Introduction

Oilseed rape (Brassica napus L., AACC) is mainly cultivated to produce edible oil from the seeds and ranks as the second-largest growing oil crop after soybean (Shi et al. 2015). As a complex quantitative trait, oilseed rape plant yield is systematically controlled by three major components, i.e., seed number per silique (SN), silique number per plant (SP), and seed weight (SW) (Wang et al. 2016). Among them, SN displays rich variations in both cultivars and germplasm resources, thus being an important breeding objective for rapeseed genetic improvement (Chen et al. 2011; Yang et al. 2017; Jiao et al. 2021). In flowering plants, ovules provide structural and ground support for the female gametophyte and develop into seeds after fertilization (Drews and Koltunow 2011; Shi and Yang 2011). Thus, the maximum SN is developmentally determined by ovule number per ovary (ON), while the final SN is also affected by the proportions of successful ovule fertilization and fertilized ovule development. Therefore, improvement in crop productivity requires understanding of the molecular pathways that control ovule initiation and development (Yuan and Kessler 2019).

In model plants like Arabidopsis and rice, the ovule initiation and developmental processes have been investigated, and more than 70 key genes have been revealed to be involved in ovule initiation and development (Qadir et al. 2021), including carpel meristem formation (CMM), ovule identity, primordia initiation, and integuments development (Skinner 2004; Shi and Yang 2011; Cucinotta et al. 2014). AINTEGUMENTA (ANT), REVOLUTA (REV), CUP-SHAPED COTYLEDON (CUC1 and CUC2), and SPATULA (SPT) regulate the CMM formation (Ishida et al. 2000; Nole-Wilson et al. 2010; Nahar et al. 2012); AGAMOUS (AG), SHATTERPROOF (SHP1, SHP2) SEEDSTICK (STK), and SEPALLATA (SEP) control the ovule identity (Favaro et al. 2003; Pinyopich et al. 2003; Skinner 2004). CUC1, CUC2 ANT and HUELLENOLS regulate the ovule initiation and boundary establishment (Liu et al. 2000; Skinner 2004; Galbiati et al. 2013; Cucinotta et al. 2014). HLL and ANT, AG, and BEL1 also play a role in the integument formation (Skinner et al. 2001; Azhakanandam et al. 2008). Hormonal signaling and interactions also play a vital role in the expression and regulation of these genes. Cytokinin-auxin interaction guides the ovule organogenesis via PIN1 under the control of cytokinin response factors (Galbiati et al. 2013; Cucinotta et al. 2016). CUC1 and CUC2 regulate cytokinin homeostasis (Cucinotta et al. 2018), suggesting a complex gene network associated with hormone signaling involved in the ovule development process. Recently, two genes, NEW ENHANCER of ROOT DWARFISM (NERD1) and OVULE NUMBER ASSOCIATED 2 (ONA2), were also identified to participate in the determination of ovule number during flower development (Yuan and Kessler 2019). However, compared to some other developmental processes or traits, the genetic factors determining ovule number remain largely elusive, especially in rapeseed..

To date, nearly hundred QTLs related to seed number (SN) were identified in rapeseed (Zhu et al. 2020; Raboanatahiry et al. 2022); among these, qSS.C9 (BnaC9.SMG7b) is the only SN-related QTL cloned to date (Li et al. 2015). By contrast, the genetics of ON remains largely unexplored, and only few QTLs related to ON were revealed (Khan et al. 2019; Qadir et al. 2022) via GWAS. Recently the integration of linkage and association mapping has exhibited powerful capability in exploring potential genetic loci for economically important traits. He et al. (2017) identified a RING-domain gene (BnaC03g63480D) with a potential role in branch morphogenesis via GA modulation. Similarly, 12 growth period related genes, including BnaTOC1.A03 and BnaFUL.A03 were detected by integrating linkage and GWAS mapping (Wang et al. 2020b). The combination of association mapping and a linkage analysis can reduce false positives from associated loci due to high LD but also facilitates fine mapping of a target region with a large QTL interval (Hu et al. 2011); it can also narrow the target location to identify fewer candidate genes and reduce the timeline to gene cloning or identification of tightly linked markers for breeding.

