玉米穗長一般配合力多位點(diǎn)全基因組關(guān)聯(lián)分析和預(yù)測

馬 娟 朱衛(wèi)紅 劉京寶 宇 婷 黃 璐 郭國俊

玉米穗長一般配合力多位點(diǎn)全基因組關(guān)聯(lián)分析和預(yù)測

馬 娟*朱衛(wèi)紅 劉京寶 宇 婷 黃 璐 郭國俊

河南省農(nóng)業(yè)科學(xué)院糧食作物研究所, 河南鄭州 450002

穗長是一個重要的農(nóng)藝性狀, 與產(chǎn)量密切相關(guān)。一般配合力(general combining ability, GCA)是評價優(yōu)異自交系的重要指標(biāo)。因此, 解析穗長GCA的遺傳基礎(chǔ), 制定相應(yīng)的育種策略對玉米雜交種產(chǎn)量的提高具有重要意義。本研究以123個玉米自交系和8個測驗(yàn)種按照North Carolina II遺傳交配設(shè)計(jì)組配的537個F1雜交種為試驗(yàn)材料, 在2個環(huán)境下進(jìn)行表型鑒定, 利用玉米5.5 K液相育種芯片鑒定的11,734個SNP (single nucleotide polymorphisms)對2個環(huán)境以及綜合環(huán)境穗長GCA進(jìn)行多位點(diǎn)全基因組關(guān)聯(lián)分析(multi-locus genome-wide association study, MGWAS)和基因組預(yù)測。利用7種MGWAS共檢測到11個穗長GCA顯著關(guān)聯(lián)SNP標(biāo)記(< 8.52E-07), 單個位點(diǎn)解釋GCA變異介于8.06%～28.23%之間。不同MGWAS共定位的SNP位點(diǎn)有5個。位點(diǎn)7_178103602在周口和綜合環(huán)境利用mrMLM (multi-locus random-SNP-effect mixed linear model)方法重復(fù)檢測到, 可解釋穗長GCA變異的26.02%～28.23%, 為環(huán)境穩(wěn)定的主效SNP。共挖掘10個候選基因, 其中和EID1-like F-box protein 2可能是控制穗長GCA的關(guān)鍵基因。5種隨機(jī)效應(yīng)模型對3個環(huán)境穗長GCA的預(yù)測準(zhǔn)確性介于0.53～0.69之間, 且模型間差異較小。在新鄉(xiāng)和周口環(huán)境, GBLUP (genomic best linear unbiased prediction)和RKHS (reproducing kernel Hilbert space)整合不同顯著位點(diǎn)作為固定效應(yīng)均可提高穗長GCA基因組估計(jì)育種值的準(zhǔn)確性, 提高率為2.34%～14.98%, 而在綜合環(huán)境中除了利用FarmCPU (fixed and random model circulating probability unification)或BLINK (Bayesian-information and linkage-disequilibrium iteratively nested keyway)鑒定的1個顯著位點(diǎn)作為固定效應(yīng)會略降低預(yù)測精度外, 其他2種MGWAS方法顯著位點(diǎn)的加入均能提高基因組預(yù)測力, 提高率為2.80%～6.84%。因此, MGWAS顯著位點(diǎn)作為固定效應(yīng)加入預(yù)測模型有利于提高穗長GCA基因組估計(jì)育種值的準(zhǔn)確性, 可用來對玉米親本穗長GCA進(jìn)行有效預(yù)測和選擇。

穗長; 一般配合力; 多位點(diǎn)全基因組關(guān)聯(lián)分析; 固定效應(yīng)模型; 基因組選擇

玉米是雜種優(yōu)勢利用最典型的作物之一, 主要體現(xiàn)在單交種的生產(chǎn)和利用上。自交系配合力鑒定和評價是玉米雜交育種的一個重要環(huán)節(jié)。配合力包括一般配合力(general combining ability, GCA)和特殊配合力(special combining ability, SCA)。GCA是評價一個自交系與其他自交系雜交后代在某個性狀上的平均表現(xiàn), 而SCA是評價雜交組合組配優(yōu)劣的參考指標(biāo)。相比SCA, GCA由加性效應(yīng)控制, 可以穩(wěn)定遺傳。因此, 研究GCA的遺傳機(jī)制對選育高配合力的親本材料更具有實(shí)際應(yīng)用價值。直觀上玉米果穗長度即穗長決定行粒數(shù)的多少, 而行粒數(shù)是產(chǎn)量構(gòu)成因子穗粒數(shù)的決定因子之一。通過基因克隆, 張人予[1]發(fā)現(xiàn)穗長基因在不顯著改變玉米株型和穗型的基礎(chǔ)上, 可以顯著增加穗粒數(shù); Jia等[2]發(fā)現(xiàn)一個編碼絲氨酸/蘇氨酸蛋白激酶的行粒數(shù)基因調(diào)控雌蕊小花數(shù)和穗長。相比單株或小區(qū)產(chǎn)量,穗長的遺傳力較高, 可作為間接選擇性狀。因此, 揭示穗長GCA的遺傳機(jī)制對玉米產(chǎn)量遺傳改良具有重要意義。

連鎖分析和全基因組關(guān)聯(lián)分析(genome-wide association study, GWAS)是挖掘穗長GCA關(guān)鍵位點(diǎn)和解析其遺傳機(jī)理的重要方法。周廣飛[3]利用玉米F2:3家系和其測交群體在7號染色體鑒定到6個控制穗長GCA效應(yīng)的遺傳位點(diǎn), 其中5個也是控制穗長本身的遺傳位點(diǎn)。Liu等[4]利用194個重組自交系與鄭58和B73、HD568和Mo17以及所有4個測驗(yàn)種組配的3個North Carolina II (NCII)雜交群體挖掘到13個控制穗長GCA的QTLs (quantitative trait loci), 其中位于4號染色體上(35.76～62.54 Mb)的1個主效QTL在3個雜交群體中均檢測到。目前, 利用單位點(diǎn)GWAS方法對穗長GCA開展了全基因組鑒定研究。監(jiān)立強(qiáng)[5]以248份玉米自交系和其組配的400個F1雜交組合為試驗(yàn)材料, 利用MLM (mixed linear model)的Q (群體結(jié)構(gòu)) + K (親緣關(guān)系)模型挖掘到13個穗長GCA顯著位點(diǎn), 其變異解釋率均大于10%。劉文童等[6]利用單位點(diǎn)方法SUPER (settlement of MLMs under progressively exclusive relationship)檢測到3個控制玉米穗長GCA的顯著關(guān)聯(lián)位點(diǎn), 其解釋GCA效應(yīng)的變異率較低, 為0.01%～4.34%, 賴氨酸和組氨酸特異性轉(zhuǎn)運(yùn)體為穗長GCA的候選基因。監(jiān)立強(qiáng)[5]和劉文童等[6]研究中沒有找到穗長本身和GCA一致的SNPs位點(diǎn), 表明GCA的遺傳基礎(chǔ)與性狀本身的遺傳基礎(chǔ)可能存在差異。以上研究明確了穗長GCA效應(yīng)數(shù)量性狀遺傳的本質(zhì), 為深入剖析其遺傳基礎(chǔ)提供了豐富的遺傳學(xué)信息。

