Combining Ability and Genetic Components of Yield Characteristics, Dry Matter Content, and Total Carotenoids in Provitamin A Cassava F1 Cross-Progeny

: Cassava is an important root crop in sub-Saharan Africa, largely cultivated for its starchy edible roots. Biofortiﬁed cassava varieties with enhanced provitamin A carotenoid content (PVAC) developed through conventional breeding provide a solution for vitamin A deﬁciency among vulnerable communities. The aim of this study was to use diallel analysis of six provitamin A cassava genotypes to determine the combining ability, genetic components, heritability, and heterosis of the most important yield characteristics and total carotenoids. Genetic variability for measured characteristics were evident. Fresh root yield was mainly determined by non-additive genetic e ﬀ ects, while dry matter content and total carotenoids were determined by additive e ﬀ ects. Total carotenoids were negatively correlated with fresh root yield, indicating that selection for higher provitamin A content could reduce yield. Mid and higher parent heterosis was seen in some of the crosses for fresh root yield, dry matter content, and total carotenoids. Narrow sense heritability was moderate for fresh root yield and dry matter content, and was high for total carotenoids. This study indicated that yield and dry matter content can be improved in provitamin A cassava but that increased provitamin A content may carry a yield penalty. was higher than the corresponding narrow sense heritability (h 2n ) for all the characteristics. Estimated values of h 2b values ranged from 22.58% for CMD to 60.58% TC whereas the h 2n ranged from 5.18% for FYLD to 37.49% for DMC. Genetic advance as a percentage of the mean was high ( > 20) for all the studied characteristics. This ranged from 24.58 for CMD severity to 53.59 for FYLD.


Introduction
Cassava (Manihot esculenta Crantz) is a shrubby perennial crop with edible storage roots, widely cultivated in the tropics, and ranked as the third largest source of calories after wheat and rice in the world [1]. Cassava grows best on well-drained light to medium soils, with a pH of 4.5-7.5. It is well adapted to various environments with low input amendment requirements and can generally withstand biotic and abiotic stress [2,3]. Storage root yield is the key objective of cassava breeding programs and farmers attach the highest importance to good storage root yield with higher dry matter content (DMC) [4][5][6]. Increasing emphasis is being placed on the nutritional value of cassava. This has led to the breeding of provitamin A cassava, which has more total carotenoids (TCs) than white cassava. Carotenoids are essential due to their conversion to vitamin A. Vitamin A deficiency (VAD) is a major global public health challenge, particularly in tropical sub-Saharan Africa [7]. Hence, provitamin A cassava is vital given the status of cassava as a major staple crop in the tropics.
The DMC trait in cassava is a component of storage root yield, a complex quantitatively inherited trait [8]. Knowledge of the mode of inheritance of quantitative characteristics helps breeders to employ suitable breeding strategies for their improvement. The diallel mating design is a popular design employed by cassava breeders to obtain information on the value of the hybrids and parents. It is used to assess the gene action involved in various characteristics [9][10][11][12][13][14] and in applying appropriate selection procedures. It also helps to understand heterotic patterns of progeny at an early stage of hybridization programs [14]. Genetic variation exists among cassava genotypes for all important characteristics, thus providing an opportunity for exploiting this variation through selection [15]. The predominance of additive gene effects in controlling the expression of DMC has been previously reported [12]. This was confirmed by another study [14] that reported additive gene effects to be double that of non-additive gene action for DMC, while non-additive genetic effects were relatively higher for root yield. The yellow pulp color, indicating beta-carotene in cassava, was found to be controlled by non-additive gene effects [3]. Hybrid vigor for root yield over better-parent values and root yield performance associated with heterosis for yield components has been reported [15], indicating that hybrids can be exploited in heterosis breeding in order to develop superior genotypes. Heritability plays a predictive role in breeding, expressing the reliability of phenotype as a guide to breeding value [16,17]. Genetic advance is an important selection parameter that explains the degree of gain obtained in a characteristic under particular selection pressure. High genetic advance coupled with high heritability estimates offers the opportunity to predict the effect of selecting superior genotypes [18,19]. This aids breeders in a selection program to identify the best responses based on a given selection intensity, for genetic improvement of characteristics [20].
The objective of this study was to do a diallel analysis of six yellow-fleshed parents to determine the combining ability and gene action controlling the expression for TC, storage root yield, and percentage DMC of storage roots, and to estimate the heterosis relative to the mid parent (MPH) and better parent (BPH) for TC, storage root yield, and percentage DMC of storage roots.

