Three accessions of B. rapa (var. yellow sarson and toria) and eight B. nigra were used for inter-specific crosses to develop resynthesized B. juncea (RBJ) (scheme in Figure 1 ), as listed in Table 1 . A total of 28 cross combinations ( Supplementary Table S1 ) were attempted using different accessions of B. rapa and B. nigra as both male and female parents. To infuse larger genetic variability in resynthesized lines, a three-way cross approach was used in which B. rapa var. yellow sarson and B. rapa var. toria were hybridized to generate F₁ seed for further crossing with B. nigra populations. B. juncea cultivars, namely, Pusa Jaikisan (PJK), Pusa Vijay (P. Vijay), Pusa Mustard 28 (PM 28) and Varuna were used as controls to compare with the synthetic lines developed.

The mature, unopened buds were chosen for emasculation to avoid any self or foreign pollen contamination. All the anthers, sepals, and petals were removed collectively from the selected flower buds using forceps, adjoining younger buds were snipped off, and the inflorescence was bagged. The male parent’s inflorescence was bagged to avoid any pollen contamination (Seyis et al., 2005). The following day, collected pollens from the male parent were used to pollinate the female parent’s emasculated buds.

To develop amphidiploids from amphihaploid inter-specific hybrids, non-absorbent cotton balls soaked in 0.2% colchicine were applied to the hybrid plants’ axillary buds and apical meristem on alternate days for 5-7 days (Rajcan et al., 2011; Manzoor et al., 2019).

The viability of pollens in the parental and interspecific hybrid plants was examined under a light microscope (Carl Zeiss Axiolab 5, Germany) using the protocol by Katche et al. (2021). The mature flower buds were collected in the ice container, and with the aid of forceps and a needle, the anther was removed and crushed in 1% acetocarmine. Three buds from each plant were inspected to determine the pollen viability, which was analyzed and expressed in terms of percentage. Round, plump, and stained pollens were regarded as viable, whereas shriveled or unstained pollens were sterile.

The procedure of Snowdon et al. (1997) was followed to count the total number of diploid (2n) chromosomes in mitotic cells from root tissue by applying DAPI (4,6-diamidino-2-phenylindole) as fluorescent dye under the microscope. The root tips were harvested early in the morning in 0.002M 8-hydroxyquinoline solution and then fixed in Carnoy’s solution (3:1 ethanol: acetic acid solution), followed by transfer into 70% ethanol. For slide preparation, enzyme solution was added to root tips on a slide and incubated at 37°C for 45 minutes, followed by the addition of 10µL 45% acetic acid after removing the enzyme solution (Rao et al., 2024). The root tip was solubilized with a needle to release the cells, and then a coverslip was placed on top. The slide was fixed in liquid N_2, and the coverslip was removed with the help of a scalpel blade and then allowed to air dry. After staining with 10µL DAPI, the chromosomes were seen under the fluorescent microscope (Carl Zeiss Axiolab 5, Germany).

Phenotypic data was collected from three selected plants in S₂ generation of each line across the three replication plots for plant height (PH), main shoot length (MSL), silique length (SL), number of siliques on main shoot (SMS), number of primary branches (PB), seeds per silique (SS), oil content (OC), yield per plant (YP), and thousand seed weight (TSW). Oil content was estimated using the Near-Infrared Spectroscopy (NIRS) by Newport NMR analyzer (Model-4000) (Shruti et al., 2023). The morphological diversity was assessed for thirty-three synthetic B. juncea lines, four B. juncea cultivars, and diploid parental species, i.e., B. rapa (3 accessions) and B. nigra (8 accessions).

A set of 94 SSR markers ( Supplementary Table S3 ) pertaining to AA (B. rapa), BB (B. nigra) and AABB (B. juncea) genome were selected for the hybridity confirmation ( Supplementary Table 3 ) in F₁ generation along with genetic diversity analysis in S₂ generation (Sudan et al., 2016). These polymorphic markers are available in the Brassicaceae database (BRAD) (http://Brassicadb.cn) (Chen et al., 2022) and are also reported by Lowe et al. (2004); Kim et al. (2009); Dhaka et al. (2017).

The total genomic DNA was isolated following DNA extraction protocol from fresh leaves (Doyle and Doyle, 1990). In a reaction volume of 20 µl, the PCR mixture contained 1 µl of template DNA (25 ng/µl), 1 µl of each forward and reverse primer (100 pmol/µl), 1µl of 10 mM dNTPs, 1.5 µl of 25mM MgCl₂, 4 µl of 10x PCR buffer, and 0.5µl of 0.5U Taq polymerase and 10 µl of nuclease-free water. The PCR cycle was designed with an initial denaturation at 94°C for 5 min, followed by 35 cycles of denaturation at 94°C for 1 min, annealing at 54°C for 1 min and 15 sec, and extension at 72°C for 1 min 30 sec, before a final extension at 72°C for 10 min. The PCR products were processed in 1x TAE buffer and separated on a 2.5% agarose gel along with the 50 bp DNA ladder as a benchmark on both sides of the gel.