由于穗長GCA是數(shù)量性狀, 受多基因控制, 因此采用單位點(diǎn)模型對其進(jìn)行遺傳解析具有一定的局限性。多位點(diǎn)GWAS (multi-locus GWAS, MGWAS)方法考慮相鄰位點(diǎn)的潛在關(guān)系[7], 比單位點(diǎn)模型更能解釋多基因性狀的遺傳基礎(chǔ)。針對單位點(diǎn)模型的局限性, 大量多位點(diǎn)GWAS模型被開發(fā)。為了解決群體結(jié)構(gòu)、親緣關(guān)系和候選標(biāo)記間的混雜問題, Liu等[8]提出了一種交替使用固定效應(yīng)模型和隨機(jī)效應(yīng)模型的多位點(diǎn)GWAS模型即FarmCPU (fixed and random model circulating probability unification), 其中固定效應(yīng)模型利用最大似然法選擇潛在關(guān)聯(lián)位點(diǎn),而隨機(jī)效應(yīng)模型則利用bin方法預(yù)測優(yōu)化顯著關(guān)聯(lián)位點(diǎn)。相比FarmCPU算法, BLINK (linkage-disequilibrium iteratively nested keyway)不僅利用連鎖不平衡代替bin方法提高統(tǒng)計(jì)功效, 還利用固定效應(yīng)模型的貝葉斯信息指標(biāo)代替隨機(jī)效應(yīng)模型中的最大似然法來提高運(yùn)算速度[9]。mrMLM (multi-locus random-SNP-effect mixed linear model)首先利用單標(biāo)記掃描策略, 根據(jù)寬松的閾值選定潛在關(guān)聯(lián)SNP, 其次將潛在SNP擬合進(jìn)多位點(diǎn)模型, 通過經(jīng)驗(yàn)貝葉斯估計(jì)和似然比測驗(yàn)最終獲得顯著關(guān)聯(lián)的SNP[10]。為了提高mrMLM方法的運(yùn)算速度, Tamba和Zhang[11]提出了FASTmrMLM (FAST multi-locus random-SNP-effect mixed linear model), 其核心是基于全基因組有效混合模型關(guān)聯(lián)算法、矩陣轉(zhuǎn)化和等式檢測顯著標(biāo)記。FASTmrEMMA (FAST multi-locus random-SNP-effect EMMA)的原理是將SNPs作為隨機(jī)效應(yīng), 并對多基因矩陣K和環(huán)境噪聲協(xié)方差矩陣進(jìn)行漂白處理, 并指定非零特征值的個數(shù)等于1[12]。pLARmEB (polygene-background-control-based least angle regression plus empirical Bayes)主要對FASTmrEMMA算法中的矩陣變換進(jìn)行拓展[13]。ISIS EM-BLASSO (iterative sure independence screening EM-Bayesian LASSO)是利用安全獨(dú)立篩選-平滑削邊絕對偏差懲罰(SIS-SCAD)算法篩選潛在SNPs標(biāo)記, EM-BLASSO算法和似然比測驗(yàn)獲得顯著SNP[14]。目前, 多位點(diǎn)GWAS被廣泛應(yīng)用于玉米等作物重要農(nóng)藝性狀的遺傳研究中[15-19], 但對玉米穗長GCA的研究鮮見報道。

盡管連鎖和關(guān)聯(lián)分析研究揭示了控制數(shù)量性狀的關(guān)鍵位點(diǎn)和候選基因, 但如何利用這些關(guān)鍵位點(diǎn)以及針對目標(biāo)性狀制定相應(yīng)的分子育種策略仍是困擾育種工作者的一個關(guān)鍵問題?；蚪M選擇(genomic selection, GS)是目前主流的一種分子育種技術(shù), 最早由Meuwissen等[20]提出。GS能夠利用影響性狀的所有變異位點(diǎn)估計(jì)育種值并進(jìn)行有效選擇, 可通過縮短育種周期來大幅度提高遺傳進(jìn)度。為提高GS的準(zhǔn)確性, 研究者發(fā)展了很多統(tǒng)計(jì)模型和算法例如基因組最佳線性無偏預(yù)測(genomic best linear unbiased prediction, GBLUP)、貝葉斯模型和非參數(shù)模型等[20-23], 其目的均是通過有效降維從而實(shí)現(xiàn)對標(biāo)記效應(yīng)的準(zhǔn)確估計(jì)。目前, GS廣泛應(yīng)用于玉米等作物親本自交系和雜交種產(chǎn)量等性狀表型預(yù)測[24-27]。而且, 利用GS對親本材料產(chǎn)量性狀GCA效應(yīng)也開展了相關(guān)預(yù)測研究。例如, Wang等[28]通過模擬和真實(shí)數(shù)據(jù)提出利用稀疏部分雙列雜交設(shè)計(jì)可以有效預(yù)測玉米單穗重的GCA效應(yīng)。Zhang等[29]利用32個玉米自交系和9個測驗(yàn)種組配的3組雜交種數(shù)據(jù)對產(chǎn)量GCA效應(yīng)進(jìn)行了預(yù)測。研究發(fā)現(xiàn), 自交系和測驗(yàn)種聯(lián)合預(yù)測的效果最好。目前利用GS策略對穗長GCA進(jìn)行基因組預(yù)測研究尚未見報道。

為進(jìn)一步解析玉米穗長GCA的遺傳機(jī)制, 提出有效的分子育種改良策略, 本研究選用123個玉米自交系和8個測驗(yàn)種按照NCII遺傳交配設(shè)計(jì)獲得的537個F1雜交組合為試驗(yàn)材料, 采用7種MGWAS方法挖掘穗長GCA顯著關(guān)聯(lián)位點(diǎn), 并利用GBLUP和RKHS (reproducing kernel Hilbert space)研究顯著關(guān)聯(lián)位點(diǎn)作為固定效應(yīng)對提高穗長GCA預(yù)測精度的影響。

1 材料與方法

1.1 試驗(yàn)材料、田間設(shè)計(jì)及配合力分析

根據(jù)NCII遺傳交配設(shè)計(jì), 利用123個玉米自選系與8個測驗(yàn)種(M189、M119、20H1419、S110T、L119A、PH4CV、昌7-2和農(nóng)系531)組配了537個F1雜交種。測驗(yàn)種昌7-2、PH4CV、M189和農(nóng)系531分別是玉米雜交品種鄭單958的父本、先玉335的父本、鄭單309的父本和農(nóng)單5316的母本, 其他4個為自選系。537個F1雜交子代于2021年種植在河南新鄉(xiāng)和周口試驗(yàn)田。采用隨機(jī)區(qū)組試驗(yàn)設(shè)計(jì), 兩次重復(fù), 每個材料每小區(qū)均種植1行。新鄉(xiāng)試驗(yàn)田小區(qū)行長、株距和行距分別是4.00、0.22和0.60 m, 而周口試驗(yàn)田小區(qū)行長、株距和行距分別是3.30、0.22和0.60 m。收獲后, 每小區(qū)每個材料選取單穗測量穗長(cm)。根據(jù)537個F1雜交子代穗長表型, 利用R語言lme4包分別計(jì)算新鄉(xiāng)和周口環(huán)境131個玉米自交系穗長的GCA效應(yīng)值, 同時將兩環(huán)境聯(lián)合計(jì)算,獲得綜合環(huán)境穗長GCA值。根據(jù)下面公式計(jì)算穗長GCA的遺傳力:

