A set of F₂ populations was generated from crossing B. juncea genotypes (DRMRIJ 31, Pusa Mustard 30 and Pusa Agrani) with B. carinata accessions (BC-4, BC-5, and BC-12). Within each of these F₂ populations, plants were selected on the basis of phenotype and were inter-mated in pairs – a method commonly called biparental mating. Phenotypic selection in the subsequent filial generations led to the development of B. carinata-derived B. juncea lines (CDLs). These cytologically stable homozygous CDLs possess 18 bivalents (2n = 36), indicating its similarity to B. juncea (upplementary Figures 1, 2). These CDLs, in F₆ generation, were evaluated in two rows of 4-m length plots under both irrigated and rainfed field condition in the Division of Genetics, ICAR- IARI, New Delhi during 2018–2019 rabi season. To select CDLs differing in metric traits and response to moisture deficit stress, row-row and plant–plant spacing were kept at 30 and 15 cm, respectively (data not presented). One row was left blank before and after every plot such that plot–plot and row–row in a plot were separated by a spacing of 60 and 30 cm, respectively. The experiment was raised in augmented design both under irrigated and rainfed conditions. No irrigation was applied in the rainfed plot, while two irrigations, each of 50 mm depth, were applied in irrigated plots at 45 and 90 days after sowing (DAS). All recommended agronomic practices were adopted for raising the crop. Based on drought tolerance indices, 15 CDLs differing for various morpho-physiological traits and phenotypical and cytological resemblance to B. juncea (Supplementary Table 1 and upplementary Figures 1, 2) were selected for generating hybrids.

During 2019, 15 diverse CDLs differing for morpho-physiological traits were intermated in half diallel design (excluding reciprocals) to generate seeds of 105 F₁ hybrids in ICAR-IARI, Regional Station, Wellington, Tamil Nadu, India in the off-season.

During rabi 2019–2020 season, the set of 105 F₁ hybrids along with 15 parental CDLs was raised under both rainfed and irrigated conditions in the research field of Division of Genetics, ICAR-IARI, New Delhi. Two recommended irrigations of 50 mm depth each were applied at 45 and 90 DAS to the irrigated plots, while no irrigation was given to rainfed plots. Hybrids along with parents were raised in a randomised complete block design (RCBD) with three replications. Each plot consisted of a paired row of 4 m in length. Row to row and plant to plant distance were kept 30 and 15 cm, respectively. One row of 30 cm was left blank before and after every pair-row plot such that plot–plot and row–row in a plot were distanced at 60 and 30 cm, respectively. Recommended agronomical practices and plant protection measures were adopted for raising the crop.

Observations were recorded on 17 quantitative traits, viz., plant height (cm), point to first branch (cm), number of primary branches, number of secondary branches, main shoot length (cm), point to first siliqua on main shoot (cm), number of siliquae on main shoot, total siliquae per plant, siliqua length (cm), seeds per siliqua, days to 50% flowering, 1,000-seed weight (g), seed yield per plant (g), biological yield per plant (g), harvest index, seed yield (kg/ha), and oil content. The data were recorded on five randomly selected competitive plants from all the plots in each replication, except for seed yield per plot (g) and days to 50% flowering where observations were recorded on plot basis. Biological yield was recorded at physiological maturity after drying the harvested plants. Data on plot yield (g) were used to calculate seed yield (kg/ha).

Data were recorded on 17 metric traits under both irrigated and rainfed conditions. Mean values of these traits were used to analyse drought susceptibility index (DSI; Fischer and Maurer, 1978), drought tolerance index (DTI; Fernandez, 1992), tolerance index (TOL; Rosielle and Hamblin, 1981), and mean relative performance (MRP) for the CDLs taken in this study (Raman et al., 2012).

Water-use efficiency (WUE) for parents and hybrids were calculated from the below given equation:

WUE [kg ha^–1mm^–1] = Seed Yield (kgha^–1)/Water received from irrigation and rainfall (mm)

and WUE [kg m^–3] = 0.1 × WUE [kg ha^–1mm^–1].