The statistical analysis was done using the metan package in R program v4.2.0 (Olivoto and Lúcio, 2020). Due to the presence of multi-collinear factors, the data was subjected to principal component analysis (PCA) based clustering. While conducting a cluster analysis with pair group distance and Euclidean similarity metrics, the factors corresponding to significant PCs were chosen. Using the DARwin software v6.0.021, the neighbor-joining tree was created (Perrier and Jacquemoud-Collet, 2006). The population structure was assessed using the Bayesian clustering model-based software STRUCTURE v2.3.4 (Pritchard et al., 2000; Falush et al., 2003). Five iterations were performed for each cluster, K = 2 to 8, with the length of the burn-in period and Markov Chain Monte Carlo (MCMC) replications set to 50,000 each. The most probable K value was determined using a web-based software StructureSelector (Li and Liu, 2018), which uses combined measures and estimators to select the best K-value (Evanno et al., 2005; Raj et al., 2014; Puechmaille, 2016) and integrates CLUMPAK program for graphical representation (Kopelman et al., 2015).

Twenty-eight crosses attempted in this study yielded thirty-three lines of RBJ in twelve cross combinations, as given in Table 2 . The crosses were successful for combinations having B. rapa as the female parent, and hence, no seeds were obtained for reciprocal crosses (B. nigra ×B. rapa). When compared with parents and control (Figures 2A–C), morphological variations for leaf architecture and size were observed in amphihaploid F₁ plants (Figures 2D–F) and S₁ generation (Figures 2G, H). In Figures 3A–K, a clear difference was observed for leaf tenderness in amphihaploid plants ( Figures 3D–G ) and robust and firm leaves in S₁ plants ( Figures 3H–K ), which resembles the natural B. juncea cultivars (Figure 3C). Furthermore, the morphology of mature plants in S₂ generation ( Supplementary Figure S1 ) clearly indicates that the resynthesized plants differ morphologically from both parental lines. The plants in the S₂ generation exhibited significant variability in terms of plant height, number of primary branches, seeds per silique, etc., and were found to be more similar to the B. juncea cultivar. This indicates that the process of resynthesis might have led to some novel genetic combinations.

The validation of hybridity in the developed amphihaploid was conducted utilizing a comprehensive set of simple sequence repeat (SSR) primers specific to the genomes involved. A total of 14 SSR primers targeting the A-genome, 18 targeting the B-genome, and 29 targeting the AB-genome were employed. The parental polymorphism was done, and the polymorphic primers were used for the hybridity confirmation. The analysis of amplification patterns obtained from these primers confirmed the hybrid nature of the developed amphihaploid. The presence of characteristic bands corresponding to the A, B, and AB genomes further substantiated the successful hybridization process ( Supplementary Figure S2 ). The list of SSR primers used and hybridity assessed for each resynthesized line is given in Supplementary Tables S2 , S3 . Figures 4A–C show pollen viability of both diploid parents and the control. Pollen sterility in F₁ hybrids was evident (nearing 100%) at the amphihaploid stage ( Figures 4D–F ), and fertility was reinstated in the S₁ generation ( Figures 4G–I ) due to chromosome doubling with over 50% pollen stainability. Upon reaching the S₂ generation, the synthesized lines displayed a discernible range of pollen fertility, spanning from 53% to nearly 100%, thereby culminating in an average of approximately 85% ( Figures 4J–M ).

The confirmed 33 true hybrids were assayed for chromosome number via mitotic configurations at the F₁ stage, revealing 18 chromosomes under the microscope ( Figure 5D ), and the mitotic analysis of diploid parents, B. rapa and B. nigra, exhibited 20 and 16 chromosomes, respectively ( Figures 5A, B ). Thirty-six chromosomes were clearly visible in synthetic B. juncea lines (S₂ generation), which was similar to the B. juncea cultivar Pusa Jaikisan ( Figures 5C, E, F ). The conducted cytogenetic studies established the successful development of synthetic B. juncea lines.

The genetic variation among the resynthesized genotypes with diploid parents and Indian mustard cultivars was evaluated through Analysis of Variance (ANOVA) for various phenotypic traits ( Table 3 ). The results highlight significant contributions from various factors in the study. The genotypes, traits and their interactions displayed significant variations of 2.16%, 89.61% and 7.90% respectively at P > 0.0001.