Here, we analyzed the variations and genetic basis of ON both in a double haploid (DH) population based on linkage analysis and a panel of natural accessions by GWAS. Our work laid a foundation for map-based cloning of the genes responsible for the ON trait and will provide molecular markers for their improvement.

Materials and methods

Plant materials

A DH population comprised of 188 lines was used for the study, which was previously developed from the F1 cross between two inbred accessions (ZY50 and 7–5) (Wang et al. 2020a). An association panel of 505 inbred lines collected from various geographic locations was previously described (Tang et al. 2021) and was utilized in this study. However, the phenotype data was collected for 374 inbred lines only in all four environments.

Field experiment

The DH lines and two parents were grown for three consecutive years (2015, 2016, and 2017) in the experimental farm of Huazhong Agricultural University, located at Ezhou, Hubei province, China (30.39° N, 114.88° E). The field was arranged in a randomized complete block design with 2 replications. Each line was grown in 2 rowed plots consisting of 24 plants. The length and width of the rows were 1.2 and 0.3 m, respectively. The ON data of the DH population was recorded from 2016 to 2018 in two environments and one in 2017. Therefore, five datasets were available for analysis designated as DH-EZ16-1, DH-EZ16-2, DH-EZ17, DH-EZ18-1, and DH-EZ18-2, respectively. The natural population was planted in two locations in the 2016–2017 growing season (Huazhong Agricultural University 30.59° N, 114.29° E and Modern Agricultural Science and Technology Innovation Demonstration Park of Sichuan Academy of Agricultural Sciences (30.65° N, 104.06° E). In 2017–2018, this population was planted in the experimental farm of Huazhong Agricultural University, located at Ezhou, Hubei province, China (30.39° N, 114.88° E). Each line was grown in two-rowed plots with 10–15 plants per row. The seeds were sown by hand, and the field management followed standard agricultural practice (Khan et al. 2020). ON data for the natural population were recorded in four environments designated as NP-WH17, NP-CD17, NP-WH18, and NP-EZ18.

Phenotyping, data collection, and analysis

To analyze the ovule number, freshly developed inflorescences (BBCH55-60) were collected from the main branches of selected plants. Three plants were randomly selected from each DH lines and inbreed accessions for sampling. These samples were fixed and stored in the formalin solution at room conditions. From each inflorescence, ten buds with close size (5–6 mm) were randomly selected and dissected, and the ovaries were excised carefully. The sampled ovaries were suspended in 90% alcohol solution for 24–48 h in 2-ml Eppendorf tubes and then washed with ddH2O. After removing the water, the ovaries were dried in the tube and were suspended in chloral hydrate solution for 12 to 72 h for clearing. Subsequently, the ovaries were pressed between two glass slides to visualize the individual ovules and photographed under a SZX2-ILLT microscope (Olympus Corporation, Japan) mounted with an Olympus DP73 camera. The number of ovules for each ovary was manually counted. Descriptive statistical analysis of the phenotypic data was carried out using Microsoft Excel. The ANOVA and H2 analysis were performed in R-software using lmerMod procedure (Bates et al. 2015).

Linkage, association mapping, and QTL analysis

The linkage map used in this study was previously described in Wang et al. (2020a). QTL analysis was performed by composite interval mapping (Zhao-Baang 1994) using WinQTL cartographer 2.5 software (http://statgen.ncsu.edu/qtlcart/WQTLCart.htm). The experiment-wise LOD threshold was determined by permutation analysis (Churchill and Doerge 1994) with 1000 permutations. LOD scores corresponding to P = 0.05 (3.1 for DH) were used for identifying significant QTL. The additive effect (A) and phenotypic variation explained (PVE) by individual QTLs were estimated. For designation and nomenclature of the detected QTLs, the recommendations of McCouch et al. (1997) were adopted. The QTL analysis was carried out on the phenotype of five datasets (DH-EZ16-1, DH-EZ16-2, DH-EZ17, DH-EZ18-1, and DH-EZ18-2). The QTLs were categorized as major and minor QTLs based on the PVE. QTLs with a PVE value of ~ 10 or ≥ 10 with LOD ≥ 3.1 were considered significant QTLs.