The effective rainfall (between October 2019 and February 2020) was calculated by using Cropwat Version 8.0 by USDA SCS method. Cropwat is a computer-based decision support tool developed by the Food and Agriculture Organization (FAO) of the United Nations to estimate effective rainfall based on climate and soil data. The maximum and minimum temperatures and rainfall (mm) recorded during 2019–2020 crop seasons from ICAR-IARI meteorological observatory, located within 500 m distance from the experimental site, is presented in Supplementary Figure 3.

The mean values obtained from 17 different quantitative traits were subjected to analysis of mean and variance. Data obtained from 105 F₁ and 15 parental CDLs were subjected to Griffing’s method II with a fixed model (model I) to determine the combining abilities (Griffing, 1956). The mathematical model for ijkth observation is as follows:

where Y_ij is the mean of i × j^th genotype over k; μ is general population mean; g_i and g_j is the general combining ability effects of i^th and j^th parent, respectively; s_ij is the specific combining ability effects of cross involving i^th and j^th parent, respectively; and b_k is the replication effect; and e_ijk is the residual error.

Heterobeltiosis expressed in the hybrids was calculated by using mean values of respective F₁ and its better parent (BP) following standard formula (Coffman, 1933). The narrow-sense heritability (h²) was calculated by using the method suggested by Mather and Jinks (1982). Analysis of variance and estimation of degree of dominance, heritability, and combining ability were done following the method suggested by Griffing (1956) using Windostat Version 9.1. The degree of dominance was estimated from variance due to dominance (σ²_D) and variance due to additive (σ²_A) gene action by using the formula [{(2 × σ²_D)/σ²_A}^1/2] given by Comstock and Robinson (1948).

The general combining ability analysis revealed that each CDL possessed significant GCA effects for one or more traits under both irrigated or rainfed conditions (Table 4). CDL89 and CDL141 were found to be a good general combiner for point to the first siliqua on the main shoot, harvest index, seed yield per plant, seed yield (kg/ha), and oil content under rainfed conditions. CDL161, CDL112, CDL186, CDL101, and CDL25 were also found to possess high GCA for seed yield (kg/ha) and some of its contributing traits under water deficit stress conditions (Table 4).

Under irrigated condition, CDL161 and CDL106 expressed good general combining ability for point to the first branch, secondary branches/plant, main shoot length, total siliquae/plant, days to 50% flowering, 1,000-seed weight, seed yield/plant, biological yield/plant, and seed yield (kg/ha) (Table 4). CDL112, CDL186, CDL101, CDL25, and CDL89 also possessed higher GCA for seed yield (kg/ha) along with some of its contributing traits under water sufficient conditions (Table 4).

Under the rainfed condition, a total of 36 hybrids exhibited significant SCA effects and high heterobeltiosis (>15%) for seed yield (kg/ha). Highest heterobeltiosis (99.55%) was observed for the cross CDL106 × CDL104 under limited water conditions. Out of 36, nine hybrids, viz., CDL102 × CDL89, CDL102 × CDL106, CDL112 × CDL89, CDL186 × CDL121, CDL182 × CDL121, CDL 121 × CDL 103, CDL25 × CDL104, CDL89 × CDL104, and CDL104 × CDL106 exhibited highly significant SCA effects and more than 50% heterobeltiosis (Table 5). Similarly, under irrigated condition, 30 hybrids expressed significant SCA effects and heterobeltiosis > 15%. Highest heterobeltiosis (82.34%) for seed yield (kg/ha) was observed for the hybrid derived from parents CDL141 and CDL161 under favourable environment. Six hybrids, viz., CDL161 × CDL186, CDL161 × CDL141, CDL141 × CDL182, CDL182 × CDL103, CDL182 × CDL106, and CDL104 × CDL106 recorded highly significant SCA effects and >50% heterobeltiosis for seed yield (kg/ha) under irrigated condition. It is also important to note that hybrid derived from cross CDL104 × CDL106 observed highly significant SCA effects and >50% heterobeltiosis under both rainfed and irrigated conditions (Table 5).

