Yuning Chen,Zhihui Wang,Xiaoping Ren,Li Huang,Jianbin Guo,Jiaojiao Zhao,Xiaojing Zhou,Liying Yan,Huaiyong Luo,Nian Liu,Weigang Chen,Liyun Wan,Yong Lei,Boshou Liao,Dongxin Huai,Huifang Jiang*

Oil Crop Research Institute,Chinese Academy of Agricultural Sciences/Key Laboratory of Biology and Genetic Improvement of Oil Crops,Ministry of Agriculture and Rural Affairs,Wuhan 430062,Hubei,China

Keywords:Peanut Number Pod Seed QTL

A B S T R A C T The inheritance of pod-and seed-number traits(PSNT)in peanut(Arachis hypogaea L.)is poorly understood.In the present study,a recombinant inbred line(RIL)population of 188 lines was used to map quantitative trait loci(QTL)for number of seeds per pod(NSP),number of pods per plant(NPP),and numbers of one-,two-,and three-seeded pods per plant(N1PP,N2PP,and N3PP)in four environments.A total of 28 consensus QTL and 14 single QTL were identified,including 11 major and stable QTL.Four major and stable QTL including qN3PPA5.2,qN3PPA5.4,qN3PPA5.5,and qN3PPA5.7 each explained 12.3%-33.0%of phenotype variation.By use of another integrated linkage map for the A5 group(hereafter referred to as INT A5 group),QTL for PSNT were located in seven intervals of 0.73-9.68 Mb in length on chromosome A05,and candidate genes underlying N3PP were suggested.These findings shed light on the genetic basis of PSNT.Major QTL for N3PP could be used as candidates for further positional cloning.

1.Introduction

The cultivated peanut is an allotetraploid (2n=4x=40)legume crop used either for edible oil or as food,owing to its high oil and protein content[1,2].It is derived from a cross of diploid ancestor species(2n=2x=20)originating primarily in South America and is cultivated in＞100 countries worldwide.Peanut is also an important crop in China,with＞17 Mt of annual production.Yield is always the predominant objective in peanut breeding and is determined by seed number per unit area and seed weight.Some quantitative trait loci(QTL)associated with yield components have been identified[3,4].Many QTL for yield traits,especially seed weight,have been identified[3-12].However,there have been few reports of research on pod-and seed-number traits(PSNT).Peanut plants produce four main types of pods,with one to four seeds.Seed number per unit area is the product of plant density,number of pods per plant(NPP),and number of seeds per pod(NSP).NPP and NSP are determined by the number of each type of pod,including numbers of one-,two-,and threeseeded pods per plant(N1PP,N2PPn,and N3PP)and the number per plant of pods containing more than three seeds.NSP is the sum of N1PP,N2PPn,N3PP and number of pods containing more seeds.To date,there has been little information in the literature about the genetic components of PSNT.

As an important criterion of peanut taxonomy,pod type is stable across years and environments[13].Limited classical genetic analysis resulted in conflicting genetic models for control of pod type.Seshadri[14]reported that one-and twoseeded pods was dominant to multiple-seeded pods.However,other authors[15,16]found that multiple-seeded pods was dominant to fewer-seeded pods.The latest report[13]states that one-seeded pods is controlled by two of three recessive genes.No further molecular research on the PSNT has been reported,and the relationship among PSNT remains unclear.

