Diallel analysis of provitamin A carotenoid and dry matter content in cassava (Manihot esculenta Crantz)

Abstract

Global efforts are underway to biofortify cassava (Manihot esculenta Crantz) with provitamin A carotenoids to help combat dietary vitamin A deficiency afflicting the health of more than 500 million resource-poor people in Sub-Saharan Africa. To further the biofortification initiative in Uganda, a 6×6 diallel analysis was conducted to estimate combining ability of six provitamin A clones and gene actions controlling total carotenoid content (TCC), dry matter content (DMC) in cassava roots and other relevant traits. Fifteen F₁ families generated from the diallel crosses were evaluated in two environments using a randomized complete block design. General combining ability (GCA) effects were significant for TCC and DMC, suggesting the relative importance of additive gene effects in controlling these traits in cassava. On the other hand, non-additive effects were predominant for root and shoot weight. MH02-073HS, with the highest level of TCC, was the best general combiner for TCC while NASE 3, a popular white-fleshed variety grown by farmers in Uganda, was the best general combiner for DMC. Such progenitors with superior GCA effects could form the genetic source for future programs targeting cassava breeding for TCC and DMC. A negative correlation was observed between TCC and DMC, which will require breeding strategies to combine both traits for increased adoption of provitamin A cassava varieties.

Introduction

Cassava, the fifth most important staple crop in the world, is a widely grown and consumed root crop in Sub-Saharan Africa (SSA) (Tan 2015). The crop is central to food and income security throughout SSA, especially among resource-poor farmers (Fermont et al. 2010). The popularity of cassava is attributed to its ability to produce reasonable yields under poor soil conditions and extended droughts, where other food crops would practically fail (El-Sharkawy 2007). In Uganda, the area harvested under cassava has steadily increased from 405,000 ha in 2003 to 440,000 ha in 2013 (FAOSTAT 2014), despite the known enormous threat from cassava brown streak disease (CBSD) and cassava mosaic disease (CMD) during this period (Alicai et al. 2007). This trend in cultivation of cassava is a reflection of the growing relevance of the crop in people’s livelihoods.

However, compared to efforts in other important food crops, genetic improvement of cassava has lagged behind, keeping the crop’s production below its potential (Ceballos et al. 2015). For example, average fresh root yield of cassava reported for Uganda is 12 t ha⁻¹, which is lower than that for Ghana (18.3 t ha⁻¹), Thailand (21.8 t ha⁻¹) and the Caribbean (22.1 t ha⁻¹) (FAOSTAT 2014). In Africa, formal cassava breeding began in the 1930s at Amani research station, Tanzania, when scientists Storey and Nichols guided the first systematic efforts to breed for CMD and CBSD resistance (Nichols 1947, Storey and Nichols 1938). Subsequent cassava breeding activities only intensified after the colonial rule when an organized program was instituted at International Institute for Tropical Agriculture (IITA), Ibadan, in the 1970s (Umanah 1977). Although efforts are now reshaping breeding strategies to exploit the yield potential of cassava, they lag considerably behind what has been achieved in, for example, maize since the 1900s after a breakthrough was recorded in use of inbred lines to exploit heterosis for increasing grain yield (Shull 1908).

Most agronomic traits including yield, dry matter content (DMC) and certain types of disease resistance are quantitatively inherited in cassava, and information about their mode of inheritance helps breeders to use methods that increase genetic gain (Calle et al. 2005, Kulembeka et al. 2012). During early stages of cassava breeding, open pollination schemes were the most predominant mating designs used to generate breeding populations upon which selection could be imposed (Kawano 2003). These designs involved establishing selected progenitors of complementary traits per se in crossing nurseries to generate half-sib progenies. Although these mating schemes are simple to implement, they are limited in that they do not provide for estimation of specific combining ability (SCA), which is important in inheritance of key traits like fresh root yield (Ceballos et al. 2004, 2015, Crossa et al. 2010). Subsequent shifts in cassava breeding schemes have seen an increased production of full-sib progeny (Ceballos et al. 2012, Nassar and Ortiz 2006). The full-sib crossing schemes employ controlled pollinations, where selected mating designs are used to generate families from specific parental combinations, facilitating genetic studies alongside production of breeding populations (Nduwumuremyi et al. 2013).

The diallel mating design, specifically, has become popular for cassava breeding because it facilitates generation of useful information on genetics of key agronomic traits and allows identification of parents with superior combining ability for developing breeding populations (Kulembeka et al. 2012, Tumuhimbise et al. 2014, Zacarias and Labuschagne 2010). It is this genetic information that guides breeders to deploy appropriate methods for crop improvement (Acquaah 2012, Nduwumuremyi et al. 2013). Knowledge of general combining ability (GCA) of parental lines is particularly helpful for predicting genetic gains in a breeding program (Falconer and Mackay 1996). Hayman (1954) and Griffing (1956) elaborated on the procedure for statistical analyses based on diallel data, which partitions total variation into GCA of the parents and SCA of crosses.

To leverage vitamin A deficiency (VAD) afflicting more than 500 million people in SSA alone, a novel effort referred to as HarvestPlus Challenge Program was initiated to biofortify staple crops (Mayer et al. 2008). Subsequent efforts to develop cassava varieties enriched with provitamin A carotenoids (pVAC) was initiated, under a joint stewardship of International Center for Tropical Agriculture (CIAT) and IITA. On this premise, the national cassava breeding program of Uganda initiated a breeding program with the objective of developing high yielding varieties with high levels of pVAC (Esuma et al. 2012). Deployment of such carotene-rich cassava would sustainably improve nutrition and reduce prevalence of VAD in communities that primarily depend on cassava (Mayer et al. 2008, Nassar and Ortiz 2010), particularly so in Uganda where the crop is the second most consumed after bananas. Ideally, pVAC cassava varieties are also expected to have high DMC. Varieties that combine these two traits are more likely to be adopted for subsistence agriculture (Njukwe et al. 2013). The strong negative correlation that has been reported between DMC and total carotenoids content (TCC) in the African cassava germplasm (Akinwale et al. 2010, Njoku et al. 2015) could present a potential challenge for cassava breeding programs tasked to improve both traits. Other studies analyzing Latin American cassava germplasm suggested that correlations between carotenoid content and DMC are not high enough to reach statistical significance (Chávez et al. 2005, Sánchez et al. 2014). Against this background, the current study was conducted with specific objectives to (1) estimate the combining ability of six cassava progenitors for TCC and DMC and (2) determine gene actions controlling these two traits in cassava roots.

