Diallel analysis of cowpea populations for resistance to Cowpea aphid-born

Cowpea aphid-born mosaic virus disease (CABMV) is one of the reasons for rejection of cowpea seed by seed inspectors in Burkina Faso. With regard to this, this study was undertaken to analyze the genetic components underlying the resistance of cowpea lines to the cowpea aphid-borne mosaic virus (CABMV) and to determine the mechanism of transmission of the resistance from parents to offspring. Therefore, crosses were made in 5x5 full diallel design. Data analysis was done following Griffing and Hayman method on disease severity and the area under disease progress curve (AUDPC) for five cowpea varieties during the 2015 off-season at Kamboinse research station. The analysis of variance associated with the general and specific combining abilities (GCA and SCA) and reciprocal effect (RCE) showed that the genetic variability was explained by additive effect. The F1 population showed that there was partial dominance and the narrow sense heritability for severity and AUDPC was high (60%). To improve cowpea for resistance to CABMV, rigorous choice of parents should be made before crosses and there was no maternal effect. 
 
 Key words: Cowpea, full diallel, severity, resistance, Cowpea aphid-born mosaic virus disease (CABMV), Burkina Faso.


INTRODUCTION
Cowpea (Vigna unguiculata, L. Walp) is a leguminous crop, self-pollinated, grown in all agro-ecological zones of Burkina Faso and has numerous advantages at both agronomical and economical levels.Its grains constitute an important source of protein and income for producers and consumers.Cowpea is also an important fodder.
However, one of the main problems in the genetic improvement of the crop to address is the choice of the parents for hybridization.This choice of parents for hybridization depends, beyond beyond resistance to diverse constraints, heavily on market and consumers' criteria.Tignegre (2010) and Batieno (2014) have reported that the market criteria were mainly based on seed size (large) and color (white).Also, the effectiveness of a method of selection depends largely on the number of genes involved in the control of the trait (Zagre et al., 1999).
Within the main constraints for cowpea production, the cowpea aphid-borne mosaic virus (CABMV) is one of the principal reasons for rejection of cowpea seeds by the seed inspectors and also by producers in Burkina Faso.Cultural practices have been used to control the disease but are weak in seed production system.Therefore, there is a need to develop resistant varieties in order to reduce losses due to CABMV.
Thus, the objective of this study was to analyze the genetic nature of resistance of cowpea lines to CABMV in order to formulate hypotheses on the possible ways of using them to improve cowpea for resistance to the disease.For this, a full diallel analysis was used following Hayman (1954) and Griffing (1956) approaches.This method has been already used in cowpea to study the genetics underlying Striga resistance (Tignegre, 2010).The Griffing's method is based on the determination of the general and the specific combining abilities.The general combining ability for (GCA) is the average of gametic effects of an individual.It provides information on combining abilities at global and individual level (Griffing, 1956).In other words, it is a measure of the value of the average gametes of a parent (Demarly 1977).It is the ability of both parents to transmit positive or negative characters to their descendants (Allard, 1999).Specific combining ability (SCA) is a deviation from the additivity of general combining.Contrary to GCA, SCA is not linked to a parent, but a cross.Statistically, while GCA appears as a primary effect, SCA is an interaction (Demarly, 1977).GCA varies depending on the additive gene action.It is therefore passed from one generation to another.SCA measures the deviation from the performance of F 1 as compared to the average of the parents.
The method of Hayman (1954) is used to estimate different genetic components for the trait and the various parameters: the additive, dominance, reciprocal effects, heterosis and heritability.It comprises four types of analysis that complement the level of interpretation: the analysis of variance of diallel tables testing the significance of the various terms that are not unlike the specific combining ability, the validity test for the model, the statistical analysis of the genetic components of the total variation and the analysis of relationships between statistical terms.

Genetic resources
Genetic resources used in this study comprised five released cowpea varieties from Burkina Faso and 20 F1 hybrids from 5x5 full diallel crosses.Lines used in these crosses were chosen based on their reaction vis-à-vis to CABMV.The five lines involved in the crosses are: KVx396-4-5-2D (resistant), KVx640 (resistant), KVx61-1 (moderately susceptible), KVx30-309-6G (susceptible) and Gorom local (susceptible) all from the long-term storage germplasm of the cowpea breeding program at Kamboinsé Research Station in Burkina Faso.

Methods
Twenty (20) F1 hybrids and their parents were planted in pots and arranged in randomized complete blocks design (RCBD) with three replications.Each replication comprised 25 entries of one pot per entry containing individual plant.Plants were sprayed to avoid contamination from aphids.The experiment was conducted under screen house at Kamboinsé Research Station (latitude 12°28N, longitude 1°32W and altitude 296m) in Burkina Faso in July 2015.To protect plants, insecticide spray was done using a mixture of PACHA (lambda-cyhalothrin 15 g/l + acetameprid 10 g/l) and TITAN (25 EC Acétamiprid 25 g/l) two weeks after planting at doses of 2 ml per liter of water per product.
Each plant received 45 kg of P2O5 per hectare from NPK fertilizer (14-23-14-6S-1B formula).One week after planting, all plants were inoculated using extract of leaves from CABMV serotype D grinded based on weight/volume proportion (p/v) =1/10.The inoculum used was from infected seedlings of Gorom local, a CABMV serotype D susceptible cowpea variety in Burkina Faso.Prior to infestation, the inoculum was homogenized in sodium phosphate buffer (0.01 M, pH 7.4).The extract was filtered through gauze and placed in melting ice.Before inoculation, the leaves of cowpea plants older than a week from the three replications were dusted with the mixture of carborundum 600 mesh, an abrasive product and inoculum using a cotton swab pestle dipped in the extract, the upper leaf surface was rubbed gently (Neya, 2011).The symptoms of CABMV were recorded between the 6 th and 21 st day after inoculation.

