.1 5.9 26.9 3.20 ± 0.06 2.69 ± 0.08
47.4 cM N99 0.0001 27.1 1.2 6.4 2.53 ± 0.05 2.36 ± 0.04
Mean 0.0006 13.8 2.7 13.7 2.98 ± 0.03 2.77 ± 0.05
Satt326 LG K(Gm9) N98 0.0001 26.3 5.4 25.5 3.20 ± 0.06 2.69 ± 0.08
49.5 cM N99 0.0082 8.6 1.6 7.8 2.52 ± 0.06 2.33 ± 0.04
Mean 0.0004 15 3 14.6 2.98 ± 0.04 2.76 ± 0.05
Satt539 LG K(Gm9) R98 0.049 5 0.8 4.5 3.39 ± 0.09 3.62 ± 0.07
2.0 cM H99 0.0006 14.7 2.6 13.4 2.45 ± 0.07 2.74 ± 0.04
N99 0.012 8.1 1.5 7.7 2.35 ± 0.05 2.54 ± 0.05
Mean 0.0008 14 2.5 13 2.77 ± 0.06 2.99 ± 0.03
TMD1 LG G(Gm18) . Mean 0.0007 9.7 3.1 27.4 2.32 ± 0.03 2.64 ± 0.05
. Also significantly associated with SCN resistance (within the intron of the RLK at rhg1).
Note: Mean involves all locations used for evaluation within each population. ‘Essex’ × ‘Forrest’ was evaluated in Carbondale, IL (C96) and Ridgway, IL (R96) in 1996 and in Desoto, IL (D97) in 1997. ‘Flyer’ × ‘Hartwig’ was evaluated in Nashville, IL (N98 and N99) in 1998 and 1999; Ridgway, IL (R98) in 1998 and Harrisburg, IL (H99) in 1999. Clustered yield QTL were found on LG K (Gm9) and G (Gm18) that could be fine mapped in NILs. Loci on C1 (Gm4) and I (Gm20) proved difficult to isolate in NILs and so might be inaccurately mapped or blends of conditional QTL. The rhg1 locus was a yield determinate in ‘Flyer’ × ‘Hartwig’ suggesting SCN pressure was significant even though SCN counts were very low.
A. RILs
B. NILs
110 105
100
Number of RILs
95 90
85
80
EF7-3 EF7-35 EF7-29
EF7-6 EF7-8
80.5 85 89.5 94 98.5 103 108 112 117 121
22. TGly 05 NIL
C. QTL map
BARCSatt251 BARC-Satt197 BARCSatt128 BARCSatt519 BARC-Satt_415
10 cM QTL LOD >10
QTL1-R4 QTL1-R6 QTL1-R3 QTL1-R7 QTL2-R8
D. Fine map QTL2-R4
QTL2-R6
EF7-3 (5 other NILs) EF7-35 (4 other NILs) EF7-29 (5 other NILs) EF7-6 (13 other NILs) EF7-8 (13 other NILs)
Recombinant classes Forrest alleles
between markers among~80 NILs
Figure 1-1 The effect of development on the apparent position of QTL underlying seed glycitein (as well as daizein and total phytoestrogen) content on LG B1 (Gm11) (see Colored Figures)
Panel A shows segregation of this polygenic trait in the RIL population EF94. Panel B shows the 4 QTL within the region Mendelized in each of 4 NILs with 1-4 beneficial alleles each (from ‘Essex’). Panel C shows the apparent position of 2 QTL for seed yield at harvest. QTL1 is actually the composite of 4 loosely clustered loci active at different reproductive stages from R3 (cell division) to R7 (cell loading). QTL2 is wrongly positioned as the composite of two closely linked QTL. Panel D shows the protein-protein interaction map for soybean and the recombination events among some NILs that might be able to separate the distantly linked QTL but would fail to separate closely linked QTL
Developmental Effects on Traits and QTL Locations
Seed traits in soybean like protein, oil, yield and seed size are controlled by multigenes that may show small or large effects (Yuan et al. 2002; Hyten et al. 2004; Sun et al. 2006; Li et al. 2010). Many QTL were discovered previously to proper analyses of harvest traits (Mansur et al. 1996; Orf et al. 1999; Yuan et al. 2002; Hyten et al. 2004). Many QTL proved to be inaccurately mapped or even impossible to introgress (Reyna and Sneller 2001). One problem many studies encountered was the strong effects of G × E on QTL. The loci underlying seed protein and oil contents in soybean were mainly based on phenotypes measured with post harvest, dried seed.