The information genomic variation map, population structure, and LD are available in Tang et al. (2021). The association analysis was carried out following single-locus GWAS (SL-GWAS), and multiple-locus GWAS (ML-GWAS) approaches simultaneously. TASSEL V5.0 software was used following generalized linear model (GLM) and mixed linear model (MLM) for the analysis. The GLM was subdivided into a naive model (only considering genotype and phenotype, genotype as an independent variable, phenotype as a dependent variable) and PCs model (adding the first five principal components as covariates to control population structure). MLM was divided into the K model (adding the relationship matrix as covariates) and the PCs + K model (adding the first five principal components and the relationship matrix as covariates). The number of valid tags and calculation threshold was evaluated by GEC software (Li et al. 2012).

Integration of QTL and GWAS results and candidate gene identification

The linkage and association mapping results were integrated according to He et al. (2017). The QTL intervals were aligned to the Darmor-bzh reference genome based on the physical location of the flanking markers for each QTL, and the corresponding QTL regions were extracted. The physical location of each association loci was mapped to the physical region of the QTL. The common regions identified by linkage and association mapping were further mined to identify the possible candidate genes. First, all the genes within 150 kb flanking region for each SNP loci were searched and extracted (He et al. 2017; Ikram et al. 2020). Next, the ZS11 homologues for each Darmor-bzh gene were identified from the BnPIR: Brassica napus pan-genome information resource (http://cbi.hzau.edu.cn/cgi-bin/bnapus/geneindex). The gene annotation and Arabidopsis homologue information were retrieved from BnTIR: Brassica napus transcriptome information resource (http://yanglab.hzau.edu.cn/BnTIR/expression_show). The expression of these genes was compared between ZY-50 and 7–5 (parental lines) at 3–4 mm, 4–5 mm, and 5–6 mm bud length. The genes showing extremely low expression or expressional difference less than two-fold between parental lines were not considered as candidate genes. Genes showing higher or more than two-fold expression differences between the parental lines were considered a putative candidate genes.

Results

Rich variations in ON were extensively observed in both the DH population and the natural inbred accessions

To investigate whether ON remained stable with the development of ovaries, we harvested the ovaries with consecutive sizes (2–7 mm) from the parental accessions. As shown in Fig. 1b, the results showed that the ovary size had no significant effects on the ON. Because large ovaries display obvious advantages in sample preparation and ON counting, 5–6 mm ovaries were used for observations in this study (Fig. 1a). The parental accession, ZY-50, had a smaller ON (25.85 ± 1.36) than 7–5 (32.92 ± 1.46) in all environments (Fig. 1c, Table S1), while ON of their derived DH lines ranged from 24.4 to 44.5, following the normal distribution with transgressive segregation (Figs. 2a and 3a and Table S1). ANOVA confirmed that the genotype of DH lines (G) and the growing environment (E) and genotype by environment (G × E) interactions have significant effects on the ON phenotype. The broad-sense heritability of ON was 0.86 (Table S3). The ON phenotype data of the DH population showed a positive correlation across the environments (Fig. 3).

