【作者】 师家勤；

【导师】 孟金陵；

【作者基本信息】 华中农业大学 ， 发育生物学， 2009， 博士

【摘要】 对作物遗传改良来讲,产量是最重要的目标性状。然而,产量又是极其复杂的性状,它是作物整个生命周期内一系列生长发育过程和环境互作的最终产物。首先,产量构成因子直接决定产量;其次,产量相关性状可以通过影响产量构成因子或其它未知的机制间接影响产量。虽然有大量的研究,但是在这些实验中产量及其杂种优势的遗传结构和遗传决定仍然是模糊不清的。首先,单个研究所检测到的QTL和互作对数目很少;其次,产量和具体的产量关联性状没有联系起来;最后,实验所设置的环境很少。本研究利用Tapidor（欧洲冬性油菜品种）×Ningyou7（中国半冬性油菜品种）组合F₁小孢子培养得到的含202个株系的DH群体在10个环境（6个半冬性和4个冬性）,及这些株系随机相互杂交衍生的RC-F₂群体在3个环境（2个半冬性和1个冬性）下的田间实验,考察了种子产量及8个产量关联性状（5个产量相关性状:一次分枝数、生物学产量、花期、成熟期和株高;3个产量构成因子:全株角果数、每角果粒数和千粒重）,结合一张高密度的分子标记遗传连锁图及其和拟南芥基因组的比较作图和候选基因的电子定位,利用元分析的方法对产量及其杂种优势的遗传结构和遗传决定在全基因组水平上进行了全面和深入的剖析。遗传图谱的构建、比较作图和候选基因的电子定位:利用TNDH群体202个株系1040个分子标记的基因型用JoinMap 3.0软件构建了一张786个分子标记的遗传连锁图:它覆盖甘蓝型油菜全基因组19个连锁群（A1-A10;C1-C9）,图谱总长2117.2cM,标记间距为2.7 cM。以已鉴定的拟南芥21个保守区段为基础,将TNDH遗传连锁图上已知序列信息的277个分子标记作为锚定标记与拟南芥基因组进行比对,共鉴定了上百个共线性区段和保守岛。从拟南芥网站和发表的文献中收集了425个与本研究考察性状（如花期、株高、分枝数和种子数目等）相关的拟南芥基因,用电子定位的方法将它们的2185个同源基因定位到TNDH遗传连锁图上。亲本和群体的表型和杂种优势:对于花期、成熟期和千粒重,不管是冬性还是半冬性环境,两个亲本的表现都是一致的;对于其它6个性状,两个亲本在冬性和半冬性环境的表现完全相反而均值相当,这说明这6个性状的可塑性更大。而且,这6个性状的杂种优势都达到了显著水平:种子产量的中亲优势最强,其次是全株角果数和生物学产量,而株高、每角果粒数和一次分枝数较弱。F₁组合的中亲优势要强于RC-F₂群体的平均中亲优势,这说明一般来讲杂合有利于杂种优势。RC-F₂群体许多组合的中亲优势超过了F₁组合,这说明并不是所有杂合位点都对杂种优势有利,而有些纯合位点反而对杂种优势有利。表型和杂种优势的遗传相关和遗传力:不同环境同一对性状的表型和杂种优势间的遗传相关都有较大差异（大部分在程度上,少部分在方向上）,这说明遗传相关的程度强烈依赖于所处的环境。性状表型及其杂种优势间的遗传相关表现出类似的规律:种子产量和绝大部分的其它性状都显著相关（和花期、成熟期负相关;和其它6个性状正相关）而且具有最高的平均决定系数,这说明产量及其杂种优势高度依赖于其它性状,它反映了产量的综合性特征。方差分析结果表明:基因型、环境和基因型×环境对所调查的性状都有显著的影响。性状中亲优势的遗传力普遍比表型遗传力低,而且在不同性状间表现相同的趋势:花期最高、千粒重和成熟期较高,株高和每角果粒数居中,生物学产量、一次分枝数和每角果粒数较低,种子产量最低。一般杂合度/特殊杂合度和杂种优势/杂种表型的遗传相关:一般来讲,特殊杂合度和杂种表型/中亲优势之间的相关性要高于一般杂合度和杂种表型/中亲优势的相关性,而杂种表型和杂合度之间的相关性和中亲优势与杂合度的相关性差别不大。不同环境间,其相关系数却表现出较大的差异,以S5最高,S6和N6差别不大。不同性状之间,其相关系数却表现出较大的差异。以种子产量各类相关系数的均值最大,其次是全株角果数、生物学产量和千粒重,再次是每角果粒数和一次分枝数,花期和成熟期最小。这一结果与这些性状杂种优势的强弱基本吻合。总的来看,这些相关系数都不高,这说明由分子标记杂合度无法预测杂种表型或中亲优势。QTL的检测及其遗传模式:两群体9性状在10个环境中共检测到了1022个QTL（其中614和408个分别在P=0.05和0.5的显著水平）,去掉其中152个不能重复的P=0.5水平的QTL后,确定了870个QTL（命名为“identified-QTL”）,其中85个控制种子产量。使花期推迟、成熟期延长、种子重量降低的identified-QTL等位基因大部分来自Tapidor,使分枝增多、生物学产量增加、植株增高、角果增多、种子增多、种子产量增加的等位基因基本上平均分布于双亲,这一结果与双亲这些性状的差异基本吻合。对RC-F₂群体来说,部分显性QTL所占的比例最高（53%）,其次是加性（24%）,显性和超显性所占的比例较少（12%和11%）。而且,种子产量QTL更多的表现显性和超显性,其次是全株角果数和生物学产量,每角果粒数、千粒重、一次分枝数和株高较少,花期和成熟期没有表现超显性的。不同实验QTL的整合及其对自然环境的响应:同一性状的大部分identified-QTL置信区间相互重叠,其中的694个用元分析的方法被整合成225个可重复的“consensus-QTL”,它们的平均置信区间也从6.5 cM缩减至3.7 cM,另外的176个identified-QTL未参与整合。这样,总的QTL数目缩减至401,它们分成两类:15个major QTL（至少在一次实验中R²≥20%或至少在两次实验中R²≥10%）和386个minor QTL（余下的）。将近一半consensus-QTL只在1个环境中出现,很少有consensus-QTL能够在一半以上的环境中出现。三分之一consensus-QTL能在冬性和半冬性环境中都出现,而三分之二的consensus-QTL只在其中一个出现。这说明产量性状及其杂种优势基因的表达强烈依赖于所处的环境。不同性状QTL的整合及指示QTL的确定:大部分consensus-QTL的置信区间相互重叠,其中的329个再次用元分析的方法被整合成111个多效性的“unique-QTL”,它们的平均置信区间缩小到2.5 cM,另外72个consensus-QTL未参与整合。所有55个种子产量consensus-QTL中只有8个是“独立”的,其它47个分别被整合成多效性的unique-QTL。事实上,这8个所谓“独立”的种子产量QTL中有6个与本研究未报道性状（叶片硫苷和菌核病抗性）的QTL重叠。这基本上表明,在我们的研究中,所有种子产量QTL都依赖于其它产量关联性状。种子产量和其它产量关联性状表现出最高程度的遗传相关,而且高比例的种子产量consensus-QTL和其它产量关联性状consensus-QTL重叠,这暗示着这些和产量consensus-QTL共定位的产量关联性状consensus-QTL是产量的潜在遗传决定。为了估计哪些产量关联性状consensus-QTL最有可能对产量consensus-QTL具有潜在的多效性效应,我们提出了指示QTL的概念及其确定标准:（1）和产量consensus-QTL重叠,（2）LOD值更高,（3）可重复性更好;（4）和产量QTL在更多环境中共同出现。