Data collection
Observations were made on: 1.The severity assessment using rating scale 6 classes (0 to 5) which is a strength criterion in CABMV.2. AUDPC: The area under disease progression curve proposed by Shaner and Finnay (1977) using the following equation AUDPC = ∑ (Xi+1 + Xi) / 2][ti+1ti] where n: total number of cases; Xi: the first observation of disease in days; Xi + 1: the second observation of disease in days; ti: time in days from the first observation of disease and ti + 1: time in days for the second observation of the disease.It is a study of a disease development rate of a given crop.This parameter selects the best lines in terms of their ability to slow down the progression of the disease.Hayman (1954) and Griffing (1956) methods were used for analysis of variance (ANOVA) from DIAL Win 98 software revised 22 September 2002.

Data analyses
The method of Griffing (1956) is based on two models: the fixed pattern and random model.The fixed model is applied to a limited number of lines set for self-pollinated crops and inbred lines of cross-pollinated species.
As for the random model, information may extend to the entire population, provided individuals are the representation of a random mating population in equilibrium.There are four methods for each model according to the use of the parents and crossing type.a. Reciprocal crosses and parents.In this experiment, the fixed model and method a were used.The statistical model is: where: μ = population mean; λ (λj) = general combining ability (GCA) of the parent i (j); Sij = specific combining ability of crossing by i j; eij = effect of the environment on the individual ij.Hayman (1954) used the following symbols for a given character to express the statistics in his model where, VP: variance of a parent; Vr: a variance r parent and his descendants; Wr: r covariance between a parent and his descendants; W'r: covariance between the value of each descendant of r parent and other descendants of that parent; Yr: r value of a parent.
The interpretation by the model of Hayman requires a certain number of conditions: homozygous parents, identical reciprocal crosses, no multi-allelism, diploid parents, absence of epistasis, no maternal effect, independent distribution of the relevant genes of the parents.
The authors can estimate the various genetic components of the change and test their significance from their own variance and the following statistical terms: E: component due to the environment; D: component due to additive effects; H1: component due to nonadditive effects; H2: component due to unweighted additive effects in terms of a possible asymmetry in the distribution of allele's dominance representative loci; F: covariance between the additive effects and non-additives.Knowledge of these components allows the following calculations: D-H1, in which sign expresses the kind of dominance.
The conformity of the model with these restrictions can be rarely achieved in practice.Most of them however, can be checked during the statistical analysis, when the results are consistent with the additive-dominance model Mather and Jinks (1982), although only the interpretation of parental values and F1 hybrids cannot fully control the factors of non-compliance with the model.Furthermore, the influence of reciprocal effect is erased by working out the average mutual boxes.

Analysis of variance for GCA and SCA and reciprocal using Griffing's method for severity
The results of the variance related to the general combining ability effects (GCA), the specific combining ability (SCA) and the reciprocal effects (RCE) are shown in Table 1.
The analysis of variance was highly significant for the SCA and non-significant for GCA and RCE.SCA effects occur very significantly in expression of severity.The calculated mean value of the GCA/SCA variance ratio is low (1.29).

Analysis of variance for severity by Hayman model
The results of different terms of Hayman variance analysis is presented in Table 2.With regards to the degree of significance of the dominance effects (SCA), the results obtained are consistent with those found using Griffing's method.The results shown in Table 2 are presented based on the different terms described by Hayman.These terms are: 1.The term b 1 is the mean deviation of the first generation F 1 hybrids relative to the average parent which is highly significant for the severity.This result shows that the dominant genes are exerted in a unidirectional manner.2. The term b 2 which is the average deviation of the F 1 as compared to the average values of each parent is not significant for the severity.This result indicates that there is no asymmetry in the distribution of alleles at loci showing dominance.3. The term b 3 deviation due to the dominance of own F 1 represents the specific combining ability.This term is highly significant for the severity.4. The term that tests the differences between reciprocal crosses is not significant for the severity.

Analysis of variance and GCA, SCA and RCE effects by Griffing's method of AUDPC
The results of the variance related to the effect of the general combining ability (GCA), specific combining ability (SCA) and the reciprocity effects (RCE) are shown in Table 3.
The analysis of variance is significant for SCA and not significant for the GCA and RCE.The calculated mean value of the variance ratio GCA / SCA is low (1.24).

Analysis of variance for AUDPC in F 1 generation by Hayman's model
The results of the different terms are presented in Table 4.The results obtained by the method of Hayman concerning the degree of significance of the dominance effects (SCA) and additive (GCA) are not consistent with those found by Griffing.These results provide the following clarifications: 1.The term b 1 which is the mean deviation of F 1 as compared to the average parent, is highly significant for AUDPC.This result shows that the dominant genes are exerted in a unidirectional manner.
2. The term b 2 which is the average deviation of the F 1 as compared to the average values of each parent is also highly significant for AUDPC.
3. The term b 3 deviation due to the dominance of own F 1 represents the specific combining ability.This term is significant for AUDPC.4. The term that tests the differences between reciprocal crosses is not significant for AUDPC.

Validity of the assumptions corresponding to the additive-dominance model
The results of the homogeneity of the expression Wr-Vr test are presented in Table 5.The test is not significant for the severity and for the AUDPC, so the model is respected and thus allows further analysis.