1.2 基因型鑒定及分析

利用河南省農(nóng)業(yè)科學(xué)院糧食作物研究所開發(fā)的玉米5.5 K液相育種芯片(5521個靶向探針位點(diǎn))對131個玉米自交系進(jìn)行基因型鑒定, 測序平臺為Illumina NovaSeq 6000。利用BWA軟件將過濾的reads與玉米B73第4版參考基因組(http://www. gramene.org/)進(jìn)行比對。利用GATK v4.1.2.0軟件檢測到33,971個原始SNPs標(biāo)記。將最小等位基因頻率<0.05, 缺失率>10%和雜合率>1%的SNPs標(biāo)記過濾后, 共獲得11,734個SNP用于后續(xù)關(guān)聯(lián)分析和基因組預(yù)測分析。

1.3 顯著位點(diǎn)和候選基因的挖掘

多數(shù)研究表明, 相比只考慮群體結(jié)構(gòu)Q或是親緣關(guān)系K的GWAS模型, 同時考慮兩者的GWAS模型具有較好的擬合性[30-32]。因此, 本研究利用7種MGWAS方法(BLINK、FarmCPU、FASTmrMLM、FASTmrEMMA、mrMLM、pLARmEB和ISIS EM- BLASSO)的Q+K模型對新鄉(xiāng)、周口和綜合環(huán)境穗長GCA進(jìn)行關(guān)聯(lián)分析。利用Structure v2.3.4[33]軟件計(jì)算群體結(jié)構(gòu)Q值, 其中亞群數(shù)設(shè)置為1～10, Burnin期的長度設(shè)置為5000, 蒙特卡羅重復(fù)個數(shù)設(shè)為50,000, 每個亞群數(shù)重復(fù)3次。根據(jù)Δ的結(jié)果, 確定最佳的亞群數(shù)為6。根據(jù)Structure軟件結(jié)果, 利用CLUMPP軟件的FullSearch方法[34]獲得最終值。親緣關(guān)系值采用TASSEL v5.0[35]軟件的Centered_ IBS算法確定。除了BLINK和FarmCPU利用GAPIT R包[36]計(jì)算, 其他方法均采用mrMLM R包[37]計(jì)算。利用< 8.52E-07 (0.01/11,734)確定標(biāo)記與穗長GCA關(guān)聯(lián)的顯著性。利用SnpEff[38]對顯著位點(diǎn)的候選基因信息進(jìn)行挖掘, 參數(shù)按默認(rèn)設(shè)置。利用MaizeGDB數(shù)據(jù)庫中qTeller工具獲取預(yù)測基因以及3個已知穗長基因([2]、[39]和[40])在B73等自交系不同組織的表達(dá)量。根據(jù)表達(dá)量, 利用R語言hclust函數(shù)對候選基因和已知基因進(jìn)行聚類, 方法為最長距離法。通過R語言的GENIE3包的隨機(jī)森林算法推測基因之間的互作網(wǎng)絡(luò), 互作關(guān)系權(quán)重閾值設(shè)置為0.1?；蜷g的網(wǎng)絡(luò)關(guān)系采用Cytoscape v3.9.1展示。

1.4 基因組選擇模型和策略

利用Bayes A、Bayes C、貝葉斯最小絕對縮減和變量選擇算子(Bayesian least absolute shrinkage and selection operator, Bayesian LASSO)、GBLUP和RHKS多核模型對2個環(huán)境以及綜合環(huán)境穗長GCA進(jìn)行基因組預(yù)測分析。5種預(yù)測方法中所有標(biāo)記的效應(yīng)均為隨機(jī)效應(yīng), 定義該模型為隨機(jī)效應(yīng)模型。為研究MGWAS顯著位點(diǎn)對GCA預(yù)測精度的影響, 本研究構(gòu)建了GBLUP和RKHS預(yù)測方法的固定效應(yīng)模型。在固定效應(yīng)模型中, 根據(jù)不同MGWAS方法挖掘的關(guān)聯(lián)位點(diǎn)信息, 將顯著SNPs標(biāo)記設(shè)定為固定效應(yīng), 其余標(biāo)記設(shè)定為隨機(jī)效應(yīng)。采用10倍交叉驗(yàn)證方式將131個自交系分為訓(xùn)練集和驗(yàn)證集, 重復(fù)100次。評價不同模型的指標(biāo)為驗(yàn)證集基因組估計(jì)育種值與表型值的相關(guān)系數(shù)均值。所有模型和方案均在BGLR R包中實(shí)現(xiàn)[41], 其中蒙特卡洛馬爾科夫鏈長為12,000, 預(yù)燒為3000, 其他參數(shù)按默認(rèn)設(shè)置。預(yù)測準(zhǔn)確性方差分析和Duncan多重比較均采用R語言計(jì)算。

2 結(jié)果與分析

2.1 穗長一般配合力效應(yīng)值統(tǒng)計(jì)結(jié)果

新鄉(xiāng)、周口和綜合環(huán)境穗長GCA效應(yīng)值分別介于1.45～1.95 cm、0.71～0.82 cm和1.24～1.33 cm (圖1-A)。新鄉(xiāng)和周口環(huán)境之間表現(xiàn)為中度正相關(guān)關(guān)系(= 0.57), 均與綜合環(huán)境高度正相關(guān)(= 0.87～0.89) (圖1-B)。方差分析表明, 自交系和測驗(yàn)種的GCA、SCA、GCA或SCA與環(huán)境互作效應(yīng)均達(dá)到顯著或極顯著水平(表1)。穗長GCA的遺傳力較高, 為0.82。這些結(jié)果表明, 雖然穗長GCA主要受遺傳因素控制, 但同時也受到環(huán)境因素的影響。

2.2 穗長一般配合力顯著關(guān)聯(lián)位點(diǎn)和候選基因

利用7種MGWAS方法共檢測到11個穗長GCA顯著關(guān)聯(lián)SNP (< 8.52E-07), 其解釋GCA變異率介于8.06%～28.23%之間(表2)。新鄉(xiāng)、周口和綜合環(huán)境均檢測到4個顯著關(guān)聯(lián)位點(diǎn)。其中, 位點(diǎn)7_178103602在周口和綜合環(huán)境利用mrMLM方法重復(fù)檢測到, 其解釋穗長GCA變異率為26.02%～ 28.23%, 為環(huán)境穩(wěn)定的主效SNP。不同MGWAS方法共定位的SNP位點(diǎn)有5個。在新鄉(xiāng)環(huán)境, 位點(diǎn)2_216138581和8_126983650利用FASTmrEMMA和pLARmEB均檢測到, 其能解釋穗長GCA表型變異的15.63%～23.09%, 為控制穗長GCA的主效位點(diǎn)。7_178327031為FASTmrMLM和pLARmEB方法共定位的SNP, 可以解釋周口環(huán)境穗長GCA變異的10.64%～11.29%, 也是控制穗長GCA的主效位點(diǎn)。

利用SnpEff共挖掘到穗長GCA候選基因10個, 有注釋信息的基因?yàn)?個(表2)。環(huán)境穩(wěn)定的主效位點(diǎn)7_178103602挖掘的候選基因?yàn)榫幋a過氧化物酶的()。不同MGWAS方法共定位的2個主效位點(diǎn)7_178327031和2_216138581對應(yīng)的候選基因分別為編碼生長素氨基合成酶的()和編碼類EID1 F-框蛋白的EID1-like F-box protein 2 (EDL2)。利用MaizeGDB數(shù)據(jù)庫中qTeller工具獲得了10個候選基因以及3個已知穗長基因(、和)在B73等自交系不同組織的表達(dá)量(附表1)。相比其他組織,和ELD2分別在胚和成熟雌穗小花中高表達(dá)(附表1)。聚類結(jié)果表明,與的歐式距離較近, 兩者可能具有相似的表達(dá)模式; ELD2也與劃分為一大類(圖2-A)。GENIE3預(yù)測的基因互作網(wǎng)絡(luò)結(jié)果表明, ELD2與、可能存在互作關(guān)系, 而與、相互關(guān)聯(lián)(圖2-B)。

圖1 穗長一般配合力分布(A)和不同環(huán)境相關(guān)性(B)

圖A中×表示均值。圖B中***表示在0.001概率水平差異顯著。

In Fig. 1-A, × represents means. In Fig. 1-B, *** represents there is significant difference at the 0.001 probability level.