47个多效性的种子产量consensus-QTL中有29个与多个产量关联性状consensus-QTL共定位,这说明除了多效性以外,产量QTL可能是多个紧密连锁的产量关联性状QTL的综合效应。因此,对其中19个多效性的种子产量consensus-QTL,我们确定了几个指示QTL,最后总共确定了63个指示QTL。多个indicatro-QTL和种子产量consensus-QTL可在相同或不同的环境中出现,这暗示着产量遗传结构的综合性、可变性和可塑性。与种子产量consensus-QTL相比,指示QTL往往LOD值和R²更大,重复性更好,候选基因更容易鉴定,性状更容易考察,因而它方便了产量QTL的克隆。有几个间接的证据支持指示QTL是种子产量consensus-QTL的潜在遗传决定:（1）指示QTL和种子产量consensus-QTL之间的遗传距离很近,平均只有1.1 cM。（2）指示QTL和种子产量consensus-QTL置信区间的重叠比例很高,平均高达80.9%。（3）对于花期和成熟期,指示QTL和种子产量QTL加性效应的方向往往是相反的,而对于其它6个性状则往往是相同的,这一结果与这些性状和种子产量遗传相关的方向基本吻合。（4）许多指示QTL加性效应所估计出来的种子产量的变化值和种子产量QTL的加性效应很接近。互作对的检测、分布、类型、效应及其对自然环境的响应:两群体9性状在10个环境中共检测到了538个互作对,它们大多数主效并不显著。这些互作位点在基因组中的分布极不均匀,可见集中分布的热点,但是和这些热点互作的位点绝大部分是在不同的环境中检测到的,并且只有1%左右的互作对能在不同的环境中重复出现,这一比例远低于玉米和水稻（10%左右）,这说明多倍体基因组中上位性的高度可变性和可塑性。对于RC-F₂群体检测到的互作对,加加互作的数目和总贡献率最大,其次是加显/显加互作,显显互作最小;而对于QTL,显性效应的总贡献率只占很小一部分。而且对于种子产量来说,互作对的总贡献率要大于QTL的总贡献率。因此,我们认为上位性（特别是加加上位性）是甘蓝型油菜杂种优势的主要遗传基础。9性状所有互作对两位点9种基因型组合的表型排序分析表明:（1）双杂合子没有排名第一和最后的,呈现两头小中间大的趋势,其它3类都有可能排名第一或最后;（2）单杂合子虽然有排名第一和最后的,但比例较小,而且其它排名的比例很均匀;（3）重组类型的纯合子和亲本类型的纯合子排名第一和最后的比例最高,呈现两头大中间小的趋势。这说明:双杂合子的稳定性＞单杂合子的稳定性＞纯合子的稳定性。本研究具有重要的科学意义和应用价值:（1）在一次实验中定位了几百个产量性状及其杂种优势相关的QTL,其中一半以上可在多个实验中得到验证,它们为分子标记辅助选择、QTL克隆和基因功能分析提供了重要的遗传资源。（2）利用元分析的方法首次准确的揭示了这些产量性状及其杂种优势QTL和互作对自然环境的响应规律,这对作物遗传育种研究具有重要的指导意义。（3）初步构建了一张较完整的产量及其杂种优势的遗传结构图,它的复杂性体现在产量QTL的多效性、综合性、可变性和可塑性以及上位性的重要性、高度可变性和可塑性。（4）在国际上首次创造性地提出了指示QTL的概念并通过它估计了产量及其杂种优势的候选基因,不但深化了我们对作物产量及其杂种优势遗传决定的认识,方便了产量QTL的克隆,而且为复杂性状的研究提供了一种新的思想和方法。更多还原

【Abstract】 Yield is the most important and complex trait for the genetic improvement of crops, reflecting the culmination of all processes of growth and development and their interaction with the environment.Crop yield is directly determined by yield-component traits.Yield-related traits may also indirectly affect yield by affecting the yield-component traits or by other unknown mechanisms.Although much research has been reported,in each such experiment the genetic architecture and determinants of yield and heterosis has remained ambiguous.Firstly,few QTL have been identified;Secondly,yield,a particularly complex factor,has not been associated with any of the yield-associated traits, relatively simple factors;Thirdly,trials were carried out in a few environments.In this study,we developed two permanent populations of rapeseed（Brassica napus） from a cross,Tapidor（a European winter cultivar）×Ningyou7（a Chinese semi-winter cultivar）.One is a double haploid population（named as TNDH） of 202 lines derived by microspore culture of F₁,and the other one is a reconstructed F₂（named as TNRC-F₂） population of 404 lines resulted from the randomly intermating of the 202 lines of TNDH population.The two populations（TNDH and TNRC-F₂） were grown in 10 and 3 year×location combinations（microenvironments）,respectively,with a randomized complete block design of 3 replicates.Seed yield and eight yield-associated traits（3 yield-component traits:pod number,seed number and seed weight;5 yield-related traits: branch number,biomass yield,flowering time,maturity time and plant height） were investigated.The genetic architecture and determinants of yield was analyzed thorough