Analysis of genetic components
The estimates of the different genetic components of the characters studied for the F 1 are presented in Table 6.These values were used to calculate the narrow sense heritability by Mather and Jinks (1982).The term D-H 1 reflects the type of dominance.When this expression is negative, there is super dominance.In that case, the variance of additive effects (D) is smaller than the variance of non-additive effects (H 1 ).When it is positive, there's partial dominance and this is the case for the severity and AUDPC with respective value of 1.56 and 85.33.When D is equal to H 1 , there is a total dominance.The expression H 1 -H 2 = 0.089 for severity is low as compared to the H 1 and H 2 estimates of dominance effects.Although, the asymmetry in the distribution of genes is significant (b 2 refers to the analysis of variance), this effect does not play a major role in non-additive effects.The same result was obtained with the area under the disease progression curve (AUDPC); H 1 -H 2 : 7.63, which is low as compared to the H 1 and H 2 estimates of dominance effects.
Table 7 shows the average values of heritability in the narrow sense obtained by Griffing and Hayman.There is a high heritability strict sense according to Griffing (68.64%) and Hayman (63.35%) for the severity parameter.By cons, it is very high according to Hayman (85.21%) and high according to Griffing (66.99%) for the AUDPC.

Graphical analysis for severity and AUDPC
The graphical representation of Wr (co-variance between a parent r and its progeny) by the Vr (variance of a parent r and its progeny) are given in Figures 1 and 2 for the severity and the AUDPC respectively.Three curves are shown on the graph: 1.A regression line; 2. A dish that cuts the regression line in two points, M and M* 3. A tangent to the parabola is almost confused with the regression line

DISCUSSION
Non-significant GCA was observed for both parameters (severity and AUDPC).This implies that non-additive gene action is operating for these parameters.This result differed from what was observed by Orawu (2007).This author found significant GCA effects in CABMV, suggesting that additive gene action is involved in the resistance of cowpea to the disease.Nevertheless, the ratio of Griffing (1956) between GCA/SCA showed that additive genes were also operating for the resistance of cowpea to CABMV disease.For this author, when the ratio is greater than 1 (one), additive effects are more important than non-additive effects.This is also in agreement with the findings of Singh and Chaudhary (1977).Additive gene action seems to be important in cowpea.Tignegre (2010) also found additive gene action for more than seven parameters under a Striga infestation study.
SCA effects were highly significant for the two parameters studied (severity and AUDPC).This implies that non-additive gene effects involving either dominance or epistasis and in some instances both, were observed for these parameters.However, where non-additive gene effects including epistasis were operative, prediction of the breeding outcome would be difficult as non-additive gene effects are not heritable for pure line cultivars (Tignegre, 2010).Dominance effects (that is, partial dominance, complete dominance or over dominance)  cannot be transferred to the progenies and might slow down the progress in selection.However, such gene action would have been useful in hybrid production.Nonetheless, the self-pollinating nature of cultivated cowpea renders difficult the production of hybrid cowpea.However, with some perennial cowpea wild relatives, the occurrence of high rates of cross pollinations (unpublished data) are new fields for hybrid production in cowpea.
There were no maternal and reciprocal effects, suggesting that there were no genetic implications in using a parent as male or female when crossing cowpea for these characters.Therefore, seeds of F 1 and reciprocal crosses can be bulked and used in studying these parameters.These results are in agreement with those of Tignegre (2010).This also implies that no genes originating from the cytoplasm are involved in the inheritance of the characters studied.
Narrow sense heritability measures the breeding value that is passed on to the progenies.Regardless of the method used, high narrow sense heritability was observed in this study.By Griffing's method, the narrow sense heritability was 68.64% for severity and 66.99% for AUDPC.By Hayman's method, the narrow sense heritability was 63.35% for severity and 85.21% for AUDPC.These rates measure the breeding progress that can be expected during selection using the type of protocol employed here.
For all parameters, based on the graphical analysis, with a regression of unit slope b Wr>0.50, a regression coefficient of approximately 50.00% or more indicated that the additive model was adequate to describe the data (Jinks and Hayman, 1953;Christie and Shattuck, 1992;Dalbholkar, 1992;Sharma, 1995).Considering Figures 1 and 2, two extremes to be taking into account are, M and M* corresponding to the intercepts between the regression line and the parabola.Theoretically, M and M* correspond to the genotypes of the parents that have respectively the parent with dominant genes and parent with recessive genes.All individuals close to M have dominant genes, those close to M* have the recessive genes and intermediate genotypes to the two points have a mixture of dominant and recessive genes.Thus, in both figures, parents 5 and 4 have dominant genes; parents 2 and 3 have both dominant and recessive genes, and parent 1 has the recessive genes for severity and AUDPC parameters.Parents 5 and 4 correspond to resistant genotypes and parent 1 is the susceptible genotype.Parents 2 and 3 are intermediate varieties.The parent 5 is very close to M and parent 1 close to M*.This means that opportunities for transgression are relatively low.The slope of the severity on the regression line is equal to 0.88 and that of the AUDPC is 1.04.These values are not significantly different from 1, showing that there is non-allelic relationship and particularly complementary gene actions between parental combinations.Only additive gene action and partially dominant action exists in the parental combinations.These results are similar to those found in 2012 by Zagre on soybeans.

