The activity and accumulation of defense-related enzymes with level of resistance against ToLCNDV in sponge gourd were assessed. Eight resistant genotypes, four genotypes each of moderately resistant, moderately susceptible, and susceptible category, were subjected for biochemical characterization. The fresh leaf samples were collected 60 days after sowing, when plants were fully infected with ToLCNDV, and stored at −20°C until the biochemical tests were performed. For catalase activity, peroxidase activity, total phenol content, and total soluble protein content, 1 g of leaves was added with 10 ml of 0.1 M phosphate buffer in a mortar and pestle. The mixture was crusted until a slurry was obtained. Similarly, 0.5 g of leaves was crushed in mortar and pestle for hydrogen peroxide analysis and 0.5 ml of 0.5 mM TCA acid and 1 M pH 8 phosphate buffer were used to prepare a slurry. 2 ml of each homogenates was taken in tubes and centrifuged at 10,000 rpm for 15 min. The supernatant was taken out for further analysis of individual biochemical parameters.

It was estimated using the protocol outlined by Bradford (1976). 140 μl of Bradford reagent was added with 60 μl plant extract till 20 samples were completed, and absorbance reading was taken at 600 nm.

The total phenol content was determined using the Folin–Ciocalteu reagent. 2 ml of plant extract was added with 2 ml of absolute ethanol in a test tube and boiled until all ethanol evaporated, leaving only the plant residue. 5 ml of distilled water was added to the plant residue, and phenol solution was prepared by mixing 2 ml FC reagent and 4 ml of 35% sodium carbonate. The 150 μl of solution was added in wells with 50 μl of diluted plant residue, and a reading of absorbance was taken at 650 nm after incubating for 1 h.

2 ml of 1 M potassium iodide was added with 1 ml of 0.1 M phosphate buffer for master mix preparation. 150 μl of mix was added along with 50 μl of hydrogen peroxide sample extract, and reading was taken at 390 nm after 20 min of incubation at room temperature.

The peroxidase activity was estimated using the Chance and Maehly (1955) method. The master mix was prepared by adding 192 μl of 1 M guaiacol, 24 μl hydrogen peroxide, and 11.38 ml of 50 mM phosphate buffer. 150 μl of master mix was added along with 50 μl of plant extract in each well of the ELISA plate till 20 samples were completed, and absorbance at 450 nm was measured before and after 1 min.

Catalase activity (CAT) was assessed using the Zhang et al. (2007) protocol. The solution for catalase activity included 11.6 ml of 0.1 M pH 7 phosphate buffer and 140 μl of hydrogen peroxide. 175 μl of solution was added to ELISA plate well, and 25 μl of plant extract until all the samples were added. The absorbance reading was taken at 270 nm for 3 min at 30-s intervals.

1 g of the leaf sample was mixed with 10 ml of potassium phosphate buffer (50 mM) pH 7 solution in a mortar and pestle to prepare the slurry. Then, the homogenate was centrifuged for 10 min at 10,000–12,000 rpm and the supernatant solution was take out into another tube. 1,100 μl potassium phosphate buffer (50 mM) was mixed with pH 7, 372.20 μl EDTA (0.1 mM), 105.99 μl sodium carbonate (50 mM), 149.21 μl, L-methionine (SOD enzyme protector), 817.64 μl NBT (50 mM), and 376.36 μl riboflavin (O-2 generator): in a tube and form a 3 ml master mix. 10 μl of supernatant was added to an ELISA plate well along with 190 μl of master mix, and absorbance was read at 560-nm wavelength with the help of a spectrophotometer.

The PAL activity was assessed following the procedure outlined by Zucker (1965). The homogenate was prepared in a chilled pestle and mortar by crushing 0.5 g of lead sample with 5 ml of sodium borate buffer (0.1 M) pH 8.5 solution. The slurry was centrifuged in a 2-ml tube at 10,000–12,000 rpm for 10 min. The supernatant was transferred to another 2-ml tube. 62.5 µl enzyme extract, 800 µl of sodium borate buffer, and 700 µl of (12 mM) phenylalanine were added in each test tube and incubated at 40°C in a water bath for 1 h. The reaction was halted by adding 200 µl of 5 N HCl, followed by the addition of 0.5 ml of 0.1 M Trans-cinnamic acid (TCA). The 62.5 µl enzyme extract and 800 µl of sodium borate buffer were added in each test tube, along with 700 µl of (12 mM) phenylalanine, and tubes were incubated at 40°C in a water bath for 1 h. 200 µl of 5 N HCl was added to stop the reaction. Then, 0.5 ml of trans-cinnamic acid (TCA) (1 M) was added, and the light absorbance was estimated at 290 nm. The absorbance reading was measured at 290 nm.