identification and meta-analysis of quantitative trait loci（QTL） for seed yield and 8 yield-associated traits from two populations and 10 year×location combinations using a high-density linkage map,comparative mapping and candidate gene in silico mapping. The main results and conclusions are summarized as follows:Construction of linkage map,comparative mapping and in silico mapping of candidate genes:A high-density linkage map of 786 molecular markers was constructed by JoinMap 3.0 software using 1040 molecular markers genotyped with TNDH population. It covered 19 linkage group identified as A1-A10（A genome） and C1-C9（C genome）, with a total length of 2117.2 cM and an average distance of 2.7 cM between markers.Based on the 21 syntenic blocks of Arabidopsis identified in previous research, 277 molecular markers with known sequence information were employed as anchored markers for the alignment between the linkage map of TNDH and the genome of Arabidopsis thaliana and more than one hundred syntenic blocks or insertion fragments （islands） were identified.A total of 425 genes of Arabidopsis with known functions relating to flowering time,plant height,branch number,and other traits investigated in this study were collected from the TAIR website and the published papers.A total of 2185 homologous of the 425 Arabidopsis genes were were aligned to the TNDH linkage map by in silico mapping approach.Phenotype and heterosis of two parents and populations:Ningyou7,a semi-winter cultivar,had larger seeds and flowered and matured earlier than the winter cultivar,Tapidor,in all environments.However,for the other six traits,the performance of the two parents in the two macroenvironments（winter and semi-winter） was reversed, such that the means in the two macroenvironments were similar.This reflects the adaptability of the two cultivars to the different macroenvironments and the phenotypic plasticity of flowering time,maturity time and seed weight were smaller than that of other 6 traits.The six traits also showed significant heterosis:seed yield＞pod number and biomass yield＞plant height,seed number and branch number.The midparent heterosis of F₁ was higher than that of the mean of RC-F₂ population,this indicated that heterozygote was generally favorable to heterosis.The midparent heterosis of many cross in RC-F₂ population was greater than that of F₁ cross,this indicated heterozygote are not always favorable to heterosis and homozygote at some loci may further improve heterosis.Genetic correlations and heritability of phenotype and heterosis among different traits:In different microenvironments,pair-wise genetic correlations among the phenotype/midparent heterosis of different traits differed considerably（mostly in degree,a few in direction）,which suggested that genetic correlations depended strongly on the environment.In general,seed yield was correlated with all investigated traits（negatively for flowering and maturity times,and positively for the other 6 traits）,notably with the highest average coefficients of determination.This indicated seed yield strongly depended on