High heterobeltiosis (>15%) and higher SCA effects for seed yield were reported for hybrids derived from different combinations of parental CDLs, irrespective of their GCA effect levels (Table 6). Furthermore, 14 hybrids expressed significant SCA effects and high heterobeltiosis (>15%) under both rainfed and irrigated conditions (Table 6).

The water requirement of rapeseed mustard is about 190–400 mm (Shekhawat et al., 2012; Sharma et al., 2018) and it is sensitive to water deficit stress at critical stages like pre-flowering and siliquae formation stages. In the presented experiment, the rainfed plots received 112.4 mm of water through rains up to physiological maturity stage, whereas two additional irrigations of 50 mm each were applied in the irrigated plots. Thus, the irrigated plots received a total of 212.4 mm of water throughout the crop developmental stages and met optimum water requirement. The data recorded on 17 morpho-physiological traits under rainfed and irrigated conditions were averaged separately for 15 parents and 105 hybrids. The results suggested that parental CDLs exhibited a higher mean value under irrigated than in rainfed condition for traits like plant height (cm), number of primary branches, number of secondary branches, main shoot length (cm), point to first siliqua on the main shoot (cm), total siliquae per plant, days to 50% flowering, and seed yield (kg/ha). Whereas, under the rainfed condition, the mean value of CDLs was higher for point to first branch (cm), siliquae on the main shoot, siliqua length (cm), seeds/siliqua, 1,000-seed weight (g), seed yield/plant (g), biological yield/plant (g), harvest index, and oil content traits (Table 7) than in irrigated condition. Under irrigated condition, the overall mean of the hybrids was higher for plant height, primary branches per plant, secondary branches per plant, main shoot length, point to first siliqua on main shoot, total siliquae per plant and days to 50% flowering than in rainfed condition. Whereas traits like point to first branch (cm), number of siliquae on the main shoot, siliqua length (cm), seeds per siliqua, 1,000-seed weight (g), seed yield per plant (g), biological yield per plant (g), harvest index, seed yield (kg/ha), and oil content (%) observed higher mean values for hybrids when evaluated under rainfed condition (Table 7). Overall performance of CDLs and hybrids was found better when evaluated under rainfed condition for important yield and yield attributing traits including plant height (cm), total siliquae on main shoot (no.), siliqua length (cm), seeds/siliqua, days to 50% flowering, 1,000-seed weight (g), seed yield/plant, biological yield/plant (g), harvest index, and oil content (%) traits (Table 7).

Drought tolerance indices, viz., drought susceptibility index (DSI), drought tolerance index (DTI), tolerance index (TOL), and mean relative performance (MRP) for seed yield (kg/ha) were estimated. Correlation coefficients between seed yield under irrigated condition (Y_IRR), seed yield under rainfed condition (Y_RF), WUE_RF, WUE_IRR, DSI, DTI, TOL, and MRP were calculated. Yield under rainfed condition (Y_RF) were showing significant and positive association with WUE_RF, WUE_IR, DTI, MRP, and TOL. Similarly, significantly positive association of Y_IR with DTI and MRP was observed (Supplementary Table 2). WUE_RF exhibited a positive and significant correlation with DSI, DTI, TOL, and MRP, while WUE_IR showed a positive and significant correlation with DTI and MRP. DTI and MRP were identified to be significantly associated with all the traits except DSI, thus, indicating the usefulness of these two indices in indirect selection for drought tolerance.

The WUE (kg m^–3) was calculated for all hybrids along with parental CDLs both under rainfed and irrigated conditions. WUE among the parental lines raised under rainfed condition varied from 1.27 to 2.59 kg m^–3 and recorded the highest and lowest values for CDL128 and CDL103, respectively (Table 8). Under irrigated condition it ranged from 0.73 to 1.50 kg m^–3 with the highest and lowest values for CDL89 and CDL103, respectively. Among 105 hybrids evaluated under rainfed condition, WUE varied from 1.02 to 3.86 kg m^–3 and the highest value was recorded for the cross CDL112 × CDL89. On the other hand, WUE varied from 0.59 to 2.13 kg m^–3 under the irrigated condition and exhibited the highest value for CDL104 × CDL106 cross (Table 8). In general, the hybrids observed higher WUE than their parental CDLs. Overall, the mean of seed yield (kg/ha) and WUE of parental CDLs and hybrids was recorded higher under rainfed condition than in irrigated condition (Figure 1).