Conclusion
From this study, it was inferred that from the pot screening, regardless of the method used, non-additive genes were predominant in the inheritance of CABMV resistance with regard to the parameters severity and AUDPC.Only non-allelic interactions (epistasis and failure of some assumption) were present with both parameters (severity and AUDPC).
Narrow sense heritability according to the methods of Griffing and Hayman for severity and area under the disease progress curve is high.This suggests that these resistance parameters are strongly passing from parents to offspring.Hayman's method is more restrictive, the heritability was retain from this model.High values of heritability indicate that additive is the major gene action phenomenon in this study.

INTRODUCTION
Rice (Oryza sativa L.) is one of the most important stable foods for more than half of world's population.It provides up to 50% of the dietary caloric supply and a substantial part of the protein intake in Asia (Muthayya et al., 2014).In Sub-Saharan Africa rice consumption among urban dwellers has steadily been grown.From 2002 to 2007, rice production in Africa had increased by an average of 3.2% per year, and from 2007 to 2012 by 8.4% per year (CGIAR, 2013).In Uganda rice production from year 2010 to 2014 increased from 93 to 95 thousand hectares, with a yield increment of 214 to 237 thousand tonnes (FAO, 2014).But, the production and productivity of the crop is hampered by a number of biotic and abiotic factors.
Rice blast, caused by Magnaporthe grisea, is one of the most devastating diseases, especially in susceptible *Corresponding author.E-mail: zelalemsafe@gmail.com.
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License susceptible varieties, causing yield losses of 50 to 90% (Hai et al., 2007;Hajano et al., 2011;Chuwa et al., 2015).
It is becoming severe under high temperature, high relative humidity (85 to 89%), presence of dew, drought stress and excessive nitrogen fertilization.This disease is a major problem in most of the rice-growing regions of the world (Onasanya et al., 2008).Since the variability of the pathogen from year to year and place to place makes its management difficult, it becomes important to give great attention to resistance breeding (Sharma et al., 2012;Kihoro et al., 2013).It is a serious concern in temperate areas as well as in tropical uplands.Even though the disease affects all the plant parts above ground, seedlings and young or tender tissues are more vulnerable than those of older ones.At optimum temperatures, new blast lesions appear within 4 and 5 days after they fall on the leaf surface.In warm and wet weather conditions, new conidia are produced within hours after the appearance of the lesions, and this continues for several days (Greer and Webster, 2001).Yield reductions due to blast are drastic when panicle itself and the panicle base are infected shortly after heading (Shim et al., 2005).Geneticaly diversified genotypes play a vital role in any breeding program for resistance to both biotic and abiotic stresses.The use of resistant varieties can not only ensure protection against diseases, but also save the time, energy and money spent on other measures of control (Sharma et al., 2012).The genetics of hostpathogen interactions are of considerable biological interest and great importance in developing diseasecontrol strategies in efforts of resistance breeding (Ribot et al., 2008).Therefore, the present study was conducted to identify rice blast resistant genotypes from a set of introduced Korean rice accessions in Uganda conditions.

Description of study area and genotypes used
In this study, the first forty-six rice genotypes introduced from South Korea though the Korea-Africa Food and Agriculture Cooperation Initiative (KAFACI) were screened with one resistant (IR-64) and one susceptible (NERICA-1) checks at the National Crops Resources Research Institute (NaCRRI) in Kampala, Uganda during the two rainy seasons of 2015 (Table 1).NaCRRI is located at 0° 31' N, 32° 35' E, with a mean altitude of 1150 m above the sea level.The soils are ferralitic (red sandy and clay loams) and have a pH range of 4.9 to 5.0.The average annual rainfall is 1300 mm and maximum and minimum temperature of 28.5 and 13.0°C, respectively.

Screening under field conditions
A nursery was raised for each genotype and the seedlings were transplanted to the main field.Twenty-one days old seedlings of 48 genotypes were transplanted in the swamp field in a 6 by 8 alpha lattice design with two replications.The spacing of 20 cm between rows and between plants and 40 cm between plots and between blocks with 1 m between replications were used.Four susceptible varieties (NERICA-1, Basimati-370, Sindano and K-85) used as spreader rows were planted between plots two weeks before raising the nursery.This helped to enhance natural infection and to minimize the chance of escape from infection (IRRI, 2014;Vasudevan et al., 2014).In order to promote development of the disease, high humidity was promoted by irrigation twice a day on rain-free days, so that soil of the field experiment was always wet.Other agronomic practices were done as recommended (Asea et al., 2010).

Screening under controlled conditions
Field screened 48 rice genotypes were further evaluated in the screen house using a single isolates of the pathogen to confirm their resistance.Seeds of test lines and the two checks (IR-64 and NERICA-1) were planted in 25 and 30 cm diameter buckets filled with forest soil (using 4 seeds/pot) in 6 × 8 alpha lattice design in three replications.

Inoculum preparation and inoculation
Blast-infected plants were collected from rice fields at NaCRRI.The infected rice plants were selected by observing the symptoms on the leaves based on the rice blast identification guide (Phadikar et al., 2012).The infected parts were cut into small pieces (0.5-1.0 cm) and then surface sterilized with 2% sodium hypochlorite for three minutes.These pieces were then washed with distilled water and placed on plates of 19.5 g L -1 Potato Dextrose Agar (PDA).The PDA plates were then incubated at 25°C for 5 days until sporulation (Hajano et al., 2011).Thereafter, single spores from sporulating lesions were transferred on 4% water agar with the use of an inoculating needle under stereomicroscope for further multiplication for 24 h and the emerging fungus was purified by isolating a single hyphal tip using a sterile needle under a stereo microscope.The resulting pure cultures were incubated at room temperature (25°C) under darkness.After four weeks, the aerial mycelia were slightly washed off by gentle rubbing with a water soaked tooth brush and spore suspension concentration of 1×10 6 spores/ml was prepared using a Neubauer haemacytometer under a compound microscope (Khan et al., 2001).Before inoculation, 0.05% Tween 20 was added to the suspension to increase the adhesion of the spores to the plants.The plants were inoculated with a hand sprayer until run off at the 3 to 4 leaf stage of the plant.High humidity was maintained by covering the area with a white plastic sheet to facilitate infestation.In addition to this, water was sprinkled on the leaves at mid-day for one week, in order to facilitate blast development (Koutroubas et al., 2009).