other traits investigated in this research,reflecting the synthesis and complexity of seed yield.Despite strong genotypic and environmental effects for all investigated traits,there were significant genotype×environment interactions.The broad-sense heritability of trait performance was higher than that of midparent heterosis:flowering time＞seed weight and maturity time＞plant height and seed number＞biomass yield,branch number and pod number＞seed yield.Correlations of general heterozygosity/special heterozygosity and heterosis and hybrid performance:In general,the correlation of specail heterozygosity and hybrid performance and midparent heterosis was higher than that of general heterozygosity and hybrid performance and heterosis.But the correlation of hybrid performance and heterozugosity were not significantly different with that of heterosis and heterozygosity.In different microenvironments,the correlation coefficients showed significant difference: the highest for S5,S6 and N6 were not significantly different with each other.For different traits,the correlation coeffecients showed great variations:seed yield＞pod number,biomass yield and seed weight＞seed number per pod and branch number＞flowering time and maturity time.This result was well accordant with the heterosis level of these traits.In general,these correlation coefficients were all low,this indicated that heterosis could not be exactly predicted by molecular marker heterozygosity.Detection of QTL and mode of inheritance:A total of 1022 QTL（614 and 408 respectively at P=0.05 and 0.5） were detected for the nine traits from two populations and 10 microenvironments using WinQTLCart 2.5 software.After deleting 152 non-overlapping QTL at P=0.5,a total of 870 QTL were identified for the nine traits（then named as "identified-QTL"）,of which nearly 10%（85） were for seed yield.The identified-QTL alleles that increased flowering and maturity times but decreased seed weight came mostly from Tapidor,and that increased branches,bioamass yield,plant height,pod number,seed number and seed yield came basically equally from the two parents,which accorded well with the performance of the two parents for these traits.For RC-F₂ population,the proportion of partial-dominant QTL was the highest,followed by additive QTL,the least was dominant and over-dominant QTL.For seed yield,more QTL showed dominant and over-dominant mode,then followed by pod number and biomass yield,followed by seed number,seed weight,branch number and plant height,the least was flowering time and maturity time:no over-dominant QTL was found.Meta-analysis of identified-QTL in different experiments and expression response of consensus-QTL in natural environments:Where the confidence intervals of identified-QTL for each trait in different experiments overlapped,the 694 overlapping identified-QTL were integrated into 225 reproducible "consensus-QTL" by meta-analysis using BioMercator 2.1 software,with reduced confidence intervals from an average of 6.5 to 3.7 cM.The other 176 non-overlapping identified-QTL were not involoved in the "actual" integration.The