PSNT are also important yield traits in other dicotyledonous crops,and some achievements have been obtained in other legume crops and rapeseed.Co-localizing QTL and genes associated with these traits have been observed and the relationships among them have been well characterized in some cases.A typical example of research on PSNT is the map-based cloning of the Ln gene in soybean[17].The pleiotropic Ln encoded Gm-JAG1,which regulated leaf shape and pod type,and an amino acid substitution in the ERF motif of Gm-JAG1 increased the number of multiple-seeded pods per plant,thereby increasing NSP and the total number of seeds per plant and in turn,seed yield[18].The GmCYP78A10b gene was selected artificially during soybean breeding because it was associated with large seed size and lower NPP;however,the selection of GmCYP78A10b did not affect the seed yield of individual plants[19].The pentatricopeptide repeat(PPR)gene was identified as a candidate underlying NPP in chickpea[20].An SSR marker associated with NSP was also identified in common bean[21].In broad bean,NSP was a component of yield and was NSP was negatively correlated with NPP[22].Co-localization of QTL for NPP,flowering date,and seed development was also observed in soybean[23].QTL for NPP and seed yield were co-located in chickpea[24].Markers associated with flowering date,NPP,and NSP were identified in common bean[25].Liu et al.[26]reported that eight metabolic pathways were involved in the formation of pods with four or more seeds in soybean.In rapeseed,the pleiotropic major QTL qSN.A6 exerted an effect on NSP and seed yield[27].In another study in rapeseed,tight linkage accounted for the co-localization of QTL for NPP and NSP[28].Investigation of the genetic basis of PSNT and the relationship among these traits could be valuable for further understanding the complexity of peanut yield components.

Multivariable conditional analysis methods have been developed[29]to dissect the genetic relationships among different traits at the QTL level.In peanut,conditional QTL mapping was used to identify genetic interdependencies among QTL for different traits that co-localized in the same genome region[3,30].By conditional mapping,Luo et al.[3]showed that pod length contributed more to 100-pod weight than pod width as a yield component of peanut,and Zhou et al.[30]found that resistance to late leaf spot was influenced strongly by the total number of branches.Thus,it is possible to dissect the genetic interrelationships among QTL for PSNT.

Recently released genome sequences of wild species of Arachis[1,31]have afforded a rough picture of the genes located in the genomic physical intervals of QTL for PSNT.Luo et al.[3,12]identified candidate genes for QTL controlling pod size and weight by comparing their locations on the genetic and physical maps of A.duranensis.

The objectives of the present study were as follows:(1)to dissect the genetic architecture of PSNT components in peanut by linkage mapping,(2)to identify QTL for PSNT,and(3)to identify candidate genes for PSNT.

2.Materials and methods

2.1.Plant material

A RIL population including 188 lines from a cross between the peanut cultivars Fuchuan Dahuasheng and ICG6375 has been previously[11]well characterized.Fuchuan Dahuasheng(subsp.hypogaea L.var.hirsuta Kohle)is a landrace of southwest China,and bears 15-20 pods per plant,approximately 50%three-seeded and 45%two-seeded.ICG6375(subsp.fastigata Waldron var.vulgaris Harz)was introduced from ICRISAT,and bears＞45-50 pods per plant,of which approximately 95%are two-seeded(Fig.1).

2.2.Field trials and trait phenotyping

The population was grown in the 2013(F5),2014(F6),and 2015(F7)seasons.Ten plants of each line were phenotyped for one-,two-,and three-seeded pods.The F5RIL population was phenotyped in Wuchang(114°34′E,30°59′N(xiāo),environment 1,E1)and Yangluo(114°52′E,30°59 N,E2)in 2013,and the F6and F7RIL populations were phenotyped in Wuchang in 2014 and 2015(E3 and E4)for two replicates each year.Five traits,NSP,NPP,N1PP,N2PP,and N3PP were scored,and they were calculated as the total number of corresponding seeds or pods divided by the number of plants.

2.3.Statistical analysis

Phenotypic data were tested for normality using the PROC UNIVARIATE procedure of SAS 9.3(SAS Institute,Cary,NC,USA).Correlation coefficients among the five traits were calculated using the PROC CORR procedure of SAS.Broadsense heritability was estimated for each trait following Wu et al.[32].

2.4.Molecular-marker and QTL mapping

QTL analysis was performed using the trait values from single replicates or the average trait values of each environment and existing map reported previously[11].An INT A5 map was constructed by integration of maps reported previously[11,33].The INT A5 group is described in Table S1.Conditional phenotypic values of T1|T2 were calculated by the mixedmodel approach for the conditional analysis of complex traits described by Wen and Zhu[29].For the conditional phenotypic value T1|T,trait 1 was conditioned on trait 2,indicating that the phenotypic value of T1 was calculated after elimination of the influence of the phenotypic value of T2.

Fig.1-Pod-type phenotype of parent lines and the RIL population.