Data collection
Data on leaf blast severity, lesion size, AUDPC for leaf blast severity and lesion size, panicle blast and yield were collected on five randomly selected plants in the field and on three plants in the screenhouse from each plot according to the standard evaluation system of rice (IRRI, 2014).In addition to these frequency distributions for leaf and panicle blast severity were calculated.Disease evaluations for leaf blast was done four times for each test line at an interval of one week after inoculation in the screenhouse and when the first symptom was observed on the susceptible lines in the field.According to IRRI (2014) standard evaluation system, severity score 0 = no lesions observed, 1 = small brown specks of pin-point size without sporulating center, 3 = small roundish to slightly elongated, necrotic grey spots, 1-2 mm in diameter, 5 = typical susceptible blast lesions 3mm or longer, infecting less than 10% of leaf area, 7 = typical susceptible blast lesions infecting 11-50% of the leaf area and 9 = more than 75% leaf area affected.

Sum of all numerical rating
Total number of rating x maximum disease rating Genotypes were classified according to Shrestha and Misra (1994), for their reaction to leaf blast as 0-15% resistant, 15.1-30% = moderately resistant, 30.1-50% = moderately susceptible and 50.1-100% = susceptible.
To compare relative levels of resistance in the genotypes, weekly assessments of disease severity was done four times.Area under the disease progress curves (AUPDC) was calculated as described by Madden et al. (2008) as; AUDPC = in which xi = blast severity at the i th observation, ti = the time in days after appearance of the disease at the i th day, and n = total number of observations.

Data analysis
The data were subjected to alpha lattice restricted maximum likelihood (ReML) analysis in GenStat 12 th edition software package.The genotypes were considered fixed while blocks, replications and season were random effects.However, the randomized complete block analysis was used when the block mean square is greater than the residual mean square.Variance components due to genotypes σ G and genotype by season 2 interactions σ GXS 2 and heritability were determined.The linear model for the across season analysis was as follows: Where, ijkl = observed value from each experimental unit, u =grand mean, i = effect of the i th season, g j = effect of k th genotype, / ik = effect of the k th replication nested within the i th season, / /b ikl = effect of r th replication and b th block nested within the i th season, g ik = interaction effect of k th genotype and the i th season and eijkl = the experimental error.

Screening result of genotypes under field conditions
Across season analysis of variance of traits showed significant differences (P≤0.05)among genotypes for final leaf blast severities, lesion size and their respective AUDPC values, panicle blast and yield (Table 2).
The across season analysis result (Table 4) showed that the lowest final leaf blast severity scores (14.3-14.4%)were obtained for three genotypes SRHB-78, SRHB-12 and SRHB-133.Moderately low final leaf blast severities (17.8 -28.9%) were recorded for ten genotypes which were grouped as moderately resistant.Twenty-four genotypes that had high final leaf blast severities (32.2 -48.9%) were classified as moderately susceptible.The remaining ten genotypes showed susceptibility levels equal to the susceptible check (Figure 1), NERICA-1 (66.7%) which was followed by .
The genotypes evaluated also showed variation in the AUDPC for leaf blast severity, with seven of them having lower values (120.6 to 182.8%) than the resistant check (IR-64) at 200.3%.Final lesion size ranged from 4.0 mm 2 for genotype SRHB-170 to 63.4 mm 2 for the susceptible check with overall mean of 19.9 mm 2 .Low AUDPC values for lesion size were obtained for four genotypes, with mean values ranging from 26.1 to 36.4 mm 2 compared to a value of 43.6 mm 2 recorded on the resistant check (IR-64).The highest lesion size AUDPC was recorded on susceptible check (413.3 mm 2 ) followed by genotype SRHB-56 (323.8.9 mm 2 ) (Table 3).

Screening result of genotypes under controlled conditions
The analysis of variance of traits under controlled condition showed significant differences (P≤0.05)among  genotypes for final leaf blast severities, lesion size and their respective AUDPC values, while it showed nonsignificant for panicle blast and grain yield (Table 4).
The frequency distribution of genotypes for reaction to leaf and panicle blast in the screen house is presented in Figure 2. In this figure five genotypes were resistant, two moderately resistant, twenty-nine moderately susceptible and twelve susceptible.Ten genotypes were resistant to panicle blast, seventeen moderately resistant and 21 were susceptible.

