Identification of Elite Pigeonpea Genotypes against Fusarium Wilt and Sterility Mosaic Disease through AMMI and GGE Biplot Analysis

Pigeonpea [Cajanus cajan (L.) Millsp.] is a significant legume crop in tropical and subtropical regions of the world (Behera et al., 2020). Globally, pigeonpea is cultivated in 6.99 M ha with production of 5.93 Mt and with productivity of 852 kg/ha. In India it is cultivated in 45 Lha, with annual production of 42 Lt and contributing nearly 90% of world’s acreage and production (FAOSTAT, 2020) and it is widely cultivated as a low-input, rain fed crop and has a straight impact on the pecuniary and economically well-being and on the nutritive status of the farmers in the country and gives high profitable returns from each part of the plant (Sharma et al., 2019). Among the diseases Fusarium wilt (FW) caused by the fungus Fusarium udum and sterility mosaic disease (SMD) by the virus pigeonpea sterility mosaic virus (and vectored by the eriophyid mite) wreak havoc on this legume crop. The pathogen Fusarium udum survives saprophytically up to eight years in the soil, causing disease and produces various symptoms i.e., loss of turgidity, inter veinal chlorosis, brown discoloration of xylem tissue and a purple band on stem extending upwards from the base. Sterility mosaic disease (SMD), often referred as “Green Plague”, as affected plants are green with excessive vegetative growth but with no flower or pod formation, during favorable conditions spreads rapidly and lead to severe epidemics. Together, these two diseases appear in all growing seasons and during favourable conditions, they are reported to cause even a complete yield loss i.e. up to 100% in susceptible pigeonpea genotypes (Saxena et al., 2021). During the conducive environmental conditions it is uneconomical to treat diseases by means of fungicides and cultural methods. Amidst of these conditions the selection of stable resistant genotypes against these diseases under varying environmental conditions is the primary management choice for minimizing yield losses (Sharma et al., 2012; Bhaskar et al., 2016). Despite considerable efforts to develop Host Plant Resistance (HPR) through various screening techniques for these diseases, there is still a gap in the identification of promising genotypes against FW and SMD under varied environmental conditions. A deeper understanding of genotype (G) × environment (E) interactions with versatile models, like additive main effects and multiplicative interaction (AMMI) model analyses the additive and multiplicative effects of genotype, environment and G × E critically. This model uses ANOVA and principal component analysis (PCA). AMMI biplots and multiplicative model are highly useful for understanding the additive ANOVA model’s interaction effect. The researchers have demonstrated the potential utility of AMMI models to understand G×E interactions and to identify resistant genotypes across varying environments (Abamu et al., 1998; Verma et al., 2016; Parihar et al., 2017). With this context, the current study used AMMI model analysis; (i) to test genotypes ability to resist FW and SMD (ii) to understand genotype output and their interactions with varied environments.

The experiment was conducted at Dr. Rajendra Prasad Central Agricultural University, Tirhut College of Agriculture, Muzaffarpur (Bihar). The sick plots required for conducting experiments were maintained with optimum pathogen population at T.C.A, Dholi and monitored under AICRP pigeonpea wilt and SMD disease screening programmes.

The study was carried out under artificial epiphytotic conditions with fifty pigeonpea genotypes against FW and SMD resistance in their sick plots respectively (Table 1 and Table 1A). The test genotypes were subjected to four different environments i.e., Kharif-2017, 2018, 2019 and 2020 using a randomized block design (RBD) with two replications and in each replication twenty five plants were maintained. Observations on wilt incidence were recorded at 30 days interval from sowing (August) to harvesting (March) by counting healthy plants and wilt diseased plants. Maximum wilt incidence on before harvesting was considered to categorize the genotypes to different disease classes as per the disease scoring scale adopted by AICRP on pulses.





Evaluation for FW and SMD resistance

Fusarium wilt resistance

The wilt sick plot was maintained by incorporating chopped wilt-infected pigeonpea plants in the soil every year. Resistance studies were carried out by planting each genotype in two rows of 5m row length with spacing of 15×40 cm in the sick plot. A susceptible cultivar ICP 2376 was included for every 5 test genotypes as an indicator/infector rows.