QTL analysis was performed by composite interval mapping(CIM)using Windows QTL Cartographer 2.5[34].The LOD value to declare a QTL significant was chosen as 2.5 based on a permutation test with 1000 runs to determine the P=0.05 genome-wide significance level.Based on the QTL detected,QTL meta-analysis was performed to further integrate QTL located close to each other using BioMercator 2.1[35].This approach can be used to determine the number of meta-QTL that best fit the expectancy value on a given linkage group based on a modified Akaike criterion[36].A two-round strategy of QTL meta-analysis was performed to integrate QTL with 95% confidence intervals.First,additive QTL expressed in different environments for the same trait were integrated into consensus QTL.Second,the consensus QTL and other QTL for different traits were further integrated into unique QTL.Both consensus QTL and single QTL involved in the generation of unique QTL were designated as stable QTL.The QTL nomenclature followed Shi et al.[37].Consensus and unique QTL names begin with q and uq respectively,followed sequentially by the trait abbreviation,the linkage group number,and the serial number.

2.5.Identification of homologous colinear loci in the INT A5 group and the A.duranensis genome by e-PCR amplication

The genomic position of SSR loci on the INT A5 group was determined using electrotic-PCR with the primer sequences against the genome sequences of A.duranensis[1,31].The e-PCR was performed with the parameters of maximum three-basepair mismatches and for a given primer pair.Amplifications with 100%matched to the genome sequences of A.duranensis were used to determine the final position of the markers on the chromosome A05 of A.duranensis.The amplicons produced by e-PCR were then analyzed to determine the colinear relationship between the INT A5 group and the physical counterpart on chromosome A05 of A.duranensis.The colinear relationship map was generated using MapChart for Windows[38].

2.6.Candidate-gene scan for major QTL on chromosome A05

QTL regions were identified based on the physical positions of markers linked to QTL on the INT A5 group.Annotated genes in these regions were used to evaluate the candidate genes for QTL.The COG and KEGG annotations of these genes were also referenced.To further identify candidate genes controlling QTL in peanut,the protein sequences of genes influencing seed number in other crops were searched against the chromosome A05 using BLASTn(E-value＜1E-5).Gene sequence information was collected for three genes in soybean:Gm-JAG1 and Gm-JAG2[18],and GmCYP78A10[19];two genes in rapeseed:BnaC9.SMG7b[39]and ARF18[40];and eight genes in rice:08SG2/OsBAK1[41],SMALL GRAIN 11[42],GNP1[43],DEP1[44],LAX1 and FZP[45],APO1[46],and GNS4[47].The protein sequences of these genes were searched with blastp against the protein database of A.duranensis genome.Based on the Blastp results,homologs of these genes could be candidates when they were located in QTL intervals for PSNT in the A.duranensis genome.

3.Results

3.1.Phenotypes of PSNT and genetic correlations among PSNT

The population was phenotyped for N1PP,N2PP,N3PP,NSP,and NPP.The average values of these five traits over 2013 and 2014 were used in variance analysis.N1PP,N2PP,NSP,and NPP showed nearly bell-shaped distributions,indicating polygenic inheritance,and N3PP showed a J-shaped distribution,indicating major-gene inheritance N3PP(Fig.2).N1PP,N2PP,NSP,and NPP showed high broad-sense heritabilities with values of 0.93,0.72,0.95,and 0.80,whereas the value for N3PP was only 0.48 and the narrow-sense heritability was 0.34.Correlations among the five PSNT are presented in Table 1.NPP showed positive correlations with N1PP and N2PP and negative correlation with NSP.N1PP showed a positive correlation with N2PP and negative correlation with NSP,and NSP showed negative correlations with NPP and N1PP,and positive correlation with N3PP.The ANOVA for NPP,NSP,N1PP,N2PP,and N3PP showed significant differences among RILs and among environments and RIL×environment interaction(Table 2).