表1 方差分析和穗長一般配合力的遺傳力

P1和P2分別表示自交系和測驗(yàn)種。

P1 and P2 represent the inbred lines and tester lines, respectively.

表2 穗長一般配合力顯著關(guān)聯(lián)SNP和候選基因

2為單個位點(diǎn)表型變異解釋率。2represents phenotypic variance explained by one locus.

2.3 不同隨機(jī)效應(yīng)模型的預(yù)測準(zhǔn)確性

利用5種GS隨機(jī)效應(yīng)模型對新鄉(xiāng)、周口和綜合環(huán)境穗長GCA進(jìn)行了基因組預(yù)測分析。方差分析和Duncan多重比較分析表明, 3個環(huán)境之間的預(yù)測準(zhǔn)確性差異顯著(附表2和附表3)。綜合環(huán)境穗長GCA的預(yù)測準(zhǔn)確性最高, 為0.67～0.69, 其次為新鄉(xiāng)環(huán)境(0.60～0.61), 周口環(huán)境的準(zhǔn)確性最低, 為0.53～0.54 (圖3)。5種隨機(jī)效應(yīng)模型對穗長GCA效應(yīng)估計(jì)育種值的準(zhǔn)確性差異較小且不顯著(附表2), 最大差值僅為0.02。3個環(huán)境中, Bayesian LASSO的基因組預(yù)測能力均最低, 而其他4個模型的預(yù)測準(zhǔn)確性基本相等。

圖3 5種隨機(jī)效應(yīng)模型對穗長一般配合力的預(yù)測準(zhǔn)確性

2.4 多位點(diǎn)全基因組關(guān)聯(lián)分析先驗(yàn)信息對預(yù)測準(zhǔn)確性的影響

考慮到模型之間的差異較小以及Bayes模型的運(yùn)算時間較長, 本研究僅利用GBLUP和RKHS方法研究了7種MGWAS方法挖掘的顯著SNP位點(diǎn)作為固定效應(yīng)對提高穗長GCA基因組預(yù)測力的影響。除了新鄉(xiāng)環(huán)境外, 周口和綜合環(huán)境不同模型之間預(yù)測準(zhǔn)確性存在顯著差異(附表4和附表5)。對于新鄉(xiāng)和周口環(huán)境, GBLUP和RKHS整合不同顯著位點(diǎn)作為固定效應(yīng)均可提高穗長GCA基因組估計(jì)育種值的準(zhǔn)確性, 提高率為2.34%～14.98% (圖4-A, B)。在綜合環(huán)境中, 利用FarmCPU或BLINK鑒定的1個顯著位點(diǎn)作為固定效應(yīng)會略降低預(yù)測的準(zhǔn)確性, 其他2種MGWAS方法均能提高預(yù)測力, 提高率為2.80%～6.84% (圖4-C)。

3 討論

3.1 顯著閾值和多位點(diǎn)GWAS模型的選擇

近年來, MGWAS方法在植物遺傳研究中得到廣泛應(yīng)用, 其在多重檢驗(yàn)、群體結(jié)構(gòu)和多基因背景控制方面的優(yōu)勢也逐漸凸顯出來[7,42]。在MGWAS方法中, 所有潛在關(guān)聯(lián)標(biāo)記和效應(yīng)能夠在一個線性模型中同時確定并估計(jì)出來, 因此無需進(jìn)行Bonferroni校正[7,42]。mrMLM軟件包的開發(fā)者建議將LOD = 3 (= 0.0002)作為多位點(diǎn)模型顯著位點(diǎn)的臨界值。由于GCA效應(yīng)值有正負(fù)之分, 相比性狀本身, 其變異較大, 因此本研究仍選擇了0.01水平下的Bonferroni矯正來控制假陽性率。盡管設(shè)置了嚴(yán)格的閾值, 本研究利用7種多位點(diǎn)模型仍檢測到11個控制穗長GCA的顯著關(guān)聯(lián)位點(diǎn)。這些結(jié)果表明, 多位點(diǎn)模型在檢測GCA關(guān)鍵位點(diǎn)方面具有較高的檢測功效。

本研究中pLARmEB方法對穗長GCA的檢測功效最高, FASTmrMLM最保守。Yang等[15]發(fā)現(xiàn)pLARmEB對小麥籽粒品質(zhì)性狀和面團(tuán)流變特性的檢測功效高于mrMLM和FASTmrEMMA, 但低于FASTmrMLM。對于水稻耐鹽性多位點(diǎn)GWAS方法研究, pLARmEB的檢測功效次于ISIS-BLASSO和FASTmrMLM, 但高于mrMLM和FASTmrEMMA[18]。Zhou等[19]采用6種多位點(diǎn)GWAS方法對玉米成熟期籽粒含水量進(jìn)行定位研究得出, ISIS-BLASSO檢測的位點(diǎn)個數(shù)最多, pLARmEB方法相對保守。An等[43]發(fā)現(xiàn)在多個環(huán)境中mrMLM模型對玉米穗行數(shù)均具有最高的檢測功效, pLARmEB居中, FASTmrEMMA最保守。由于不同多位點(diǎn)GWAS方法在統(tǒng)計(jì)模型、潛在關(guān)聯(lián)位點(diǎn)選擇策略以及顯著位點(diǎn)檢驗(yàn)方法等方面的差異, 可能導(dǎo)致了其對不同性狀檢測功效的不同。多數(shù)研究表明, 利用多種GWAS方法有助于挖掘穩(wěn)定的變異位點(diǎn)[15-19]。本研究發(fā)現(xiàn)不同模型共定位的位點(diǎn)有5個, 而且多為控制穗長GCA的主效位點(diǎn)。因此, 利用多種MGWAS方法有助于提高位點(diǎn)檢測的可靠性, 為后續(xù)功能驗(yàn)證提供可靠的基因信息。

圖4 GBLUP和RKHS整合顯著位點(diǎn)作為固定效應(yīng)的預(yù)測準(zhǔn)確性

A、B和C分別表示新鄉(xiāng)、周口和綜合環(huán)境。GBLUP和RKHS表示隨機(jī)效應(yīng)模型。GBLUP和RKHS作為后綴的模型表示固定效應(yīng)模型。

A, B, and C represent Xinxiang, Zhoukou, and combined environment, respectively. GBLUP and RKHS represent random effect models. Models with GBLUP and RKHS as suffixes represent fixed effect models.

