Adaptability and Phenotypic Stability of Resistance to Two Viral Diseases and Yield Traits in Cassava

Cassava productivity is hampered by pests and diseases including cassava mosaic disease (CMD) and cassava brown streak disease (CBSD). The main objective of this study was to identify stable superior genotypes that combine disease resistance and high yield. Sixteen cassava genotypes were planted in a randomized complete block design with three replications for six planting seasons (years) at five sites in Tanzania. The genotypes were assessed using the additive main effect and multiplicative interaction (AMMI) analysis, and highly significant (P < 0.001) effects of genotype, environment, and genotype-by-environment (G * E) interactions were observed for all traits studied. Percent sum of squares (SS) due to environment (12.66% - 85.23%) was the highest followed by G * E (14.12% - 39.56%) for CMD foliar symptoms, root weight and dry matter. On the other hand, % SS due to genotype (52.14% - 69.14%) was highest followed by G * E (26.14% - 35.91%) for CBSD foliar and root symptoms indicating that the environment and G * E greatly influenced trait expression. The most stable genotypes which combined disease resistance and high yield were NDL 2003/31 and NDL 2003/111. The findings of this study will give impetus for the release of new cassava varieties that are not only high yielding but are also dually resistant to both CMD and CBSD in different locations and sites.


Introduction
Cassava (Manihot esculenta Crantz) is a vital food staple in sub-Saharan Africa (SSA), ranked as the number one root crop, followed by sweet potato and yam [1]. With over 300 million MT of annual root production [1], cassava is a major source of carbohydrates in the diet of millions of people in SSA and is grown as a famine reserve crop owing to its tolerance of harsh environmental conditions [2] [3]. The crop also has industrial application as it is used to produce high-quality flour, starch, beverages, animal feeds, alcohol, biofuel, detergents, textiles, plastics and pharmaceutical products [4] [5] [6] [7].
Although Tanzania has the largest area (885,091 ha) under cassava production in East Africa, its average yield is low at 5.7 t/ha [1], which is far below the estimated yield potential of cassava (50 -60 t/ha) [8]. This is due to many biotic and abiotic factors including the two viral diseases: cassava mosaic disease (CMD), and cassava brown streak disease (CBSD) [9] [10] [11]. Cassava roots affected by CBSD have a brown necrotic rot and are unfit for consumption. By contrast, storage roots of cassava plants severely affected by CMD fail to bulk because their leaves become chlorotic and mottled, thus having impeded photosynthesis and leading to stunted growth [12]. Dual infections of CMD and CBSD are common and a serious threat to cassava production and food security as losses more than 80% have been reported in susceptible varieties [13].
Deployment of cassava varieties with dual resistance to both diseases is currently being pursued as the most effective and sustainable way to manage the devastating effects of the viral diseases in Eastern and Southern Africa [14]. CMD, CBSD and yield traits expression in cassava can be influenced by the environment leading to varied phenotypes in different environments [15] [16] [17]. This is defined as genotype-by-environment (G * E) interaction [18] and it can result from differences in the sensitivities of genotypes to the conditions in the target environment [19]. This leads to inconsistent performances across different environments; therefore, limiting the efficiency of selection of superior genotypes.
The objective of most cassava improvement programmes is to identify and select diseases free, high yielding and stable genotypes across several environments and seasons. The efficiency and success of such selections depend on the consistency of the performances of genotypes in varying environments [20] [21]. For this reason, genotypes are tested in diverse environments to assess their adaptability and stability. Genotypes whose G * E effects are not significant are said to be stable [22]. Several methods have been used to assess the G * E effect and stability in crop performances including the additive main effect and multiplicative interaction (AMMI) model [23] [24].
The AMMI model fits the sum of several multiplicative terms rather than only one multiplicative term in assessing the performance of genotypes in different environments [25]. AMMI analysis can be used to determine the stability of the genotypes across locations using the PCA (principal component axis) scores and AMMI stability value (ASV) [26]. The ASV is based on the AMMI model's IPCA1 The main aim of this research was to analyze the effects of G * E interaction on resistance to CMD, CBSD and yield traits on 16 cassava genotypes using the AMMI model. The specific objectives were to 1) Identify superior genotypes that exhibit high stability which combine CMD and CBSD resistance and high yield; 2) Identify environments that best represent the target environment for high expression of the traits.

Study Location and Germplasm
The study was done in five sites: Chambezi, Mtopwa, Nachingwea, Naliendele and Mtopwa for six planting seasons (2013,2014,2015,2016,2017 and 2018) ( Table 1). Advanced breeding lines including released improved varieties and local landraces were evaluated in the study ( Table 2). The advanced breeding lines and improved varieties were obtained from the TARI-Naliendele or the International Institute for Tropical Agriculutre's breeding programmes, while the local landraces were obtained from farmers' fields.