Materials and Methods

Experimental sites

The crossing block for the study was established in November 2011 at the National Crops Resources Research Institute (NaCRRI) located within the Lake Victoria crescent at 32°37′36.0″E and 0°31′13.7″N, 1.134 m above sea level. The seedling trial was established at Abi Zonal Agricultural Research and Development Institute (Abi-ZARDI) located at 31°1′28.4″E and 2°36′33.3″N, 1.060 m above sea level, in a region with low CBSD pressure. This consideration was important to ensure that the planting materials for clonal trials were free of CBSD to minimize the necrotic root symptoms of the disease (Hillocks and Jennings 2003) for analysis at clonal stage. Clonal trials were conducted at NaCRRI and Abi-ZARDI during 2014–2015. Both locations experience a bimodal rainfall pattern, with two rainy and dry seasons of almost equal length. The first season of rainfall peaks between March and mid-June and second season rains peak from August to November. Temperature and rainfall data were recorded during the experimentation period as well as the soil nutrient profile of the fields before planting the trials, and these varied within the range reported to be ideal for cassava production (Cadavid 2012, Hauser et al. 2014) (Table 1).

Table 1 Description of weather and soil conditions at experimental sites for clonal evaluation

Parameter	Unit of measure	Critical value	Abi-ZARDId	Namulonge
pH		4.0–8.0*	6.1	6.0
OMa	%	3.0	6.6	7.8
N	%	0.2	0.3	0.4
P	ppm	10.0*	3.0	2.4
Ca	ppm	50.0*	1,141.6	1,248.9
Mg	ppm	14.3*	336.2	734.2
K	ppm	58.5*	336.2	443.2
Rainfall	mm		884	1264
T (min)b	°C		15.9	17.4
T (max)c	°C		31.4	29.8

^a  Organic matter;

^b  Minimum temperature;

^c  Maximum temperature;

^d  Abi Zonal Agricultural Research and Development Institute;

Values with asterisks are critical for cassava (Cadavid 2012).

Progenitors and hybridization

Six genetically diverse clones in advanced selection stages were used as progenitors (Table 2). The six progenitors were planted in the crossing block at Namulonge under rain-fed conditions in paired rows to facilitate generation of the 15 F₁ families of a 6×6 diallel design. Planting was done at a spacing of 1.5 m between pairs of progenitors for crossing and 2 m alleys between subsequent progenitor pairs, which collectively provided the additional space to ease movement during the pollination process. Controlled pollinations were performed by hand following the standard procedures described by Kawano (1980), with special care given to cover mature flowers with nylon-meshed pollination bags 2–3 days before and after pollination. Botanical seeds were harvested from mature fruits within 2.5–3 months after pollination and stored in paper bags for two months to break dormancy. At least 100 seeds from each of the 15 F₁ families were germinated in plastic pots filled with natural forest soil, in a screen house. After germination, seedlings were watered routinely whenever deemed necessary to ensure vigorous growth.

Table 2 List of progenitors used in the 6×6 diallel study

Genotype	Code	Source	RFCa	Salient traits
NASE 3	P1	IITAb	White	High DMCc; CMDd and CBSDe tolerance
CPCR24B-10	P2	CIATf	Light yellow	pVACg, CMD resistance
MH05-2870	P3	IITA	Yellow	pVAC, CMD resistance
MH05-0233	P4	IITA	Yellow	pVAC, CMD resistance
CPCR15B-26	P5	CIAT	Yellow	pVAC, CMD resistance
MH02-073HS	P6	IITA	Deep yellow	pVAC, CMD resistance

^a  Root flesh color;

^b  International Institute for Tropical Agriculture;

^c  Dry matter content of roots;

^d  Cassava mosaic disease;

^e  Cassava brown streak disease;

^f  International Center for Tropical Agriculture;

^g  Provitamin A carotenoids.

Field evaluation

Seedlings from the 15 F₁ families were grown in the field during 2013–2014 for the purpose of generating planting materials for replicated clonal trials. At 12 months after planting (MAP), 20 genotypes with the ability to generate ≥12 standard-size cuttings (i.e. 4–6 nodes) were randomly selected from each F₁ family to constitute the population for clonal evaluation. At each site, trials were laid out in a randomized complete block design with two replications. A single row plot of three plants represented each entry, with 2 m alleys between blocks. Planting was done at a spacing of 1 × 1 m, giving a total population of 10,000 plants ha⁻¹. These clonal trials were planted in May 2014 and harvested for root measurements in April 2015. Progenitors were evaluated alongside the progeny in the clonal trials, following similar planting pattern. Weeding was done as necessary and trials rain-fed.

Measurement of root yield, dry matter content and total carotenoid content

During harvesting, all plants in a plot were uprooted and the biomass bulked to estimate yield components by separately weighing the fresh roots (kg plant⁻¹) and foliage (kg plant⁻¹) using a Salter Brecknell suspended weighing scale [model: 23510S(SHFSB-0404)] calibrated in kilograms. Harvest index (HI) was computed from the measure of fresh root weight (FRW) and fresh shoot weight (FSW) using the formula:

  

$$\text{HI}=\frac{\text{FRW}}{(\text{FRW}+\text{FSW})}$$

Three roots (of an average size) were randomly selected from each plot, labelled and processed for measurement of DMC and TCC. These roots were peeled, washed under running water and dried with a paper towel. The dried roots were cut longitudinally into four quarters. The opposite quarters of each of the three roots were pooled, chopped into small pieces and mixed manually to get a homogeneous sample. Approximately 200 g of each homogenous sample were taken for measurement of DMC by drying the sample in an oven to constant weight at a temperature of 105°C for 24 hours. The dried samples were reweighed to obtain their DMC, calculated as:

  

$$\text{DMC}(\%)=\frac{\text{DSW}}{\text{FSW}}\times 100$$

where DSW = dry sample weight and FSW = fresh sample weight.