DISCUSSION
Identifying sources of resistance to rice blast has been a major objective for many researchers involved in rice breeding programs (Rama Devi et al., 2015;Biotica et al., 2014;Vasudevan et al., 2014).In this study, 46 introduced genotypes from KAFACI with two checks were evaluated in order to identify resistant sources.The analysis of results revealed that genotypes were significantly different for final leaf blast severity, lesion size, AUDPC values panicle blast severity and yield in both field and screen house conditions.This indicated that genetic variability exists among the screened genotypes, an advantage for improved breeding for blast resistance in rice.Of the genotypes used in this study, none was immune to leaf and panicle blast either in the field or screen house but there were resistant genotypes in these screening conditions.
In the first season's screening for final leaf blast severity under field conditions, four genotypes (SRHB-     2015) for screening rice genotypes against resistance to rice blast.Their results revealed that while none of the varieties were immune to blast, genotypes were grouped as resistant, moderately resistant and susceptible.These variations may be attributed variously to genetic difference for resistance to blast, or to variation in environment from season to season and screening conditions.These findings indicate that screening under both field and screen house conditions and in several seasons could be effective for getting genotypes with resistant genes for rice blast disease.The significant effect of season that produced variation in values for leaf blast, lesion size and their AUDPC values could be due to variable weather conditions.Environmental factors, relative humidity, temperature and amount of rainfall could strongly affect the sporulation, release and germination of blast conidia (Park et al., 2009;Yang et al., 2011).
Variation for panicle blast severity, shown in the analysis of the overall field screening indicates the presence of genetic variation among genotypes.None of the genotypes showed immunity to panicle blast severity, though 36 genotypes were resistant and 12 were found susceptible.However, in the screen house condition 10 genotypes showed resistance, 17 were moderately resistant and the remaining was susceptible.A similar result was reported by Pasha et al. (2013b), Chuwa et al. (2015), Lee et al. (2015).Nagaraju et al. (2008) also reported in screening 265 genotypes, none of them was immune for leaf and panicle blast, eight genotypes were resistant and 138 moderately resistant to leaf blast and 18 genotypes were resistant, and 82 moderately resistant to panicle blast.

Conclusion
In general, this study showed the value of testing the reaction of the introduced Korean rice genotypes to the Ugandan situation, even when they were introduced by the source as being resistant.In this study the acrossseason field screening results showed that three genotypes were resistant, eleven moderately resistant, 24 moderately susceptible and ten susceptible to rice leaf blast.In the screen house five genotypes were shown to be resistant, two moderately resistant, 29 moderately susceptible and 12 susceptible, again indicating genetic variation among genotypes.
Results from the two screening environments showed that genotypes SRHB-133, SRHB-93 and SRHB-78 were more consistent for resistance to rice blast and good performance for yield.So, these genotypes be either used by farmers after intensive evaluation for production or used to introgress the resistant genes into the locally-adapted elite materials of Uganda.Therefore, genotypes that consistently showed resistance to rice blast disease under both screening conditions can be used as a source for resistance in the rice blast breeding program.From this study, it is possible to conclude that screening in both the field across seasons and in the screen house helps the breeder to identify the genotypes that are truly resistant for further utilization as resistant sources.Additionally, large populations could be screen in the screen house at reduced cost.

INTRODUCTION
The cacao industry is driven by the major international chocolate manufacturing in Europe and USA.However, all the raw materials are produced in the tropical south and Central America, Africa and the Caribbean (Motamayor et al., 2002).Commercial cacao (Theobroma cacao L.; formerly Sterculiaceae family; reclassified Malvaceae family] (Alverson et al., 1999) is a tropical tree [3 to 5 m] which is derived from varieties belonging to three major groups viz: Criollo, Forastero and Trinitario (Lachenaud et al., 1997).
Author(s) agree that this article remain permanently open access under the terms of the Creative Commons Attribution License 4.0 International License The crop growth is highly influenced by environmental conditions viz.temperature (Daymond and Hadley, 2004), flooding (Sena and Kozlowski, 1986), and water stress (Almeida and Valle, 2007).The bi-modal seasons influence the phenological stages of flowering, fruiting and pod growth (Cazorla et al., 1989).The plant produces caulescent flowers with the non-pollinated flowers abscising 24 to 36 h after anthesis (Garcia, 1973).The cacao flower is hermaphrodite and is pollinated by insects, mainly Forcipomyia sp.(Diptera: Ceratopogonidae (Dias et al., 1997)).The flowers setting to pods are very low [0.5 to 5%] (Aneja et al., 1999).
The quality of pollination can depend on two factors, the degree of pollen compatibility and the number of pollen grains deposited on the stigma (Lanaud et al., 1987).It is assumed that with increased pollen grains pod set is improved (Hasenstein and Zavada, 2001) and more pollinations result from the visit of a single pollinator (Yamada and Guries, 1998).The increase in Forcipomyia larvae and pupae associated with rotten banana stems had shown to produce more cocoa flowers (Young, 1986).The pod yield is influenced by photosynthesis and partition of photo-assimilate (Sounigo et al., 2003).
It is assumed that midge population can be a limiting factor in the pollination of cocoa in addition to the environmental conditions.However, populations of insect pollinators are often severely disturbed by hurricanes through flooding of essential habitat and the widespread loss of existing flowers.Small, poor-flight insects such as midges are likely to be swept away by high winds.Climate variation, particularly changes in rainfall leading to sporadic or less rain, may also affect midges which normally thrive in moist humid environments.
Understanding these ecological dynamics can lead to ways of conserving midge populations and mitigating the effects of global climate change and extreme climatic events.The objective of this study is to examine the relationship between the midge population, flower pollination in Trinidad Selected Hybrids (TSH) cacao, and selected weather variables in several different Caribbean cocoa producing islands.