3.2.Unconditional and conditional QTL mapping of PSNT

Unconditional analysis yielded 157 QTL for NSP,NPP,N1PP,N2PP,and N3PP.Discarding QTL detected in only one environment left 107 QTL for the five traits on groups A2,A4,A5,A6,A10,B2,B3,B5,B6,B8,and B9(Fig.3,Table S2).Each QTL accounted for 4.0%-33.0%of phenotype variation explained(PVE).Eleven were major QTL on A5 group accounting for＞10%PVE,and eight QTL were detected across four environments(Table 3,Fig.4,Table S2,Table S3).In the first round of meta-analysis,a total of 89 QTL were integrated into 28 consensus QTL,with the remaining 18 left as single QTL.For example,six QTL for N3PP were integrated into qN3PPA5.4.In the second round of meta-analysis,16 consensus QTL and 10 single QTL were integrated into 10 unique QTL(Table S4).For example,uqA5-1 was the result of integrating the three consensus QTL qN1PPA5.1,qN2PPA5.1,and NSPA5.1.To further clarify whether pleiotropy,tight linkage,or intergenic interaction accounted for the co-localization of the QTL for PSNT,conditional QTL analysis was performed to detect QTL when the phenotypic value of these traits were conditioned on one another.In all,18 conditional QTL were detected with no＞10%of PVE(Table S5).Three QTL effective alleles were donated by the female parent FCD when NPP was conditioned on NSP,N2PP,and N3PP,and all other QTL alleles were donated by male parent ICG6375.

Table 1-Correlations among pod and seed traits in RILs across three environments.

For N1PP,nine QTL were identified by unconditional mapping with 5.6%-12.5% PVE and additive effects of 0.05-1.32.Of the nine,three and two were on groups A5 and B3 and one each on groups B2,B6,and B8,and six were consensus QTL.The major QTL qN1PPA5.1 accounted for 12.5%of PVE across four environments.The additive effective alleles of qN1PPA5.1,qN1PPB2.1,and qN1PPB3.1 were donated by the female parent FCD and the others by the male parent ICG6375.Three QTL were identified by conditional mapping.conqN1PPB8.1 was detected when N1PP was conditioned on N2PP in E1 and conqN1PPB9.1 was detected when N1PP was conditioned on NSP and N2PP in E3.

Fig.2-Phenotypic variations of five PSNT in the RIL population.

Table 2-Analysis of variance for pod and seed traits in RILs across three environments.

For NPP,five QTL on groups A5,B3,B5,and B8 were identified by conditional mapping.They explained 5.2%-7.7%of phenotype variation with absolute values of additive effects ranging from 1.26 to 2.48.qNPPA5.1 and qNPPB8.1 were consensus QTL.The male parent ICG6375 donated the additive effective alleles of qNPPA5.1,qNPPB5.1,and qNPPB8.1.con qNPPB2.1,2.2,and 2.3 were detected when NPP was conditioned on NSP,N3PP,and N2PP in E1 or E2.

For NSP,eight QTL were identified by unconditional mapping on groups A5,A10,B3,B6 and B8 and with 4.6%-17.4%of PVE.Of them,six were consensus QTL,and the four major QTL qNSPA5.1,qNSPA5.4,qNSPA5.5,and qNSPA5.6 explained respectively 7.1%-12.7%,12.4%-17.4%,7.8%-12.1%,and 10.8%-14.6%of phenotype variation.Except at qNSPA5.1,the female parent FCD donated alleles to NSP.For conditional mapping,two QTL were detected when NSP was conditioned on N1PP in E1 and E3 and two QTL were detected when NSP was conditioned on N1PP and NPP in E3.In particular,conqNSPA6.2 was detected when NSP was conditioned on NPP in E2 and E3.

For N2PP,five QTL were located on groups A5,B2,and B8.They explained 5.3%-11.4%of phenotype variation with absolute size of additive effect ranging from 0.30 to 2.12,and three of the five were consensus QTL.qN2PPA5.1 and qN2PPB2.1 explained 6.8%-10.4%and 5.8%-11.4%of phenotype variation.The male parent ICG6375 donated all the additive effective alleles to the N2PP.For N2PPA5.1,the PVE was 10.9%when the average values in the four environments were used,indicating a candidate stable QTL for N2PP.For conditional mapping,two QTL were detected on group B1 when N2PP was conditioned on NSP and N3PP in E2 and three QTL were detected on group B8 when N2PP was conditioned on N1PP and NPP in E3.

Fig. 3 - QTL detected for five NPST in a RIL population.