The iCheck analytical kit developed by BioAnalyt Laboratory (http://www.bioanalyt.com) was used for measuring TCC (Ceballos and Parkes 2014). Briefly, 5 g of the homogenous root sample was pounded and ground into a smooth and fine paste using a mortar and pestle. To aid grinding of the sample, 20 ml of distilled water was added gradually and the resulting solution transferred into a 50 ml calibrated falcon tube. The falcon tube content was shaken thoroughly and 0.4 ml of the solution injected into the iEx™ CAROTENE vial using the syringe and needle provided with the kit. Vials were placed on a solid surface for approximately 5 min, shaken again and allowed to stand until two solution phases appeared inside the vial: a clear upper phase and a turbid lower phase. At this point, the absorbance of the vial content (the upper solution phase) was measured using the iCheck™ CAROTENE device. TCC was calculated as:

  

$$\text{TCC}\left(\mathrm{\mu }\text{g\hspace{0.17em}}{\text{g}}^{-1}\right)=\frac{{\text{V}}_{\text{s}}}{{\text{W}}_{\text{s}}}\times \text{A}$$

where V_s = volume of solution transferred to the falcon tube, W_s = weight of a sample and A = absorbance of the iEx™ CAROTENE vial content at a wavelength of 450 nm. Each sample was extracted and measured for TCC once. All procedures for carotenoid quantification were performed in a dark room. All harvested root samples were analyzed for both TCC and DMC within 12 hours. Root flesh color was also scored for all genotypes using visual inspection following the standard color scale developed by CIAT.

Statistical analysis

Analysis of variance (ANOVA) was done using the Plant Breeding Tools software (PBTools) (PBTools 2014). Diallel analysis was conducted according to the Griffing (1956) method 2, model I for fixed effects, to estimate the GCA and SCA effects. Briefly, method 2 is a diallel, where parents and one set of F₁s, but not reciprocals, are included in the statistical analyses. Model I considers parents as fixed effects and interpretations of the genetic effects are limited to the specific set of parents used. Thus, GCA and SCA effects were estimated as:

  

$$Y_{ijk}=\mathrm{\mu }+g_i+g_j+s_{ij}+e_{ijk}$$

where: Y_ijk = observed value for the ij^th cross grown in the k^th replication/environment combination, μ = overall mean, g_i = GCA effect for the i^th parent, g_j = GCA effect for the j^th parent, s_ij = SCA of the cross between the i^th and j^th parents and e_ijk = error term associated with the ij^th cross evaluated in the k^th replication/environment. The relative importance of GCA and SCA effects for each trait was determined from the proportions of a family’s sum of squares (SS) due to GCA and SCA. Further statistical inference was made using Baker’s ratio (BR) calculated from estimates of the genetic parameters as:

  

$$\text{BR}=\frac{2{\text{MS}}_{\text{GCA}}}{(2{\text{MS}}_{\text{GCA}}+{\text{MS}}_{\text{SCA}})}$$

where MS_GCA and MS_SCA are mean squares due to GCA and SCA effects, respectively (Baker 1978). Pearson’s correlation coefficients for phenotypes from the combined data across environments and replications were computed using the following method: r_p = cov_xy/σ_xσ_y, where r_p is the phenotypic correlation, cov_xy is phenotypic covariance between traits x and y while σ_x and σ_y are the standard deviations of the phenotypes x and y respectively. Genetic correlation coefficients among traits were calculated from the GCA effects as described by Hohls and Clarke (1995).

Results

Mean performance of the 15 F₁ families and their progenitors

For all genotypes evaluated across the 15 F₁ families, TCC values varied from 0 to 11.0 μg g⁻¹, with the highest mean (5.8 μg g⁻¹) recorded for family P5×P6 and the lowest (2.1 μg g⁻¹) for family P1×P2 (Table 3). Individual DMC values for the evaluated genotypes ranged from 15.9 to 45.9%. At family level, DMC ranged from 22.2% for family P5×P6 to 34.4% for P1×P2. Family P5×P6 recorded the highest mean FRW (2.4 kg plant⁻¹), while family P2×P3 recorded the lowest mean FRW (1.2 kg plant⁻¹). The highest mean value of FSW (5.5 kg plant⁻¹) was recorded for family P1×P5 and the lowest value (2.4 kg plant⁻¹) for P2×P4. Meanwhile, family P2×P3 had the lowest mean value (0.28) for HI and family P5×P6 had the highest mean value (0.74) for HI. At progenitor level, genotype MH02-073HS recorded the highest level of both TCC (10.4 μg g⁻¹) and FRW (6.3 kg plant⁻¹), but had the lowest DMC (22.2%). NASE 3 recorded the highest DMC (37.3), but had very low TCC (0.2 μg g⁻¹). Fig. 1 illustrates phenotypic distribution of DMC and TCC for all genotypes evaluated, with a negative correlation observed between the two traits.

Table 3 Performance of progenitors and their respective F₁ progeny across two locations in Uganda during 2014–2015

Parent/Family	Numbera	TCCb	DMCc	FRWd	FSWe	HIf
NASE 3 (P1)	–	0.2	37.3	2.5	2.6	0.51
CPCR24B-10 (P2)	–	4.4	32.3	2.0	2.9	0.44
MH05-2870 (P3)	–	4.3	33.4	1.5	5.7	0.32
MH05-0233 (P4)	–	4.9	29.0	2.4	1.7	0.58
CPCR15B-26 (P5)	–	5.3	30.7	2.3	2.1	0.49
MH02-073HS (P6)	–	10.4	22.2	6.3	2.5	0.72
P1×P2	17	2.1	34.4	1.5	4.1	0.33
P1×P3	18	3.1	31.0	1.8	4.0	0.42
P1×P4	19	3.1	31.3	1.9	3.7	0.45
P1×P5	20	2.4	33.9	1.9	5.5	0.34
P1×P6	16	2.6	28.1	1.9	5.3	0.56
P2×P3	15	2.7	32.4	1.2	3.8	0.28
P2×P4	15	3.1	32.1	1.5	2.4	0.37
P2×P5	20	3.2	32.1	1.8	3.4	0.43
P2×P6	17	3.7	29.6	1.5	4.1	0.54
P3×P4	20	2.3	33.0	2.1	3.7	0.43
P3×P5	16	3.2	32.9	2.0	2.9	0.38
P3×P6	20	3.7	29.7	2.1	4.1	0.57
P4×P5	20	4.3	33.6	2.0	2.8	0.52
P4×P6	20	4.4	29.0	2.0	3.8	0.41
P5×P6	19	5.8	22.2	2.4	3.4	0.74
Minimumg		0.0	15.9	0.0	0.1	0.00
Maximumg		11.0	45.9	13.5	52.0	0.85
Mean		3.8	30.9	2.1	3.5	0.44
SEh		0.062	0.150	0.150	0.322	0.025
LSD0.05i		0.127	0.304	0.303	0.650	0.050
Number of F1j		261	255	272	272	272

^a  Number of genotypes evaluated per F₁ family;

^b  Total carotenoid content (μg g⁻¹);

^c  Dry matter content of roots (%);

^d  Fresh root weight (kg plant⁻¹);

^e  Fresh shoot weight (kg plant⁻¹);

^f  Harvest index;

^g  Values based on all F₁ genotypes evaluated;

^h  Standard error;

ⁱ  Least significant difference at 5% confidence level;

^j  Number of F₁ genotypes evaluated: reduction from the total population (272) indicates proportion of genotypes whose roots were not sufficient for measuring TCC and/ or DMC.