3.2 不同定位研究結(jié)果比較

本研究中絕大多數(shù)位點(diǎn)對穗長GCA變異的解釋率大于10%, 這與監(jiān)立強(qiáng)[5]利用單位點(diǎn)混合線性模型得出的研究結(jié)果一致。檢測的9個顯著位點(diǎn)與前人利用雙親群體定位的穗長、行粒數(shù)、產(chǎn)量和單株產(chǎn)量QTL或MQTL存在重疊。位點(diǎn)1_192956360同時被BLINK和FarmCPU方法檢測到, 其位于Zhou等[44]利用四交群體定位的一個控制穗長的QTL置信區(qū)間內(nèi)。位于7號染色體上的3個顯著位點(diǎn)均位于一個測交群體定位的單株產(chǎn)量QTL[45]和一個對產(chǎn)量和單穗重具有多效性的QTL區(qū)間內(nèi)[46]。3個位點(diǎn)(6_84476172、7_106761824、8_126983650)均位于Chen等[47]利用Meta-QTL整合的產(chǎn)量、穗部性狀和籽粒性狀的MQTL內(nèi)。位點(diǎn)10_149653708位于一個控制行粒數(shù)和穗行數(shù)的MQTL區(qū)間內(nèi)[48]。位點(diǎn)5_194917385與一個同時控制產(chǎn)量、單穗重的QTL和一個控制小區(qū)產(chǎn)量GCA的QTL存在重疊[49-50]。不同群體間重疊的基因組區(qū)域說明影響穗長GCA的關(guān)聯(lián)位點(diǎn)可能對性狀本身、產(chǎn)量以及行粒數(shù)具有多效性。

3.3 穗長一般配合力候選基因預(yù)測

生長素氨基合成酶和類EID1 F-框蛋白EDL2是不同MGWAS方法共定位主效位點(diǎn)預(yù)測的候選基因。編碼生長素/吲哚乙酸蛋白的基因BARREN INFLORESCENCE1和BARREN INFLORESCENCE4是生長素信號途徑調(diào)控玉米花序形成不可或缺的關(guān)鍵因子[51], 說明生長素在玉米穗部小花的發(fā)育調(diào)控制中具有重要作用。已知穗長和行粒數(shù)基因編碼絲氨酸/蘇氨酸蛋白激酶, 通過介導(dǎo)Arf GTPase-activating protein的磷酸化來調(diào)控生長素依賴的花序發(fā)育, 從而影響玉米穗長和產(chǎn)量[2]。玉米穗長基因通過參與糖和生長素信號途徑影響玉米雌穗花序分生組織的發(fā)育, 進(jìn)而影響玉米穗長和單穗產(chǎn)量[39]。本研究利用一種隨機(jī)森林集成算法發(fā)現(xiàn)與、可能存在互作關(guān)系。擬南芥中, 過表達(dá)和條件敲除突變體研究發(fā)現(xiàn)EDL3是脫落酸依賴的信號級聯(lián)反應(yīng)的正向調(diào)節(jié)因子, 而脫落酸信號傳導(dǎo)途徑在控制種子萌發(fā)、開花轉(zhuǎn)換等發(fā)育過程起著重要作用[52]。穗長基因通過控制玉米花序中內(nèi)源乙烯生物合成水平影響小花敗育率, 從而調(diào)節(jié)玉米穗長和穗粒數(shù)。玉米EDL2是擬南芥EDL3的同源子, 其與穗長基因、存在互作。而且,和EDL2分別在胚和成熟雌穗小花中高表達(dá)(附表1)。因此,和EDL2可能是調(diào)控玉米穗長GCA的關(guān)鍵基因。

3.4 穗長一般配合力全基因組預(yù)測策略

本研究中5種隨機(jī)效應(yīng)模型對玉米穗長GCA具有相似的基因組預(yù)測力, 與玉米產(chǎn)量及產(chǎn)量相關(guān)性以及小麥黃銹病預(yù)測研究結(jié)論一致[53-54]。由于穗長GCA的遺傳力較高, 利用5種隨機(jī)效應(yīng)模型均取得了中等及以上的預(yù)測準(zhǔn)確性。但相比遺傳力, 穗長GCA的預(yù)測精度仍有提高的空間。根據(jù)MGWAS定位的結(jié)果, 本研究將1～2個顯著位點(diǎn)作為固定效應(yīng)加入GBLUP和RKHS模型對穗長GCA效應(yīng)開展了基因組預(yù)測分析。結(jié)果發(fā)現(xiàn), 相比隨機(jī)效應(yīng)模型, 除了綜合環(huán)境FarmCPU或BLINK方法挖掘的1個顯著位點(diǎn)作為固定效應(yīng)不能提高預(yù)測精度外, 其余情況均能提高GCA基因組估計(jì)育種值的準(zhǔn)確性, 提高率為2.34%～14.98%。Ma和Cao[55]將單位點(diǎn)模型CMLM和多位點(diǎn)模型FarmCPU挖掘的4～8個控制玉米穗行數(shù)和穗長的顯著SNPs作為固定效應(yīng)整合到5種預(yù)測模型(GBLUP、RKHS、Bayes A、Bayes B和Bayes C)中也得到了相同的結(jié)論。此外, 將主效的生長習(xí)性基因作為固定效應(yīng)整合到預(yù)測模型也能提高小麥產(chǎn)量的預(yù)測準(zhǔn)確性[56]。Odilbekov等[57]發(fā)現(xiàn)將3個單位點(diǎn)模型(GLM、MLM和SUPER)和2個多位點(diǎn)模型(FarmCPU和MLMM)檢測的1～3個顯著關(guān)聯(lián)位點(diǎn)作為固定效應(yīng)能夠提高小麥葉枯病的預(yù)測力。將連鎖分析定位的顯著QTL作為固定效應(yīng)也可提高性狀本身的預(yù)測準(zhǔn)確性。在玉米BC3F5群體中, 相比完全隨機(jī)效應(yīng)模型, 將1～2個QTL作為固定效應(yīng)整合到6種預(yù)測模型中可以提高穗行數(shù)、穗粒數(shù)、行粒數(shù)和葉夾角基因組估計(jì)育種值的準(zhǔn)確性[58]。Arruda等[59]發(fā)現(xiàn)將控制小麥赤霉病的QTL作為固定效應(yīng)擬合到ridge regression BLUP (RR-BLUP)模型中能夠提高小麥赤霉病的基因組預(yù)測力。通過模擬, Bernardo[60]認(rèn)為將表型變異解釋率大于10%的主效基因作為固定效應(yīng)加入RR-BLUP模型總會提高性狀的預(yù)測力。本研究中除了BLINK和FarmCPU檢測的位點(diǎn)外, 多數(shù)位點(diǎn)對GCA變異的解釋率大于10%, 這可能是固定效應(yīng)模型可以提高穗長GCA預(yù)測力的原因之一。

此外, 研究者還發(fā)現(xiàn)相比隨機(jī)選擇的位點(diǎn), 將GWAS挖掘的顯著位點(diǎn)全部作為隨機(jī)效應(yīng)也可以提高性狀本身的預(yù)測準(zhǔn)確性。對于干旱和高溫環(huán)境的玉米產(chǎn)量和開花期, RR-BLUP利用8～339個顯著關(guān)聯(lián)的SNP獲得的預(yù)測準(zhǔn)確性高于隨機(jī)選擇的覆蓋全基因組的10,108個SNP[61]。在熱帶玉米種質(zhì)中, Liu等[62]發(fā)現(xiàn)僅利用GWAS顯著關(guān)聯(lián)位點(diǎn)作為標(biāo)記集對玉米莖腐病的準(zhǔn)確性高于覆蓋基因組的所有標(biāo)記。在小麥中, Cericola等[63]也認(rèn)為相比隨機(jī)選擇的標(biāo)記, GWAS衍生的標(biāo)記有助于提高產(chǎn)量、倒伏和淀粉含量的預(yù)測準(zhǔn)確性。因此, GS整合MGWAS挖掘的性狀關(guān)聯(lián)標(biāo)記可以提高GCA全基因組預(yù)測的準(zhǔn)確性, 可以用來預(yù)測和選擇穗長配合力較高的親本材料。