Fig. 1
figure 1

Observation of the ON. a A 6-mm ovary excised from unopened floral bud with visualized ovules. b Comparison of ON in ovaries of various length ranging from 2 to 7 mm ZY50 and 7–5. c Comparison of ON between ZY-50 and 7–5 in different environments, average ON and SN of ZY50 and 7–5

Full size image
Fig. 2
figure 2

Phenotypic variations and distribution of ON in the a DH and b natural population. Violins and box plots depict the phenotypic distribution of DH lines (5 environments) and natural population (4 environments)

Full size image
Fig. 3
figure 3

Correlation of ON among different environments; a DH population and b natural population

Full size image

Similarly, we also found extensive variations for ON in the natural population across all the four environments, varying from 17.13 to 38.8 (Table S2). Normal distribution of ON was observed across the environments (Figs. 2b and 3b and Table S2), and the phenotypic correlation of ON was significantly positive among the four growing locations (Fig. 3). The ANOVA revealed that the genotype (G) and environment (E) had significant effects on the ON phenotype in the natural population (Table S4), and the broad-sense heritability of ON reached 0.93, indicating the stability of the ON and small effects of the environment on the trait.

Identification of QTLs linked with ON from the DH population

QTL mapping based on the DH population in five environments detected five loci for ON located on chromosomes A03, A07, A10, and C06, respectively (Fig. 4), with the detail information of each QTL given in Table 1. These QTLs explained 2.00–17.40% of the phenotypic variations for ON, respectively. qONA-07–2 was detected in DH-EZ16-1, DH-EZ16-2, DH-EZ18-1, and DH-EZ18-2, and qON-C06 was detected in three environments (DH-EZ16-2, DH-EZ18-1, and DH-EZ18-2). qON-A07-1 was detected in DH-EZ16-1 and DH-EZ16-2, while qON-A10 was detected in the DH-EZ16-1 and DH-EZ17 environments. qON-A03 was observed only in the DH-EZ16-1 environment (Table 1). The linkage and physical locations of these QTLs are also visualized on Circos (Fig. 4).

Fig. 4
figure 4

Genetic linkage map of ON based on ZY50-75 DH population. Different environments are indicated by different color backgrounds on the cycle. From outside to inside, five cycles represent five environments, EZ16-1, EZ16-2, EZ17, EZ18-1, and EZ18-2, respectively. The two outer-most cycles represent comparison of linkage map and physical map of B. napus. Red bars within the cycles indicate QTL regions on chromosomes

Full size image
Table 1 Summery of the detected QTLs for ON in the ZY-50–7-5 DH population
Full size table

From these QTL, we found that qONA-07–2 and qON-A10 show larger effects, with the PVE of 17.38% and 9.32%, respectively. The confidence intervals for these QTLs were individually 27 cM and 26 cM for qONA-07–2 and qON-A10 on the linkage map. Correspondingly, the candidate physical interval of qONA-07–2 is about 5.75 Mb on chromosome A07, while the qON-A10 has a 3.65 Mb physical interval on chromosome A10.

Association mapping analysis

GWAS was carried out for four environment and BLUP using GLM model (Fig. 5, Fig. S1). The GLM method detected 247, 757, 148, 80, and 214 significant QTNs in NP-WH17, NP-CD17, NP-WH18, NP-EZ18, and BLUP, respectively. The significant QTNs in NP-WH17 were detected on chromosomes A02, A07, A08, A10, C02, C03, C06, and C08. The QTNs detected in NP-CD17 were located on chromosomes A02, A03, A06, A08, A09, A10, C02, C03, C06, and C08. In NPWH18, the significant QTNs were located on chromosomes A01, A02, A07, A08, A09, C02, C03, and C04, while in NP-EZ18, the significant QTNs found on chromosomes A02, A06, A07, A08, C02, C08, and C09. The significant QTNs in BLUP were located on chromosomes A02, A07, A08, A09, C02, C03, C04, C06, and C08 (Fig. S1, Supp File S1). The PVE by these QTNs ranged between 5.03 and 7.33%, suggesting that ON is controlled by multiple genes with a small effect in the association panel.