Fig. 1

Scatter plot and histograms of total carotenoids content (TCC) vs. dry matter content (DMC). Color represents pigmentation of the root parenchyma such that 1-WT = white; 2-LC = light cream; 3-CM = cream, 4-LY = light yellow; 5-YL = yellow; 6-DY = deep yellow; red lines on histograms indicate means.

Analysis of variance

Based on mean squares (MS), differences in performance of families were very significant (P < 0.01) for all the studied traits (Table 4). Mean squares of the crosses by environment interaction were significant for DMC, FRW and FSW, but such interaction was non-significant for TCC and HI. Meanwhile, GCA effects were highly significant (P < 0.001) for TCC, very significant (P < 0.01) for DMC, significant (P < 0.05) for HI, but non-significant for FRW and FSW. SCA effects were highly significant for FRW and FSW, very significant for DMC and HI and non-significant for TCC. The coefficient of variation (CV) associated with traits ranged from 7.5% for TCC to 21.2% for HI; these are within the acceptable range, and thus give confidence in the generated datasets. TCC had the highest value for Baker’s ratio (0.89) followed by DMC (0.69), while FRW had the lowest value (0.36). A Baker’s ratio above 0.5 implies that additive genetic effects are proportionately more important than dominance effects (Baker 1978). When total SS were partitioned into SS due to parents (GCA effects) and interaction between parents (SCA effects), the GCA effects accounted for more than 64% of the total variation expressed by the families for TCC and DMC, while FRW and FSW contributed less than 22% of variability in GCA effects. The SCA effects were important for FRW and FSW.

Table 4 Mean squares of crosses and combining ability effects of five traits evaluated in 15 F₁ families and six parents at two locations

Source of variation	DFa	TCCb	DMCc	FRWd	FSWe	HIf
Environment (E)	1	0.03	9.31**	4.63***	31.29***	0.13**
Crosses (C)	14	4.23***	16.87**	1.75**	2.73**	0.07**
C × E	14	0.05	3.09***	0.26*	1.12*	0.01
GCAg	5	42.26***	16.66**	2.86	3.85	8.49*
SCAh	14	10.81	15.24**	10.30***	12.66***	16.21*
GCA × E	5	0.06	15.71***	0.73*	9.08**	0.05
SCA × E	14	1.01	6.47*	8.81*	15.95**	0.48*
% SSi due to GCA		71.31	64.34	21.29	12.84	51.27
% SS due to SCA		24.48	31.28	56.84	48.11	28.51
Residual	34	0.02	0.36	0.36	1.66	0.01
CVj (%)		7.50	16.31	11.21	19.42	21.20
Baker’s ratio		0.89	0.69	0.37	0.38	0.56

^a  Degrees of freedom;

^b  Total carotenoid content (μg g⁻¹);

^c  Dry matter content (%);

^d  Fresh root weight (kg plant⁻¹);

^e  Fresh shoot weight (kg plant⁻¹);

^f  Harvest index;

^g  General combining ability;

^h  Specific combining ability;

ⁱ  Sum of squares;

^j  Coefficient of variation.

*, ** and ***  significant at P < 0.05, P < 0.01 and P < 0.001, respectively.

Combining ability and estimates of genetic parameters

Traits evaluated in this study were measured with preference for high scores (i.e. higher positive values for combining ability estimates were preferred for each of the evaluated traits). Genotype MH02-073HS, with the highest TCC, showed the highest GCA effect of 1.93 for the trait (Table 5). On the other hand NASE 3, a white-fleshed progenitor with negligible TCC, had significant negative GCA effect of −1.59 for TCC. Negative GCA’s for TCC indicate unsuitability of specific progenitors as combiners when targeting high carotenoid content in the progeny. However, NASE 3 was the best general combiner for DMC with positive and significant GCA of 1.75, while progenitor MH02-073HS had negative and significant GCA of −3.72 for DMC. Only progenitor MH02-073HS had positive and significant GCA for FRW. Meanwhile, MH02-073HS was the best general combiner for HI, with a GCA effect of 0.16.

Table 5 GCA effects of cassava progenitors used in a 6×6 diallel analysis of five traits

Progenitor	TCCa	DMCb	FRWc	FSWd	HIe
NASE 3	−1.59***	1.75***	−0.11	0.38*	−0.12
CPCR24B-10	−0.38*	0.87**	−0.42**	−0.14*	−0.04
MH05-2870	−0.27	0.82**	−0.33*	0.63**	−0.05
MH05-0233	0.03	−0.17*	−0.06	−0.63**	0.03
CPCR15B-26	0.28*	0.44**	−0.03	−0.34**	0.02
MH02-073HS	1.93***	−3.72***	0.95**	0.12	0.16*
LSD0.05f	0.253	0.607	0.605	1.300	0.020
SEg	0.041	0.097	0.097	0.208	0.014

^a  Total carotenoid content (μg g⁻¹);

^b  Dry matter content (%);

^c  Fresh root weight (kg plant⁻¹);

^d  Fresh shoot weight (kg plant⁻¹);

^e  Harvest index;

^f  Least significant difference at 5% confidence level;

^g  Standard error.

*, ** and ***  significant at P < 0.05, P < 0.01 and P < 0.001, respectively.

Three families (P1×P5, P3×P4 and P4×P6) showed positive significant SCA effects for TCC (Table 6). The other 12 families had either negative or non-significant positive SCA effects. Overall, there was high variation in SCA effects for DMC, with five families showing positive significant SCA effects. For this trait, family P1×P2 had the most positive and significant SCA effect of 1.60. For FRW, there were four F₁ families with positive significant SCA effects, with P1×P5 as the best specific combination. Meanwhile, none of the 15 F₁ families showed positive significant SCA effects for FSW. On the other hand, three families (P2×P5, P3×P5 and P4×P6) had positive and significant SCA effect for HI.