4 結(jié)論

本研究利用7種MGWAS共檢測到11個控制穗長GCA的顯著關(guān)聯(lián)SNP, 不同MGWAS共定位的SNP有5個, 環(huán)境穩(wěn)定的有1個。生長素氨基合成酶和類EID1 F-框蛋白ELD2可能是控制穗長GCA的關(guān)鍵基因。5種隨機(jī)效應(yīng)模型對3個環(huán)境穗長GCA的預(yù)測準(zhǔn)確性介于0.53～0.69之間, 可以有效預(yù)測穗長GCA效應(yīng)值。將本研究檢測到的MGWAS顯著位點(diǎn)作為固定效應(yīng)加入預(yù)測模型有利于提高穗長GCA基因組估計(jì)育種值的準(zhǔn)確性。

附表 請見網(wǎng)絡(luò)版: 1) 本刊網(wǎng)站http://zwxb. chinacrops.org/; 2) 中國知網(wǎng)http://www.cnki.net/; 3) 萬方數(shù)據(jù)http://c.wanfangdata.com.cn/Periodical- zuowxb.aspx。

[1] 張人予. 玉米穗長基因EL3的克隆及我國優(yōu)良自交系基因組變異分析. 中國農(nóng)業(yè)大學(xué)博士學(xué)位論文, 北京, 2018. Zhang R Y. Cloning of EL3 for Ear Length in Maize and Patterns of Genomic Variation in Chinese Maize Inbred Lines. PhD Dissertation of China Agricultural University, Beijing, China, 2018 (in Chinese with English abstract).

[2] Jia H T, Li M F, Li W Y, Liu L, Jian Y N, Yang Z X, Shen X M, Ning Q, Du Y F, Zhao R, Jackson D, Yang X H, Zhang Z X. A serine/threonine protein kinase encoding generegulates maize grain yield., 2020, 11: 988–998.

[3] 周廣飛. 一個控制玉米行粒數(shù)、穗長其一般配合力的多效性QTL(qKNR7.2)鑒定. 華中農(nóng)業(yè)大學(xué)碩士學(xué)位論文, 湖北武漢,2014. Zhou G F. Identification of A Pleitropic QTL (qKNR7.2) for Kernel Row Number Per Row, Ear Length, and General Combining Ability of Maize. MS Thesis of Huazhong Agricultural University, Wuhan, Hubei, China, 2014 (in Chinese with English abstract).

[4] Liu X G, Hu X X, Li K, Liu Z F, Wu Y J, Feng G, Huang C L, Wang H W. Identifying quantitative trait loci for the general combining ability of yield-relevant traits in maize., 2021, 71: 217–228.

[5] 監(jiān)立強(qiáng). 玉米產(chǎn)量相關(guān)性狀及其一般配合力的關(guān)聯(lián)分析. 河北農(nóng)業(yè)大學(xué)碩士學(xué)位論文, 河北保定, 2017. Jian L Q. Genome-Wide Association Study of Yield-Related Traits and General Combining Ability in Maize (L.). MS Thesis of Hebei Agricultural University, Baoding, Hebei, China, 2017 (in Chinese with English abstract).

[6] 劉文童, 監(jiān)立強(qiáng), 郭晉杰, 趙永鋒, 黃亞群, 陳景堂, 祝麗英. 玉米穗部性狀及其一般配合力的關(guān)聯(lián)分析. 植物遺傳資源學(xué)報, 2020, 21: 706–715. Liu W T, Jian L Q, Guo J J, Zhao Y F, Huang Y Q, Chen J C, Zhu L Y. Association analysis of ear-related traits and their general combining ability in maize., 2020, 21: 706–715 (in Chinese with English abstract).

[7] 溫陽俊, 馮建英, 張瑾. 多位點(diǎn)關(guān)聯(lián)分析方法學(xué)的研究進(jìn)展. 南京農(nóng)業(yè)大學(xué)學(xué)報, 2022, 45: 1–10. Wen Y J, Feng J, Zhang J. Research progress of mulit-locus genome-wide association study., 2022, 45: 1–10 (in Chinese with English abstract).

[8] Liu X L, Huang M, Fan B, Buckler ES, Zhang Z. Iterative usage of fixed and random effect models for powerful and efficient genome wide association studies., 2016, 12: e1005767.

[9] Huang M, Liu X, Zhou Y, Summers RM, Zhang Z W. BLINK: a package for the next level of genome-wide association studies with both individuals and markers in the millions., 2019, 8: 1–12.

[10] Wang SB, Feng JY, Ren WL, Huang B, Zhou L, Wen YJ, Zhang J, Dunwell JM, Xu S, Zhang YM. Improving power and accuracy of genome-wide association studies via a multi-locus mixed linear model methodology., 2016, 6: 19444–19453.

[11] Tamba CL, Zhang YM. A fast mrMLM algorithm for multi-locus genome-wide association studies., 2018. https://doi.org/10.1101/341784.

[12] Wen YJ, Zhang H, Ni YL, Huang B, Zhang J, Feng JY, Wang SB, Dunwell JM, Zhang YM, Wu R. Methodological implementation of mixed linear models in multi-locus genome-wide association studies., 2018, 19: 700–712.

[13] Zhang J, Feng JY, Ni YL, Wen YJ, Niu Y, Tamba CL, Yue C, Song Q, Zhang YM. pLARmEB: integration of least angle regression with empirical Bayes for multilocus genome-wide association studies., 2017, 118: 517–524.

[14] Tamba CL, Ni YL, Zhang YM. Iterative sure independence screening EM-Bayesian LASSO algorithm for multi-locus genome-wide association studies., 2017, 13: e1005357.

[15] Yang Y, Chai Y M, Zhang X, Lu S, Zhao Z C, Wei D, Chen L, Hu YG. Multi-locus GWAS of quality traits in bread wheat: mining more candidate genes and possible regulatory network., 2020, 11: 1091–1109.

[16] Peng Y C, Liu H B, Chen J, Shi T T, Zhang C, Sun D F, He Z H, Hao Y F, Chen W. Genome-wide association studies of free amino acid levels by six multi-locus models in bread wheat., 2018, 9: 1196–1204.

[17] Su J J, Wang C X, Hao F S, Ma Q, Wang J, Li J L, Ning X Z. Genetic detection of lint percentage applying single-locus and multi-locus genome-wide association studies in Chinese early-maturity upland cotton., 2019, 10: 964–974.

[18] Cui Y R, Zhang F, Zhou Y L. The application of multi-locus GWAS for the detection of salt-tolerance loci in rice., 2018, 9: 1464–1472.

[19] Zhou G F, Zhu Q L, Mao Y X, Chen G Q, Xue L, Lu H H, Shi M L, Zhang Z L, Song X D, Zhang H M, Hao D R. Multi-locus genome-wide association study and genomic selection of kernel moisture content at the harvest stage in maize., 2021, 12: 697688–697700.