Further, drought tolerance index (DTI) and mean relative performance (MRP) were calculated from average values of individual parent and hybrid for all the 17 traits. All the traits, except point to first branch and point to first siliqua on the main shoot, exhibited higher DTI and MRP for hybrids than their corresponding parents (Figure 2). DTI for seed yield (kg/ha) varied from 0.25 to 2.55 in hybrids, whereas it varied from 0.31 to 1.12 in CDLs (Table 8). Some of the hybrids such as CDL112 × CDL89 (DTI: 2.55), CDL104 × CDL106 (DTI: 2.49) CDL182 × CDL121 (DTI: 2.23), CDL186 × CDL121 (DTI: 2.0), CDL101 × CDL106 (DTI: 1.99), CDL102 × CDL25 (DTI: 1.95), CDL25 × CDL104 (DTI: 1.93), CDL161 × CDL101 (DTI: 1.91), and CDL186 × CDL101 (DTI: 1.91) has a DTI value of >1, indicating their superiority under both moisture-stress and non-stress conditions. Among the CDLs highest MRP was recorded for CDL89 (2.09) and the lowest was observed for CDL103 (1.09). Among the hybrids, CDL112 × CDL89 recorded with the highest (3.11) MRP (Table 8). Further, nineteen hybrids, viz., CDL102 × CDL186, CDL102 × CDL25, CDL102 × CDL89, CDL102 × CDL106, CDL161 × CDL128, CDL161 × CDL141, CDL161 × CDL186, CDL161 × CDL101, CDL128 × CDL141, CDL186 × CDL121, CDL186 × CDL101, CDL182 × CDL121, CDL121 × CDL25, CDL101 × CDL104, CDL101 × CDL106, CDL25 × CDL89, CDL25 × CDL104, CDL89 × CDL104, and CDL104 × CDL106 were identified to be significantly better in productivity and possessing higher values of DTI and MRP (Table 8).

Combining ability analysis partitions the genotypic variability into GCA (σ²gca) and SCA (σ²sca) variances representing additive and dominance effects, respectively. The higher value of σ²sca than σ²gca, along with an average degree of dominance more than unity observed for almost all the traits studied under irrigated and rainfed conditions, demonstrates the predominance of non-additive gene effects in the manifestation of heterosis (Table 3). For siliqua length and seeds per siliqua evaluated under both irrigated and rainfed condition, σ²gca were higher than σ²sca and average degree of dominance were less than unity demonstrating the role of additive gene effects in governing them. Similar results were also reported by Yadava et al. (2012) and Singh N. et al. (2015) in Indian mustard. Since additive and non-additive gene action were governing the yield contributing traits under irrigated and rainfed conditions, multiple or reciprocal recurrent selection could be useful for the improvement of these traits. The narrow-sense heritability (h²) found low to moderate for almost all the traits under both the water regimes, suggesting the importance of interaction effects (Sayar et al., 2007). Since the expected genetic gain per cycle of selection will be less important, the development of hybrids involving CDLs carrying B. carinata genomic segments holds promise for exploitation of heterosis under rainfed conditions.