Characteristics of the study area
A multi-location study during the project period of 2013 to 2016 was conducted on several farms in the islands of Trinidad and Tobago (10.667°N, 61.567°W), and Jamaica (18.1824°N, 77.3218°W) in the Anglo-Caribbean which were previously under natural forest (tropical Montane Crappo-guatecare, fine leaf cocorite, black heart) in altitude 120 to 330 m (Nelson, 2004).The areas experienced annual average temperatures of 26.5 ± 2.09°C, relative humidity of 86.1 ± 12. 6%, and mean monthly rainfall ranging between 19.1 and 235.1 mm (Anon, 2016).
The 4 farms/estates were in Trinidad: Jude  (Maharaj et al., 2011), and the trees were in full reproductive phase.The first flowerings were in early January over a 3 month period, and a second period, depending on the rains, in June.Harvesting usually occurred over a 2 month period around 6 months after the first flowering.
All the islands experienced a bimodal rainfall distribution, with peaks in June and November.The first and second growing seasons typically last from mid-March to mid-July and from mid-August to end of November, respectively.However, this is separated by a short dry spell of about two weeks in September and referred as petite careme.The major dry season starts in mid-December and lasts till end of May, and the climate is marked by high incidence of solar radiation and relatively little variation in day length.All data on temperature and relative humidity were measured using the Data Davis Wireless Vantage Weather Pro [Model E14062 Rainfall data, were taken from the meteorological records from the National Water Resources Agency.

Experimental
Four separate studies were conducted during the period 2013 to 2016 in which the European Union COCOAPOP was executed in the following areas: 1. Insect population dynamics.2. Cocoa floral phenology.3. Substrate augmentation trials for culture of cocoa midges (Diptera: Ceratopogonidae), and 4. Generalized linear modelling of weather, midge dynamics and floral phenology.

Study 1: Insect population dynamics
The cocoa insect population dynamics survey was conducted in the 3 islands on 2 well established and managed farms that cultivated the cacao TSH variety under similar agronomic practices.The selected farms were of similar altitude (120 m) and agronomic conditions.The study was conducted over a minimum of fifteen (15) months duration (2013)(2014)(2015).However, the data analysis was confined to 2 complete flowering seasons over 1 year period.
Insect suction traps (Arnold and Chittka, 2012) were set up in 9 representatives transects within each cocoa estate of the different territories.These traps were secured onto branches of cocoa tress, powered by 9-volt batteries and insects were sucked into vials containing 90% ethanol.Insect samples were collected for 2 days/month for each sample site, labelled, stored properly for analysis in the insectary for other insects and midge count.Collection was timed to the midge life cycle (Figure 1).

Study 2: Cocoa floral phenology
The cocoa floral phenology was conducted on the same cocoa farms for each island.Over 20 mature cocoa trees95 to 12 m tall] with 5 cushions/ tree were randomly selected and labelled within an experimental area not exceeding 500 m 2 .The study ensured that data were collected from a minimum 100 plants over 3 consecutive flowering years (2013 to 2015).The observations were conducted monthly on each tree using the modified BCCH (Bleiholder et al., 1991) on counts of flowers, buds, number of mature flower buds, open flowers, new pods or cherelles, small pods (5 to 10 mm),  The BBCH scale was amended to include days from the first day buds become visible [FBV] for each stage and was used to compute the length of each reproductive phase (Figure 2).

Principal growth Code Description
Stage 5: Inflorescence emergence 52 Flower buds expanded, emergence of sepal primordia (bud <1 mm long).59 Flower bud growth complete (buds 6 mm long and 3 mm large; pedicle 14 mm), buds l closed The BBCH Scale (Bleiholder et al., 1991) and the extended BBCH scale (Hack et al., 1992) covered the 10 principal growth stages numbered 0 to 9 (Table 1).However, for the purpose of this study, only 4 of these stages were considered; namely 'macro stages' numbered from 5 to 7.
Study 3. Substrate augmentation trials for culture of cocoa midges (Diptera: Ceratopogonidae) Two (2) separate studies were conducted on 3 commonly found substrates within the fields to determine if they can augment the midge population as suitable breeding sites (Figures 3 and 4).These studies were confined to Trinidad farms only, as the insectary was located there.The substrates assessed over the 2 crop seasons in 2015 were as follows: 1. Field substrate in-situ assessment, and 2. Field augmentation and insectary evaluation.
Field substrate in-situ assessment: During the cropping season of 2015, four (4) cocoa farms were designated for field manipulation to determine if the substrates had any effect on the midge population dynamics.Three substrates were assessed in heaps viz: Rotted cocoa pod (15 kg) (Figure 4), banana pseudo-stem slices (Figure 2) (15 kg) and cocoa leaf litter, all of which replicated three times per farm.All treatments were moistened (5 L water/heap/weekly).The experimental sites (25 m 2 /substrate) were laid out as a Latin square (3 × 3) design.During the first 2 months, insect populations were monitored for 2 days per month using a standard suction trap placed in the approximate centre of each area.Cocoa floral phenology was also monitored during the duration of the study which lasted over 6 months.

Field augmentation and insectary evaluation:
The field experimentation was conducted at one farm (Gran Couva, Trinidad) The Ceratopogonid midge larvae after developing in the organic matter were collected using the Berlese Funnel Traps (Dietick et al., 1959).The substrates were inspected for larger midge larvae (Forcipomyia spp.) which are removed from the substrates and placed in a ball of well-decomposed cocoa pod husk with 100 larvae/vial and adequate air-flow and temperature (26°C).