Fig.4-QTL scan profiles for the A05 linkage group for PSNT.The horizontal and vertical axes represent respectively genetic distance(cM)and LOD value.The line and curves indicate respectively threshold and true LOD values.The different traits and experiments are represented using different colors and line styles.

For N3PP,15 QTL were identified with 5.3%-33.0%of PVE and the absolute size of the additive effect ranged from 0.22 to 0.65.Of the 15,three,two,and seven were on groups A2,B3,and A5 and one each on groups A4,A6,and B6,nine were consensus QTL,and five were major QTL with 9.9%-33.0%of PVE.qN3PPA4.1,qN3PPA5.1,qN3PPA5.2,qN3PPA5.3,qN3PPA5.4,and qN3PPA5.7 were each detected across four environments.The male parent ICG6375 donated three additive effective alleles on A2 and one each on groups A4,A5,and A6,and the female parent FCD donated nine other alleles to N3PP.For N3PP,unexpectedly,no new QTL were detected when N3PP was conditioned on other PSNT.

3.3.Comparative mapping of QTL on the INT A5 group

The major QTL identified in the present study were mainly located close on A5 group,which contained only 27 markers mapped at low resolution.Only 11 of the 27 markers were assigned to the chromosome A05 of A.duranensis,and poor marker sequence information prevented accurate localization of the QTL.Accordingly,the INT map was employed to improve the mapping resolution of the QTL and to explore more valuable genomic position of them.The QTL were located on the INT A5 group,containing 161 markers with a mean interval of 0.61 cM(Table S1).Of the 27 markers on the existing map,24 were assigned to the INT map.High synteny was observed between the A5 and the INT A5 groups(Fig.S1),and the QTL were distributed in similar orders on the two groups based on the markers tightly linked to the QTL.The seven QTL covered an interval of 41.42 cM on the INT A5 group(Table S6).Interestingly,uqA5-2 and uqA5-3 overlapped completely and were accordingly integrated into uqA5-2+3.Owing to the changed position of markers ARS702 and AGGS1335,and the absence of marker AHGA44691,which was linked to uqA5-6 on the INT A5 group,three unique QTL,uq A5-5,-6,and-7,were located together in an interval of 14.64 cM.Use of the INT map greatly increased the mapping resolution of the QTL.For the region harboring these seven QTL,the mapping resolution was 1.92 cM/marker on A5 and 0.39 cM/marker on the INT A5 group.

3.4.Comparative mapping of QTL on the INT A5 group and chromosome A05 of Arachis duranensis

To get the specific genomic intervals of the major QTL,the INT A5 group was compared to the physical map of A.duranensis.Of 161 markers,92 of were assigned to chromosome A05 of A.duranensis and covered a 103.98-Mb genomic region with an average of 1.142 Mb between adjacent markers.General synteny with frequent inversion was observed between INT A5 and chromosome A05 of A.duranensis(Fig.S1,Table S7).

3.5.Candidate genomic regions for QTL

QTL were mapped to chromosome A05 based on the putative physical positions of markers located in the confidence intervals of QTL on INT A5.A total of 43 markers covered a 75.5-Mb region on chromosome A05.The QTL were located based mainly on markers close to the peak position.Further,for determination of the position of uqA5-7,the genomic positions of 10 markers linked to uqA5-7 on the existing A5 or the INT A5 group that the primer sequences did not show a 100%match with the chromosome A05 were also referenced(Fig.5,Table S8).Finally,uqA5-1 was located in the 2.25-cM interval between markers AGGS351 and ARS777 covering a 1.09 Mb interval on the A5 genome;qN3PPA5.1 was located in a 5.17-cM interval covering 6.34 Mb on the A5 genome;uqA5-2 and uqA5-3 were located in a 3.75-cM interval covering 0.73 Mb;uqA5-4 was located in a 2.58-cM interval covering 2.32 Mb on the A5 genome;uqA5-5 was located in a 9.68-Mb interval;uqA5-6 was located in a 1.1-Mb interval;and uqA5-7 was located in a 3.40-Mb interval(Fig.5,Table S9).

