New cassava germplasm for food and nutritional security in Central Africa

Cassava is a key food security crop in Central Africa, but its production depends largely on the use of local farmers’ varieties characterized by inherently low yield which is compounded by generally high susceptibility to various growth and yield-limiting pests and diseases. Improved cassava genotypes have demonstrated the potential to substantially improve cassava’s contribution to food security and the development of the cassava industry and the improvement of nutrition status elsewhere in Western Africa. Eleven improved cassava genotypes were compared with a local landrace (LMR) used as a check under field conditions over two years in eight locations, grouped in four agro-ecologies in Cameroon. Pest and disease abundance/incidence and damage severity were evaluated. At harvest, root yield and carotenoid content were measured. Best linear unbiased predictors showed the lowest breeding value for LMR with the cassava mosaic virus disease (+ 66.40 ± 2.42) compared with 1.00 ± 0.02% for the most susceptible improved genotype. Two genotypes (I010040-27 and I011797) stood out for having higher predicted fresh root yield means which were at least 16 times greater compared with LMR. Predicted total carotenoid content was the highest (+ 5.04 ± 0.17) for improved genotype I070593 compared with LMR which showed the lowest (− 3.90 ± 0.06%) and could contribute to the alleviation of vitamin A deficiency from cassava-based food systems. Diffusion of high-yielding and nutritious genotypes could alleviate food and nutritional security in Central Africa.


Best linear unbiased predictors (BLUPs) for the diseases: CMD and CAD. The predicted breed-
ing values of CMD ranged from − 7.79 ± 2.42 for genotype I090521, which did not show any disease symptoms, to + 66.40 ± 2.42 for the local landrace LMR, which had the highest CMD incidence and severity (+ 1.36 ± 0.10) ( Table 2). The most infected improved variety was I010040-27 with a BLUP of + 1.00 ± 0.02%. Other genotypes performed below the grand mean and with a very small difference among them (Fig. 1A).
The quadrants in Fig. 2 represent the four classes of cassava genotypes/location for a joint interpretation of CMD performance and stability, using the weighted average of absolute scores for the BLUP of the genotypes and location interaction (WAABS) in the eight locations. The 1st quadrant shows that no genotype contributed much to the genotype and location interaction, but there were 2 locations-Foumbot and Meyomessala-that displayed a high discriminative ability for CMD. The 2nd quadrant includes two genotypes, I010040-27 and LMR, which were the most infected and unstable. The two locations included on that quadrant (Gamboula and Njombe), in addition to high disease incidence, also had good discrimination ability for the genotypes. Ten genotypes fell in the 3rd quadrant and are considered as low-infection genotypes, because of the low WAABS values for CMD. No location was included in 3rd quadrant, which was the environment with low infection and Table 1. Average soil physical and chemical characteristics (± SE). Means in a column followed by the same letter are not significantly different. SE standard error, CV coefficient of variation. www.nature.com/scientificreports/ discrimination ability. There were no broadly adapted genotypes to be displayed in the 4th quadrant; however, there were 2 locations (Ekona and Meiganga) that had high CMD incidence but with low discriminative ability. For CAD, the BLUP breeding values ranged from − 4.43 ± 2.61 for improved genotypes (I070557), which showed the lowest CAD incidence,-to + 4.90 ± 3.21 for LMR which also had the highest CAD severity (+ 0.07 ± 0.01) ( Table 2). Six genotypes had CAD incidence above the grand mean, but the predicted disease incidence was below 15% for all genotypes (Fig. 1B).