There were 50 diverse germplasms including commercial varieties and advanced breeding lines that were subjected to three consecutive years of natural screening in the kharif season. The pooled vulnerability index score of 3 years varied from 3.23% to 96.10% ( Supplementary Table 1 ). Eight genotypes showed resistant reaction against ToLCNDV with the lowest VI value, which was recorded in DSG-7 (3.23%), followed by DSG-6 (4.44%) and DSG-29(8.00%). There were 11 genotypes that showed a moderately resistant type of reaction, whereas 14 showed moderately susceptible reaction. There were 17 genotypes including all the commercially grown varieties that were found to be susceptible with VI values ranging from 76.50% to 96.10% against ToLCNDV. Furthermore, susceptible check DSG-55 showed a pooled mean VI value of 95.06%, thus indicating high disease pressure across the years of preliminary screening.

The average environment coordination (AEC) view of the mean vs. stability biplot based on genotype-focused singular value portioned (SVP = 1) was conducted to evaluate the mean performance and stability of 50 genotypes across the tested environments ( Supplementary Figure 1 ). The mean vs. stability biplot revealed that genotypes, DSG-7(G-2), DSG-6(G-1), DSG-29(G-3), DSGVRL-22(G-4), DSGVRL-23(G-5), DSGVRL-18(G-6), DSGVRL-3(G-7), and DSGVRL-2 (G-8), were categorized as stable resistant genotypes against ToLCNDV over the 3 years. These genotypes were positioned in a descending order with respect to farthest genotype at the left side on the AEC abscissa from the biplot origin, expressing low VI values. Moreover, all these genotypes were present on the AEC ordinate, thus showing resemblance in their response toward disease severity against ToLCNDV. These genotypes, except DSGVRL-22(G-4) and DSGVRL-3(G-7), showed stability regarding the number of fruits and fruit yield per plant. DSG-29(G-3), DSG-7(G-2), and DSGVRL-23(G-5) were most desirable with respect to earliness viz., DFMA, DFFFA, and DFFH. Regarding fruit length, DSG-512 (G-30) and DSG-51-1 (G-47) had higher and stable performance whereas DSG-7(G-2), DSGVRL-22(G-4), and DSGVRL-3(G-7) had higher and stable performance for fruit diameter. Similarly, DSG-512 (G-30) was the most ideal and stable for average fruit weight during all the years.

Correlation analysis revealed that 30 pairs of nine traits evaluated were significantly correlated considering all the three harvesting seasons ( Figure 2 ). Among them, 27 pairs exhibited significant correlation at a significance level of p< 0.001, whereas three pairs were significant at p< 0.05. The vulnerability index showed a highly significant positive correlation with DFMA (r = 0.59), DFFFA (r = 0.666), and DFFH (r = 0.713). In contrast, it had a highly significant negative correlation with fruit yield per plant and number of fruits per plant. In addition, all these traits showed a similar trend for all 3 years. Fruit yield and number of fruits per plant exhibited the highest positive correlation of r = 0.984 with similar results for all the three harvesting seasons. Furthermore, the yield is negatively correlated with flowering traits, viz., DFMA (r = −0.562), DFFFA (r = −0.621), and DFFH (r = −0.679). This negative correlation suggests that earliness leads to a higher yield in genotypes screened against ToLCNDV. The significant positive correlation of the vulnerability index with flowering traits indicates that high VI values increase the duration of appearance of first male and female flowers and, ultimately, the harvesting of first fruits. Thus, susceptibility against ToLCNDV delays the flowering. Similarly, VI showed a significant negative correlation with fruit number and fruit yield per plant, severely affecting yielding capacity of the genotypes under ToLCNDV infection in sponge gourd.