Experimental Design
A randomized complete block design with three replicates was used for this study. Cassava cuttings (about 25 cm long with 4 to 5 nodes and viable buds) from each of the genotypes were planted in 4 rows with 10 cuttings each at a spacing of 1.0 m × 1.0 m, resulting in a total of 40 plants/plot/replicate. To increase disease inoculum pressure, susceptible cassava varieties Albert and Limbanga were planted as spreader rows for CBSD and CMD, respectively [30]. Albert and Limbanga cutting were planted alternately after every 8 plots and as a border row around each replicate. Released varieties and landraces including Albert, Kiroba, Pwani, Mahiza, Mkumba and Naliendele 134 were planted as controls in the experiment. Neither fertilizer nor irrigation was applied; the field was rain-fed throughout the growing period but was kept weed-free.

Data Collection
Data on several parameters were collected including CMD and CBSD foliar severity at 3, 6, and 9 MAP; root necrosis; root weight (t/ha), and dry matter content during harvest at 12 MAP. CMD foliar severity was scored on a 1 -5 scale where: 1 = no visible symptoms; 2 = mild distortion only at the base of leaflets with the rest of leaflets appearing green and healthy/mild chlorotic pattern over entire leaflets; 3 = conspicuous mosaic pattern throughout the leaf, narrowing and distortion of lower 1/3 of leaflets; 4 = severe mosaic, distortion of two-thirds of leaflets and general reduction of leaf size; and 5 = severe mosaic, distortion of ¾ of leaflets, twisted and malformed leaves [31]. CBSD foliar severity was scored on a 1 -5 scale where: 1 = no visible symptoms; 2 = mild foliar mosaic on some leaves and no stem lesions; 3 = foliar mosaic with mild stem lesions and no die back; 4 = foliar mosaic and pronounced stem lesions and no dieback; and 5 = defoliation with pronounced stem lesions and dieback [32]. At 12 MAP, plants were harvested, and roots were examined for CBSD root symptoms. Roots from each plant were chopped longitudinally and transversely to identify the presence of necrotic patches on the starch bearing tissues. Scoring for root necrosis severity was also done based on a 1 -5 where: 1 = no clear symptoms; 2 = <5% of root necrotic; 3 = 5% -25% of root necrotic; 4 = 25% -50% root necrotic and mild root constriction; and 5 = >50% of root necrotic [32] [33] [34]. Roots from each plant were harvested and chopped longitudinally and transversely to check for root necrosis on the starch bearing tissues. Root weight in tonnes per hectare (t/ha) was estimated according to Masinde et al. [16] while root dry matter content using the specific gravity method [35].

Data Analysis
The AMMI model was used to determine the stability of the genotypes across environments. The AMMI model first fits the additive effects for the genotypes and the growing environments (five growing sites and six seasons) and multiplicative term for G * E interactions. The AMMI model according to Gauch [36] and Farshadfar et al. [37] is presented as The ASV was calculated for each genotype according to the relative contributions of IPCA1 and IPCA2 to the interaction sum of squares. The ASV has been defined as the distance from the coordinate point to the origin in a two-dimensional scatterplot of the first IPCA1 scores against the second IPCA2 [27] [38].
The IPCA1 accounts for most of the G * E variation. The IPCA1 scores are weighted by the ratio of IPCA1 SS (from the AMMI ANOVA) to IPCA2 SS in the ASV formula as The larger the IPCA score is, either negative or positive, the more adapted a genotype is to a certain environment. Smaller ASV scores indicate a more stable genotype across environments [39]. Genotype stability index (GSI) was also calculated using the sum of the ranking based on trait and ranking based on the AMMI stability value. GSI incorporates both the mean and stability of the trait being studied in a single criterion. Low values of both parameters show suitable genotypes for example those with high mean yield and stability [29] [40]. Both AMMI and biplot analysis were computed using the R package Agricolae [41].
where RASV = Rank of the genotypes based on the AMMI stability value, RY = Rank of the genotypes based on yield across environments.

CMD Foliar Symptoms
The results of the combined AMMI analysis of variance revealed highly significant (P ≤ 0.001) effects of genotype, environment and G * E for CMD foliar symptoms at 3, 6, and 9 MAP (  higher mean and % SS due to IPCA1 and IPCA2 were observed at 6 MAP in comparison to at 3 and 9 MAP. This indicated that there were fewer interactions, therefore, more stable symptoms expression at 6 MAP. ASV ranked the genotypes based on the least scores where low scores represented the most stable genotypes. Low ASV coupled with low disease severity resulted in the selection of stable genotypes with minimal CMD symptoms. All the genotypes had low CMD severity of ≤1.8 (Table S1). Based on CMD foliar symptoms at 6 MAP, the most stable genotypes with regards to low ASV values and their position relative to the biplot origin (0.0) were Albert, NDL 2003/111, KBH 2002/66 and NDL 2003/31 with a means ≤ 1.27 ( Figure 1, Table S1). The GSI ranking combines both stability and higher scores of a trait. Accordingly, site Chambezi's environments had moderate stability with the highest mean CMD  Table S1).

