Identification of sources resistant to a virulent Fusarium wilt strain (VCG 0124) infecting Cavendish bananas

Bananas are vital for food security in many countries, and half of banana production relies solely on ‘Cavendish’ (AAA), which is presently threatened by the fungal pathogen Fusarium oxysporum f. sp. cubense (Foc) tropical race 4. This particular virulent Foc strain was also found to attack other banana varieties of commercial importance. As there is no single effective management practice available so far, this study was undertaken to determine resistant sources from the genotype collection available at the ICAR-National Research Centre for Banana, Tiruchirappalli, Tamil Nadu, India for direct use by farmers and/or in breeding programmes to develop resistant hybrids. A total of 258 genotypes of different ploidies and genomic constitutions were tested against Foc race 1 (VCG 0124). In total, 19 genotypes (AA Unique-6, BB type-2, AAA Unique-1, AAA Cavendish-1, AAB Mysore-3, AAB Pome-1, AAB Plantain-4 and AAAB-1) were found to be immune; eight genotypes (AA Unique-1, BB type-3, AAA Cavendish-1, AAB Mysore-1, AAB Unique-1, AAB Plantain-1) were highly resistant; and nine genotypes (AA Unique-1, AAA Cavendish-3, AAB Silk-1, AAB Pome-4) were resistant. The genotypes that are resistant to the virulent Foc race 1 (VCG 0124) strain can be exploited directly for commercialization and/or in breeding programs to develop resistant hybrids.

Globally, banana is one of the most important fruits and the eighth-most important food, with ~ 148 million tons produced from 135 countries. It provides staple food for > 400 million people worldwide and is vital for food security. Although there are more than 1000 varieties of banana 1 , the entire global banana industry relies solely on Cavendish varieties (subgroup AAA) with ~ 50 million tons of annual production 2 . India is the largest producer of banana, with 29.7 million tons per annum from an area of 0.802 million ha, contributing 20.1% to global production 3 . Although more than 20 varieties are commercially grown in different parts of India, Cavendish alone occupies 57% of the total area under banana cultivation 4 .
Among the production constraints, Fusarium wilt caused by the fungal pathogen Fusarium oxysporum f. sp. cubense is considered one of the most devastating diseases in banana worldwide, including India 5 . The pathogen causes yellowing of leaves, pseudostem splitting, internal vascular discolouration and eventually death of the plants 6 . It survives in the soil for decades and has wide genetic diversity within three physiological races infecting banana 4 . Foc race 1 attacks Gros Michel (AAA), Silk (AAB), Pisang Awak (ABB), etc., race 2 attacks ABB cooking bananas and race 4 attacks Cavendish, in addition to the varieties susceptible to race 1 and race 2. In India, 12 different vegetative compatibility groups (VCGs), 0124, 0125, 0124/0125, 0128, 0129, 01211, 01212, 01214, 01217, 01218, 01220, 01221 and 01213/01216, have been recorded in different banana varieties, and VCG 0124 of race 1 was found to attack the Cavendish banana varieties in the Theni district of Tamil Nadu 7,8 . This virulent strain has now spread to the entire banana-growing belt of Theni district, Tamil Nadu, resulting in field abandonment or the cultivation of less remunerative crops. In addition, the strain has spread to other major banana growing states of West Bengal, Madhya Pradesh, Gujarat and Maharashtra (Thangavelu, unpublished). Foc race 1 could also cause serious damage to the banana industry of other countries in the near future if it crosses the border.
Different management practices exist against Foc pathogens, but complete protection is possible only through cultivating resistant cultivars 4,6 . Hence, there is the need to identify resistant banana genotypes either for direct OPEN 1 ICAR-National Research Centre for Banana, Tiruchirappalli 620102, India. 2 Laboratory of Tropical Crop Improvement, KU Leuven, Leuven, Belgium. 3 International Institute of Tropical Agriculture, Arusha, Tanzania. 4 These authors contributed equally: R. Thangavelu, M. S. Saraswathi and S. Uma. * email: rtbanana@ gmail.com cultivation or for improving commercial varieties through breeding. A cross between the Foc race 1-susceptible genotypes 'Sukali Ndizi' and 'TMB2 × 8075-7′ resulted in complete resistance against Foc race 1 (VCG 0124), with the single recessive gene responsible for resistance named Panama disease 1 9 . The present study aimed to identify sources of resistance to Foc race 1 (VCG 0124) to use in breeding for developing resistant varieties.