The heatmap and hierarchical clustering based on earliness, yield and its attributing components, and ToLCNDV vulnerability index were carried out to show a chromatic evaluation of genotypes ( Figure 3 ). The heat map analysis produced two dendrograms; the one in the horizontal direction represents the genotypes, and the second in the vertical direction represents the traits influenced by this diffusion. The relationship and strength among variables and genotypes determines the color of each box. The genotypes were grouped to identify genotypes with common characteristics in terms of resistance, earliness, and higher yield. A clustered heat map distinguished the genotypes into seven groups based on resistance level and yield-related traits. Group I consisted of five genotypes (G-2, G-3, G-4, G-5, and G-6), exhibiting resistance with low VI values, higher yield traits, and earliness. Group II included eight genotypes, mostly moderately resistant and three moderately susceptible genotypes based on yield and other traits. Group III comprised eight genotypes, two as resistant and the remaining moderately resistant. The resistant check (DSG-6) belonged to this cluster based on fruit-related traits, delayed anthesis, and harvesting time. Furthermore, Groups IV, V, VI, and VII consisted of 8, 4, 10, and 7 genotypes, respectively, and were classified as moderately susceptible and susceptible based on their high VI values.

Considering 3-year screening results, 20 diverse genotypes, including identified resistant genotypes, commercially grown varieties, and promising genotypes, were evaluated across four diverse environments during 2021 ( Table 1 ). Significant genotypic variations (P<0.001) were observed through individual environment ANOVA for traits such as vulnerability index against ToLCNDV, days to first female flower anthesis, number of fruits per plant, average fruit weight, and fruit yield under disease incidence during 2021. Combined ANOVA also revealed significant genotypic differences, as well as significant environment and genotype × environment interaction effects for all the traits including vulnerability index ( Table 2 ). The contribution of genotypic variance was recorded higher as compared to GEI and environmental variance for all the traits.

All eight resistant genotypes were positioned on the left side of the AEA abscissa, whereas the remaining genotypes were placed toward the direction of the AEA abscissa from the biplot origins ( Figure 4A ). The genotypes positioned toward the direction reveal higher VI values, which indicate susceptibility, whereas genotypes in the opposite direction show lower VI values, thus signifying resistant reaction. Genotypes G-48(DSG-55) and G-2(DSG-7) were positioned at two extremes of AEA abscissa, respectively, showing the highest and lowest values of VI across the environments. The stability of genotypes, which is represented by projection on the AEA abscissa, reveals that genotype DSGVRL-2(G-8), followed by DSG-7(G-2) and DSG-29(G-3), was the most stable across the environments studied. Therefore, genotype DSG-7 (G-2) was considered as the ideal genotype based on mean performance and stability. According to Yan and Tinker (2006), Euclidian distance was used to estimate the distance between genotypes; thus, genotypes closer to ideal genotypes are regarded as favorable. Therefore, genotypes DSG-6(G-1), DSG-29(G-3), DSGVRL-22(G-4), DSGVRL-23(G-5), DSGVRL-18(G-6), and DSGVRL-3(G-7), were considered as desirable and stable performers with respect to low VI values. For earliness observed by days to first female flower anthesis, genotypes, DSG-29(G-3) and PSG-9(G-46) displayed the lowest and highest values, respectively ( Figure 4B ). Genotype DSG-29(G-3) was considered as the most ideal and stable performer, followed by DSG-7(G-2), DSGVRL-23(G-5), and DSGVRL-18(G-6), across the environments. Similarly, genotypes DSG-29(G-3) and DSG-55 (G-48) were positioned at two extremes of AEA abscissa, indicating the highest and lowest values concerning the yield traits, viz., number of fruits per plant and fruit yield per plant ( Figures 4C, D ). Genotype DSGVRL-18(G-6) was the most ideal and stable, followed by DSG-29(G-3) and DSG-6(G-1). Genotype DSG-6(G-1) was observed as the most ideal genotype, followed by DSG-43(G-20) and CHSG-2(G-40) for average fruit weight ( Figure 4E ).