total 401 consensus-QTL were classified into two types:15 major QTL（those occurring at least once with R²≥20%or at least twice with R²≥10%） and 386 minor QTL（the remainder with relatively small effect）.Nearly half consensus-QTL,particularly those of seed yield,was specifically detected in one of the 10 microenvironments and few were detected in more than half microenvironments.Two thirds of the consensus-QTL was detected in either the winter or semi-winter macroenvironment,and only one third appeared in both.The high proportion of environment-specific QTL indicates the large impact the natural environment has on the genes underlying seed yield and yield-associated traits.Meta-analysis of consensus-QTL for different traits and the identification of indicator-QTL:Most of the consensus-QTL determined for each trait overlapped with those determined for other traits,the 329 overlapping consensus-QTL were integrated into 111 pleiotropic "unique-QTL" by meta-analysis again,with reduced average putative confidence interval of 2.5 cM.Another 72 consensus-QTL were not invooved into the "actual" integration.The 55 consensus-QTL for seed yield were classified into two types: （ⅰ） 8 non-overlapping QTL,and（ⅱ） 47 overlapping QTL,which were integrated into pleiotropic unique-QTL with shorter average confidence interval of 2.5 cM.Additionally, six of eight "independent" QTL for seed yield overlapped other QTL for traits not published in this paper.Seed yield showed the highest correlation with the other 8 traits, and a high percentage（85%） of QTL for seed yield colocalized with QTL for other yield-associated traits,which indicated that the QTL of yield-associated traits were potential contributors to the colocalized QTL for seed yield.To estimate which QTL of yield-associated trait（s） were more likely to have pleiotropic effect on seed yield at a particular locus,the idea of indicator-QTL was proposed to identify the probable genetic determinant of the colocalized QTL for seed yield.The indicator-QTL are determined by their larger LOD scores,better reproducibility,overlapping confidence intervals and presence in common environment（s） with the QTL for seed yield.Twenty-nine of 47 pleiotropic QTL for seed yield colocalized with more than two consensus-QTL,which indicated that,in addition to pleiotropy,the effect of the QTL for seed yield could be a synthetic effect of several underlying tightly linked QTL of different yield-associated traits.Multiple indicator-QTL were chosen for 14 pleiotropic QTL for seed yield and a total of 63 indicator-QTL were determined.Multiple indicator-QTL came from the same or different environments,indicating the synthesis,variability and plasticity of QTL for yield which was strongly dependent on the environmental conditions.Indicator-QTL, from easily measured traits,usually had more stable expression,higher LOD scores,larger R² values,and identifiable candidate genes than the colocalized QTL for seed yield and thus will facilitate the cloning of yield QTL,assuming that pleiotropy and not linkage is the cause of the colocalization of the indicator-QTL