Results from all genomic predictions showed higher heritability for CMD incidence (0.90) and severity (0.64) compared with CAD. The linear mixed model test indicated highly significant effects (p < 0.001) in genotype x environment interactions for all the parameters ( BLUP for fresh root yields, biomass, dry matter, and total carotenoids content (TCC). Two genotypes (I010040-27 and I011797) stood out for having the highest predicted fresh root yield means among all the tested genotypes ( Table 3). The yields of genotypes I090590 and I0701026 were also above the grand mean (Fig. 3B). The predicted breeding values for fresh root yield ranged from − 12.2 ± 2.23 for genotype I070557, which showed the lowest root yield, to + 13.5 ± 1.89 for genotype I010040-27, which had the highest root yield. For cassava above-ground biomass, which included stems and leaves, the predicted breeding values ranged from − 3.73 ± 1.62 for genotype I071026, which showed the lowest value, to + 8.51 ± 1.57 for genotype I090616, which had the highest aboveground biomass. The biomass of five genotypes was above the grand mean (Fig. 3A) The predicted breeding values for dry matter ranged from − 4.03 ± 0.02 for genotype I070593, which showed the lowest DM content, to + 9.36 ± 3.03 for the local landrace LMR, which had the highest DM content (Table 3). Pre- Table 2. Best linear unbiased predictors (BLUPs) of breeding values with standard errors for cassava mosaic disease (CMD) incidence (%) and severity, cassava anthracnose disease (CAD) incidence (%), and densities of, whiteflies (WF), cassava green mite (CGM). *P < 0.05, **P < 0.01. The statistics listed for every trait are broadsense heritability, genotype variance, residual variance, grand mean, SD = Standard Deviation, the coefficient of variation (CV %), the number of replications (n Replicates), and the number of genotypes (n Genotypes). The statistics shown are the estimates derived.  (Table 3). Overall, the highest heritability (0.60) among the traits was estimated from TCC ( Table 3). The linear mixed model indicated highly significant effects (p < 0.001) for genotype and environmental interactions for all tested traits ( Table 3).
The quadrants in Fig. 4 represent the 4 classes of cassava genotypes/location for a joint interpretation of fresh root yield and stability using the weighted average of absolute scores for the BLUP of the genotype and location interaction (WAABS) in the 8 locations. The 1st quadrant shows that genotype I070557 was the most unstable and contributed much to the genotype and location interaction. The Njombe environment, which is displayed in that quadrant, had a high discriminative ability for fresh root yield. The 2nd quadrant includes two genotypes I010040-27 and I011797 which are the most productive but unstable genotypes. The two locations included on that quadrant, Ekona and Bambui, in addition to providing high performance, also provided good discrimination ability for the genotypes. Six genotypes fall in the third quadrant and are considered as low productive genotypes because of the low WAABS values. The Mbalmayo and Meyomessala environments included in this quadrat had Black and grey circles represent the genotypes that had BLUP above and below of BLUP mean, respectively. Horizontal error bars represent the 95% confidence interval of a prediction considering a 2-tailed t test. (R V3.6.2, https:// cran.r-proje ct. org/ web/ packa ges/ metan/). www.nature.com/scientificreports/ low production and low discrimination ability. Two genotypes I090590 and I071026 were broadly adapted and are displayed in the fourth quadrant. Three locations in this quadrant, Foumbot, Gamboula, and Meiganga, had high yield but with low discriminative ability.

Discussion
Multi-environment experiments are a primary focus in plant breeding programs; therefore, their prediction accuracy, compared with observed value, is crucial for selection recommendation of cultivars, and the identification of mega-environments. This study showed variable responses of a set of new cassava genotypes for food and nutritional value in contrasting environments in Cameroon, thus justifying the necessity to evaluate genotype by environment interactions. This study, together with the previous study conducted in the same locations, but with different genotypes 8 , provide a highly important set of data on the effects of genotype and environment on 28 cassava genotypes that are in advanced stages of development. Moreover, the two studies together, are a rarity in evaluations that covers a very wide spectrum of pest/disease resistance, yield, and nutritional value, all with a significant potential contribution to food and nutritional security in Central Africa. The four agroecological zones in which the genotypes were tested are representative of the agro-ecologies of the Congo Basin, including nearly all states in Central Africa, except the Sahelian zone 33 . Importantly, we found that the environment was not the main determinant of a genotype response to cassava mosaic virus (CMD) as the heritability was close to one; this opens the possibility of introducing genotypes with good performance in terms of productivity, yield stability, and resistance to pest and disease into areas with similar characteristics as in Cameroon. Indeed, as with varieties for industrial processing, the tested genotypes were initially selected based on their resistance