The correlation between different agro-morphological traits was analyzed using Pearson’s correlation coefficient, and the results are presented in Figure 6A . Several significant correlations were observed among the traits. Firstly, YP exhibited positive correlations with PH (r = 0.52, p < 0.001), MSL (r = 0.51, p < 0.001), and TSW (r = 0.57, p < 0.001), indicating potential associations between yield-related parameters. TSW also demonstrated strong positive correlations with PH (r = 0.46, p < 0.01), MSL (r = 0.74, p < 0.001), SL (r = 0.72, p < 0.001), SS (r =0.38, p < 0.01) and oil content (r = 0.58, p < 0.001) suggesting their potential influence on seed weight. This suggests that taller plants with longer main shoots and siliqua length tend to have higher yields. Additionally, oil content showed a positive association with SL (r = 0.60, p < 0.001) and SS (r = 0.56, p < 0.001), indicating their potential contribution to oil accumulation. This indicates that longer main shoots and siliqua length are associated with higher oil content. PH and MSL are positively correlated with r = 0.53 at p < 0.001. The weak negative correlations observed between oil content and PH (r = -0.04, p = ns) and SMS (r = -0.18, p = ns) are found to be non-significant in our study. This may imply that either there is no correlation between these two variables or this may be a random association. The positive correlation between oil content and YP (r = 0.30, p < 0.05) suggests that higher seed yield may be associated with increased oil accumulation. Oil content and YP may not be directly associated with each other but may be dependent on a complex agronomic trait, seed size or TSW which is positively correlated with both YP (r= 0.57, p< 0.001) and oil content (r= 0.58, p< 0.001). Oil content is an economically important trait and, lines with a higher oil content can thus be selected by opting for lines with higher TSW or YP. These findings provide insights into the interdependence of these traits and can guide future breeding and selection strategies to enhance specific desirable traits in Indian mustard cultivars.

Principal component analysis (PCA) ( Figure 6B ; Supplementary Table S4 ) was computed to show each trait’s contribution to the overall morphological variations. Categorized under various clusters, along with their associated morphological traits, the PCs represent axes of variation that capture the morphological diversity within the genotypes. The importance of each PC is measured by its standard deviation, percentage of variance, and cumulative proportion of variance. These statistics indicate how much each PC contributes to the overall morphological diversity captured by the entire set of PCs. The cumulative proportion of variance demonstrates that the first four PCs contribute more than 80% to the total variance, with PC1 alone accounting for about 41.2% of the total variance. It highlights the significance of specific PCs in explaining variations in morphological traits and underscores the potential implications for crop enhancement and breeding strategies.

The cluster analysis ( Figure 7A ) based on SSR markers showed that the RBJ lines were effectively distributed across all three discrete clusters. Within cluster IIb, a subset of resynthesized lines (specifically, RBJ 175, RBJ 179, RBJ 186, and RBJ 188) demonstrated close affinity with genotypes of parent B. rapa. This observation underscores a genomic-level congruence among these resynthesized lines and the B. rapa genotypes. Notably, the RBJ lines underwent subsequent sub-clustering, revealing the emergence of genomic-level diversity as a result of the resynthesis process. Moreover, the composition of cluster IIa included B. juncea cultivars exhibiting a more proximate phylogenetic alignment with B. nigra accessions. This proximity is likely attributed to historical processes of natural or artificial selection over extended timeframes.

The outcomes of the morphological cluster analysis also unveiled the presence of three prominent clusters ( Figure 7B ). The clusters I and II encompassed a substantial proportion of the synthetic B. juncea lines, along with the cultivar Varuna. The third cluster comprised B. nigra and B. rapa genotypes. In this cluster, a notable proximity was observed between the morphology of RBJ 170 and the cultivars Pusa Jaikisan, Pusa Vijay, and PM 28. Additionally, this cluster highlighted the close morphological proximity of RBJ 102 and RBJ 186 to the Nigra 2 genotype.

The population structure of the Brassica genotypes was analyzed using STRUCTURE software. The optimal K value was determined by plotting the values of ΔK against the number of clusters ( Figure 8A ; Supplementary Table S5 ) that show the highest ΔK value at K=3. This indicates that the genotypes used in the study are divided into three subpopulations (pop 1, pop 2 and pop 3). The estimated Ln probability of data was -18391.7 ( Figure 8B ) with mean ln likelihood, variance of ln likelihood, and mean alpha value at -18118.9, 545.6, and 0.0381, respectively. Table 4 summarizes the overall proportion of membership (inferred clusters), mean fixation index (Fst), divergence among the two subpopulations, and number of genotypes in each population. The genotypes were assigned to either of the populations based on the Q values from both clusters. The inferred ancestry of all individuals can be seen in Figure 8C ; Supplementary Table S5 .

A clear overview of variations observed in the developed RBJ lines is presented in Figure 9 . The synthetic lines are seen to perform better than the four B. juncea cultivars in terms of plant height, main shoot length, no. of siliqua on main shoot, no. of seeds per siliqua, thousand seed weight and yield per plant. RBJ 163 and RBJ 142 recorded the highest plant height of 290 cm and 284 cm, while PM 28 was the shortest at 159.7 cm. RBJ 122 has the longest main shoot length (87.7 cm) with a higher number of siliquae on the main shoot, i.e., 81.7. RBJ 106 (18.7) and RBJ 135 (18) have more seeds per siliqua followed by RBJ 149, RBJ 102 and Varuna having the same mean of 17.7. RBJ 128 observed the highest thousand seed weight (4.8) as well as yield per plant (75.1 g), while RBJ 170 has the lowest values for both traits (1.40 and 7.3 g, respectively). Yield and TSW for B. juncea cultivars are as follows: PJK (20.7 g, 3.3 g), P.vijay (11.5 g, 2.8 g), PM 28 (10.3 g, 2.5 g) and Varuna (67.3 g, 2.9 g).