Sterility mosaic disease resistance

Resistance studies of SMD carried out through leaf staple technique in isolated disease nursery using SMD-infested susceptible cultivar ICP 8863. At two-leaf stage, SMD infected leaf was pinned to the primary leaf of the test seedling using a small paper stapler in such a way that the lower surface of infested leaf was in contact with the primary leaf thus facilitating in successful disease infection.

Each genotype was planted in two rows of 2 m length with 15×40 cm spacing. A SMD susceptible cultivar ICP 8863 was included for every 5 test genotypes as an indicator row. The ICP 8863 was planted in the sick plot one month in advance of the regular planting time to serve as an infector rows in order to have a good source of virus inoculum (Sharma et al., 2019).

Per cent disease incidence (PDI)

PDI of both FW and SMD recorded in four environments was pooled and categorized into resistant (0-10%), moderately resistant (10-30%) and susceptible (>30%). 


Weather data was taken for crop standing period i.e., sowing (August) to harvesting (March) in weekly intervals from Meteorological Station, RPCAU.

AMMI and GGE biplot analysis

The performance of fifty pigeonpea genotypes and four testing environments for FW and SMD resistance and the G×E interactions were analyzed in this study. The disease incidence data from four test environments were structured to fit the AMMI models. The AMMI statistical model and computational methods described in Gauch (2013) were used. Analysis of variance was utilized to partition the variation into genotype (G) and environment (E) main effects and the G×E interaction effect. The genotype and genotype by environment (GGE) biplot analysis was utilized to graphically ascertain the stability of genotypes over the four test environments (Yan, 2001; Parihar et al., 2017). Both the AMMI and GGE biplots analyses were carried out using the softwares GEA-R developed by ‘CIMMYT’ and ‘R’ package Agricolae.

AMMI analysis of FW and SMD

AMMI ANOVA of FW revealed that among total sum of squares (SS), 82.21% of the SS was observed for genotype effect, 0.47% of SS provides environment effect and 17.31% of SS was observed for interaction effect (G×E). G×E interaction effect was 4.74 times smaller than that for the genotype SS and it was also 36.85 times larger than that for environment SS thus indicating that the variation in the genotypes across the environments were significant. The G×E was further divided into Interaction Principal Component Axis (IPCA) and residuals, in which IPCA1 has contributed 55.85% of SS followed by IPCA2 which contributed 27.96% of SS and IPCA1 and IPCA2 cumulatively contributed to 83.81% of the total SS (Table 2). Similar trend of results were observed for SMD AMMI ANNOVA, i.e., among the total sum of squares genotype effect was 47.99%, environment effect was 15.50% and G×E interaction effect was 36.49%. G×E interaction effect was 1.31 times smaller than genotype and 2.35 times larger than environment. IPCA1 and IPCA2 collectively contributed 89.79% of the total SS, in which IPCA1 contributed 67.70% SS and IPAC2 contributed 22.08% SS (Table 2).



The high SS for genotypes obtained in the AMMI ANOVA analysis for both FW and SMD indicated the diverse nature of the pigeonpea genotypes with significant differences in the mean PDI causing most of the variations within the reactions of the genotypes (Persaud and Saravanakumar, 2018; Persaud et al., 2019). It also indicated that the resistance has also been influenced by the G×E effect. In the current research for both FW and SMD, IPCA1 and IPAC2 collectively contributed more than 80% of interaction effect. It explains that IPCA1 and IPAC2 will be sufficient for studying interaction between 50 pigeonpea genotypes over four environments (Yan et al., 2000; Nayak et al., 2008).

AMMI biplot analysis

AMMI biplot explains the relationship between the per cent disease incidence of genotypes, test environments and IPCA1 scores. Genotypes or environments present left side to perpendicular line are resistant genotypes (lower disease incidence) and environments will be less favourable for disease screening. If they are present on the right side of perpendicular line genotypes will be highly susceptible (high disease incidence) and environments will be favourable for disease screening (Persaud et al., 2019; Srivastava et al., 2021).