Results
A total of 258 genotypes belonging to different genomic groups of bananas were included in the disease screening under glasshouse and field conditions against Foc race 1 to accomplish the formulated objectives. Before assigning the disease reaction categories of the tested banana genotypes, the distribution of the IWDS (Internal Wilt Disease Score) was assigned based on the percentage of discoloured area in the corm on a 0-5 scale: where 0 = corm completely clean, no vascular discolouration; 1 = 1 to 5%, 2 = 6-25%, 3 = 26-50%, 4 = 51-75%, and 5 = over 75% discoloured corm) 22 was analysed for its normality and variance in addition to mean comparisons of genotypes within a genomic group under glasshouse and field conditions by ANOVA and t-test. The IWDSs of all genotypes evaluated are presented in terms of minimum, maximum, mean and mode values with DMRT (Duncan's multiple range test) and pairwise mean comparison t-test in Supplementary Table 1 Table 1), and 160 (62%) genotypes did not show a shift in resistance level between glasshouse and field conditions. The distribution of IWDSs in pooled data showed that 22.5% of genotypes (36) were classified as resistant, with maximum scores ranging from 0 to 2. Among them, 52.8% (19 genotypes) were immune to the disease, showing a maximum score of 0, 22.2% (8 genotypes) were highly resistant, with a maximum score reaching 1, and 25% (9 genotypes) had a maximum score of 2 and were classified as resistant. The remaining genotypes (77.5%; 124) were found to be susceptible, reaching a maximum score between 3 and 5. Among them, 80.6% of genotypes (124) were highly susceptible to disease, showing a maximum disease score of 5, 8.9% (11 genotypes) had a maximum score of 4 and were classified as susceptible, and 10.5% (13 genotypes) were moderately resistant, with a maximum score reaching 3. Highly significant variations were observed in the IWDSs among the genotypes (F = 169.9; df = 257; P > 0.001) and genomic groups (F = 637.9; df = 6; P > 0.001) tested. The overall analysis of the IWDSs showed that the AA genomic group registered disease scores ranging from 0 to 3, while AAA, AAB, AB, ABB, BB and Tetraploid (AAAB &ABBB) groups registered IWDSs of 2 to 3, 2 to 5, 4 to 5, 5, 2 to 4, and 2 to 5, respectively, under glasshouse conditions ( Supplementary Fig. 2). Except for the AAA (t = 0.460, P > 0.0001), AB (t = 0.424, P > 0.001) and BB (t = -0.386, P > 0.01) genomic groups, all other genomic groups exhibited similar disease reactions (P < 0.05) under field conditions. To establish the relationship between glasshouse and field experiments, the pooled data were subjected to Pearson rank correlation analysis. The correlation coefficient of the glasshouse IWDSs (r = 0.90, df = 256, P < 0.001) and DI (r = 0.90, df = 256, P < 0.001) showed a highly significant correlation ( Fig. 1) with the field IWDSs and DI.
As there was a significant shift in the IWDS and DI of banana genotypes, multivariate techniques, hierarchical clustering heatmaps (HCHs) and principal coordinate analysis (PCoA), were employed to classify and visually select the genotypes based on the colour scale of disease reaction categories (i.e., Immune-I, Highly resistant-HR, Resistant-R, Moderately resistant-MR, Susceptible-S, and Highly susceptible-HS) derived from glasshouse (GDI) and field (FDI) disease indices. Based on the HCH (Fig. 2), all the banana genotypes evaluated were separated into two major clades. Clade 1 (C1, Fig. 2), consisting of 48 genotypes, was categorized into the I (blue), HR (green) and R (dark green) genotypes, while clade 2 (C2, Fig. 2) consisted of 210 genotypes, which were categorized into the MR (grey), S (orange) and HS (red) genotypes. The shift in resistance levels of certain genotypes (38%) might be due to the difference in the DI in field and glasshouse experiments. It was observed that 3. To further confirm these results, PCoA was conducted to visually determine the disease reaction categories of the 258 banana genotypes. The data of individual genotypes were shifted to be centred at zero and then normalized to obtain a unitary variance. Eigenvalues were examined to determine the extent of variation explained by the principal coordinates where FDI was principal coordinate-1 (PC1) and GDI was principal coordinate-2 (PC2). PCoA showed that PC1 accounted for 95% of the variation and that PC2 accounted for 5% of the variation. Figure 3 clearly shows different clusters that separate banana genotypes into I, HR, R, MR, S, and HS categories based on PC1 and PC2. Individual genotypes with I, HR, R and MR reactions are shown in the left part of the graph, with loads from -0.07 to -0.15 for PC1 and from -0.01 to 0.18 for PC2. Individual genotypes with S and HS reactions are in the centre to the right side of the graph, with loads from -0.01 to 0.05 for PC1 and from -0.01 to 0.28 for PC2. Red points (labelled with I) correspond to genotypes with an immune reaction (19), brown points (labelled with HR) correspond to genotypes with a highly resistant reaction (11), green points (labelled with www.nature.com/scientificreports/ R) correspond to genotypes with a resistant reaction (18), copper blue points (labelled with MR) correspond to genotypes with a moderately resistant reaction (30), sky blue points (labelled with S) correspond to genotypes with a susceptible reaction (42) and pink points correspond to genotypes with a highly susceptible reaction (138). Therefore, the PCoA further strengthened the results observed in the dendrogram in determining the disease reaction categories of the banana genotypes tested.

