Phenotypic variation for SN-related traits in Parents, F1 and DH population
The DH lines and their corresponding parents as well as F1 were grown in the winter-type growing seasons. The phenotypic performances of six SN-related traits across the years were investigated (Table 1; Fig. 1). The average SN of C4-146 was nearly two times that of C4-58B in each investigated environment. However, the ON of the two parents was not signiﬁcant in the three environments (Table 1), suggesting that the decrease of SN in C4-58B mainly result from defects after ovule initiation. Also, C4-146 had higher SL, lower AON, PAO and TSW than C4-58B in different environments (Table 1). The SN of F1 plants from the reciprocal crosses was all significantly lower than C4-146 but similar to C4-58B (Table 1), indicating the dominant nature of low SN over high SN in the cross between C4-146 and C4-58B. The other SN-related traits of F1 plants from the reciprocal crosses also displayed a similar trend with SN (Table 1).
The phenotypic performances of SN-related traits in the DH population were shown in Table 2 and Fig. 2. Continuous distributions and transgressive segregations in the DH population suggested a quantitative inheritance pattern for these traits. The distributions of SN, AON and PAO deviated from normality and appeared to have a bimodal pattern, indicating the involvement of a major gene and some minor genes for each of these traits. Whereas, SL, TSW and ON showed nearly normal distribution across all environments, indicating the quantitative inheritance of these traits in the present population. Two-way ANOVA of these traits across environments indicated that the genotype of each line (G), environment (E) and genotype-environment interactions (G × E) had signiﬁcant effects respectively, except for the environmental effect on PAO (Table S3). In this population, SN showed the highest broad-sense heritability (0.97), followed by TSW/PAO (0.91), and AON (0.85), suggesting that these traits are stably inherited. While the SL and ON have broad-sense heritability of 0.69 and 0.66, respectively, indicating that SL and ON are relatively sensitive and plastic to environmental variations.
Signiﬁcant phenotypic correlations were observed for most of these traits across three years (Table 3). The only exception is the ON trait which shows no significant correlations with SN, SL and TSW, indicating that ON is not influencing these three traits in this population. SN showed a highly negative correlation with PAO, AON and TSW, but had a highly positive correlation with SL, suggesting that the ovule abortion is a key negative regulator controlling SN and longer silique will produce more but smaller seeds.
Primary mapping of SN‑related QTLs in the DH population
Because the two parents C4-146 and C4-58B were selected from our previously reported HJ DH population which was used to construct a high-density genetic map (Cai et al. 2014). We first analyzed the polymorphism distribution between C4-146 and C4-58B at the whole-genome scale using more than 2,000 markers (Fig. S1). Subsequently, 179 molecular markers that covered all the polymorphic regions between the two parents were selected to genotype the DH population. A genetic map containing 17 chromosomes spanning a total of 684 cM of the B. napus genome was constructed, with the lack of chromosomes A6 and C1 because of lack of polymorphism between the parents (Fig. 3).
A total of 40 QTLs for six examined traits were identified in the three environments with LOD values ranging from 2.78 to 48.48, which were distributed on 12 chromosomes (Table S4; Fig. 3). The phenotypic variation explained by individual QTL varied between 2.43% and 56.57% (Table S4). These identified QTLs were integrated into 19 reproducible consensus QTLs for each trait by QTL meta-analysis (Table 4). Among the identified QTLs, 6, 3, 4, 2, 1 and 3 consensus QTLs were detected for SL, SN, TSW, ON, AON and PAO, respectively. Eight consensus QTLs, including cqSL.A8, cqSN.A8, cqTSW.A7-1, cqTSW.A7-2, cqTSW.A8, cqON.C6-2, cqAON.A8 and cqPAO.A8, showed main effects and the others exhibited minor effects. Positive alleles for SN and ON and negative alleles for TSW, AON and PAO both originated from the C4-146 parental line while positive alleles for SL were dispersed between the two parents, in consistency with the trait performance of the two parents (Table S4; Table 4).
