An effective strategy in developing B. juncea-type hybrids, i.e., the hexaploid bridging strategy was developed in our previous study (Mei et al., 2022) and employed in the present study to generate new-type B. juncea germplasm. In brief, 32 B. nigra accessions were randomly cross-pollinated with three A^jA^jA^rA^rB^jB^j hexaploid genotypes (designated MS2, MS3 and MS4) to produce A^jA^rBⁿB^j-type F₁ hybrids, and 20 B. rapa accessions were crossed with one A^jA^jB^jB^jBⁿBⁿ (designated MS6) hexaploid to produce A^jA^rBⁿB^j-type F₁ hybrids. Resultant F₁ hybrids were self-pollinated to produce F₂, F₃ and other generations. From the above B. rapa panel, the European fodder turnip ECD04 (AA) — a known donor harboring multiple resistance genes against Plasmodiophora brassicae — was specifically crossed with the hexaploid bridge line MS6 to initiate CR gene transfer. Phenotypic observations (e.g., leaf morphology) and imaging were conducted concurrently at designated growth stages (e.g., seedling stage) to ensure direct comparability across all genotypes. Fertility was assessed by counting the number of seeds per pod from 30 consecutive pods collected from the primary inflorescence of each plant. All materials, including the parental lines (B. nigra, B. rapa, hexaploids MS2-MS6 and ECD04), the derived new-type lines (F₁–F₅), and ten control accessions of current B. juncea (vegetable and oil types) from Southwest University (SWU), were sown and transplanted on the same dates in the experimental field at SWU. A uniform planting design was maintained with 30 cm between rows and 25 cm within rows. Detailed information for all accessions and crosses is provided in Supplementary Table S1.

To understand the genetic characteristic of new-type B. juncea germplasm, 61 new-type B. juncea individuals were randomly selected from the F₁ and F₂ generations for whole-genome resequencing, together with 20 B. nigra parents, the four hexaploid parents, three B. rapa parents, and ten current B. juncea accessions. Young leaf samples from these materials were submitted to Wuhan Benagen Biotech Co., Ltd. for whole-genome resequencing using BGISEQ-500. Sequencing yielded an average depth ranging from 9.3× to 16.6× per sample. Following whole-genome resequencing, the genomic sequence data were subjected to quality assessment and filtering using fastp, which included: removal of adapter sequences, trimming of poly G and poly X tails from reads, discarding reads containing more than 5 N bases, removing reads in which the proportion of bases with quality scores below 15 exceeded 40%, eliminating reads with an average base quality score lower than 20, and discarding reads shorter than 50 bp after filtering.

Clean reads were aligned to the reference genome of B. juncea (BjuIR: Brassica juncea information resource) using BWA. Variants were called using Genome Analysis Toolkit (https://wiki.rc.usf.edu/index.php/Genome_Analysis_ToolKit_GATK), with SNPs and InDels filtered based on specific criteria. Genetic distances among materials were calculated based on the SNPs and InDels using Powermarker V 3.25 (https://brcwebportal.cos.ncsu.edu/powermarker/). Phylogenetic trees were constructed using MEGA software (https://www.megasoftware.net/), and principal component analysis (PCA) was performed using Rstudio.

Genomic DNA was isolated from fresh young leaf tissues of all plant materials using a modified cetyltrimethylammonium bromide (CTAB) method. Approximately 100 mg of tissue was ground in liquid nitrogen and incubated in pre-warmed CTAB extraction buffer at 65 °C for 45 min. After chloroform: isoamyl alcohol (24:1) extraction, DNA was precipitated with isopropanol, washed with 70% ethanol, and dissolved in nuclease-free water. DNA concentration and quality were assessed using a NanoDrop spectrophotometer and agarose gel electrophoresis.