[20] Meuwissen T H, Hayes B J, Goddard M E. Prediction of total genetic value using genome-wide dense marker maps., 2001, 157: 1819–1829.

[21] Vanraden P M. Efficient methods to compute genomic predictions., 2008, 91: 4414–4423.

[22] de los Campos G, Naya H, Gianola D, Crossa J, Legarra A, Manfredi E, Weigel K, Cotes J M. Predicting quantitative traits with regression models for dense molecular markers and pedigree., 2009, 182: 375–385.

[23] González-Recio O, Forni S. Genome-wide prediction of discrete traits using bayesian regressions and machine learning., 2011, 43: 7–18.

[24] Guo Z G, Tucker D M, Lu J, Kishore V, Gay G. Evaluation of genome-wide selection efficiency in maize nested association mapping populations., 2012, 124: 261–275.

[25] Lian L, Jacobson A, Zhong S Q. Genome wide prediction accuracy within 969 maize biparental populations., 2014, 54: 1514–1522.

[26] Technow F, Schrag T A, Schipprack W, Bauer E, Simianer H, Melchinger A E. Genome properties and prospects of genomic prediction of hybrid performance in a breeding program of maize., 2014, 197: 1343–1355.

[27] de Oliveira A A, Resende M F R Jr, Ferr?o L F V, Amadeu R R, Guimar?es L J M, Guimar?es C T, Pastina M M, Margarido G R A. Genomic prediction applied to multiple traits and environments in second season maize hybrids., 2020, 125: 60–72.

[28] Wang X, Zhang Z L, Xu Y, Li P P, Xu C W. Using genomic data to improve the estimation of general combining ability based on sparse partial diallel cross designs in maize., 2020, 8: 819–829.

[29] Zhang A, Pérez-Rodríguez P, Vicente F S, Palacios-Rojas N, Dhliwayo T, Liu Y B, Cui Z H, Guan Y, Wang H, Zheng H J, Olsen M, Prasanna B M, Ruan Y Y, Crossa J, Zhang X C. Genomic prediction of the performance of hybrids and the combining abilities for line by tester trials in maize., 2021, 10: 109–116.

[30] 熊雪航, 段海洋, 李文龍, 李建新, 孫莉, 孫巖, 秦永田, 湯繼華, 張雪海. 玉米穗長全基因組關(guān)聯(lián)分析. 分子植物育種, 2022, https://kns.cnki.net/kcms/detail/46.1068.S.20220630.1359. 004.htmlXiong XH, Duan HY, Li WL, Li JX, Sun L, Sun Y, Qin YT, Tang JH, Zhang XH. Genome-wide association study of ear length in maize.,2022, https://kns.cnki.net/ kcms/detail/46.1068.S.20220630.1359.004.html (in Chinese with English abstract).

[31] 秦文萱, 鮑建喜, 王彥博, 馬雅杰, 龍艷, 李金萍, 董振營, 萬向元. 玉米葉夾角性狀的全基因組關(guān)聯(lián)分析與關(guān)鍵位點(diǎn)優(yōu)異等位變異挖掘. 作物學(xué)報, 2022, 48: 2691–2705. Qin WX, Bao JX, Wang YB, Ma YJ, Long Y, Li JP, Dong ZY, Wang XY. Genome-wide association study of leaf angle traits and mining of elite alleles from the major loci in maize., 2022, 48: 2691–2705 (in Chinese with English abstract).

[32] 彭勃, 趙曉雷, 王奕, 袁文婭, 李春輝, 李永祥, 張登峰, 石云素, 宋燕春, 王天宇, 黎裕. 玉米葉向值的全基因組關(guān)聯(lián)分析. 作物學(xué)報, 2020, 46: 819–831. Peng B, Zhao XL, Wang Y, Yuan YW, Li CH, Li YX, Zhang DF, Shi SY, Song CY, Wang TY, Li Y. Genome-wide association studies of leaf orientation value in maize., 2020, 46: 819–831 (in Chinese with English abstract).

[33] Pritchard J K, Stephens M, Donnelly P. Inference of population structure using multilocus genotype data., 2000, 155: 945–959.

[34] Jakobsson M, Rosenberg NA. CLUMPP: a cluster matching and permutation program for dealing with label switching and multimodality in analysis of population structure., 2007, 23: 1801–1806.

[35] Bradbury P J, Zhang Z W, Kroon D E, Casstevens T M, Ramdoss Y, Buckler E S. TASSEL: software for association mapping of complex traits in diverse samples., 2007, 23: 2633–2635.

[36] Wang J B, Zhang Z W. GAPIT Version 3: boosting power and accuracy for genomic association and prediction., 2021, 19: 629–640.

[37] Zhang Y W, Tamba C L, Wen Y J, Li P, Ren W L, Ni Y L, Gao J, Zhang Y M. mrMLM v4.0.2: an R platform for multi-locus genome-wide association studies., 2020, 18: 481–487.

[38] Cingolani P, Platts A, Wang L, Coon M, Nguyen T, Wang L, Land S J, Lu X, Ruden D M. A program for annotating and predicting the effects of single nucleotide polymorphisms, SnpEff: SNPs in the genome of Drosophila melanogaster strain w1118; iso-2; iso-3.(Austin), 2012, 6: 80–92.

[39] Luo Y, Zhang M L, Liu Y, Liu J, Li W Q, Chen G S, Peng Y, Jin M, Wei W, Jian L, Yan J, Fernie A R, Yan J B. Genetic variation incontributes to ear length and grain yield in maize., 2022, 234: 513–526.

[40] Ning Q, Jian Y N, Du Y, Li Y F, Shen X M, Jia H T, Zhao R, Zhan J M, Yang F, Jackson D, Liu L, Zhang Z W. An ethylene biosynthesis enzyme controls quantitative variation in maize ear length and kernel yield., 2021, 12: 5832–5842.

[41] Pérez P, de los Campos G. Genome-wide regression and prediction with the BGLR statistical package., 2014, 198: 483–495.

[42] Zhang Y M, Jia Z, Dunwell J M. The applications of new multi-locus GWAS methodologies in the genetic dissection of complex traits., 2019, 10: 100–105.

[43] An Y X, Chen L, Li Y X, Li C H, Shi Y S, Zhang D F, Li Y, Wang T Y. Genome-wide association studies and whole-genome prediction reveal the genetic architecture of KRN in maize., 2020, 20: 490–500.

[44] Zhou B, Zhou Z J, Ding J Q, Zhang X C, Mu C, Wu Y, Gao J Y, Song Y X, Wang S W, Ma J L, Li X T, Wang R X, Xia Z L, Chen J F, Wu J Y. Combining three mapping strategies to reveal quantitative trait loci and candidate genes for maize ear length., 2018, 11: 1–8.

[45] Li D D, Zhou Z Q, Lu X H, Jiang Y, Li G L, Li J H, Wang H Y, Chen S J, Li X H, Würschum T, Reif J C, Xu S Z, Li M S, Liu W X. Genetic dissection of hybrid performance and heterosis for yield-related traits in maize., 2021, 12: 774478–774496.

[46] Su C F, Wang W, Gong S L, Zuo J H, Li S J, Xu S Z. High density linkage map construction and mapping of yield trait QTLs in maize () using the genotyping-by-sequencing (GBS) technology., 2017, 8: 706–719.

[47] Chen L, An Y X, Li Y X, Li C H, Shi Y S, Song Y C, Zhang D F, Wang T Y, Li Y. Candidate loci for yield-related traits in maize revealed by a combination of MetaQTL analysis and regional association mapping., 2017, 8: 2190–2203.