Fig. 5
figure 5

Manhattan and QQ plot in the detection of SNPs for ON using single-locus GWAS based on BLUP values (green dashed line, suggestive threshold; red dashed, significant threshold)

Full size image

Further, we performed ML-GWAS approaches, including MrMLM and FAST-MrMLM, on the same datasets as individual environments and the BLUP. Using mrMLM, we detected 46 QTNs in NP-WH17 and NP-CD17 each, 41 in NP-WH18, 37 in NP-EZ18, and 48 in BLUP. The PVE of these QTNs ranged between 3.92 and 21.22%. The FAST-MrMLM identified 44 QTNs in NP-WH17, 47 in NP-CD17 CD, and 46 in NP-WH18, while 40 SNPs in NP-EZ18 and 40 QTNs in BLUP (Fig. S2, Supp File S1). The explained PVE of these SNPs ranged from 3.73 to 21.20%.

Comparison of the QTLs between DH and association mapping

The physical location of each QTL from the DH population and SNP loci from the natural population was compared to mine the common loci between the two populations. The linkage analysis detected QTLs on chromosomes A03, A07, A10, and C06, while the association analysis identified loci on chromosomes A02, A03, A07, A08, A10, C02, C03, and C06 for ON. By integrating linkage and association mapping loci, four common genomic regions were identified on chromosomes A03, A07 and A10. Within these genomic regions, 31 significant SNPs were distributed (Table 2). Eight significant SNPs on chromosome A03 detected in GWAS were located in the CI of qON-A03 (2.48–6.44 Mb). Five SNPs detected by GWAS fall within the CI of qON-A07-1 (0.28–8.99 Mb). Twelve significant SNPs were detected in the CI of qON-07–2 physical region. In qON-A07-2, five SNPs are clustered within 400-bp interval (17,249,274 ~ 17,249,692). Six SNPs on chromosome A10 detected in GWAS overlapped with the CI of qON-A10 (9.6–13.25). Among these six SNPs, three are clustered in a ~ 240 bp region (11,839,872 ~ 11,840,110). Furthermore, C06 also harbored three loci detected in GWAS and one QTL, qON-C06; however, the physical locations of these loci and QTLs were separated by large genomic interval (23.95 Mb and 5–15 Mb, respectively).

Table 2 Consensus loci between DH and natural population for ON
Full size table

Candidate gene prediction

We then analyzed the annotated genes in the consensus regions identified commonly by linkage mapping and GWAS. A total of 269 genes distributed in the flanking regions of 12 significant SNPs located in qON-A07-2, while 145 genes flanked around the 6 SNPs located in qONA-10. According to the expression data derived from the developing flower buds of ZY-50 and 7–5, we further selected 54 and 37 putative genes from the candidate regions of qON-A07-2 and qON-A10, respectively (Fig. 6, Supp File 2).

Fig. 6
figure 6

Differential expression of genes predicted in qON-07–2 and qON-A10 regions (FPKM)

Full size image

Based on the functional annotations of the putative candidate genes located in the qON-A07-2 interval, we speculated that several genes might be involved in ON determination, including BnaA07g22900D, BnaA07g27510D, BnaA07g25810D, BnaA07g27570D, and BnaA07g27740D. BnaA07g22900D is an orthologue of Arabidopsis gene HTH (AT1G72970), which has been reported to be involved in floral organ development. The mutants showed an aberrant embryo sac development, floral organ fusion, and defective ovules (Lolle et al. 1998; Krolikowski et al. 2003; Pagnussat et al. 2005). The Arabidopsis CML23 (AT1G66400) is orthologue to BnaA07g25810D. CML23 reported to be involved in plant development and transition to reproductive phase. CML23 in conjugation with CML24 regulated floral organ development (Tsai et al. 2007; Nie et al. 2017; He et al. 2020). BnaA07g27510D is an orthologue of Arabidopsis gene SOFL (AT1G68870). SOFL1 and SOFL2 are positive regulator of cytokinin homeostasis and CK-mediated development (Zhang et al. 2006, 2009). CIB1 (AT1G68920) is the Arabidopsis orthologue for BnaA07g27570D. CIB1 is basic-helix-loop-helix (bHLH) transcription factor that regulate floral initiation (Liu et al. 2013). BnaA07g27740D is the orthologue of the Arabidopsis CRC (AT1G69180). Previous studies report CRC to be an ovule development and floral organ development regulator (Kuusk et al. 2002; Orashakova et al. 2009; Skinner and Gasser 2009; Liao et al. 2020). Knockdown of CRC orthologue in E. californica and P. sativum caused defective carpel and ovule initiation (Orashakova et al. 2009; Fourquin et al. 2014). BnaA07g31240D is orthologue to Arabidopsis EXT3/RSH (AT1G21310). EXTINSINs reportedly functions in call wall assembly in the rapidly growing cell in meristems (Saha et al. 2013; Choudhary et al. 2015). We also found some genes that showed high expressional abundances at the bud level but uncharacterized functions, including BnaA07g27560D (AT1G68910), BnaA07g27670D (AT1G69050), and BnaA07g27750D (AT1G69200).