Experimental Site
The experiment was conducted at the International Institute of Tropical Agriculture (IITA), at Ibadan and Ikenne research stations in 2012-2015. Ibadan, Oyo State is located at latitude 7 • 24 N, longitude 3 • 55 E, and altitude 273 m above sea level while Ikenne, Ogun State is located on latitude 6 • 53 N, longitude 3 • 42 E, and altitude 60 m above sea level. Both locations are in the southern rainforest vegetation belt of Nigeria and are characterized by nitosol soils with an annual rainfall range of about 1200-1515 mm and temperature range of about 22-32 • C.

Generation of F 1 Progenies
The six parental genotypes were planted in a 6 × 6 half diallel design to generate 15 F 1 full-sib families ( Table 2) through controlled pollination of immature flowers following standard procedures [4]. F 1 botanical seeds were harvested three months after pollination. The seeds were placed in a labeled paper bag and stored for three months to break seed dormancy. Seeds were sown directly into a nursery field at 0.25 m × 1 m intra-and inter-row spacing in a family row plot that was partitioned into a block ridge 20 m long and arranged in a serpentine pattern with no replicates. An alley of 1 m was maintained between families. The resulting F 1 seedlings were harvested at 12 months after planting (MAP). Selection of plants from the nursery field trial for inclusion in the clonal evaluation trial was based on the ability of each plant to produce five to six good plant stakes. A total of 1581 healthy F 1 plants were selected from the nursery trial and advanced to the clonal trial in the second year. For the clonal trial, the F 1 genotypes were cut into stakes of about 20 cm with approximately four to five nodes and planted at a spacing of 0.5 m × 1 m arranged in an augmented design with no replicates. Parents and checks were included in the trials. Morphological evaluation was carried out at different growth stages (1, 3, 6, 9, and 12 MAP). Selection of plants from the clonal trial for inclusion in the preliminary yield trial stage was based on performance of each clone based on a selection index rating that considers four key characteristics (TC, DMC, cassava mosaic disease resistance, and plant architecture). Forty clones per family (for most families) were obtained, but there were cases where some clones failed to develop or died and thus could not be evaluated.

Experimental Design
The preliminary yield trial was arranged in a randomized complete block design with two replicates. Planting was done in September 2014. A total of 379 clones consisting of the 15 cross progenies with the six parents were included in the respective plots of each replicate. Each plot consisted of a single row of 10 plants for each of the clones. The spacing between plants was 0.8 m × 1 m. The field was kept free of weeds for the duration of the growing cycle and neither fertilizer nor irrigation was applied.

Data Collection
Data were collected at harvest (12 MAP). Eight central plants, excluding the first and last plants in a plot, were assessed for their number of storage roots per plant, root weight (kg plant −1 ), and total shoot weight (kg plant −1 ). Storage root yield (t ha −1 ), DMC (%), and harvestable index were estimated [21]. Total carotenoids (µg g −1 ) in fresh weight were estimated following the IITA standard laboratory procedure, using the iCheck Device [22].