Discussion
Among the production constraints, Fusarium wilt caused by Foc is a serious problem for banana worldwide 10 . Epidemics of Fusarium wilt due to Foc race 1 caused severe damage to Gros Michel (AAA) plantations in the twentieth century, after which the Foc race 1-resistant alternative Cavendish banana rescued the global banana industry 11,12 . Grand Naine, a Cavendish variety, contributes to 47% of global production, with an export of 22.3 million tons worth US$ 13.8 billion per annum. The production of Cavendish in Asia 13 is also seriously threatened by the virulent Foc TR4 14 . This lethal strain has spread to 19 different banana-growing countries 15,16 and is The Foc race 1 strain (VCG 0124) was found to be more virulent, as it affected more commercial banana cultivars and resulted in disease symptoms ten days earlier, than other Foc race 1 strains. The Cavendish banana grown in Nampula, Mozambique (pers. comm. Altus Viljoen, Stellenbosch University, South Africa) and in New South Wales, Australia 18 also succumbed to Foc race 1 (VCG 0124) during 2013 and 2010-2011, respectively. This indicates that Foc race 1 along with TR4 can also seriously challenge Cavendish banana production worldwide. In addition, this particular Foc race 1 is already present in most banana-growing regions of the world, causing serious damage to many other banana varieties. Among the available strategies to combat this fungal disease 19 , cultivation of resistant varieties is considered to be the most effective method 20 . Therefore, we screened banana genotypes available in India belonging to various genomes, groups and subgroups both in glasshouse and field conditions (for main and ratoon crops) to identify sources of resistance to Foc race 1 for utilization either directly or for incorporation into the banana improvement programme.
In the present study, 258 genotypes consisting of diploids (AA, AB, BB group and wild species), triploids (Cavendish, Plantain, Pome, Silk, Mysore, Pisang Awak, Monthan, etc.), and tetraploids (AAAB, ABBB) were evaluated, and their IWDSs and DI were statistically analysed. The results covered a wide diversity (Fig. 4). Based on the statistical analyses, 19 and 22 genotypes were immune, 17 and 11 were highly resistant and 24 and 22 were resistant to Fusarium wilt race 1 under glasshouse and field conditions, respectively. Many genotypes were consistently immune (19), highly resistant (8) and resistant (9) under both glasshouse and field conditions. They belonged to the AA Unique (8), AAA Cavendish (5), AAB Pome (5), AAB Plantain (5), AAB Mysore (4), BB type (5), AAA Unique (1), AAB Silk (1), AAB Unique (1), and AAAB groups (1). However, 98 of 258 genotypes evaluated showed a significant difference in IWDSs between glasshouse and field conditions (Supplementary Table 1). This shift in the disease reaction categories might be due to (i) the different ages of the planting material used Although certain genotypes showed variation between the screening conditions, the correlation analysis showed a highly positive and significant relationship based on the IWDS and DI at the 95% confidence level. This result indicates the consistency of genotype performance in addition to the possibility of the early selection method for resistance to Fusarium wilt under glasshouse conditions. The heatmap (Fig. 2) generated in the present study based on the DI of 258 genotypes evaluated against Foc race 1 (VCG 0124) under both GDI and FDI conditions showed the presence of two major clades, viz., C1 and C2. In general, C1 contained immune, highly resistant and resistant genotypes, and C2 contained moderately resistant, susceptible and highly susceptible genotypes. This is in accordance with the results of the PCoA, indicating that the results observed in the dendrogram were similar to those of the PCoA in determining the disease reaction categories (Fig. 3). As both analyses clearly distinguished all the tested banana genotypes into different resistance categories, one can visually select the genotypes of interest based on the colour scale and progression of the symptom values 26 .