Similarly, we also identified several genes with known functions that could be related to ON formation from the candidate region of qON-A10. BnaA10g14760D is an orthologue of AT5G20730 that encodes auxin response factor (ARF7). The ARF7 create auxin responsiveness in MND1 and regulate plant development (Li et al. 2021).. BnaA10g14910D is the orthologue of Arabidopsis AT5G20570 that encodes RING-BOX / RBx1. AtRBx1 is part of the SCF-complex an E3-Ubiquitine ligase. Downregulation of AtRBx1 impair developmental aspects including floral development (Ni et al. 2004; Bernhardt et al. 2006; Chen et al. 2006). AT5G17300 is the Arabidopsis orthologue of BnaA10g17370D that encodes REVEILLE (RVE1). RVE1 is Myb-like transcription factor that regulate the auxin level. RVE1 is reported to be expressed in ovule primordia (Skinner and Gasser 2009), hypocotyl growth, auxin response, and seed development (Rawat et al. 2009; Jiang et al. 2016). BnaA1017060D is orthologue of AT5G17690 that encodes LHP1. LHP1 cooperate with BPC and MADS-domain factors to orchestrates the SKT activity during floral development (Petrella et al. 2020). BnaA10g16730D is orthologue to AT5G18090 that encodes an AP2/B3 like transcription factor that belongs to REM family. The REM family genes are involved in early stage of floral and ovule primordia development (Kelley and Gasser 2009; Mantegazza et al. 2014). In wheat AP2-like transcription factors (AP2L2 and AP2L5) are redundantly involved in floral development and regulate MAD-Box floral genes (Debernardi et al. 2020). In the qON-A10 interval, some genes showed high expressional abundances at the bud level but uncharacterized functions, including BnaA10g16530D (AT5G18310), BnaA10g17120D (AT5G17620), BnaA10g17220D (AT5G17510), BnaA10g17250D (AT3G03341), and BnaA10g17390D (AT5G17280).

Integration of the known QTLs for SY-related traits

To assess whether the QTLs detected in the present study are in the genomic regions important for oilseed rape breeding, we compared the physical interval of our QTLs with the previously reported QTLs. The comparison was restricted to SY-related traits only, like seed number (SN), the number of silique/pods (NP), and seed weight (SW) in different mapping populations following Raboanatahiry et al. (Raboanatahiry et al. 2018, 2022). We searched the previously published QTLs within the physical interval of the QTLs detected in our study (Table 1). Several loci related to SN, SW, NP, SL, biomass, and SY were found within the interval of our QTLs (Supp File S3). Interestingly, all the overlapping loci were found to be located on the A genome.