3.6.The putative candidate genes of major QTL on chromosome A05

The seven genomic intervals on chromosome A05 harbored major QTL for N1PP,NSP,N2PP,and N3PP.To identifypossible candidate genes underlying these QTL,the genes in these seven intervals were extracted and analyzed.Of the 686 genes,583 were functionally annotated and 103 genes encoded unknown proteins(Table S10,Table S11).

Table 3-QTL identified for PSNT by unconditional mapping.

To identify possible candidate genes controlling PSNT,the genome annotation of A.duranensis was referenced.Based on literature information,a total of 41 genes associated with reproductive development were included(Table S12,Table 3).For N3PPA5.1,uqA5-2+3,uqA5-4,uqA5-5,and uqA5-7,there were 15,1,2,9,and 14 genes involved in ubiquitination of protein(11 genes),cell division(4 genes),cell growth(1 gene),cell proliferation(1 gene),petal differentiation(2 genes),auxin response(6 genes),gibberellin signal pathway(1 gene),plant self-incompatibility(8 genes),and others.

A gene cluster was observed in each QTL interval.For uqA5-2+3,NThA5.1,uqA5-1,uqA5-4,uqA5-5,uqA5-6,and uqA5-7,respectively 3,21,2,5,9,10,and 18 gene clusters were observed.For all seven QTL intervals,a total of 216 genes composed 64 clusters(Table S13,Table S14).The cluster containing the most genes was observed in the uqA5-7 interval,where 20 genes encoded disease resistance proteins,nine genes encoded pentatricopeptide repeat(PPR)superfamily proteins,and eight genes encoded putative plant selfincompatibility S1.It was noteworthy that genes with the same molecular function were shared by these QTL intervals.A total of 206 genes located in these intervals encoded 45 proteins(Table S15,Table S16).For example,13 genes:one each of uqA5-2 and-3 and uqA5-4,four of uqA5-5,and six of uqA5-7,encoded the same cytochrome P450 protein.Five genes,one of NThA5.1 and two each of uqA5-5 and uqA5-7,encoded E3 ubiquitin-protein ligase.

Fig.5-The comparative locations of major QTL on the INT A5 group and chromosome A05.

KEGG annotation was also used to identify metabolic pathways potentially involved in the formation of PSNT.Of the 686 genes,110 were assigned to 60 pathways of which 76 were involved in 27 pathways(Table S17).Interestingly,genes located together in the same QTL interval often showed identical function and were involved in the same pathways.For example,for uqA5-7,three genes,Aradu.3MQ4D,Aradu.FUJ5L,and Aradu.FL5Y4,were involved in plant hormone signal transduction.A total of 26 genes located in four QTL intervals(uqA5-4,uqA5-5,uqA5-7,and N3PPA5.1)were involved in 10 pathways.Even genes located in different QTL intervals often showed identical function and were involved in the same pathways.A total of 54 genes located in different QTL intervals were involved in 17 pathways.For example,10 genes,of which one each was located in the uqA5-2+3,NThPA5.1,and uqA5-5 intervals and two and five were located respectively in the uqA5-6 and uqA5-7 intervals,were assigned to the ko00940 term and involved in phenylpropanoid biosynthesis(Table S17).Interestingly,four genes,one located in the uqA5-1 interval and three in the uqA5-7 interval,were involved in plant hormone signal transduction.Two genes(Aradu.Y2Q2L located in the uqA5-5 interval and Aradu.5HP19 located in the uqA5-6 interval)were involved in cell growth and death.Two genes(Aradu.A1TRQ and Aradu.JR6BU located in the N3PPA5.1 interval)were involved in ubiquinone and other terpenoid-quinone biosynthesis.The most highly shared pathway was the biosynthesis of other secondary metabolites(10 genes in five QTL intervals),followed by carbohydrate metabolism(nine genes in three QTL intervals)and translation(eight genes in three QTL intervals)(Table S17,Fig.S2).As for the molecular function of GO annotation,the GO term 0005515(protein binding)was highly represented(63 genes in seven QTL intervals),followed by term 0003824(catalytic activity,25 genes in six QTL intervals),term 0003676(17 genes in six QTL intervals),and others(Table S18,Fig.S3,S4,S5).