The “which-won-where” view of the GGE biplot revealed that resistant genotypes were positioned toward the left side of the origin and did not share the sector with any location. Thus indicating that their performance remained relatively consistent across the environments studied ( Figure 5 ). For VI, the vertex genotypes, viz., DSG-7(G-2), PSG-9(G-46), Pusa Supriya(G-49), and Pusa Sneha(G-35), were considered as best or the poorest ( Figure 5A ). The test environments were grouped into two sectoral regions, i.e., mega environments. The first mega environment contains the test environment, E1 and E3, with a winning genotype, Pusa Supriya(G-49), which is susceptible to ToLCNDV due to high VI values. The second mega environment consists of test environments, E2 and E4, with winning genotype PSG-9(G-46), which is also susceptible, as indicated by high VI values. The genotypes DSG-7(G-2) and DSG-29(G-3), followed by DSG-6 (G-1), were plotted farthest on the left side, indicating the lowest VI scores across the environments. Genotypes DSG-29 (G-3), DSGVRL-23(G-5), Pusa Supriya(G-49), Kalyanpur Hari Chikni(G-50), PSG-9(G-46), and JSLG-55(G-42) were vertex genotypes indicating their best or poorest performance in the particular mega environment for days to first female flower anthesis ( Figure 5B ). E1 and E2 were present in one mega environment with the winning genotype Kalyanpur Hari Chikni(G-50), whereas genotype PSG-9(G-46) was observed as the winning genotype for the second mega environment, which consisted of E3 and E4. These genotypes were considered poor performers, as revealed by the delayed first female flower anthesis. However, genotypes DSG-29(G-3) and DSGVRL-23(G-5) presented on the extreme right side of the origin were the best performers with respect to earliness. The genotypes DSG-29(G3), DSGVRL-22(G-4), DSG-55(G-48), DSG-31(G-22), DSG-43(G-20), and DSGVRL-18(G-6) were vertex genotypes whereas environments E1, E2, and E3 were present in a single mega environment with DSG-29(G-3) as the winning genotype for number of fruits per plant ( Figure 5C ). For fruit yield per plant, genotypes DSG-29(G-3), DSGVRL-22(G-4), DSG-55(G48), DSG-43(G-20), and DSGVRL-18(G-6) were the vertex genotypes ( Figure 5D ). All the test environments are grouped in two sectors, thus indicating two mega environments. The first mega environment consisted of three test environments (E1, E2, and E3) and genotypes DSG-29(G-3), followed by DSG-7 (G-2), DSGVRL-22 (G-4), and DSGVRL-18 (G-6) were the best performers. Meanwhile, test environment E4 comes under the second mega environment, with genotype DSG-43(G-20) followed by DSG-31(G-22), which were considered as the best performers for that particular mega environment. In the case of average fruit weight, vertex genotypes were DSGVRL-3(G-7), DSGVRL-22(G-4), DSGVRL-23(G-5), DSG-55(G-48), and CHSG-1(G-39) ( Figure 5E ). The equality line divides the four test environments into two mega environments, E1 and E2 in the first and E3 and E4 in the second. DSGVRL-3(G-7) and DSGVRL-22(G-4) were the winning genotypes of the first and second mega environments, respectively.

Among the four test environments, the longest vector length was recorded in E1, followed by E2 and E3, whereas the shortest was in E4 for VI ( Figure 6A ). Thus, the test environment E1 was the most discriminating location for ToLCNDV screening compared with other locations, whereas E4 had less discriminating ability. Among the test environments, E1 and E2 were the most representative environments revealed by a small angle with AEA abscissa compared with other two environments for disease screening against ToLCNDV. The test environment E1 and E3 displayed acute angles with E2, thus indicating a close association and consensus on the genotypic response to ToLCNDV disease severity. Similarly, E1 was identified as the most discriminating location for fruit yield per plant, and E3 was the representative environment ( Figure 6B ).