In total, 118 previously reported loci were identified that overlap with four QTL identified in our study. Further observation revealed that majority of these loci is related to SW and SN. It is crucial to identify regions that influence multiple traits, especially the closely related traits, i.e., SW and SN. The qON-A03 CI includes 18 previously reported loci. These loci are related to SW (7), NP (2), and SY (9 loci) (File S3). Forty-three previously reported loci were found in the qON-A07-1 interval. Among these, five loci are related to SN, twenty-seven related to SW, three related to biomass, and four loci are SY related. The qON-A07-2 interval harbors thirty-four loci. Among these, fourteen loci are related to SN, and two (SW), five (biomass), eight (SP), and four are SY-related loci. Within the qON-A10 interval, 23 previously reported SY-related overlaps (File S3). The qON-A07-1 interval was found to have the highest number of overlapping regions (34) with previous studies. In combination with the previous studies and present results, it can be concluded that these genomic regions might be of great potential for oilseed rape breeding.

Discussion

QTL mapping has been proved to be a potential tool to unveil the genetic mechanisms of complex agronomic traits (Wang et al. 2020a) and is frequently utilized in B. napus for several traits. The combination of linkage and association mapping approaches further aids the detection by identifying common and stable loci with strong genetic control of the trait (Hu et al. 2011; He et al. 2017). In the present study, we investigated the phenotypes and genetics of ON through a biparental population and in a natural population consisting of comprehensive inbred accessions in different environments. We found that ON showed a wide range of variations in both populations inheriting as a typical quantitative trait. The ON data was subjected to linkage and association analysis. This study identified five ON-related QTLs via linkage mapping while 214 significant loci via association mapping (SL-GWAS). MrMLM and FASTMrMLM (ML-GWAS) also detected 48 and 40 significant loci associated with ON. Interestingly, both approaches identified common genomic regions that control the ON. This ascertains the accuracy of our mapping results as both approaches augment each other. To further intuit the accuracy of the identified loci, the genomic regions underlying these loci were compared to the previously reported SY-related QTLs. Since previous studies associated with yield-related traits in oilseed rape majorly focused on SN, SP, SL, and SW, our findings will strengthen understanding of the genetic basis of yield components in oilseed rape.

The strategy of combining linkage mapping with association mapping has been proposed to promote the identification of the causal genes for a quantitative trait (He et al. 2017; Wang et al. 2020a). Linkage mapping utilizes and associates recent recombination events (biparental populations) to the traits, while association mapping relies on the historic recombination (natural populations) accumulated over the course of time (Nordborg and Weigel 2008; Li et al. 2014). Therefore, the combination of both approaches might be helpful in uncovering the consensus loci based on recent mutations or recombination. In the present study, consensus loci were found on ChrA03, A07, and A10 in linkage mapping and GWAS (Table 2). Four QTLs on the A genome (qON-A03, qON-A07-1, qONA-07–2 qON-A10) overlapped with the SNPs detected in the association analysis. Among these consensus genomic regions, the major effect QTLs (qON-A07-2 and qON-A10) were searched for the putative candidate genes. These results suggest that common loci detected in linkage and association are stable loci and provide strong genetic control of the traits (He et al. 2017). Linkage and association approaches have been used simultaneously to identify QTLs in oilseed rape and cotton (He et al. 2017; Liu et al. 2018) to detect branch morphogenesis and fiber quality-related loci. However, the possibility of consensus loci identification is low because of the genetic backgrounds of biparental and natural populations. The consensus genomic regions identified in our study carry important genes involved in the floral and morphological development, cell fate, and specification and hormonal response, particularly auxins and cytokinin. The role of auxins and cytokinin and their crosstalk in ovule initiation and development is well elucidated in Arabidopsis (Galbiati et al. 2013; Cucinotta et al. 2016). Functional analysis of these genes in oilseed rape will open up new ways towards understanding the molecular mechanism of ovule development and number determination.