Data Analysis
The data were analyzed using SAS version 9.3 [23]. Diallel analysis was done following method 2, model 1 of Griffing [24] to determine the combining ability and separate components of genetic variance. The mixed model procedure based on the restricted maximum likelihood (REML) estimation method was fitted to the data, where the replication effect is considered fixed, while parent and cross effects were considered random, to estimate the GCA and SCA variance components, respectively [25].
The distribution of crosses in relation to GCA and SCA effects was determined by denoting significant positive combining ability effects as high, non-significant as average, and significantly negative as low, for fresh root weight, DMC, and TC [15]. The GCA/SCA ratio of mean squares for the studied characteristics was determined to predict the gene effects for a trait [14,26]. Baker's prediction ratio [27] was used to determine progeny performance. The closer this ratio is to one, the better the chances of predicting progeny performance based on GCA. A value of less than 1 was taken as predominance of non-additive gene action, and higher than 1 as predominance of additive gene action. Heterosis relative to mid-parent and better parent values were estimated as deviation of the F 1 value from the mid-parent (MP) and the better-parent (BP) values, respectively [28].
Phenotypic correlations (r P ) were estimated for each pair of characteristics as the Pearson product moment correlation coefficient using the CORR procedure of SAS version 9.3 [23]. Genetic correlations (r G ) were estimated as previously described [29].
Estimates of variance components for genotypic and phenotypic variances were determined from the estimated values of the mean squares obtained from the ANOVA [19]. Phenotypic coefficient of variation (PCV) and genotypic coefficient of variation (GCV) were also estimated [30]. PCV and GCV values of more than 20% were regarded as high, whereas values less than 10% were considered as low and values between 10% and 20% as average [31]. Broad and narrow sense heritability (h 2 b and h 2 n ) were calculated as the ratio of genotypic variance to the phenotypic variance. The broad sense heritability estimates were classified as low (0-30%), moderate (30-60%), and high (60% and above) [1]. Expected genetic advance with one cycle of selection and expected genetic advance as a percentage of the mean was calculated to compare the extent of predicted genetic advance of different characteristics under selection [32]. Genetic advance as a percent of the population mean was categorized as follows: less than 10% as low, 10-20% as moderate, and above 20% as high.

Analysis of Variance
Highly significant (p < 0.001) genotype effects were observed for all the characteristics (Table 3). Significant SCA mean squares were observed for FYLD and TC. The GCA/SCA ratio was higher than unity for DMC (12.46) and TC (1.17) but less than one for FYLD (0.16). In addition, the Baker's prediction ratio was low for FYLD (0.24) but closer to 1 for DMC (0.96) and TC (0.70), respectively.

Mean Performance of Parents and Their Hybrid Progenies
Parent IITA-TMS-IBA070337 had the highest mean values for root number, root weight, FYLD, DMC, and harvest index while parent IITA-TMS-IBA070593 had the highest mean value for TC (Table 4)

Estimates of the Phenotypic and Genetic Correlations
Genetic correlations (r G ) were higher than phenotypic correlations for all measured characteristics (Table 8). HI showed a positive and high correlation with the number of roots per plant and FYLD. A negative correlation was observed for DMC and TC with FYLD.    3.6. Estimates of Variance Components, Broad and Narrow Sense Heritability, and Genetic Advance FYLD, DMC, TC, and cassava mosaic disease (CMD) exhibited high genotypic (σ 2 G ) and phenotypic (σ 2 P ) variances (Table 9). The genotypic variance was separated into the additive (σ 2 A ) and non-additive/dominance (σ 2 D ) variance components. Additive variance (σ 2 A ) values were higher for DMC, TCC, and CMD than their corresponding non-additive variance (σ 2 D ) except for FYLD, where the non-additive variance was higher than its corresponding additive variance. The difference between the σ 2 G and σ 2 P was high for all the characteristics evaluated except for CMD. High genotypic (GCV) and phenotypic (PCV) coefficient of variances were observed for all characteristics except for DMC, which recorded high PCV but average GCV (Table 9). Phenotypic coefficient of variability (PCV) values ranged from 31.24% for TC to 69.24% for FYLD, while the genotypic coefficient of variability (GCV) ranged from 18.50% for DMC to 42.39% for FYLD. PCV values were generally higher than their corresponding GCV values for all the characteristics considered. Broad sense heritability (h 2 b ) was higher than the corresponding narrow sense heritability (h 2 n ) for all the characteristics. Estimated values of h 2 b values ranged from 22.58% for CMD to 60.58% TC whereas the h 2 n ranged from 5.18% for FYLD to 37.49% for DMC. Genetic advance as a percentage of the mean was high (>20) for all the studied characteristics. This ranged from 24.58 for CMD severity to 53.59 for FYLD.