To track the introgression of CR genes, genotyping was performed using four PCR-based markers (Powell et al., 1996): CR-m090a (linked to PbBa8.1), CR-BSA3 (linked to PbBa1.1), and CR-S14R14 and CR-S17R17 (both linked to BraA.CR.b) (Chen et al., 2013a; Hirani et al., 2018; primer sequences in Supplementary Table S2). All F2 individuals from the cross MS6 × ECD04 were screened using these markers, and those carrying the maximum number of ECD04-specific alleles were self-pollinated Selection was carried out in subsequent generations (F3 to F5) by genotyping ten plants per line with the same markers. PCR was carried out in 10 µL reactions containing approximately 50 ng genomic DNA, 0.5 µL each of forward and reverse primer (10 µM), 5 µL of 2× PCR Master Mix (Vazyme, China), and nuclease-free water to volume. The thermal cycling conditions were as follows: initial denaturation at 94 °C for 5 min; 10 cycles of touchdown PCR (94 °C for 30 s, annealing from 62 °C to 57 °C decreasing by 0.5 °C per cycle for 30 s, 72 °C for 30 s); followed by 25 cycles of standard PCR (94 °C for 30 s, 55 °C for 30 s, 72 °C for 30 s); and a final extension at 72 °C for 5 min. Reactions were held at 16 °C post-amplification. The products were separated by electrophoresis on a 2% agarose gel. Clear bands exceeding the background intensity were scored as 1 (presence of the resistance allele), while faint or absent bands were scored as 0. The ECD04 allele frequency (AF, %) was calculated as.

where Np indicates the number of plants identified as positive for the molecular marker, and Nt represents the total number of plants assayed. The line graph depicting ECD04 allele frequencies for each marker was generated using GraphPad Prism.

To evaluate clubroot resistance, the plants were inoculated with Plasmodiophora brassicae race 4 using the injection method in lab (Jubault et al., 2013) under controlled conditions (22–25 °C, 60% relative humidity). Three replicates were set up for each tested line, with 30 seeds sown per replicate. Disease incidence (DI) and severity were assessed 42 days post-inoculation. Based on root symptoms and gall size, disease severity was classified on a 0–4 scale (Chen et al., 2021), with 0 indicating no symptoms and 4 representing the most severe root deformation. The DI was calculated as follows (Ludwig-Müller et al., 2017):

where N1-N4 indicate the number of plants in each disease severity category, and Nt represents the total number of inoculated plants. The correlation between the DI and the ECD04 AF was calculated and graphically presented using GraphPad Prism.

Through hybridization between hexaploids and diploids, followed by successive selfing, we developed 180 new-type B. juncea lines, comprising 76 F₁ hybrids (from direct crosses), 104 advanced selfing lines (46 F₂, 26 F₃, 23 F₄ and 9 F₅ generation). To evaluate the fertility of the new-type B. juncea lines, we quantified the seed set rate, measured as seeds per pod, for the F₁, F₂, and F₃ generations. Statistical analysis across various generations (Figure 1A) revealed that in the F₁ and F₂ generations, hybrids derived from MS2, MS3, and MS4 (A^jA^jA^rA^rB^jB^j) exhibited higher seed set rate compared to those derived from MS6 (A^jA^jB^jB^jBⁿBⁿ), suggesting that the different hybridization combinations significantly affect the fertility of hybrids, i.e., the cross of hexaploid A^jA^jA^rA^rB^jB^j with B. nigra is easier to produce hybrids with better fertility in early generations than the cross between hexaploid A^jA^jB^jB^jBⁿBⁿ and B. rapa. More importantly, the data demonstrated that the seed set rate of the hybrids significantly improved with successive generations, particularly noticeable in the hybrids derived from MS6. This indicates that successive selfing can effectively enhance the fertility of all hybrids, regardless of whether they are hybrids of A^jA^jA^rA^rB^jB^j × A^rA^r or A^jA^jB^jB^jBⁿBⁿ × BⁿBⁿ.

Consistent with fertility restoration, the new-type B. juncea lines showed dynamic morphological changes from the F₁ to F₃ generations (Figures 1B-M). The F₁ plants exhibited considerable phenotypic variation, including pinnately lobed leaves, leaf trichomes, purple venation and stem, and prominent venation. In the F₂ generation, while phenotypic variation persisted, a few plants began to converge toward current B. juncea traits, with certain F₁ features such as trichomes and serrated leaf margins becoming less distinct. By the F₃ generation, some plants clearly converged toward the morphology of current B. juncea cultivars. These morphological changes demonstrate that successive selfing generations enabled the selection of plants with a high degree of phenotypic identity to current B. juncea from the new-type lines.