Although a differential (immune to susceptible) reaction was observed in certain genome groups evaluated, only a susceptible reaction was recorded in all AB genotypes (Kunnan and NeyPoovan types), all ABB genotypes (Pisang Awak, Monthan and Unique) and one tetraploid ABBB tested under both glasshouse and field conditions. Hence, one could hypothesize that genotypes with 50% or more of the B genome (B-and B-rich genotypes) might be especially susceptible to Foc race 1.
As India is extensively under Cavendish banana production, 14 Cavendish ecotypes have been selected, among which the important ecotypes are Shrimanti, Manjahaji, Singapur, Borjahaji, Jahaji, Williams, Madhukar, Pedda Pacha, etc., 27 . Out of the 14 Cavendish genotypes evaluated, five genotypes, Shrimanti (immune), Manjahaji (highly resistant; Fig. 5), Williams, Borjahaji, and GCTCV-119 (resistant), showed differences in the degrees of resistance under both glasshouse and field conditions. Based on earlier reports, the dwarf variant of Cavendish, Dwarf Parfitt and GCTCV 119 14,28 were resistant to Foc TR4, while the present study showed resistance to Foc race 1. The AAA Cavendish cultivars Shrimanti (immune), Manjahaji (highly resistant), Williams and Borjahaji (resistant) grouped together in one clade in the diversity studies of Cavendish types using IRAP (Fig. 4) and RAPD markers 27 . Red Banana (AAA-Red Dacca), which is being grown in different parts of the world, including India, was found to be consistently immune under both glasshouse and field conditions. Due to Cavendish www.nature.com/scientificreports/ succumbing to Foc race 1 (VCG 0124) in India, this variety is being widely cultivated to replace Grand Naine in the hot spot area of Theni district, Tamil Nadu. Among the 23 AB diploid genotypes (Kunnan-14 and NeyPoovan-9), none were found to exhibit Foc resistance. Their morpho-molecular characterization 29 suggested that they might have evolved from the same ancestors (Fig. 4), which lacked Foc race 1 resistance. NeyPoovan (AB) has much commercial importance and is grown extensively in all southern states of India and most of the East African countries where it is called Sukali-Ndizi. Breeding NeyPoovan is difficult due to its female sterility 30 , but Kunnan has residual fertility 31 .
Out of the 15 Mysore subgroups among the AAB genotypes, three genotypes, Cheeni Champa, Karpura Chakarakeli and Kottavazhai, were immune, and Terabun was highly resistant to Foc race 1 under both glasshouse and field conditions. Of these, Kottavazhai is not commercially cultivated, as it exhibits enlarged ovules resembling seeds that are not preferred by consumers. Thus, Cheeni Champa and Karpura Chakarakeli, which are generally high yielders, could replace the other Mysore types of bananas.
Among the 22 genotypes belonging to the Silk subgroups (AAB), only Sabri showed resistance to Foc race 1 under both glasshouse and field conditions. As Sabri is an ecotype of Rasthali, a niche cultivar of the northeastern state, Tripura, this genotype can replace the other Silk genotypes grown in the southern states of India. Among www.nature.com/scientificreports/ the 37 genotypes of the Pome subgroup, only Atrusingan showed an immune reaction but has no commercial value due to its poor bunch quality. Its residual female fertility offers scope for improvement through breeding. Among the AAB Pome types, Ladies Finger and Pacha were found to be resistant and clustered together in one microcluster sharing > 80% similarity based on the IRAP diversity analysis 29 . None of the AAB Unique genotypes exhibited resistance against Foc race 1, except Popoulu, an introduced banana genotype. These genotypes could be used for varietal rotation. Interestingly, all five plantain types were found to be immune to highly resistant. AAB plantains such as Nendran (296 and 615), which were immune in the present study, clustered together (Fig. 4) in the IRAP-based diversity study 29 . Their morpho-molecular characterization 29 suggested the possibility of common ancestors to plantains. The resistant plantains are being grown commercially and have replaced Grand Naine in the Foc race 1 endemic area of Tamil Nadu in India.