It is worth noting that cqSN.A8, flanked by BM1355 and BM1360 on chromosome A8, was a newly identified QTL and had the largest LOD scores (from 35.60 to 48.48) and explained as much as 56.57% of the SN variation in all environments. To evaluate the effect of cqSN.A8, the DH lines were classiﬁed into two groups (AA and BB) based on the genotype of the two flanking markers BM1355 and BM1360. Lines carrying the C4-146 alleles (AA) had an average SN of 19.20 ± 2.04, 18.45 ± 2.46 and 20.42 ± 2.26 in 2016, 2017 and 2018 environments, respectively, which were signiﬁcantly higher than those carrying the C4-58B alleles (BB) with an average SN of 11.99 ± 2.19, 12.5 ± 2.25 and 12.88 ± 2.25 across the three environments. Clearly, the two genotypes of the DH lines at the cqSN.A8 locus corresponded exactly to the two distinct phenotypic classes, exhibiting bimodal distribution with 15-18 SN as the boundary (Fig. 2). Similarly, two consensus QTLs on chromosome A8 (cqAON.A8 and cqPAO.A8) also displayed a large effect (which explained about half of the trait variation in all environments), justifying the bimodal distribution exhibited by the two traits in the DH population (Fig. 2). The remaining two major consensus QTLs on chromosome A8 (cqSL.A8 and cqTSW.A8), which explained 36.74% and 15.82% of the phenotypic variance of SL and TSW on average, exhibited relatively minor effects, consistent with the nearly normal distribution of these traits in the DH population (Fig. 2). Interestingly, the confidence intervals of the five consensus QTLs on chromosome A8 for SN, AON, PAO, SL and TSW overlapped perfectly (Table 4).
Further meta-QTL analyses integrated these overlapped QTL into a unique QTL un.A8 (Table 5), which has a major pleiotropic effect on all these traits except for ON. The C4-146 alleles at un.A8 causes an increase in SN and SL and a simultaneous decrease in AON, PAO and TSW respectively, justifying the significant correlations observed among these traits. Additionally, another two other unique QTLs were identified on chromosomes A4 and A7 (Table 5). The unique QTL un.A7 showed a major effect on TSW but had a minor effect on SN, SL, and PAO. Also, the C4-146 alleles at un.A7 causes an increase in SN and SL, and a simultaneous decrease in PAO and TSW, which is consistent with the correlations observed among these traits. For the minor unique QTL un.A4, the increasing allele for SN was inherited from C4-146 and the increasing allele for PAO was derived from C4-58B, reﬂecting the signiﬁcant negative correlations between SN and PAO.
Fine mapping of the un.A8 locus
Due to the ease of assessment, SN was selected as the target trait for the fine mapping of un.A8 locus. To narrow down the genomic region containing un.A8, 24 polymorphic INDEL markers from the candidate region were developed and 18 recombinants between the BM1355 and BM1360 intervals in the DH population were screened. Based on the allelic composition and recombination breakpoints, the recombinants were grouped into three genotypes (Fig. 4a). The mean phenotypic value for each recombinant genotype was compared with the two parents to locate un.A8. Using these data, un.A8 was narrowed down to a region between BM1663 and BM1646, corresponding to an 820-kb region on chromosome A8 in the reference genome of B. napus cultivar Darmor-bzh (Chalhoub et al. 2014; Fig. 4a).