Significant GCA effects were observed for CDL161, CDL141, CDL112, CDL186, CDL101, CDL25, and CDL89 for seed yield (kg/ha) along with some of its contributing traits under rainfed condition, whereas under irrigated condition, CDL161, CDL112, CDL186, CDL101, CDL25, CDL89, and CDL106 expressed good general combining ability for seed yield (kg/ha) and its contributing traits (Table 4). The results suggested the importance of additive or additive × additive gene effects which are governing these traits. Similar results have also been reported earlier using a different set of Indian mustard genotypes (Mahak et al., 2008; Singh et al., 2010, 2013; Yadava et al., 2012; Singh N. et al., 2015). Furthermore, the good general combiners for seed yield traits were not showing significant GCA effects for some of its contributing traits, hence, there is a scope of improving the combining ability among parents for yield contributing traits under both irrigated and rainfed conditions. The findings of the present study suggest that CDLs, viz., CDL161, CDL112, CDL186, CDL101, CDL25, and CDL89 possessing variable drought tolerance and expressing high GCA for seed yield along with other component traits under both rainfed and irrigated conditions can be incorporated in the breeding programme for accumulating desirable alleles for developing improved mustard cultivars suitable to water-scarce regions.

In the present study, significant SCA effects and high heterobeltiosis (>15%) were exhibited by 36 and 30 hybrids under rainfed and irrigated condition, respectively. Out of these, nine and six hybrids were expressing more than 50% heterobeltiosis when evaluated under rainfed and irrigated conditions, respectively (Table 5). This suggests that CDLs taken in this study can be exploited for developing hybrids for water-scarce regions. It is also important to note that 14 hybrids exhibited higher SCA effects and heterobeltiosis (>15%) under both water regimes (Table 6). These hybrids, expected to carry different introgression segments from B. carinata, have better resilience to changing water regimes and, thus, expressed better stability of performance. Earlier studies have recorded similar observations with a different set of conventional B. juncea genotypes (Singh et al., 2006; Srivastava et al., 2009). Teklewold and Becker (2005) reported significant SCA effects along with high heterobeltiosis for seed yield in B. carinata inbred lines, indicating substantial potential of Ethiopian mustard genotypes and their significance in the exploitation of hybrid vigour in B. juncea breeding programmes. Considerable heterosis was also reported in inter-subgenomic hybrids between B. carinata-derived introgression lines and B. juncea accessions (Wei et al., 2016). Therefore, the development of hybrids using B. carnata derived lines will be an ideal approach in realizing higher yields under drought conditions.

The higher SCA effects and heterobeltiosis (>15%) for seed yield were exhibited by hybrids derived from CDLs, irrespective of their GCA effect values. Thus, hybrids generated from CDLs with significant × significant, significant × non-significant, and non-significant × non-significant GCA effects expressed high SCA effects and heterobeltiosis (Table 6). It demonstrates that joint effects of native and alien heterotic loci are responsible for the expression of such SCA effects and heterobeltiosis in hybrids derived from CDLs. Langham (1961) and Lal et al. (2006) have explained this anomaly where parents with other than high × high GCA effects also resulted in better heterosis and concluded that cross between parents with significantly high × low trait value resulted in superior transgressive segregants and, thus, the genes governing high trait values may express under homozygous or heterozygous conditions. Further, the genes governing lower trait values can also be expressed to their full potential in the homozygous state, but in heterozygote state, they can express very high performance due to complementation of alleles or overdominance. The existence of heterotic loci in these B. carinata derived lines can thus be exploited in the B. juncea breeding program for improving the level of heterosis and its commercial use.

Brassica carinata, in general, adapted to tolerate drought conditions (Thakur et al., 2021), performed better under less availability of water and lodges under irrigated conditions due to its poor ability to withstand accumulated higher biomass. It has also been observed that irrigations in B. carinata genotypes also exhibits heavy regeneration, resulting in reduced plant height, poor seed set, and reduced seed size, which ultimately leads to reduction in seed yield per unit area. Conversely, bolder seed size is realised under rainfed than in irrigated conditions. In the present study, some of the B. carinata derived lines, along with their hybrids, observed reduced performance for seed yield under irrigated than in rainfed conditions. This could be attributed to the introduction of genomic segment(s) or substituted chromosomes from the C genome of B. carinata to B. juncea. Although, the possibility of later is remote when the biparental mating approach is being used for the development of such CDLs.