Consequently, the QTL detected could not show the genetic effects of plant developmental stages, just their mean. To clarify the area of study the net genetic effect of a harvest trait or pre-harvest trait was called an unconditional genetic effect (underlain by unconditional QTL) whereas the loci detected at a specific growth stages with conditional genetic effects were conditional QTL (Yan et al. 1998; Sun et al. 2006). The association of developmental behaviors of quantitative traits with molecular markers had been reported in soybean (Sun et al. 2006; Li et al. 2007b, 2010). Therein, QTL analysis was adapted to include the effects of developmental state. Re-evaluation of yield QTL reported previously with the new method should provide better tools for marker assisted selection of harvest trait QTL.
In an alternate approach several researchers have tried to exclude the effects of clustered conditional QTL by the use of closely related inbred lines (Triwitayakorn et al. 2005; Ruben et al. 2006; Bolon et al. 2010). This approach can succeed when genetic power is sufficient to separate linked QTL (Ashfield et al. 2003; Srour et al. 2012) but will struggle or fail where genes underlying traits are clustered as in seed quality (Figure 1-1), within heterochromatin as with insect resistance (Zhu et al. 2009) or pleiotropic as at rhg1 for SCN resistance and seed yield (Table 1-1).
Selective Genotyping
The detection of quantitative trait loci (QTL) in RIL populations requires large sample sizes (400 < n < 4000) to attain reasonable power and avoid sampling biases (Soller et al. 1976). Such large numbers are difficult to plant, maintain and score accurately with sufficient replication. In marker-QTL linkage experiments for reduction of the number of individuals needed to be genotyped, a procedure termed “selective genotyping” has been proposed. Selective genotyping (Darvasi and Soller 1992) is a method in which only individuals from the high and low phenotypic extremes are genotyped to get the most informative, quantitative trait values. It has been shown that the number of individuals genotyped to attain a given power can be decreased significantly. The major limitation of this approach is that if the experiment is aimed at analyzing a number of traits, then by selecting the extremes of each trait one would select most of the population and thus no reduction in genotyping can be obtained. Selective genotyping is thus most appropriate for the cases where only one trait is being analyzed. This conclusion is valid when selective genotyping is applied to QTL
detection.
Bulked Segregant Analysis
Bulked segregant analysis (BSA) is a specialized form of selective genotyping (Darvasi and Soller 1992) that provides a rapid procedure for identifying markers in specific regions of the genome (Michelmore et al. 1991; Darvasi and Soller 1994). The only prerequisite for BSA is the existence of a population whose members contrast for a trait; either resulting from a single cross (standard QTL mapping) or from serial intercrosses (association mapping). In soybean, the method can be used for both qualitative traits and for detecting major genes underlying quantitative traits, even polygeneic traits like seed yield (Mansur et al. 1996; Yuan et al. 2002).
The BSA method involves comparing two pooled DNA samples from a segregating population (Meksem et al. 2001). Within each pool, or bulk, the small number of individuals selected is expected to have identical genotypes for a particular genomic region (target locus or region) but random genotypes at loci unlinked to the selected region. Pools can be selected from phenotype for locus discovery (Mansur et al. 1996) or by genotype for marker saturation of a particular region (Meksem et al. 2001; Figure 1-2). In the latter case the pools can be fine tuned to saturate a region by the inclusion of recombination events flanking the target region. The two pools contrasting for a trait
(e.g. resistant and susceptible to a particular disease) are analyzed to identify markers that distinguish them. Markers that are polymorphic between the pools may be genetically linked to the loci determining the trait used to construct the pools or may be error associations caused by sampling error. Therefore, the size of pools and their accuracy of phenotypes or genotypes underlying their member genotype composition are critical factors in the success of the technique (Table 1-1).
Bulked segregant analysis has two immediate applications in developing genetic maps. ①BSA provides a method to focus on regions of interest or areas sparsely populated with markers. ②BSA is a method for rapidly locating genes that do not
segregate in populations initially used to generate the northern US germplasm based composite genetic map (Song et al. 2004) or other high density maps (Harada 2012; Lightfoot 2008). The BSA technique is advantageous in identifying markers associated with new traits without the need for full map construction.