[48] Zhou Z P, Li G L, Tan S Y, Li D D, Liu W X. A QTL atlas for grain yield and its component traits in maize ()., 2020, 139: 562–574.

[49] Zhao Y M, Su C F. Mapping quantitative trait loci for yield-related traits and predicting candidate genes for grain weight in maize., 2019, 9: 16112–16121.

[50] Lu X, Zhou Z Q, Yuan Z H, Zhang C S, Hao Z F, Wang Z H, Li M S, Zhang D G, Yong H J, Han J N, Li X H, Weng J F. Genetic dissection of the general combining ability of yield-related traits in maize., 2020, 11: 788–802.

[51] Galli M, Liu Q J, Moss BL, Malcomber S, Li W, Gaines C, Federici S, Roshkovan J, Meeley R, Nemhauser JL, Gallavotti A. Auxin signaling modules regulate maize inflorescence architecture., 2015, 112: 13372–13377.

[52] Koops P, Pelser S, Ignatz M, Klose C, Marrocco-Selden K, Kretsch T. EDL3 is an F-box protein involved in the regulation of abscisic acid signalling in., 2011, 62: 5547–5560.

[53] Zhang H H, Yin L L, Wang M Y, Yuan X H, Liu X L. Factors affecting the accuracy of genomic selection for agricultural economic traits in maize, cattle, and pig populations., 2019, 10: 189–198.

[54] Tehseen M M, Kehel Z, Sansaloni C P, Lopes M D S, Amri A, Kurtulus E, Nazari K. Comparison of genomic prediction methods for yellow, stem, and leaf rust resistance in wheat landraces from Afghanistan., 2021, 10: 558–572.

[55] Ma J, Cao Y Y. Genetic dissection of grain yield of maize and yield-related traits through association mapping and genomic prediction., 2021, 12: 690059–690069.

[56] Lozada D N, Mason R E, Sarinelli J M, Brown-Guedira G. Accuracy of genomic selection for grain yield and agronomic traits in soft red winter wheat., 2019, 20: 82.

[57] Odilbekov F, Armoniené R, Koc A, Svensson J, Chawade A. GWAS-assisted genomic prediction to predict resistance toBlotch in Nordic winter wheat at seedling stage., 2019, 10: 1224–1233.

[58] 馬娟, 朱衛(wèi)紅, 丁俊強(qiáng). 玉米重要農(nóng)藝性狀的基因組預(yù)測分析.玉米科學(xué), 2022, 30(1): 48–52. Ma J, Zhu W H, Ding J Q. Genomic prediction analysis for maize important agronomic traits., 2022, 30(1): 48–52 (in Chinese with English abstract).

[59] Arruda M, Lipka A, Brown P, Krill A, Thurber C, Brown-Guedira G, Dong Y, Foresman B J, Kolb F L. Comparing genomic selection and marker-assisted selection for Fusarium head blight resistance in wheat ()., 2016, 36: 1–11.

[60] Bernardo R. Genomewide selection when major genes are known., 2014, 54: 68–75.

[61] Yuan Y, Cairns J E, Babu R, Gowda M, Makumbi D, Magorokosho C, Zhang A, Liu Y B, Wang N, Hao Z F, San V F, Olsen M S, Prasanna B M, Lu Y L, Zhang X C. Genome-wide association mapping and genomic prediction analyses reveal the genetic architecture of grain yield and flowering time under drought and heat stress conditions in maize., 2019, 9: 1919–1933.

[62] Liu Y B, Hu G H, Zhang A, Loladze A, Hu Y X, Wang H, Qu J T, Zhang X C, Olsen M, Vicente F S, Crossa J, Lin F, Prasanna B M. Genome-wide association study and genomic prediction of Fusarium ear rot resistance in tropical maize germplasm., 2021, 9: 325–341.

[63] Cericola F, Jahoor A, Orabi J, Andersen J R, Janss L L, Jensen J. Optimizing training population size and genotyping strategy for genomic prediction using association study results and pedigree information: a case of study in advanced wheat breeding lines., 2017, 12: e0169606.

Multi-locus genome-wide association study and prediction for general combining ability of maize ear length

MA Juan*, ZHU Wei-Hong, LIU Jing-Bao, YU Ting, HUANG Lu, and GUO Guo-Jun

Institute of Cereal Crops, Henan Academy of Agricultural Sciences, Zhengzhou 450002, Henan, China

Ear length is an important agronomic trait, which is closely related with yield. General combining ability (GCA) is an important index to evaluate excellent inbred lines. Therefore, the dissection of genetic basis of ear length GCA and formulation of corresponding breeding strategies is of great significance to improve maize yield. In this study, 537 F1hybrids as the experimental materials were obtained from 123 maize inbred lines and eight tester lines according to North Carolina II genetic mating design, and phenotyped under two environments. A total of 11,734 single nucleotide polymorphisms (SNPs) identified using the maize 5.5 K liquid breeding chip were used to conduct multi-locus genome-wide association study (MGWAS) and genomic prediction for ear length GCA in two environments and combined environment. A total of 11 SNPs significantly associated with ear length GCA were detected using seven MGWAS, and the variation of GCA effect explained by a single locus was 8.06%?28.23%. Five SNPs were co-located using different MGWAS. Locus 7_178103602 was repeatedly detected using mrMLM (multi-locus random-SNP-effect mixed linear model) in Zhoukou and combined environment, explaining 26.02%?28.23% of variation of ear length GCA, which was an environment-stable and major-effect SNP. 11 candidate genes were identified, among whichand EID1-like F-box protein 2 may be key genes for GCA of ear length. The accuracy of five random effect models for predicting ear length GCA ranged from 0.53 to 0.69 in the three environments, and there were minor differences among these models. In Xinxiang and Zhoukou environments, GBLUP (genomic best linear unbiased prediction) and RKHS (reproducing kernel Hilbert space) incorporating different significant loci as fixed effects could improve the accuracy of genomic estimated breeding value for GCA of ear length, with a percentage increase of 2.34%?14.98%. In the combined environment, except that the accuracy was slightly reduced using one significant locus derived from FarmCPU (fixed and random model circulating probability unification) or BLINK (Bayesian-information and linkage-disequilibrium iteratively nested keyway) as fixed effects, the addition of significant loci derived from the other two MGWAS methods could improve the genomic prediction ability, with a percentage increase of 2.80%?6.84%. Therefore, the incorporation of significant loci from MGWAS into the prediction models as fixed effects is helpful to improve the accuracy of the genomic estimated breeding value for ear length GCA, which could be used to effectively predict and select GCA of maize parental ear length.

ear length; general combining ability; multi-locus genome-wide association study; fixed effect model; genomic selection

10.3724/SP.J.1006.2023.23042

本研究由河南省科技攻關(guān)項(xiàng)目(222102110043)和河南省農(nóng)業(yè)科學(xué)院優(yōu)秀青年基金項(xiàng)目(2020YQ04)資助。

This study was supported by the Science and Technology Project of Henan Province (222102110043) and the Science-Technology Foundation for Outstanding Young Scientists of Henan Academy of Agricultural Sciences (2020YQ04).

馬娟, E-mail: majuanjuan85@126.com

2022-05-18;

2022-11-25;

2022-12-05.

URL: https://kns.cnki.net/kcms/detail//11.1809.S.20221205.1216.001.html

This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).