Table 6 SCA effects for a 6×6 diallel analysis of five traits evaluated at two locations in Uganda

Family	TCCa	DMCb	FRWc	FSWd	HIe
P1×P2	0.77	1.60*	−0.43	−0.49*	−0.04
P1×P3	−0.62	−1.17*	−0.87*	0.03	0.06
P1×P4	−0.98	−1.09	−0.23	0.12	−0.09
P1×P5	1.65*	1.22*	1.90**	0.16	−0.10
P1×P6	−0.82	−0.56	−0.37	0.17	0.16
P2×P3	−0.04	−0.43	−0.33	0.02	−0.25*
P2×P4	0.46	−0.04	0.52	0.06	−0.08
P2×P5	−0.18	−1.34*	0.83*	0.20	0.32*
P2×P6	−1.01*	0.21	−0.58	0.22	0.05
P3×P4	1.00*	0.54	0.71*	−0.04	0.12
P3×P5	−0.19	0.16	0.00	0.03	0.24*
P3×P6	−0.15	0.90	0.49	−0.04	−0.16
P4×P5	−1.87*	0.55	−2.09**	−0.08	−0.18
P4×P6	1.39*	0.04	1.09*	−0.05	0.22*
P5×P6	0.59	−0.59	−0.64	−0.31*	−0.28*
SEf	0.77	1.60*	−0.43	−0.49*	−0.04

^a  Total carotenoid content (μg g⁻¹);

^b  Dry matter content (%);

^c  Fresh root weight (kg plant⁻¹);

^d  Fresh shoot weight (kg plant⁻¹);

^e  Harvest index;

^f  Standard error. P1, P2, P3, P4, P5 and P6 are progenitors defined in Table 2.

*, ** and ***  significant at P < 0.05, P < 0.01 and P < 0.001, respectively.

Phenotypic correlation among traits

The most positive and highly significant phenotypic correlation was between root flesh color (RFC) and TCC, with a coefficient of 0.94 (Table 7). This relationship indicated that higher intensities of root pigmentation reflected higher levels of TCC in roots. There was negative and significant correlation (r = −0.44) between the RFC and DMC. Similarly, a negative correlation was noted between TCC and DMC (r = −0.45). These relationships indicate that cassava roots with higher levels of TCC have low DMC. HI correlated positively and significantly with FRW but had significantly negative correlation with FSW, which is expected, given that HI is derived from the two traits. The most positive and highly significant genetic correlation was between RFC and TCC (r = 0.87). DMC had significant negative genetic correlation with TCC (r = −0.82). These genetic correlations (RFC with TCC and DMC with TCC) were consistent with patterns of phenotypic correlations between the same trait pairs.

Table 7 Phenotypic (lower diagonal) and genetic (upper diagonal) correlation coefficients for six traits in 6×6 diallel families evaluated at two locations in Uganda

Trait	RFCa	FRWb	DMCc	FSWd	TCCe	HIf
RFC		0.04	−0.62**	0.31	0.87***	0.02
FRW	0.14		−0.23	−0.03	0.83**	0.81**
DMC	−0.44**	0.07		0.08	−0.82**	−0.08
FSW	−0.02	0.07	−0.01		−0.05	−0.13
TCC	0.94***	0.132	−0.45**	−0.02		0.86**
HI	0.09	0.57**	0.09	−0.47**	0.08

^a  Root flesh color;

^b  Fresh root weight;

^c  Dry matter content;

^d  Fresh shoot weight;

^e  Total carotenoid content;

ⁱ  Harvest index.

*, ** and ***  significant at P < 0.05, P < 0.01 and P < 0.001, respectively.

Discussion

TCC in the F₁ progeny evaluated for this study varied from 0.0–11.0 μg g⁻¹, with a mean of 3.8 μg g⁻¹. Values for TCC in the progenitors varied from 0.2–10.4 μg g⁻¹. The observed TCC mean for the F₁ progeny was comparable to 3.6 μg g⁻¹ and 5.0 μg g⁻¹ reported, respectively, by Maroya et al. (2012) and Ssemakula and Dixon (2007) for breeding populations evaluated at IITA, but quite lower than the mean (14.7 μg g⁻¹) reported for populations at CIAT (Ceballos et al. 2013). The high level of TCC in cassava breeding populations at CIAT is a result of 10-year cyclic selection process imposed to primarily advance carotenoid-rich clones, targeted to attain levels above 15 μg g⁻¹. The population evaluated in the current study arose from a single round of recombination and fell short of this nutrient level targeted by HarvestPlus program (Pfeiffer and McClafferty 2007). It is worth noting that this study was strategically designed to generate pVAC clones expressing both CMD and CBSD resistance by including the white-fleshed NASE 3 as a progenitor. Nonetheless, the wide level of segregation for TCC observed in the current breeding population could provide a basis for implementing a recurrent selection scheme for developing cassava varieties with increased provitamin A carotenoid content in future. The rapid cycling scheme used by CIAT for increasing carotenoids content in the roots (Ceballos et al. 2013) could be adopted for this endeavour. With this scheme maximum carotenoids levels was increased from 10 to 26 μg g⁻¹ within a period of nine years.

DMC varied from 15.9–45.9%, with a mean of 30.9%. The mean DMC in the pVAC populations developed in this study is somewhat below the DMC levels (mean = 35%) of most improved white-fleshed varieties currently grown by farmers in Uganda (Kawuki et al. 2011). Importantly, some of the carotenoids-rich genotypes evaluated here had higher DMC than that of varieties commonly grown by farmers, for example, 81 genotypes had DMC values ≥35%. These genotypes are of interest to breeding, as they qualify to be used as progenitors for the next round of recombination. High DMC in cassava roots is a trait highly preferred by farmers in SSA (Njukwe et al. 2013, Tumuhimbise et al. 2012). To enhance adoption of pVAC varieties in future, breeders will need to focus deliberate efforts on increasing DMC in the carotene-rich breeding populations. Fresh root yield is another critical trait that influences adoption of new varieties by farmers; however, in the current study, this trait was measured in form of FRW and HI. FRW potential was assessed on plant basis and it varied from 0 to 13.5 kg plant⁻¹. Genotypes with the ability to yield ≥3 kg plant⁻¹ would be preferred for cultivation, as current varieties popularly grown by farmers in Uganda have average yield of 25 t ha⁻¹ (2.5 kg plant⁻¹) (Kawuki et al. 2011). However, such genotypes identified from the current study will require further verification using larger plot sizes.