In B napus, 2,438 QTLs have been identified for 79 yield-related traits (Raboanatahiry et al. 2022). Several of these identified QTLs overlap or coincide with each other. Further, these QTLs were also found to affect other traits (Raboanatahiry et al. 2022). QTL comparison or colocalization from different mapping population having diverse genetic backgrounds aids the identification and validating stable loci (Li et al. 2018). Several QTLs reported in the previous were found to overlap or correspond to the QTLs identified in this study. However, we only selected the direct yield component traits for comparison to our QTLs. For instance, the qONA-3 likely corresponds to SW QTLs; qSW.A03-1, qSW012, and DHqSW06 (Shi et al. 2009, 2011; Wang et al. 2020a). We identified and co-located 118 previously reported loci to our present QTLs. Among these 27 QTLs for SW, 22 for SN, 12 for SP, and 12 were collated for SY. The qONA-3 and qSW.A03-1 were detected in the same DH population. Similarly, the qON-A07-1 and qON-A07-2 are likely to complement to TSWA7a, cqSW.A07-1, TSWA7b, and cqSW.A07-2, (Fan et al. 2010; Wang et al. 2020a). The overlapping loci that affect multiple traits might be suitable for selection to improve the desired traits, simultaneously (Raboanatahiry et al. 2022). In B. napus, several QTLs have been reported that could influence more than one trait simultaneously (Jiao et al. 2021; Raboanatahiry et al. 2022; Liu et al. 2022). Integrating QTLs with overlapping intervals for different traits obtained significant co-localization of QTLs or pleiotropic QTLs (Zhao et al. 2016, 2019). This colocalization of QTLs from different traits indicates a strong inter-relationship or dependence on each other (Wang et al. 2010; Xin et al. 2021). This also suggests that these loci contain many tightly linked trait-specific genes or genes that affect multiple traits (Hall et al. 2006). This further suggests that the selection of these loci might aid the simultaneous improvement in more important traits. The qON-A07-2 and qON-A10 control ON, however its overlapping QTLs in other populations the control SN, SW and SP. Selection of these loci in breeding program will be helpful for simultaneous imprudent in other yield component traits, i.e., SN.

In recent GWAS studies, Khan et al. (2019) and Qadir et al. (2022) analyzed SN, SW, and ON and reported 8 and 18 significant SNPs associated with ON. However, comparing these results, no common QTN was found for ON. By contrast, Khan et al. (2019) identified five SW and five SN-related SNPs that overlap in the QTL intervals qON-A03 (3), qON-A07-1 (1), qNO-A07-2 (1), and qON-A10 (5). Interestingly, one of these SNP on chromosome A03 corresponds to BnaA03g55500D (GA20OX3: Gibberellin 20-oxidase 3). Two SNPs on the A10 corresponds to BnaA10g12800D (GASA10: Gibberellin-regulated family protein) and BnaA10g16730D, respectively. BnaA10g16730D is a homologue of the Arabidopsis gene AT5G18090 that encodes an AP2/B3-like transcriptional factor family protein. Since AP2 was previously reported to regulate the floral organ patterning, including ovule and ON number in Arabidopsis (Modrusan et al. 1994; Elliott et al. 1996; Krizek 2009; Huang et al. 2013), it is likely to speculate that BnaA10g16730D is a candidate of qON-A10. At unopened bud level, ZY-50 and 7–5 shows 5–tenfold expression differences (Fig. 6, Supp File S2). The expression differences between ZY-50 and 7–5 for BnaA10g16730D possibly suggest its involvement in ON development. However, the functional validation of these genes in oilseed rape is important to confirm the possible role and elucidate the understanding of ON development and number control.

Conclusion

We performed linkage mapping and association analysis based on a DH population and a panel of inbred accessions, respectively, to preliminarily determine the genetic structure of ON. The results showed that ON could inherit with a high broad inheritability. Linkage and association mapping co-detected consensus in four genomic regions on chromosomes A03, A07 and A10. These loci contain 8, 5, 12, and 6 SNPs. Two QTL, qON-A07-2 and qON-A10, show a relatively major effect. Based on the results from the linkage and association mapping, it can be concluded that several loci control the ON with small effects on the phenotype. Conceivably, these common loci may be conserved among the genetically diverse population causing variations in the phenotype. The putative genes underlying these loci are related to floral development, hormonal signaling, and carbohydrate metabolism. Understanding the role of these loci in the ovule development and determination will contribute to the yield potential of oilseed rape.