The roles in peanut of 13 genes controlling seed number in other crops were also investigated.This approach failed to find genes with＞70%sequence identity in the genomes of A.duranensis and A.ipaensis.CYP78A10[19]showed 66.73%identity with Aradu.X48YR and was located in the physical interval of uqA5-5.SMALL GRAIN 11,GNP1,and 08SG2/OsBAK1 showed several matches with＜60%identity in the physical intervals of uqA5-2+3,N3PPA5.1,uqA5-5,uqA5-6,and uqA5-7.In particular,two soybean Ln gene[18],Gm-JAG1(Glyma.JX119212.1)and Gm-JAG2(Glyma.JX119216),showed weak identity(40.50%,42.86%)with Aradu.VY0Q0 and Araip.S3BW6,the rapeseed gene ARF18 also showed weak identity(40.91%)with Aradu.UX5R5,and these putative homologs were adjacent to the physical interval of uqA5-7(Table S19).

4.Discussion

4.1.The complexity of the genetic basis of PSNT in peanut

The large variation of PSNT in present RILs indicated typical quantitative inheritance(Fig.1,Fig.2).A total of 28 consensus QTL and 14 single QTL were identified on the 11 linkage groups for the five PSNT,and strong interaction among them was observed by unconditional and conditional mapping,thus indicating a complex genetic basis for PSNT(Fig.3).The finding that 19 of them and all 11 major QTL were located on the A5 linkage group indicates an important role of A5 in the determination of PSNT.Similar results have been observed in other dicotyledonous crops.In soybean,PSNT were controlled by 12 major QTL distributed on six linkage groups,and the A1 linkage group harbored five major QTL and played an important role in PSNT[48].In rapeseed,＞20 QTL for NSP were distributed on 15 chromosomes,and the chromosome A9 contributed much to the determination of NSP[28,49].PSNT are the product of both evolution and domestication in most crops.The ancestral diploid wild species of peanut often had one-seeded pods,whereas the tetraploid wild species of peanut often had two-seeded pods,and the tetraploid cultivated peanut typically has pods of two or more seeds.It appears that the transformation from one-to two-seeded pods was a consequence of polyploidization during peanut evolution.Although the female parent FCD and the male parent ICG6375 belong to two different subspecies of peanut,the distant genetic relationship did not affect their cocontribution to PSNT on the A5 group.The finding that the two parents together donated most QTL alleles for PSNT on the A5 group indicates that the A5 group is functionally highly conserved with respect to the genetic information of PSNT in peanut.

As the main objective of the present study,the genetic architecture of NSP and NPP was dissected.Four of six QTL controlling NSP were major QTL with 10%PVE,suggesting that major QTL are the main genetic basis of NSP architecture and also suggesting a potential for molecular selection in NSP breeding in peanut.In rapeseed and soybean as well,a major QTL or single gene controlled NSP[18,27,28].Stable minor QTL controlling NPP were donated by the male parent ICG6375.Because in soybean and chickpea,NPP was controlled by major QTL on several chromosomes[20,48,50],it was reasonable to expect that NPP in peanut should originally have been controlled by major QTL.The difficulty of collecting all pods from each plant in the field may account for the nondetection of major QTL for NPP.The finding that NSP was negatively correlated with NPP indicates a tradeoff between NSP and NPP.The QTL for NSP were donated mainly by the female parent FCD,whereas the stable QTL for NPP were donated mainly by the male parent ICG6375,accounting at the QTL level for the negative correlation of NSP with NPP.The tradeoff between NSP and NPP could be conventionally explained by antagonistic pleiotropy of co-localized QTL,as reflected by the negative coefficients of phenotypic correlation between NSP and NPP[51].It is consistent with the opposite additive-effect direction of co-localized QTL on the A5 linkage group.Conditional mapping also suggested an interaction between NSP and NPP,making it unlikely that antagonistic pleiotropy was the genetic basis of colocalization of QTL for NSP and NPP on the A5 group,given that conditional mapping generated new QTL when NSP was conditioned on NPP or NPP was conditioned on NSP(Table S5).In rapeseed,the tight linkage of co-localized QTL qPN.Ao6-1 and qSN.Ao6-1 with opposite additive effect underlay the tradeoff between NSP and NPP[28].The present study provides some molecular evidence to account for the negative correlation between NSP and NPP in peanut.The potential prospect of QTL for NSP and NPP on the A5 group in peanut breeding should be further studied.