Two phenotypic correlations in this study are of special importance for developing pVAC cassava varieties. Firstly, the strong positive correlation between root flesh color and TCC is a good incentive for screening large early-generation breeding populations by visual assessment. At such stage of breeding, genotypes with higher intensity of root pigmentation can be selected for advancement, which saves time and the high costs associated with quantification of carotenoids. Iglesias et al. (1997) and Chávez et al. (2000) have previously demonstrated the effectiveness of using root color to select for high carotenoid content in cassava, especially if increasing carotenoids content is the sole breeding objective. However, Sánchez et al. (2014) suggest that the intensity of pigmentation, though useful for identifying high pVAC clones, may result in lower DMC. Perhaps a simplistic explanation is that high DMC has lots of white starch which tends to dilute the intensity of pigmentation. Thus, it would be important to make a mild initial selection based on color intensity and then a stronger one based on quantified carotenoids levels.

Continuous selection for both DMC and carotenoid content in Latin America has been underway much longer than in Africa and weak initial negative correlations could have been broken during the several cycles of recombination. For example, Ceballos et al. (2013) reported an interesting result of increased pVAC levels in breeding populations upon which several cycles recurrent selection were imposed for high levels of carotenoids content, yet DMC increased along with this selection strategy. In that study, the authors suggested that continuous recombination in a recurrent selection scheme could generate clones that combine high levels of both DMC and TCC, with values of up to 37% and 25 μg g⁻¹, respectively. Compared to other crops, Ortiz et al. (2014) reported a strong negative correlation between carotene content and fruit DMC in butternut squash (Cucurbita moschata D.), which is similar to reports by Vimala et al. (2011) for studies on sweetpotato (Ipomea batatas L.). As further studies are undertaken, more useful genetic information relevant for cassava breeding will be generated and hopefully, this discrepancy clarified. For now, it is important to continue to pursue this approach of combining both traits.

Interpretation of the genetic nature of traits evaluated in this study was based on the mean squares for GCA and SCA effects. The proportion of SS for crosses explained by GCA components gives an estimate of the relative importance of additive effects in expression of traits (Falconer and Mackay 1996). GCA accounted for a significantly larger SS than SCA for TCC, DMC and HI, explaining more than 64% of the total variation for DMC and TCC. These results suggest that additive gene effects are more important in controlling accumulation of TCC and DMC in cassava. Similarly, 51.3% of the variation in HI was attributed to GCA. The relative importance of additive genetic factors for TCC, DMC and HI was reflected by the higher Bakers’s ratios for these traits. Baker’s ratio above 0.5, as was the case with these three traits, means these traits are under the control of additive genetic factors. The implication of these findings for cassava breeding is that a recurrent mass selection method can be an efficient breeding method for improving TCC and DMC, as it would enhance the exploitation of additive genetic effects for the traits (Ceballos et al. 2013).

FRW had non-significant GCA effects, suggesting that this trait is largely under control of non-additive genetic factors. This deduction is consistent with previous reports by Calle et al. (2005), Zacarias and Labuschagne (2010) and Kulembeka et al. (2012). In this case, breeders targeting the increase of fresh root yield in cassava would consider crossing progenitor combinations with superior SCA for FRW, a strategy that would increase the chances of selecting high-yielding clones from the segregating progeny (Chipeta et al. 2013). However, the relatively high importance of GCA effects for HI suggests a possibility of increasing FRW by crossing progenitors with high HI. Although Ojulong et al. (2010) demonstrated the effectiveness of using progenitors with high HI for improving FRW in cassava, the practice does not necessarily result in gains for root yield. In reality, it is common to find a genotype with low plant vigor (low foliage weight) showing a high HI when its actual root yield is very low (Hay 1995). Perhaps it would be a better practice to use HI as a complementary trait to FRW when selecting for high yield in cassava, which agrees well with suggestions by Hay (1995) that selections based on HI alone can be less effective.

Generally, progenitors with higher levels of TCC showed positive GCA, suggesting their contribution towards enhancing TCC in the progeny. Progenitor MH02-073HS had the highest positive and significant GCA effect for TCC, FRW and HI. Therefore, an appropriate breeding design would be to cross MH02-073HS to a genetic background of high DMC to increase chances of generating clones that accumulate favorable alleles for expressing higher levels of both traits (Ceballos et al. 2013). Similar relative importance of GCA in inheritance of carotenoid content has been reported in Cucumis sativus L. (Navazio and Simon 2001) and Zea mays L. (Senete et al. 2011), suggesting the trait is controlled by genes that act additively. The highest negative GCA effects for TCC shown by NASE 3 is unsurprising, given it is a white-fleshed genotype with negligible levels of TCC. Conversely, the same genotype had the highest positive GCA effect for DMC, which also correlates well with its high level of DMC. NASE 3, also referred to as TMS 30572, is an IITA bred variety that is characterized with high levels of DMC (mean of >35%). This clone was officially released in Uganda in the 1990s and has remained popular among farmers, largely because of the high DMC and tolerance to both CMD and CBSD.

The non-significant SCA effects for TCC indicate that SCA is less important than GCA for inheritance of TCC in cassava, which is similar to deductions made by Senete et al. (2011) for inheritance of carotenoids in maize. Meanwhile, five F₁ families with positive and significant SCA effects for DMC could present possibility for generating clones with high DMC when such genotypic combinations are used for hybridization. Overall, SCA effects were more important than GCA effects for inheritance of fresh root yields, which agrees with previous reports by Kulembeka et al. (2012) and Tumuhimbise et al. (2014).

In summary, this study reports on a pioneer effort to breed cassava for high pVAC content, targeting to benefit the entire eastern Africa. Both breeding products in the form of genotypes and genetic information have been generated, all of which are useful for follow-up studies and future breeding programs. The generated datasets suggested that GCA effects are more important than SCA effects in the genetic control of carotenoid content in cassava, indicating that additive genetic effects largely control inheritance of carotenoids in cassava. The additive nature of carotenoids content provides scope for its improvement under the recurrent selection scheme. The negative correlation between root carotenoid content and DMC is important situation to deal with as a matter of priority. In particular, future breeding efforts that can uncover the genetic basis of the negative correlation between DMC and TCC are warranted. A breakthrough in this case could offer more realistic prospects for developing pVAC varieties acceptable to farmers. Such a breakthrough can be attained by at least two different approaches (that can hopefully be pursued together): 1) introducing germplasm from Latin America for combining high TCC and DMC and 2) strengthening the selection of segregating progenies based on the quantification of carotenoids as opposed to visual pigment selection.

Acknowledgements

The research was supported by funds from the East African Agricultural Productivity Project (EAAPP), a grant from the World Bank to the Government of Uganda.