AMMI1 biplot (Trait vs IPCA1)

AMMI1 biplot analysis of FW (Fig 1) revealed that genotypes 21, 12, 19, 7, 36, 4, 13, 33, 11, 27, 35, 42, 45, 39, 14, 44, 6, 43, 10, 46, 32, 47, 15, 22, 5, 17, 1, 40, 2, 9, 37, 30, 3 and 16  were resistant for FW and environments Kharif-2018 and Kharif-2020 are less favourable to FW due to the presence on the left side of perpendicular line. While, genotypes 18, 8, 23, 29, 50, 38, 26, 24, 20, 28, 49, 41, 34, 25, 48 and 31, environment Kharif-2019 on the right side of perpendicular line which are susceptible and favourable to FW respectively and Kharif-2017 was present on perpendicular line, it shows that it is a mean environment. Whereas, AMMI1 biplot analysis of SMD shows that genotypes 1, 25, 46, 10, 19, 49, 12, 30, 33, 47, 7, 48, 9, 39, 20, 50, 15, 31, 13, 36, 14, 16, 42, 35, 34, 38, 6, 44 and 26 are resistant for SMD disease and environment Kharif-2019 was less favourable for disease screening, however genotypes 32, 11, 23, 4, 45, 3, 27, 21, 5, 22, 24, 37, 2, 8, 28, 29, 41 and 17 were susceptible to disease and environments Kharif-2017, Kharif-2018 and Kharif-2020 were favourable for SMD disease screening. While genotypes 18, 43, 40 present exactly on the perpendicular line i.e., representing mean genotypes (Fig 2) (Sharma et al., 2019).





Similarly, genotypes or environment with low IPCA1 scores have slight interactions, good stability and better adaptation over test environments, while large IPCA1 scores have big interaction effect and reflect more specific stability and adaptation to specific environments (Persaud and Saravanakumar, 2018). From our experiment it is observed that Kharif-2018, Kharif-2017 and larger proportion of genotypes recorded low IPCA1 scores and showed small interactions, which led to clustering of the genotypes on the FW biplot (Fig 1), While environments Kharif-2019, Kharif-2020 and genotypes 34, 16, 28 recorded the highest IPCA1 scores. In SMD, Kharif-2020, Kharif-2017 observed with low IPCA1 scores had small interactions, while a greater effect was observed for Kharif-2018, Kharif-2019. Similar to FW IPCA1 scores, in SMD also most of the genotypes exhibited low IPCA1 scores which leads to clustering. However, genotypes 11 and 22 recorded the highest IPCA1 scores (Table 3).



AMMI2 biplot (IPAC1 vs IPAC2)

AMMI2 biplot analysis differentiates environments and responsive genotypes. Genotypes which are placed near to test environments have better specific adaptions to test environments, while close to origin shows stable performance in all test environments, while away from origin show differential response to test environments. This analysis also illustrates an interaction between the genotypes and environments with reference to sector, if both are present in same sector then they interact positively, while in opposite sector interacts negatively and in adjacent sectors show complex interactions (Persaud and Saravanakumar, 2018; Persaud et al., 2019).

AMMI2 biplot analysis of FW illustrates genotypes 2 and 25 falling close to Kharif-2017; Genotype 32 falling close to Kharif-2018 and genotype 16 and 31 falling close to Kharif-2020, were particularly suitable and showed stable FW response in that environment. The genotypes 28 and 29 that were between Kharif-2019 and Kharif-2020 indicated highly stable disease response in those environments. While, Kharif-2019 and Kharif-2020 are the most differentiating environments; 16, 31, 28 and 34 are most responsive genotypes (Fig 3). Similarly, AMMI2 biplot analysis of SMD envisages that Kharif-2018, Kharif-2019 and Kharif-2020 were the most differentiating environments; while genotypes 11, 45 and 21 were the most responsive genotypes. From the biplot it is understood that genotypes 42, 6, 13, 15, 26, 50, 48, 47, 7 and 35 falling close to Kharif-2017; Genotypes 22 and 27 falling close to Kharif-2018 were particularly suitable and showed stable SMD response in that environment. The genotype 31 was between Kharif-2017 and Kharif-2020; the genotype 36 was between Kharif-2017 and Kharif-2019 indicated highly stable disease response in those environments (Fig 4).