to CMD and yield, as well as provitamin A content for the yellow root genotypes, traits in which variability comes from genetic differences, with very little contribution from environmental factors 34 . We noted, however, two genotypes, I070593 and LMR, that had CMD severity scores of 2.8 and 2.6 respectively on average for both years, although the incidence on the same genotypes was nearly nil (less than 2% on average). Our two checks, I010040-27 and LMR, displayed on average the same incidence and severity as in the previous study 8 ; however, the other improved varieties (10 in total) were five-fold more resistant to CMD compared with the 16 of the previous study, thus confirming the improvement made on the current set of varieties with respect to CMD resistance and the necessity to disseminate those genotypes. The significant interaction observed between environment and genotype for CMD infection could be related to the virus strains present in the locations as previous studies established the presence of various CMD strains in Cameroon, particularly the Ugandan variant which is the most virulent and present in the east region (Gamboula) 35 . Typical CMD symptoms on cassava plants are misshapen leaves which hamper the growth of the plant and reduce root yields, hence lowering productivity and profitability 21,36 . Contrary to CMD, Cassava Anthracnose Disease (CAD) was found mostly in the forest areas where relative humidity and rainfall significantly influence the levels of CAD inoculum pressure, in addition to favoring the development of its vector Pseudotheraptus devastans Dist (Het. Coreidae) 23 . CAD severity was statistically similar on all genotypes, but with low incidence. Genotype I071026 did not show CAD symptoms during the 2nd year. CAD has become an economic threat to www.nature.com/scientificreports/ cassava production as severe outbreaks have been reported 22 . Affected plants display necrotic lesions on leaves and stem, reducing planting material availability 37 . Careful selection of clean planting material and good field maintenance (timely weeding and field aeration) could help in reducing CAD pressure in the forest zone where it is most prevalent. Interestingly, the genotype I090521 did not show any CMD symptoms over the two trial years despite harboring the highest number of whiteflies per plant as previously demonstrated with four genotypes (01/0098, 01/1086-55, 95/0211, and 98/0581) 8 ; however, the lack of CMD symptoms could mean that whiteflies did not carry enough virus loads to efficiently transmit the virus, as an increase in whitefly numbers can lead to the higher efficiency of CMD transmission on cassava 38 and also could likely be explained by the resistance mechanism of this cassava line. This contradicts other findings 39 which reported that 10 whiteflies per plant constituted an adequate population for the spread of CMD due to their persistent mode of transmission.
The observed resistance to major disease was translated into a higher yield for the improved genotypes.These varieties could also be carriers of genes for higher yield. Current cassava production in Cameroon (and in the Central Africa region respectively) stands at 5,798,909 tons (resp. 52,019,756 tons) with an average yield of 14.5 t/ha (resp. 8.9 t/ha) 3 . Therefore, promoting the new genotypes with an average yield of 31 t/ha could potentially increase production in Cameroon to 12,353,035 tons with the current production areas (398,485 ha) which would represent an increase from 11 to 24% of the cassava production in Central Africa. It is worth noting the sharp increase in yield of the check I010040-27 which almost doubled within two years after its first evaluation. Cassava's potential to alleviate food and nutritional security in Central Africa would substantially increase if neighboring countries sharing similar ecologies with Cameroon (Gabon, Equatorial Guinea, Central African Republic, Congo)-where cassava is an essential food security crop-also adopt the new genotypes.
Two genotypes (I090590, and I0701026) produced the highest and most stable fresh root yield across locations, which are qualities that favors them for dissemination throughout the targeted environments to improve cassava yields and hence food security. Overall, tested cassava genotypes performed best at Ekona and Bambui, probably because of the high K and Organic C in their soils 40 . Dry matter content was constant for all genotypes across all environments. As a polygenic trait, dry matter varies from one genotype to another (20-40%) and is usually stable across locations 41 . Genotypes with high dry matter (> 30%) are generally mealy with high starch Table 3. Best linear unbiased predictors (BLUPs) of breeding values with standard errors for fresh root yield (FRY), biomass yield (BY), dry matter (DM), total carotenoid content (TCC) of cassava genotypes across eight locations. *P < 0.05, **P < 0.01. The statistics listed for every trait are broad-sense heritability, genotype variance, residual variance, grand mean, SD = Standard Deviation, the coefficient of variation (CV %)), the number of replications (n Replicates). The statistics shown are the estimates derived.  [42][43][44] , making them suitable for processing into flour and starch which have been shown to improve the potential for cassava adoption and commercialization for income generation and livelihoods improvement 15,45 .