Literature Cited

•  Acquaah,  G. (2012) Principles of Plant Genetics and Breeding. Second edn. John Willey and Sons Ltd, UK.
•  Akinwale,  M.G.,  R.D.  Aladesanwa,  B.O.  Akinyele,  A.G.O.  Dixon and  A.C.  Odiyi (2010) Inheritance of β-carotene in cassava (Manihot esculenta Crantz). Int. J. Genet. Mol. Biol. 2: 198–201.
•  Alicai,  T.,  C.A.  Omongo,  M.N.  Maruthi,  R.J.  Hillocks,  Y.  Baguma,  R.  Kawuki,  A.  Bua,  G.W.  Otim-Nape and  J.  Colvin (2007) Re-emergence of cassava brown streak disease in Uganda. Plant Dis. 91: 24–29.
•  Baker,  R.J. (1978) Issues in diallel analysis. Crop Sci. 18: 533–536.
•  Cadavid,  L.F. (2012) Soils and fertilizers for the cassava crop. In:  Ospina,  B. and  H.  Ceballos (eds.) Cassava in the Third Millennium: Modern Production, Processing, Use, and Marketing Systems. CIAT, Cali, Colombia, pp. 113–137.
•  Calle,  F.,  J.C.  Pérez,  W.  Gaitán,  N.  Morante,  H.  Ceballos,  G.  Llano and  E.  Alvarez (2005) Diallel inheritance of relevant traits in cassava (Manihot esculenta Crantz) adapted to acid-soil savannas. Euphytica 144: 177–186.
•  Ceballos,  H.,  C.A.  Iglesias,  J.  Perez and  A.G.O.  Dixon (2004) Cassava breeding: opportunities and challenges. Plant Mol. Biol. 56: 503– 516.
•  Ceballos,  H.,  C.H.  Hershey and  L.A.  Becerra-López-Lavalle (2012) New approaches to cassava breeding. Plant Breed. Rev. 36: 427–504.
•  Ceballos,  H.,  N.  Morante,  T.  Sánchez,  D.  Ortiz,  I.  Aragón,  A.L.  Chávez,  M.  Pizarro,  F.  Calle and  D.  Dufour (2013) Rapid cycling recurrent selection for increased carotenoids content in cassava roots. Crop Sci. 53: 2342–2351.
•  Ceballos,  H. and  E.Y.  Parkes (2014) Measuring provitamin A content in crops. The 2^nd Global Conference on Biofortification: Getting Nutritious Foods to People. Kigali, Rwanda.
•  Ceballos,  H.,  R.S.  Kawuki,  V.E.  Gracen,  G.C.  Yencho and  C.H.  Hershey (2015) Conventional breeding, marker-assisted selection, genomic selection and inbreeding in clonally propagated crops: a case study for cassava. Theor. Appl. Genet. 128: 1647–1667.
•  Chávez,  A.L.,  J.M.  Bedoya,  T.  Sánchez,  C.A.  Iglesias,  H.  Ceballos and  W.  Roca (2000) Iron, carotene, and ascorbic acid in cassava roots and leaves. Food Nutr. Bull. 21: 410–413.
•  Chávez,  A.L.,  T.  Sánchez,  G.  Jaramillo,  J.M.I.  Bedoya,  J.  Echeverry and  E.A.  Bolaños (2005) Variation of quality traits in cassava roots evaluated in landraces and improved clones. Euphytica 143: 125–133.
•  Chipeta,  M.M.,  J.M.  Bokosi,  V.W.  Saka and  I.R.M.  Benesi (2013) Combining ability and mode of gene action in cassava for resistance to cassava green mite and cassava mealy bug in Malawi. J. Plant Breed. Crop Sci. 5: 195–202.
•  Crossa,  J.,  L.  Campos,  P.  Pérez,  D.  Gianola,  J.  Burgueno,  J.L.  Araus,  D.  Makumbi,  R.P.  Singh,  S.  Dreisigacker,  J.  Yan et al. (2010) Prediction of genetic values of quantitative traits in plant breeding using pedigree and molecular markers. Genetics 186: 713–724.
•  El-Sharkawy,  M.A. (2007) Physiological characteristics of cassava tolerance to prolonged drought in the tropics: implications for breeding cultivars adapted to seasonally dry and semiarid environments. Braz. J. Plant Physiol. 19: 257–286.
•  Esuma,  W.,  P.  Rubaihayo,  A.  Pariyo,  R.  Kawuki,  B.  Wanjala,  I.  Nzuki,  J.J.W.  Harvey and  Y.  Baguma (2012) Genetic diversity of provitamin A cassava in Uganda. J. Plant Stud. 1: 60–71.
•  Falconer,  D.S. and  T.F.C.  Mackay (1996) Introduction to Quantitative Genetics. Fourth edn. Longman, Essex, UK.
• FAOSTAT (2014) Food and Agriculture Organization of the United Nations Statistics. http://data.fao.org/ref/262b79ca-279c-4517-93de-ee3b7c7cb553, 4^th September 2015.
•  Fermont,  A.M.,  A.  Babirye,  H.M.  Obiero,  S.  Abele and  K.E.  Giller (2010) False beliefs on the socio-economic drivers of cassava cropping. Agron. Sustain. Dev. 30: 433–444.
•  Griffing,  B. (1956) Concept of general and specific combining ability in relation to diallel crossing systems. Aust. J. Exp. Biol. Med. Sci. 9: 463–493.
•  Hauser,  S.,  L.  Wairegi,  C.L.A.  Asadu,  D.O.  Asawalam,  G.  Jokthan and  U.  Ugbe (2014) Cassava System Cropping Guide: Africa Soil Health Consortium CAB International, Nairobi.
•  Hay,  R.K.M. (1995) Harvest index: a review of its use in plant breeding and crop physiology. Ann. Appl. Biol. 126: 197–216.
•  Hayman,  B.I. (1954) The theory and analysis of diallel crosses. Genetics 39: 789–809.
•  Hillocks,  R.J. and  D.L.  Jennings (2003) Cassava brown streak disease: a review of present knowledge and research needs. Int. J. Pest Manag. 49: 225–234.
•  Hohls,  T. and  G.P.Y.  Clarke (1995) Additive and non-additive genetic correlations. I. properties of sampling distributions. Biomed. J. 37: 919–935.
•  Iglesias,  C.,  J.  Mayer,  L.  Chavez and  F.  Calle (1997) Genetic potential and stability of carotene content in cassava roots. Euphytica 94: 367–373.