Parental CDLs and their hybrids exhibited higher mean values for seed yield (kg/ha) and other yield attributes under rainfed than in irrigated condition (Table 7). It is most probably due to presence of different introgression segments in CDLs which alter their performance in water deficit and surplus condition, whereas complementation of distinct genomic segments and expressed overdominance leads to heterosis in hybrids. Therefore, comparatively better performance of these hybrids under rainfed compared to irrigated conditions suggested that B. carinata-derived lines hold promise for the development of hybrids or cultivars for water-scarce regions.

The present study revealed that yield under rainfed condition (Y_RF) was showing positive correlation with WUE_RF (r = 0.99), WUE_IR (r = 0.65), and stress tolerance indices, viz., DTI (r = 0.90), MRP (r = 0.92), and TOL (r = 0.54). Similarly, yield under irrigated conditions (Y_IR) was positively correlated with DTI (r = 0.89) and MRP (r = 0.90). In addition, DTI was found positively correlated with MRP (r = 0.99) (Supplementary Table 2). In the present study, a higher mean seed yield was recorded in rainfed than in irrigated conditions. Therefore, DSI would not be a reliable criterion for selection when mean values under stress condition is higher than that of normal irrigated condition. Better performance of CDLs and their hybrids under rainfed condition compared to irrigated plots attributed to excessive lodging of plants in irrigated conditions. Although the level of this lodging in CDLs and hybrids is much less than what normally happens in B. carinata after irrigation or rainfall. Therefore, DTI and MRP were found to be better predictors of Y_IR and Y_RF than TOL and DSI. This study is in accordance with previous reports suggesting DTI as a better parameter for identifying drought-tolerant genotypes in mungbean (Fernandez, 1992), corn (Shirinzadeh et al., 2010), wheat (Farshadfar et al., 2012), chickpea (Pour-Siahbidi and Pour-Aboughadareh, 2013), durum wheat (Mohammadi, 2016), and barley (Khalili et al., 2016). Mahajan et al. (2018) also reported that DTI is a more desirable index for stress tolerance than SSI in rice under moisture deficit stress.

The test hybrids expressed improved WUE as compared to parental CDLs. Overall, the mean of seed yield (kg/ha) and WUE of parental CDLs and hybrids was recorded higher under rainfed condition than in irrigated conditions (Figure 1), indicating the presence of putatively different loci/genomic segments in the CDLs for improved WUE and yield heterosis. Expression of overdominance and/or complementation of favourable alleles in hybrids appear to be the reason for improved WUE and established genetic elite in the hybrids.

Higher MRP and DTI confirmed better tolerance/adaptation of hybrids to moisture deficit stress over the parental CDLs for most of the traits under study (Figure 2). Hybrids expressing DTI value of >1, indicating their superiority under both moisture-stress and non-stress conditions. Lower DTI value, on the other hand, did not precisely explain susceptibility to moisture deficit stress conditions as mean yield under rainfed was higher than irrigated in present study. CDLs carry introgressions from B. carinata lodged after irrigation or rain and, thus, lead to a reduction in seed yield. In general, higher MRP of hybrids than CDLs indicates better performance of hybrids under rainfed and irrigated conditions. Based on higher DTI and MRP, highly productive and drought-tolerant hybrids, viz., CDL102 × CDL186, CDL102 × CDL25, CDL102 × CDL89, CDL102 × CDL106, CDL161 × CDL128, CDL161 × CDL141, CDL161 × CDL186, CDL161 × CDL101, CDL128 × CDL141, CDL186 × CDL121, CDL186 × CDL101, CDL182 × CDL121, CDL121 × CDL25, CDL101 × CDL104, CDL101 × CDL106, CDL25 × CDL89, CDL25 × CDL104, CDL89 × CDL104, and CDL104 × CDL106 were identified (Table 8). These hybrids can be commercially exploited by converting the parental CDLs into CMS or restorer lines based on their seed production suitability. Hence, hybrid vigour expressed under moisture deficit stress conditions is well demonstrated through the deployment of B. carinata-derived B. juncea lines.