with the corresponding QTL for seed yield.There is indirect evidence that supports the idea that pleiotropy is likely to be the genetic cause of the co-localization of indicator-QTL and seed yield QTL:（1） the peak positions of the two kinds of QTL are very close to each other;（2） the proportion of the overlapping confidence intervals of the two kinds of QTL is very high;（3） the additive-effect directions of the seed yield QTL are generally opposite（-） to that of indicator-QTL for two traits（flowering time and maturity time） while mostly the same（+） for the other six traits,which accorded well with the signs of the genetic-correlation coefficients of seed yield with the above eight yield-associated traits;（4） in many cases, the additive effect of the seed yield QTL was very near to the putative change in the value of seed yield that was estimated from the additive effect of the indicator-QTL.Detection,distribution,type,effect and environmental response of epistatic interactions:A total of 538 E-QTL were found in two populations in all microenvironments for nine traits,most of them did not show significant main-effect.The loci involved into E-QTL showed uneven distribution in the genome with significant hotspot that was identified in different microenvironments.Only 1%of the E-QTL was reproducible and this proportion was significantly lower than that of maize and rice（10%）. This reflected the high variability and plasticity of the epistatic interactions in polyploid. For the E-QTL identified in RC-F₂ population,the total number and R² of AA interaction was the largest,followed by AD/DA interaction,the least was DD interaction.For the M-QTL,dominant effect was responsibole for a minor part of variance explained.For seed yield,the variance explained by E-QTL was higher than that of M-QTL.This strongly suggested that epistasis,especially AA epistasis,was the major genetic basis of heterosis in Brassica napus.Double homozygote was most likely to be the best and worst genotype.Double heterozygote was most unlikely be the best and worst genotype.Our research has great significance for science and application:（1） We identified hundreds of QTL for yield traits and their heterosis,more than half of them were confirmed by different experiments and provided an important genetic resource for marker-assisted selection,QTL cloning and gene function analysis.（2） The mode of expression of QTL for yield traits and heterosis in natural environments was revealed by meta-analysis and this may have significant implications for crop genetics and breeding. （3） The complexity of the genetic architecture of yield and heterosis was demonstrated by meta-analysis,illustrating the pleiotropy,synthesis,variability,and plasticity of yield QTL and the importance,high variability and plasticity of epistasis.（4）The idea of estimating indicator-QTL for yield QTL and identifying potential candidate genes for yield not only deepens our understanding of the genetic determinants of yield and facilitates the cloning of yield QTL but also provides an advance in methodology for complex traits.更多还原