4.2.Stable and major QTL on linkage A5 group controlled N3PP

The J-shape distribution of N3PP shown in the RILs indicates major-gene inheritance of N3PP(Fig.2).QTL mapping also revealed that major QTL controlled N3PP in the RILs(Fig.3,Table 3).For N3PP,the five consensus QTL qN3PPA5.2,qN3PPA5.4,qN3PPA5.5,qN3PPA5.6,and qN3PPA5.7 were stably expressed and explained respectively 14.2%-17.0%,18.2%-33.0%,9.9%-17.2%,9.1%-9.2%,and 12.3%-20.5%of phenotype variation,.This finding was consistent with the importance of pod type as a character in peanut taxonomy.However,the low broad-sense heritability of only 0.48 may have been the result of interaction between genotype and location and year(Table 2),and also of the occasional occurrence of three-seeded pods.Some lines produced only a few three-seeded pods in a trial.Some lines stably produced three-seeded pods,but their N3PP varied widely over locations and years.The female parent FCD belongs to the botanical variety hirsute,which typically has three-seeded pods and is endemic to China[52,53].The molecular marker of the major QTL may be the molecular fingerprint of the variety hirsute.Interestingly,there were seven QTL located together in an interval of 24.4 cM on the A5 linkage group,indicating that this region was important for N3PP.In soybean,the three-and four-seeded pod type was governed mainly by the Ln gene and other modifying genes with minor effect[17,18].However,it was not clear whether the QTL for N3PP on the A5 group was responsible for fourseeded pods,which were only occasionally observed in the RILs.Also in soybean,three-and four-seeded pods were controlled mainly by other major QTL[48,54].A similar trait in rapeseed,the number of seeds per silique,was also controlled by a single gene or major QTL depending on the genetic background[28,39].

4.3.The comparative location and annotation of candidate genes for major QTL on chromosome A05

In the present study,the employment of INT A5 group greatly improved the mapping location of QTL on the chromosome A05 and facilitated to identify the candidate genes for PSNT.The strategy of employing the INT group was useful for improving mapping resolution while the genome sequencing of tetraploid species of Arachis is still in progress.

In plants,the numbers of fruits and seeds are controlled by many biological processes and have a complex genetic basis.These numbers are determined by the number of female and male reproductive organs,the fertility of female and male gametophytes,and the fate of zygote differentiation[55,56].In the present study,genes located in the physical intervals of several major QTL on chromosome A05 may have been involved in reproductive development as shown in Table 3.These genes determine seed number in other crops and are involved in protein ubiquitination,cell division,cell growth and cell proliferation,petal differentiation,auxin response,gibberellin signal pathway,plant self-incompatibility,and other developmental events[57,58].Furthermore,known genes involved in the determination of PSNT were commonly shared by these QTL(Table S12).For example,one,two and two genes encoding E3 ubiquitin-protein ligase were located in the interval NThA5.1,uqA5-5,and uqA5-7.It was difficult to directly identify the genes underlying the QTL for PSNT based on their functional annotations.Further study may reveal whether a gene family or several single genes underlie the QTL for PSNT.

Genes underlying QTL for PSNT on chromosome A05 could not be identified by the homology-based method.The BLASTp discovered the lower homology of 13 genes controlling seed number in soybean,rice and rapeseed with genome of A.duranensis.The weak sequence similarity of peanut genes in QTL intervals with known genes in other species provided insufficient evidence for their acceptance as candidate genes underlying QTL for PSNT in peanut.

Supplementary data for this article can be found online at https://doi.org/10.1016/j.cj.2018.09.002.

Conflict of interest

The authors declared that they have no conflicts of interest.

Acknowledgments

This study was supported by the National Natural Science Foundation of China(31271764,31371662,31471534,31601340,31461143022),the China's Agricultural Research System(CARS-14),the National Key Technology R&D Program of China(2013BAD01B03),and the National Infrastructure for Crop Germplasm Resources(NICGR2017-036).