GGE biplot (discriminative vs representative)

Discriminating power of the environment is proportional to its length from the origin. Closer environments have less discriminative power and by increasing length from origin discriminative power of environment increases and it will explain more about the genotypes. Average environment axis will explain the ideal test environment for testing the genotypes (Persaud and Saravakumar, 2018; Srivastava et al., 2021). FW biplot shows that Kharif-2019 was the most discriminating environment and Kharif-2020 was the least discriminating environment (Fig 5). Kharif-2017 followed by Kharif-2018 were ideal test environments for FW testing because in biplot they were close to the “average environment” and “ideal test environment” and Kharif-2019 and Kharif-2020 were least representative because they were away from AEA (Fig 7).

The SMD biplot showed that Kharif-2018 to be the most discriminating (informative) environment, whereas Kharif-2019 was the least discriminating environment (Fig 6). Average environment axis explains Kharif-2020 was the most representative environment and Kharif-2019 least representative environments (Fig 8) (Tekalign et al., 2017; Tekdal and Kendal 2018; Sharma et al., 2019).









Which-won-where biplot 

FW biplot revealed that two mega environments existed in the study (Fig 9), first includes 3 test environments (Kharif-2017, Kharif-2018 and Kharif-2020) and remaining test environment (Kharif-2019) befitted second mega environment. Genotypes with constant susceptibility in all test environments will be treated as winners. Each mega environment sector showed different winning genotypes (Yan and Tinker, 2006). From our investigation genotype 18 was identified as the winner in first mega environment and genotype 8 was the winner in the second mega environment. Genotypes moving away from origin and distanced from each other reveals their diverse in FW resistant status, having less stability and contributing great in Genotype and G×E interactions, while some of the genotypes (1, 2,  3,   4,  5,  6 , 7,  9,  10, 11,  14,  15,  17,  19,  20,  21,  22,  25,  30,  32,  33,  35,  36,  37,  40,  42,  43,  44,  45,  46,  47, 48  and  49) clustered close to the origin explains the similarity in FW resistant status (Fig 5).



Similarly, the SMD biplot demonstrated that two mega environments existed in the study. First mega environment includes (Kharif-2017, Kharif-2019 and Kharif-2020) while second mega environment includes (Kharif-2018). Genotypes 32 and 27 were winning genotypes in first and second mega environments respectively (Fig 10). Furthermore, the genotypes clustered towards the origin of the biplot (1, 6, 7,  9,  11,  12,  14,  16,  20,  25,  26,  30, 33, 34, 35, 42, 42, 44, 46, 47, 48, 49  and  50)  have been referred to as consistent and stable resistant lines to SMD (Fig 6) (Sharma et al., 2015; Tekalign et al., 2017; Tekdal and Kendal, 2018).



Weather data      
SMD

In AMMI1 biplot of SMD (Fig 2) it was observed that Kharif-2017, Kharif- 2018 and Kharif-2020 were present on the right side of the biplot and they were favourable for disease screening. Weather data also supports the above findings; temperature and wind speed were more in the Kharif-2017, Kharif-2018 and Kharif-2020 when compared to Kharif-2019, the temperature and wind speed may help in the mite population development and spread (Table 4).



FW

In AMMI1 FW biplot (Fig 1), Kharif-2019 on the right side of the biplot and it is more favourable for disease screening. Weather data also revealed that average rainfall also more and temperature was low in the year Kharif-2019 which may help in the pathogen development and spread.

The current research concluded that G×E interaction influencing FW and SMD incidence in pigeonpea and demands for the next studies to know how G×E interaction influences their incidence. The high SS for genotypes obtained in the AMMI ANOVA analysis for both FW and SMD indicated the diverse nature of the pigeonpea genotypes with significant differences in the mean PDI causing most of the variations within the reactions. It also indicated that the resistance has also been influenced by the G×E effect. In the current research for both FW and SMD, IPCA1 and IPAC2 collectively contributed more than 80% of interaction effect. It explains that IPCA1 and IPAC2 will be sufficient for studying interaction between 50 pigeonpea genotypes over four environments. The environment Kharif-2019 suitable for FW disease screening and Kharif-2017, Kharif- 2018 and Kharif-2020 was suitable for SMD screening, the weather data also supports to above findings. AMMI model in integration with GGE biplot, enabled us to identify stable and resistant genotypes to FW and SMD (1, 6, 7, 9, 11, 14, 20, 25, 21, 30, 33, 35, 42, 44, 46, 47, 48 and 49) based on their performance across diverse environments.