•  Kawano,  K. (1980) Cassava. In:  Fehr,  W.R. and  H.H.  Hadley (eds.) Hybridization of Crop Plants. America Society of Agronomy and Crop Science Society of America, WI, USA, pp. 225–233.
•  Kawano,  K. (2003) Thirty years of cassava breeding for productivity—biological and social factors for success. Crop Sci. 43: 1325–1335.
•  Kawuki,  R.S.,  A.  Pariyo,  T.  Amuge,  E.  Nuwamanya,  G.  Ssemakula,  S.  Tumwesigye,  A.  Bua,  Y.  Baguma,  C.  Omongo,  T.  Alicai et al. (2011) A breeding scheme for local adoption of cassava (Manihot esculenta Crantz). J. Plant Breed. Crop Sci. 3: 120–130.
•  Kulembeka,  H.P.,  M.  Ferguson,  L.  Herselman,  E.  Kanju,  G.  Mkamilo,  E.  Masumba,  M.  Fregene and  M.T.  Labuschagne (2012) Diallel analysis of field resistance to brown streak disease in cassava (Manihot esculenta Crantz) landraces from Tanzania. Euphytica 187: 277–288.
•  Maroya,  N.G.,  P.  Kulakow,  A.G.  Dixon,  B.  Maziya-Dixon and  M.A.  Bakare (2012) Genotype × environment interaction of carotene content of yellow-fleshed cassava genotypes in Nigeria. J. Life Sci. 6: 595–601.
•  Mayer,  J.E.,  W.H.  Pfeiffer and  P.  Beyer (2008) Biofortified crops to alleviate micronutrient malnutrition. Curr. Opin. Plant Biol. 11: 166–170.
•  Nassar,  N.M.A. and  R.  Ortiz (2006) Cassava improvement: challenges and impacts. J. Agric. Sci. 145: 163–171.
•  Nassar,  N.M.A. and  R.  Ortiz (2010) Breeding cassava to feed the poor. Sci. Am. 302: 78–82.
•  Navazio,  J.P. and  P.W.  Simon (2001) Diallel analysis of high carotenoid content in cucumbers. J. Am. Soc. Hortic. Sci. 126: 100–104.
•  Nduwumuremyi,  A.,  P.  Tongoona and  S.  Habimana (2013) Mating designs: helpful tool for quantitative plant breeding analysis. J. Plant Breed. Genet. 1: 117–129.
•  Nichols,  R.F.W. (1947) Breeding cassava for virus resistance. East Afr. Agr. J. 12: 184–194.
•  Njoku,  D.N.,  V.E.  Gracen,  S.K.  Offei,  I.K.  Asante,  C.N.  Egesi,  P.  Kulakow and  H.  Ceballos (2015) Parent-offspring regression analysis for total carotenoids and some agronomic traits in cassava. Euphytica 206: 657–666.
•  Njukwe,  E.,  R.  Hanna,  H.  Kirscht and  S.  Araki (2013) Farmers perception and criteria for cassava variety preference in Cameroon. Afr. Study Monogr. 34: 221–234.
•  Ojulong,  H.F.,  M.T.  Labuschagne,  L.  Herselman and  M.  Fregene (2010) Yield traits as selection indices in seedling populations of cassava. Crop Breed. Appl. Biotechnol. 10: 191–196.
•  Ortiz,  S.,  M.P.  Valdés,  F.A.  Vallejo and  D.  Baena (2014) Genetic correlations and path analysis in butternut squash Cucurbita moschata Duch. Rev. Fac. Nal. Agr. Medellin 68: 7399–7409.
• PBTools (2014) Biometrics and Breeding Informatics, Version 1.4. PBGB Division. International Rice Research Institute, Los Banos, Laguna.
•  Pfeiffer,  W.H. and  B.  McClafferty (2007) HarvestPlus: breeding crops for better nutrition. Crop Sci. 47: 88–105.
•  Sánchez,  T.,  H.  Ceballos,  D.  Dufour,  D.  Ortiz,  N.  Morante,  F.  Calle,  T.  Zum Felde,  M.  Dominguez and  F.  Davrieux (2014) Prediction of carotenoids, cyanide and dry matter contents in fresh cassava root using NIRS and hunter color techniques. Food Chem. 151: 444–451.
•  Senete,  C.T.,  P.E.O.  Guimarães,  M.C.D.  Paes and  J.C.  Souza (2011) Diallel analysis of maize inbred lines for carotenoids and grain yield. Euphytica 182: 395–404.
•  Shull,  G.H. (1908) The composition of a field of maize. J. Hered. 4: 296–301.
•  Ssemakula,  G. and  A.G.  Dixon (2007) Genotype × environment interaction, stability and agronomic performance of carotenoid-rich cassava clones. Sci. Res. Essays 2: 390–399.
•  Storey,  H.H. and  R.F.W.  Nichols (1938) Studies of the mosaic disease of cassava. Ann. Appl. Biol. 25: 790–806.
•  Tan,  S.L. (2015) Cassava—silently, the tuber fills: the lowly cassava, regarded as a poor man’s crop, may help save the world from the curse of plastic pollution. Utar Agric. Sci. J. 1: 12–24.
•  Tumuhimbise,  R.,  R.  Melis,  P.  Shanahan and  R.  Kawuki (2012) Farmers’ perceptions on early storage root bulking in cassava (Manihot esculenta Crantz) in east and central Uganda and their implication for cassava breeding. World J. Agric. Sci. 8: 403–408.
•  Tumuhimbise,  R.,  R.  Melis and  P.  Shanahan (2014) Diallel analysis of early storage root yield and disease resistance traits in cassava (Manihot esculenta Crantz). Field Crops Res. 167: 86–93.
•  Umanah,  E.E. (1977) Cassava research in Nigeria before 1972. In:  Cock,  J.,  R.  Machntyre and  M.  Graham (eds.) Proceedings of the Fourth Symposium of the International Society for Tropical Root Crops. IDRC, Ottawa, pp. 137–141.
•  Vimala,  B.,  A.  Sreekanth,  H.  Binu and  G.  Wolfgang (2011) Variability in 42 orange-fleshed sweet potato hybrids for tuber yield and carotene and dry matter content. Gene Conserve 40: 190–200.
•  Zacarias,  A.M. and  M.T.  Labuschagne (2010) Diallel analysis of cassava brown streak disease, yield and yield related characteristics in Mozambique. Euphytica 176: 309–320.