The GT biplot facilitates the depiction of relationships between genotype and traits based on trait-standardized data and singular value partitioning focused on traits. The GT biplot analysis showed a strong positive correlation with fruit yield and number of fruits per plant. However, both traits showed a negative correlation with vulnerability index and days to first female flower anthesis as indicated by an obtuse angle among them ( Figure 7A ). Similarly, we recorded a strong positive correlation between VI values and delayed flowering. These results are in conformity with the correlation graph of 3 years of preliminary screening. On the contrary, average fruit weight shows no relation with VI values. Genotype DSG-29(G-3) had a higher fruit yield and number of fruits per plant, low VI values, and earliness. However, DSG-55(G-48) had the highest VI values and delayed flowering. Genotypes DSGVRL-3(G-7) and DSGVRL-22(G-4) had high average fruit weight. Although the GT biplots reveal the association among the traits and genotypes’ trait profiles, they lack the practical utility in decision making on cultivar selection and rejection. Therefore, to address this, the proposed GYT biplot analysis was carried out ( Figure 7 ). Higher days to first female flower anthesis and higher vulnerability index value are undesirable traits for sponge gourd improvement. Thus, the combination of yield*earliness (FYPP/DFFF) and yield*disease score (FYPP/VI) had a division operator (“/”). It reflects that their values were reversed before multiplication to the yield values. The which-where-won view helps in visualizing the genotypes’ trait profiles, and the perpendicular lines divide the yield-trait combinations into two sectors ( Figure 7B ). Genotype DSG-29(-3), followed by DSGVRL-18(G-6), had the largest values for FYLP/DFFF, FYLP*FPP, and FYLP*AFW, thus making them as the best genotype for combining fruit yield with earliness, number of fruits per plant and average fruit weight. Similarly, genotype DSG-7(G-2) had the highest level for FYLP/VI, indicating that it was superlative for combining fruit yield with a low vulnerability index. Figure 7C displays the superiority ranking of genotypes, along with their strengths and weaknesses, as assessed through the average tester coordinator (ATC) view of the GYT biplot. The average tester axis (ATA) is a single arrow line that passes through the biplot’s origin and signifies the direction of superior mean performance. Consequently, it aids in ranking genotypes according to their overall superiority. The genotypes were separated into better and worst performers based on the average overall performance by the perpendicular lines that pass through the biplot origin on ATA. It also suggests the balanced trait profile or strength and weakness for a trait or group of traits of genotypes. Thus, based on the average yield-trait combinations, the best-ranked genotype was DSG-7(G-2), followed by DSG-29(G-3), DSGVRL-18(G-6), DSGVRL-23(G-5), DSGVRL-22(G-4), and DSG-6(G-1), and so on as we move left to right on the ATA. Meanwhile, genotypes ranked poorest are plotted on the far right side of the ATA. The projection of genotypes suggests the strengths and weaknesses; thus, genotype DSG-7(G-2) has a strong performance level with regard to FYLP/VI, and DSG-29(G-3) was strong in FYLP/DFFF, FYLP*FPP, and FYLP*AFW, followed by DSGVRL-18 (G-6). However, the genotypes that are closer to ATA, i.e., short projection length, have a balanced trait profile; thus, DSG-29(G-3), followed by DSGVRL-18 (G-6) and DSGVRL-23 (G-5), had a more balanced overall trait profile. The ideal genotype was identified by genotype ranking plot view of the GYT biplot ( Figure 7D ). DSG-29(G-3) is located in the first concentric circle, thus making it an ideal genotype followed by DSG-7(G-2) and DSGVRL-18 (G-6) with less VI value, earliness, and higher yield.

The MGIDI relies on evaluating the distance between genotypes and specified ideotype aligned with the breeders’ preference (Olivoto and Nardino, 2021). The broad sense heritability ranged from 54.3% for fruit weight to 88.05 for vulnerability index. The studied traits were reported in a single factor (FA1) with most yield-related traits. Fruit yield and number of fruits had positive loadings in FA1, whereas vulnerability index, days to first female flower, and fruit weight had negative loadings. Genotypes ranking based on MGIDI values are depicted in Figure 8 . The selected genotypes based on selection pressure (~15%) were G-3(DSG-29), G-2(DSG-7), and G-6(DSGVRL-18). The selected genotypes resulted in desirable positive selection gains for average performance of yield-attributing traits, viz., fruit yield per plant, number of fruits per plant, and fruit weight. These genotypes also recorded desirable negative gains for vulnerability index and earliness ( Supplementary Table 2 ).