To further narrow down the position of QTL un.A8 and identify the candidate genes responsible for SN, a large F2 population (containing 16,421 individuals obtained from the cross between C4-146 and C4-58B) was constructed and surveyed using the two flanking markers BM1663 and BM1646. Considering the dominant nature of low SN over high SN in the cross between C4-146 and C4-58B, only recombinants between the C4-146 homozygous genotype (AA) and the other two genotypes (BB or Heterozygous) are informative for fine-mapping. Then, 114 recombinants were identiﬁed and used for further genotyping with 10 markers within the candidate region. These recombinants were grouped into 12 genotypes according to the allelic composition and recombination breakpoints (Fig. 4b). The recombinants information indicated that the BM1638, BM1641, BM1642, BM1357, BM1664, BM1667 and BM1668 markers were located on one side of un.A8, while the BM1672 marker was located on the other side. Also, BM1669 and BM1670 were found to co-segregate with the un.A8 locus (Fig. 4b). Therefore, the genomic region containing the un.A8 locus was narrowed down to the DNA fragment bounded by BM1668 and BM1672, with an approximate length of 80.1-kb in the Darmor-bzh genome (Fig. 4c).
Candidate gene analysis of un.A8
Based on the released rapeseed genome information, the candidate interval contained six predicted genes in the Darmor-bzh genome (designated BnaA08g07900D through BnaA08g07950D) corresponding to 3 and 4 genes in the B. napus ZS11 and Shengli genomes, respectively (Song et al. 2020; Fig. 4c; Table 6). Among these putative genes, one gene encodes expressed protein without annotation while the other genes encode homologs of Arabidopsis ATPase family AAA domain-containing protein, F-box/kelch-repeat protein, related to ubiquitin 1, photosystem I subunit F, and zinc transporter 10 precursor (Table 6). None of the six genes has been previously reported to be involved in SN regulation. Therefore, un.A8 might be a novel QTL responsible for SN control.
To identify the candidate gene, the transcript levels of the six genes in both parents across five stages during ovary and silique development (1-2 mm, 2-4 mm and 4-5mm pistil, 3DAP and 5DAP ovary) were quantified using qRT-PCR. The transcript levels of BnaA08g07940D and BnaA08g07950D were significantly higher in C4-58B than C4-146 during ovary and silique development (Fig. 5d), which supports the dominant nature of the low SN phenotype. However, BnaA08g07930D was expressed in C4-146 but not in C4-58B during ovary and silique development. Through sequence comparisons, we identiﬁed a 1,445-bp deletion starting from 1,015-bp upstream of the transcription start site to the mid region of the second exon of BnaA08g07930D in C4-58B compared with the C4-146, which resulted in the formation of a pseudogene gene in C4-58B (Fig. 5a; Fig. S2). The remaining three genes, BnaA08g07900D through BnaA08g07920D, were not expressed in the developing ovary and siliques. These results suggested that both BnaA08g07940D and BnaA08g07950D may contribute to the observed variations in SN, hence we focused on them for further analysis.
To characterize the relationship between the candidate genes and the variations in SN, the promoter sequences, full-length genomic DNA, and coding sequences of BnaA08g07940D and BnaA08g07950D in the parental lines were amplified. A total of 10 exonic SNPs were detected in the open reading frame of BnaA08g07940D with 2 non-synonymous SNPs which resulted in amino acid variations in the signal peptide (Fig. 5b; Figs. S2 and S3). In the promoter region of BnaA08g07940D, abundant polymorphisms were also identified between the parental lines, including a 6,159-bp insertion (starting from 1,019-bp upstream of the transcription start site), 7 small INDELs and 7 SNPs in C4-58B (Fig. 5b; Fig. S5). For another candidate gene BnaA08g07950D, a 3-bp insertion and an exonic SNP were detected in the first exon of C4-58B, which resulted in an addition of Valine acid residue at the 21st amino acid (aa) and substitution of phenylalanine to leucine at the 28th amino acid (Fig. 5c; Figs. S6-S8). The BnaA08g07950D allele also showed alternative splicing, which terminated at the first intron and resulted in 8-aa insertion and 150-aa truncation in the C-terminus of the predicted protein in C4-58B (Fig. 5c; Figs. S6-S8). Additionally, 11 SNPs and 2 small INDELs were identified in the promoter interval (~ 3.0 kb before the start codon) between the two parents (Fig. 5c; Fig. S9). Thus, these results further supported the idea that BnaA08g07940D and BnaA08g07950D were the most promising candidate genes underlying the un.A8 locus.