Discussion
The mean performance of the clones within each family showed significant variations. This confirmed the possibility of selection for these characteristics through hybridization of respective parents. A similar trend was reported previously [13,14]. The significant variation in values observed for the genotypes for the different characteristics (Table 4) indicate the presence of genetic variability among the genotypes. High genetic variability for different characteristics has also been reported in other studies [8,13]. The significant estimates of SCA mean squares for FYLD and TC presented in this study indicates that non-additive gene effects are more important than additive gene effects in the control of these characteristics. The significant SCA effects for FYLD observed in this study agreed with previous findings [9,13] while that of TC was in agreement with a previous study [3]. Higher values than unity for the GCA/SCA ratio for DMC and TC is an indication that the additive gene effects make a larger contribution in the control of DMC and TC than non-additive effects, while for FYLD, the non-additive gene effects predominated. Similarly, a Baker's ratio close to one for DMC and TC is an indication that progeny performance is based on GCA [16]. Nonetheless, the predicted GCA/SCA ratio and Baker's prediction ratio may not always project the true picture of the gene action for a trait due to the differential parental ability to combine well. The performance of the crosses with respect to the parental genotypes suggests that progeny performance cannot be predicted from the performance of the parents per se. Performance also depends upon the complex interaction between the genotype and environment [9,16].
GCA and SCA effects show both the magnitude and direction of genetic effects. The significant positive and negative GCA effects recorded for parent IBA070337 for DMC and TC, respectively, indicate that the genotype contributed highly to DMC but very little to TC. Thus, this genotype was the most superior and best general combiner for DMC but a poor combiner for TC. For the cross progenies, the significant positive SCA effects obtained for IITA-TMS-IBA011368/IITA-TMS-IBA070337 and IITA-TMS-IBA070593/IITA-TMS-IBA011368 for FYLD and TC, respectively, indicate that these cross progenies are the best specific combiners for these characteristics. More than one-third of the cross progenies showed desirable positive (though not significant) SCA effects for DMC. This suggests average SCA for DMC. SCA of contrasting parent genotypes used for genetic improvement of a trait can transmit the desirable specific trait to the next generation [10].
There was good mid and better parent heterosis for the studied characteristics in some of the cross progenies. This may indicate that additive and non-additive genetic effects in the crosses are acting in the same direction. It also suggests that these cross progenies can be exploited in heterosis breeding to develop superior genotypes to take full advantage of these specific characteristics. Heterosis over mid and better parents were previously reported for cassava characteristics, which varied according to the cross combinations and measured characteristics, and it played a very important role in cassava performance [9,12,14].
Phenotypic correlation is the correlation of the genotype means for characteristics and this is partitioned into genetic and environmental sources. However, genetic correlation is heritable and of more practical importance. Higher genetic than the phenotypic correlations observed for all evaluated characteristics is an indication that genetic factors were major contributors to correlations among characteristics, and this is favorable for the selection process. Positive correlation between the harvest index and number of roots per plant (r G = 0.63) and FYLD (r G = 0.80) is an indication that the increase of one trait will simultaneously increase the other. These results agree with previous findings [6], where positive and high genetic correlations between tuberous root yield and shoot biomass yield were reported. The presence of a positive correlation may indicate a balance in the source sink relationship. Negative correlations of TC with DMC and FYLD were observed in this study as was previously reported [33,34], where an inverse relationship of β-carotene content with DMC as well as FYLD was seen.
The higher phenotypic coefficient of variance than the genotypic coefficient of variance observed for all characteristics agrees with a previous study [31]. Considering the benchmark for broad sense heritability (0-30% as low, 30-60% as moderate, and 60% and above as high), h 2 b for CMD severity was low (22.58%), moderate for FYLD (37.52%) and DMC (40.5%), and high for TC (60.58%). The relatively high estimates of broad sense heritability observed for FYLD, DMC, and TC gives an indication that these characteristics are primarily under genetic control. Selection for these characteristics should lead to genetic advance [18]. High heritability of a characteristic makes selection more effective. High heritability (>30) coupled with high genetic advance (>20) as a percent of the mean for FYLD, DMC, and TC gives an indication that selection for genetic improvement of cassava is reliable. High heritability with high genetic advance as a percentage of the mean for plant height and root fresh weight in cassava was previously reported [35].

Conclusions
The role of additive and non-additive gene action in the determination and control of FYLD, DMC and TC presented in this study suggests the possibility of effective selection for genetic improvement of cassava as applicable mainly to the genetic makeup of the parents presented. Families from crosses among elite cassava clones were very promising for genetic improvement for multiple important characteristics, taking full advantage of genetic variance inherent in cassava clones for high yield potential. High heritability (>30) with high genetic advance (>20) is important for selection and the genetic improvement of cassava. It is therefore necessary to conduct further studies with the test genotypes in bigger trials and in several locations and seasons to confirm the results presented in this study.