CBSD Foliar Symptoms
There was a highly significant (P ≤ 0.001) effect of genotype, environment and G * E interaction for CBSD foliar symptoms at 3, 6, and 9 MAP (Table 3). Percent SS due to genotype (60.90% -69.14%) was higher than due to environment    Figure 4, Table S2).

CBSD Root Necrosis
The effect of genotype, environment and G * E interaction was highly significant (P ≤ 0.001) for root necrosis (  Figure 5, Table S3). All the genotypes had low root necrosis severity (≤1.9) below the grand mean

Root Weight
The effect of genotype, environment and G * E interaction was highly significant  (P ≤ 0.001) for root weight (  (Table  S3). Chambezi 2013 to 2018 environments had the highest mean root weight (45.21 t/ha) with moderate to high stability based on GSI ranking ( Figure 6, Table S3). Higher yields were observed in favourable environments for example Chambezi which had higher rainfall than other sites ( Figure 7, Table 1, Table  S3). Site Segera having received the least rainfall was one of the sites with a lower combined mean root weight of 33.47 t/ha.

Dry Matter Content
The effect of genotype, environment and G * E interaction was significant (P ≤  (Table 4). Percent SS due to environment was very high at 85.23% followed by G * E (14.12%) and very low SS due to genotype (0.68%).
Four IPCAs had significant (P ≤ 0.05) mean squares and IPCA1 and IPCA2 accounted for a total SS of 65.43% of the G * E variation. The mean dry matter contents for genotypes were close ranging from 26.95% -28.77% ( Figure 6, Table  S3).

Discussion
The performance of cassava is subject to the strong influence of genotype, environment and G * E interactions [15] [29] [47]. TARI-Naliendele has been developing improved genotypes, however, only a few varieties have been released. The newly developed breeding lines are in their final stages of breeding. Therefore, evaluating them in diverse environments and providing recommendations for suitable ones will contribute to increasing cassava production and improved food and nutrition security.
The AMMI model was used in this study and the effects of genotype, environment, and G * E interactions were significant. Percent SS due to environment was the highest followed by G * E interaction in CMD foliar symptoms severity, root weight and dry matter content showing that environment and G * E interaction greatly influenced the variations observed. On the other hand, % SS due to genotype was the highest followed by G * E interaction in CBSD foliar symptoms and root necrosis. A considerable percentage of G * E interaction was explained by Mean CMD and CBSD foliar symptoms severity increased from 3 to 6 MAP then dropped at 9 MAP. The total % SS due to both IPCA1 and IPCA2 was the highest at 6 MAP for both CMD and CBSD. A possible explanation for this is that at 3 MAP, some plants may still have low viral titre [49] and may not express symptoms thus causing significant variations in the replications and environments. This may result in the representation of substantial % SS by other IPCAs apart from IPCA 1 and 2. CBSD foliar symptoms are more difficult to recognize in older plants as the lower leaves with prominent symptoms senesce and fall off, causing variation in symptoms expression among the plants partic-ularly at 9 MAP [50]. Additionally, younger leaves are more susceptible to CMD resulting in a decrease in CMD symptoms in some plants with increasing plant age [51]. In our earlier study we reported a higher heritability at 6 MAP for CMD and CBSD foliar symptoms thus emphasising the importance of assessment at this time point [20].
Stability analysis methods are often used by breeders to identify genotypes that have stable performance and respond positively to improvements in environmental conditions [39] [40]. With regards to CMD and CBSD, suitable genotypes would have low ASV and low disease severity. Further, genotypes with CMD foliar severity scores (<2.0) are classified as resistant while those with (≥2.0) as susceptible [52]. In this study, all genotypes had low foliar severity (>1.35) apart from Mahiza which was slightly higher at 1.77. The most stable genotypes with low CMD foliar severity (≤1. 28  Chambezi had the highest combined root weight means of 45.21 t/ha. Chambezi had higher rainfall particularly during the first six months, a critical period for root initiation and development [56] [57]. Most of the genotypes had moderately high dry matter content ranging from 26.95% -28.77%. Low rainfall results in high dry matter content as was observed in the environments in site Segera which had the highest combined dry matter content mean of 30.43 [16] [58].
Additionally, the environments with the least CBSD root necrosis symptoms had the highest dry matter content and vice versa, indicating that presence of root symptoms can affect key agronomic traits leading to loss of farmer preferred traits [16] [58].

Conclusion
G * E was significant (P ≤ 0.05) for CMD foliar symptoms, CBSD foliar and root symptoms, root weight and dry matter content. This emphasised the importance of testing genotypes in multiple environments before an effective selection is made. Besides, variations were also significant among the test environments. Site Chambezi had the highest mean CMD and CBSD severity; therefore, it has been empirically confirmed as the most suitable environment for evaluation for dis-