Nutritionally-biofortified genotypes evaluated in this study contain up to 6-folds higher provitamin A β-carotene compared with the local genotype, underlining the potential contribution of the biofortified genotypes to nutritional security in Cameroon and central Africa. These varieties have almost 3 times (up to 11.1 μg/g) the level of carotenoids contents of our previous set of yellow varieties 8 , thus justifying the improvement of their nutritional content and the need to promote them for nutritional security in central Africa. Total carotenoid content is known to be affected by the location in which the genotype is grown, especially in sweet potato (Ipomoea batatas (L.) 46 , but it appears not to be the case for cassava as the levels of provitamin A (β-carotene) of a genotype were stable across all the different environments used in the present study and elsewhere 47 . These results are very encouraging, but it would be necessary to conserve as much of the provitamin A as possible when cassava is transformed into different products-including cooking-to harness the full potential of biofortified cassava's contribution to reductions in vitamin A deficiency 48 . Best linear unbiased prediction for biomass (A) and fresh root yield (B) for 12 genotypes. Black and grey circles represent the genotypes that had BLUP above and below of BLUP mean, respectively. Horizontal error bars represent the 95% confidence interval of a prediction considering a two-tailed t test. (R V3.6.2, https:// cran.r-proje ct. org/ web/ packa ges/ metan/).

Conclusion
This study identified high-yielding cassava genotypes with higher levels of resistance/tolerance to pests and diseases and with elevated provitamin A-improved nutritional content. Disseminating the selected genotypes for cassava production in Cameroon-and by extension elsewhere in Central Africa-would improve cassava production for food and nutritional security. Despite the good performance of most of the tested genotypes, there is also a strong need to conduct consumer preference studies to match the usability of these newly developed varieties to actual use of cassava in the area of introduction to ensure proper adoption by the end-users in the various agro-ecologies.

Materials and methods
Cassava genotypes. IITA had the permission to collect all the plants used in this study. The improved cassava genotypes used were selected in Nigeria by the IITA cassava breeding unit and sent to IITA Cameroon. Accession names and their pedigree are available in the cassava database (Table 4) (www. cassa vabase. org). All improved genotypes were tested along with two cassava genotypes from our recent work 8 which were used here as references (1) for high yield (I010040-27) and (2) for high susceptibility to cassava mosaic virus disease (LMR). Including reference genotypes is a common practice in such studies. The present and previous 8 study were conducted in the same locations (not same fields) in Cameroon and followed a similar methodology to   I070593  IITA-TMS-IBA070593 IITA-TMS-IBA011277/IITA-TMS-IBA990067  Yellow   I010040-27  IITA-TMS-IBA010040 IITA-TMS-IBA010040/?*  Yellow   I011797  IITA-TMS-IBA011797 IITA-TMS-IBA950379/TMEB117  Yellow   I070557  IITA-TMS-IBA070557 IITA-TMS-IBA011663/IITA-TMS-IBA940330  Yellow   I070738  IITA-TMS-IBA070738 IITA-TMS-IBA011649/IITA-TMS-IBA051237  Yellow   I071026  IITA-TMS-IBA071026 IITA-TMS-IBA011277/IITA-TMS-IBA011412 (4X) Yellow   I090521  IITA-TMS-IBA090521 IITA-TMS-IBA974763/MAUNJILI  White   I090537  IITA-TMS-IBA090537 IITA-TMS-IBA961569/IITA-TMS-IBA961569  White   I090574  IITA-TMS-IBA090574 IITA-TMS-IBA961632/CM5306-8  White   I090590  IITA-TMS-IBA090590 IITA-TMS-IBA972205/MCOL 1468  White   I090616  IITA-TMS-IBA090616 IITA-TMS- Soil sampling. At each location, 5 soil samples were collected from each plot and mixed thoroughly in a plastic basin to obtain a composite soil sample. A subsample of 250 g of soil from each plot was used for the determination of various soil physical and chemical properties at the IITA-Cameroon Analytical Laboratory. Soil samples were air-dried and ground to pass through a 2-mm sieve. The sample was further ground to pass through a 0.5 mm sieve for C and N analysis. Soil pH in water was determined in a 1:2.5 (w/v) soil: water suspension. Organic C was determined by chromic acid digestion and spectrophotometric analysis 49 . Total N was determined from a wet acid digest 50 and analyzed by colorimetric analysis. Exchangeable Ca, Mg, K, and P were extracted using the Mehlich-3 procedure 51 , and the cations were determined by flame atomic absorption spectrophotometry. Exchangeable P from the resulting extract above was determined with the molybdate blue procedure 52 . Particle size was determined with a hydrometer.