The observed convergence in both fertility and phenotype suggested underlying genetic changes. To elucidate the genetic characteristics of the new-type B. juncea, whole-genome resequencing was conducted on the 61 new-type B. juncea lines (24 F₁ and 37 F₂), alongside 23 parental B. rapa and B. nigra accessions, 10 current B. juncea accessions (including vegetative and oilseed types), and four hexaploid parents. After strict filtering, 24.6 million high-confidence SNPs and small indels were retained, spanning the A and B sub-genomes, with a mean density of 1.36 million variants per chromosome.

As revealed by phylogenomic analysis (Figure 2A), both F₁ and F₂ generations exhibited substantial genetic variation as compared to current B. juncea accessions. Most F₁ individuals occupied a long-branch lineage diverging prior to the clade of current B. juncea, whereas F₂ individuals formed a paraphyletic group partially overlapping with current B. juncea. Consistent with this phylogenetic pattern, the principal component analysis separated most of the F₁ progeny from current B. juncea along Principal Component 1 (PC1) (19.2% of variance) and Principal Component 1 (PC2) (8.6%), while more F₂ individuals were notably shifted toward the current B. juncea cloud along PC1 (Figure 2B).

Genetic distance (GD) analysis further quantified these relationships (Table 1). The mean within-group GD was lowest in current B. juncea (GD_Cj = 0.090), intermediate in the F₂ (GD_F2 = 0.152), and highest in the F₁ (GD_F1 = 0.168). The between-group GD to current B. juncea was slightly lower for the F₂ (GD_F2-Cj = 0.136) than for the F₁ (GD_F1-Cj = 0.146), suggesting that the F₂ lines are still markedly different from current B. juncea. However, it also suggests that genetic recombination had already produced some individuals with genomic compositions more closely aligned with current B. juncea. Collectively, these results confirm that the new-type lines constitute a unique genetic group, harboring a spectrum of diversity that includes both distinct lineages and F₂ individuals with genomic compositions closer to current B. juncea.

To validate the practical application value of the hexaploid bridge hybridization strategy for trait improvement, we adopted a MAS in the hybrids between 6MS6 (A^jA^jB^jB^jBⁿBⁿ) with the European fodder turnip ECD04 which is well known B. rapa accession harboring multiple CR loci. Four PCR-based markers tightly linked to these CR loci were assayed in F₂-F₅ progeny (Supplementary Table S3), and plants carrying the highest number of ECD04-alleles were advanced to the next generation.

According to the four CR-linked markers (Figure 3A), a progressive enrichment was presented since the mean frequency of ECD04-alleles rose from 31.3% in the F₂ to 54.4% in the F₅, representing a 73.8% rise over the selection period (Figure 3B). More than half of the F₅ families (e.g., 5G06, 5G09, 5G10, 5G11 and 5G13) retained ≥ 50% donor alleles, while a subset (5G16-5G19) still carried< 50% ECD04 alleles (Table 2). Laboratory phenotyping for CR was carried out on F₅ lines with AF ≥ 50%, alongside parental controls. A strong negative linear correlation was observed between ECD04 AF and disease index (r = -0.98, p < 0.0001; Figure 3C), indicating that the allele dosage is a reliable predictor of resistance levels.

Concerning the detailed resistance (Figure 4, Table 2), the hexaploid parent and current B. juncea exhibited nearly 100% incidence and DI > 60, whereas the resistant parent ECD04 showed near immunity (disease incidence = 4.3%, DI = 1.7). Two F₅ families with AF = 77.5% (5G06 and 5G11) were virtually homozygous for the target haplotype (especially markers S14R14 and m090a) and recorded the lowest DI values (11.5 and 12.5, respectively), indicating the successful introgression and concentration of ECD04-derived CR loci into the B. juncea background. Three other F₅ families with AF between 50% and 75% displayed moderate resistance, with DI of 23.1-31.2 and disease incidence of 63%-75%, indicating the need for further selfing to achieve full homozygosity.