Within the ABB genotypes of the Pisang Awak, Monthan and Unique subgroups, only Karthobiumtham was highly resistant and resistant to Foc race 1 in the glasshouse and field conditions, respectively. As this genotype is also resistant to the root-lesion nematode Pratylenchus coffeae 32 , it can be used as a good source of resistance to Fusarium wilt and P. coffeae.
Among the 17 AA Unique genotypes, eight genotypes consistently showed either immune, highly resistant or resistant reactions to Foc race 1 under glasshouse and field conditions. However, the use of Pisang Berlin, Tongat and Hatidat in breeding is limited due to their poor bunch and fruit qualities. Wild AA diploids such as Balukpong and M. acuminata-Assam, which were immune in the present study, were grouped in the same clade in the IRAP diversity studies (Fig. 4) 29 . Among the wild genotypes (seeded) with an immune reaction to Foc race 1, M. acuminata ssp. burmannicoides (also called Calcutta 4) is extensively used in breeding programmes worldwide 33,34 . Pisang Jari Buaya with multiple resistances is also used extensively across various banana breeding programs, resulting in several promising hybrids 35,36 . In the Indian banana breeding programme, however, it produces few seeds and progenies. Matti, an important AA parthenocarpic diploid, and a landrace in southern India showed immune and highly resistant reactions under field and glasshouse screenings, respectively. Many improved diploids have been developed using Matti as one of the parents in the breeding programme 37,38 .
In the present study, five out of 22 M. balbisiana genotypes showed consistently resistant reactions against Foc race 1. Wild BB diploids Manohar and Sasrabale were immune in the present study and clustered in an IRAPbased diversity study 29 . Among these, Bhimkol, a soft seeded type valued for its rich nutritional and therapeutic value, is grown extensively in the backyards of the hilly northeastern region of India. Considering the importance of this genotype, efforts are being made to develop a seedless Bhimkol, but progress is slow due to its long crop duration (18-24 months). Bhimkol is also less susceptible to stem weevils 39 ; hence, this genotype could be used in breeding to develop improved parthenocarpic diploids with multiple resistance.
In general, this screening study of diverse banana genotypes against Foc race 1 infecting Cavendish and, more particularly, the evaluation of AA and BB diploid genotypes indicates that wild (M. acuminata and M. flaviflora) and AA parthenocarpic genotypes exhibit much better resistance than M. balbisiana types. Approximately 64% was isolated from dried vascular strands of wilt-infected Cavendish banana (cv. Grand Naine-AAA) using 25% strength potato dextrose agar (PDA) medium 7 . The single spore culture obtained was maintained on carnation leaf agar medium 40 for immediate use, and for long-term use, the culture was stored on dried filter papers at 4°C 41 . The pathogenicity of the fungus was tested by inoculating Foc race 1 into potted tissue-cultured Grand Naine plants under glasshouse conditions. The pathogen was multiplied in sand maize (19:1) medium and used for inoculation of plants in the glasshouse.
Glasshouse evaluation. Disease-free suckers weighing approximately 1.5 to 2 kg were extracted from healthy mother plants of all banana genotypes and subjected to paring and pralinage (dipping the pared suckers in Triazophos-40% EC at 0.5% conc. for 30 min) treatments to protect them from nematodes and insect pests. These treated suckers were planted in mud pots (30 × 30 × 15 cm) filled with a sterilized potting mixture containing red earth, sand and cattle manure in equal proportions. Suckers grew for 30 days with regular watering. Thereafter, approximately 30 g of sand maize Foc inoculum (Foc R1 VCG 0124) containing approximately 10 6 cfu g -1 was applied to the soil around the plants in each pot. The experiment was conducted in a completely randomized design and with ten plants (i.e., replications) per genotype. Five months after Foc inoculation, the plants were pulled out, and disease evaluation was carried out based on the percentage of discoloured area in the corm on a 0-5 scale, where 0 = corm completely clean, no vascular discolouration; 1 = 1 to 5%, 2 = 6-25%, 3 = 26-50%, 4 = 51-75%, and 5 = over 75% discoloured corm 22 . Field evaluation. All genotypes were evaluated for two cycles in 2016-17 and 2017-18 in the Theni district of Tamil Nadu, India. The selected commercial plantation site was abandoned, as more than 90% of the Cavendish plants there exhibited Fusarium wilt. The red loamy soil of the study area had a pH of 6.2. The average annual rainfall of the area was 1,000 mm with an ambient temperature of 20.3-38.6 °C and RH of 