Pests and diseases assessment. Cassava disease incidence and severity were evaluated at 3, 6, and 9 months after planting on 10 plants per plot, excluding border rows, while alternating plants within a row. On www.nature.com/scientificreports/ each plant, the number of whitefly adults (Bemisia tabaci Gennadius) was counted on the top (the first five apical leaves of the plant tilting the apex) of the plant, while the number of nymphs was counted on the 14th fully developed leaf 53 . The cassava green mite (CGM) was counted on the fifth fully developed leaf using a head lens (Opti-VISOR Optical Glass Binocular Magnifier, 10× ). CMD incidence (presence/absence) was scored on the whole plant foliage, while CAD incidence and severity were scored based on disease symptoms on the stems. Disease and damage scoring was done on a scale of 1 (no symptoms observed) to 5 (severe symptoms on plant parts) 54 . At harvest (12 months after planting), above-ground biomass (stems and leaves) and storage roots were weighed using Macro scales type with optional accessories (Macro-Line 800050, PRESOLA). Samples of 500 g of the fresh root of each genotype were collected and oven-dried at 60 °C for 48 h to measure their dry matter content. Total carotenoid content (TCC) in storage roots was determined within 24 h of harvest using iCheck Carotene following the BioAnalyt protocol 53 . Only yellow flesh genotypes were considered for the carotenoids content analysis as the yellow coloration of the parenchyma is closely related to carotenoid content 55 . The local landrace LMR was included as a check.
Statistical analysis. Incidence data was calculated for CMD and CAD for each plot by calculating the percentage of sampled plants that showed symptoms of the diseases.
We used the metan package of the R 3.6.2 software 56 to perform a stability analysis of multi-environment trial data (MET 57 ) using parametric and non-parametric. MET allows the identification of genotypes that display a small temporal variability-which is desired by breeders and is beneficial to growers, or cultivars that are consistent from location to location-which is desired by and is beneficial to seed companies and breeders 57 . Incidence data were calculated for CMD and CAD by calculating the percentage of sampled plants that showed symptoms of the diseases. Whitefly and CGM densities were log-transformed to reduce heteroscedasticity inherent in insect and mite counts. The best linear unbiased prediction (BLUP) was used to predict Breeding Values (BV) of each genotype for cassava mosaic disease incidence and severity, cassava anthracnose disease incidence and severity, whitefly and cassava green mite densities, fresh root yield (FRY), biomass (BY), dry matter (DM), total carotenoid content (TCC), CMD and CAD severity. The BLUP was performed using the linear mixed model's approach (ANOVA) that considered cassava genotype as fixed factor and locations and year as random factors. The model also included genotype by location (G × L) interactions.
To select genotypes that combine high performance and stability, we introduced the weighted average of absolute scores from the singular value decomposition of the matrix of best linear unbiased predictions for the genotype × environment interaction effects generated by a linear mixed effect model index (WAASB), which is a superiority index that allows weighting between performance (in our study, fresh root yield and CMD incidence) and stability (WAASB index). The first step is rescaling FRY and WAASB; CMD and WAASB that can be directly used to compare genotypes. The best values for FRY and CMD are the highest value and for WAASB is the lowest value 58-60 .

Data availability
The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.