The strong reproductive barriers that impede traditional breeding, such as hybrid sterility and embryo abortion, have severely restricted the introgression of novel alleles from diploid progenitors into the cultivated B. juncea gene pool (Liu et al., 2018; Zhu et al., 2022). In this study, synthetic hexaploids (AABBBB/AAAABB), proposed as a bridge for gene transfer (Mei et al., 2022), were employed successfully to achieve efficient introgression of genomic components from both progenitor species into B. juncea. The newly developed materials represent a distinct and unique gene pool, not merely a random genetic mixture.

Phylogenetic, principal component, and genetic distance analyses consistently demonstrated that the new-type lines, particularly the F₁ generation, formed a cluster separate from natural B. juncea. Although a genetic shift toward natural B. juncea was observed in the F₂ generation, the genetic distance between the new-type lines and natural B. juncea remained substantially larger than that within natural B. juncea itself. These lines therefore retain a unique set of alleles from the progenitor species, collectively establishing a valuable and novel genetic reservoir. This population can serve as a diverse foundation set for future gene discovery and breeding, directly addressing the issue of genetic constriction in current cultivars.

The practical utilization of this novel gene pool first requires overcoming a common challenge in newly synthesized polyploids, i.e., meiotic instability, which often leads to reduced fertility (Tian et al., 2023). The significantly lower seeds per pod in the F₁ generation in our study confirmed such instability, reflecting a form of genomic shock characterized by rapid genomic reorganization in response to aberrant homologous chromosome pairing (McClintock, 1984; Xiong et al., 2011; Martin and Jouve, 1992). Nevertheless, the fertility was rapidly restored in subsequent generations. This recovery is consistent with a similar scenario (Xiong et al., 2011), wherein meiotic stability is gradually established through restricted homologous pairing, epigenetic silencing of conflict regions, and selective elimination of cytologically unstable lineages. Concurrently, phenotypic convergence from diploid-parent-like morphological variation in the F₁ to current B. juncea type traits in the F₃ was possibly also driven by recurrent selfing. The increase in homozygosity through selfing not only facilitated stable chromosome pairing but also purged deleterious genomic interactions, thereby accelerating the emergence of a cytologically and phenotypically stable allopolyploid.

Building upon this rapid stabilization, MAS was effectively applied within the complex polyploid background. Four cycles of targeted selection substantially enriched CR alleles from ECD04, demonstrating the high efficiency and precision of the hexaploid bridge strategy for trait improvement. Notably, a strong linear relationship was observed between allelic dosage of the resistance loci and disease severity, indicating that the introgressed CR genes likely act in a dominant or additive manner (Li et al., 2016). This relationship enabled reliable phenotypic prediction in early generations based on genotypic data, significantly improving selection efficiency.

However, our findings also reveal a limitation of the dosage model. Family 5G13, which was fixed for the major locus S17 and carried 75% donor alleles across the four target loci, still exhibited 63% disease incidence. We hypothesize that the following factors may be responsible: (i) linkage drag, in which undesirable genomic segments tightly linked to the target CR loci were co-introgressed, potentially carrying alleles that compromise resistance or plant fitness (Jabran et al., 2023); (ii) residual heterozygosity at loci BSA3, S14, and m090a (87.5% donor alleles, not yet fixed), or the presence of tightly linked deleterious segments near S17, may have compromised the expected resistance; (iii) background effects from the recipient B. juncea genome, where interactions between the introgressed alleles and the genetic background may modulate the final resistance level. This case illustrates a general limitation of relying solely on foreground selection in marker-assisted breeding (Dakouri et al., 2013). Therefore, incorporating routine background selection is recommended to break such unfavorable linkages and accelerate recovery of the recurrent parent genome (Nannuru et al., 2025), ultimately facilitating the development of agronomically superior lines with stable and complete clubroot resistance.