37-80%. The suckers were prepared for glasshouse planting and were planted at a spacing of 2 × 2 m. The experiments were conducted in a completely randomized block design, with each genotype represented by ten replications for each cycle. Timely application of fertilizers, manures, water and other cultural operations were followed according to standard production practices 42 . The observation of the IWDS was taken at the end of each cycle based on the percentage of discoloured area in the corm at the time of harvest or when the plants were severely infected with wilt disease as described earlier. Based on these observations, the wilt index was calculated as follows: Disease Index (DI) = (average disease score of each plant/the highest disease index value) × 100 22 . The disease index was calculated for both the main and ratoon crops, and the average values of pooled samples are presented. Based on the DI, the genotypes were classified into six disease reaction categories as described in Zuo et al. 22  Genetic diversity of banana genotypes using IRAP marker. The genomic DNA of the 258 banana genotypes was extracted from cigar leaf samples following the CTAB method 43 and inter-retrotransposon amplified polymorphism (IRAP) analysis was performed using long terminal repeat (LTR) primers for banana as reported by Saraswathi et al. 29 . The PCR-amplified products were scored as the presence (1) or absence (0) of bands in a binary format and analysed using NTYSYS 2.01i. A genetic similarity (GS) matrix was then used in www.nature.com/scientificreports/ hierarchical cluster analysis using the unweighted pair group method with arithmetic averages (UPGMA) and sequential agglomerative hierarchical and nested (SAHN) clustering methods (NTSYS statistical package, Rohlf, 1990) to produce a dendrogram. The reliability and robustness of the dendrogram were tested by bootstrap analysis using DARwin software ver. 6.0.021 (https ://darwi n.cirad .fr/produ ct.php) with 1000 replications.
Statistical analysis. All analyses and plot generation took place in R, version 3.6.1, using the "Rbase 3.6.1" 44 "dplyr 1.8.6" 45 and "ggplot2 3.3.2" 46 packages. To describe the difference in the IWDSs of the banana genotypes against Foc race 1, the distribution of the IWDSs (0-5 scale) under both glasshouse and field experiments was analysed separately. To confirm the assumption of normality in the disease scale within a genotype, the Shapiro-Wilk test was used with untransformed data using the R function "shapiro.test()". The homogeneity of the variance across experiments and genomic groups was assessed using Bartlett's test, as implemented in the R function "var.test()". The mean IWDS of a genotype was assumed if a genotype had a significantly (P > 0.05) higher frequency of occurrence analysed by χ 2 (chi square) test using the R function "chisq.test()". In the case of field experiments, the data from the first and ratoon crops were pooled and analysed in every instance. Furthermore, the difference between the genotypes within a genome group was elucidated by comparing the mean IWDS and DI of the genotypes by one-way analysis of variance (ANOVA) using the R function "aov()", where replications were included as a random variable to account for the nonindependence of successive counts from the same genotype followed by DMRT using the "agricolae 1.3-3" 47 package with the "duncan.test" function. The difference in the mean IWDS of banana genotypes versus experimental conditions (i.e., glasshouse and field) was tested with two-sample paired Tukey's HSD test using the R function "Tukey HSD()", where P > 0.05 is considered significant, which indicates a shift in the resistant reaction. The concurrence between the glasshouse and field in terms of IWDS and DI was established separately by the Pearson correlation coefficient (r) using the "ggstatsplot 0.6.1" 48 package with the "grouped_gghistostats" function since the data were normally distributed (according to the Shapiro-Wilk test for normality). The multivariate techniques HCH and PCoA were used to classify and visually select the genotypes based on the colour scale of the disease reaction category in a threedimensional space (x, y and z axes), where the DIs measured under both the glasshouse (GDI) and field (FDI) conditions were centred and scaled before performing the PCoA. The HCH was drawn using the "heatmap.2" function of the "gplots 3.6.3" 49 package, and the PCoA was drawn using the "prcomp" followed by the "autoplot" function of the "ggfortify 0.4.11" 50 packages based on the distance matrix of the genotypes. The colour code for the HCH of the DI was blue (immune-I), green (highly resistant-HR), dark green (resistant-R), grey (moderately resistant-MR), orange (susceptible-S), and red (highly susceptible-HS).