Study 4. Generalized linear modelling of midge dynamics, floral phenology and weather variable
The approach was to determine the relative role of the midge population dynamics and cocoa floral and reproductive phenology, and its interaction under the prevailing weather variables (rainfall and temperature).This study was conducted over the period 2014 to 2015 in the three countries (Trinidad, Tobago, and Jamaica) on two estates per country.The data was collected from previous midge collection and the floral phenology trials and daily weather data (Table 11) for each location.Best fit generalized linear models were developed to determine the interactions and significance.

Data analysis
The count of flowers and other parameters taken were pooled together on each farm, but separate for each location.All count data were transformed when necessary using the square-root (√x + 0.1) before analysis.Regression analysis were used to determine the relationship between weather variables (temperature, relative humidity, rainfall and light intensity) and flower production, and insect population dynamics using the MINTAB statistical package.

Study 1. Insect population dynamics
There were significant differences between the monthly midge and other insect's population and farms over the 3 territories.There were two distinct and observable high populations May/June and November/January.These periods coincided with the new flushes of cocoa flowers (Figure 4) and the higher rainfall patterns.In Trinidad, the seasonal midge population was 19 ± 3.65 and 53.5 ± 8.47 compared to Tobago which varied between 27.1 ± 3.37 and 22.6 ± 6.47, and Jamaica 21 to 28 ± 4.39/ transect site (Table 3).In all territories, the low midge population varied between 2 to 6 midges/transect site.Jamaica (82) and Tobago (72) had higher midge populations compared to Trinidad (45).The other insect's population was significantly higher than midges and varied between 1067 and 1547 insects/transect site.This indicated that the midge population was less than 2% of the insect trapped (Table 4).

Study 2. Cocoa floral phenology
The cocoa floral and reproductive phenology followed a similar pattern (Figure 4) as outlined on the modified model developed by Bleiholder et al. (1991).In Trinidad, the mean number of flowers was 33.6 ± 6.1/cushion, with the highest ranging between 40 to 96 flower/cushion  The percentage of flowers that were pollinated and successfully fertilized i.e. (Flowers → Cherelles (0" -2.0")) were higher in Jamaica This manifested with a similar pod/cushion yield between countries, with Jamaica (1.5) having a higher pollination/ fertilization, compared to Trinidad (1.0) and Tobago (<1), and was very low for that season (Table 9).

Trial 1: Field substrate assessment
The field trials (Table 6) indicated that there were no variations between the 3 substrates (5.0 to 5.4 ± 1.27) during the experimental period.However, during the wet months of July/August, 2014, the number of midges caught in the suction traps located in the areas of the banana pseudo-stem, and cocoa pod increased, compared to the litter substrate.Similarly, the cocoa leaf litter was not significantly different from pods or pseudostems in August.
The number of midges per suction trap in this trial was consistent to the results obtained in the cocoa insect population dynamics studies (2013/14).The study demonstrated that regardless of the quality of the substrate to improve on the feeding and fecundity of midges, the suction trap appeared to have a determining factor, and may not actually reflect the substrate suitability.

Trial 2: Field manipulative and laboratory evaluations
In this study, no suction traps were used, but samples of the substrate were removed and incubated in the insectary, where the emerging larva were counted, and reared to adult.The results in this study are different from Trial 1, and reflected the potential midge population when interventions of substrates are manipulated in the field.
The fresh cocoa pod (Table 7) left to decay was the preferred substrate for the adult midge to feed and continue its reproductive cycle (Figure 1).The total midge population in the cocoa pod was 3 to 4 times higher than the banana pseudo-stem.The data suggested that increasing the breeding sites with augmentation of cocoa pod substrates can increase the midge population (Table 7) dynamics in the field and new pods development (Table 8).Further, the use of suction traps are not effective or a reliable indicator of the true insect population dynamic in the cocoa estates.

Study 4. Generalized linear modelling of midge dynamics, floral phenology and weather variable
This study involved data transformation and statistical manipulation of observations on the cocoa crop reproductive phenology (Table 9), and midge population dynamics (Table 10) during a one year period, and taking into consideration the prevailing weather variables (Rainfall and Temperature at the different Farm locations) (Table 11).
The generalized linear model revealed that there were variations between farms which influenced the yield of flowers and cherelles (Table 12).Also, the variation in rainfall between months, confirmed the bimodal (wet/dry) season which affected flower emergence and pollination into cherelles.The other main variables in the model;  midge, other insects, and temperature, were not significant and had no impact on flower and pollination.Additionally, the analysis did not reveal any interactions between any of the independent variables on flower and cherelles (Table 13).The analysis showed that the ratio of flowers to cherelle per cushion varied between territories: Jamaica (33:10), Trinidad (33:0.7),and Tobago (18:0.3).However, this data has to be interpreted in the light of the limitations of the suction trap and the true midge population as reported in Study 3. Further, the numbers of flowers were similar

Figure 1 .
Figure1.Graphical representation of Wr depending on the severity parameter to Vr. Wr: covariance between a parent r and its progenies; Vr: variance between a parent r and its progenies.

Figure 2 .
Figure2.Graphical representation of Wr on Vr function for setting the area under the disease progression curve (AUDPC).Wr: covariance between a parent r and its progeny; Vr: variance between a parent r and its progeny.

*Figure 1 .
Figure 1.Frequency distribution of rice genotypes for resistance to leaf and panicle blast across seasons under field conditions at NaCRRI, Kampala, Uganda.R = Resistant; MR = moderately resistant; MS = moderately susceptible; S = susceptible.

Figure 2 .
Figure 2. Frequency distribution of rice genotypes for resistance to leaf and panicle blast in the screenhouse at NaCRRI, Kampala, Uganda.R = Resistant; MR = moderately resistant; MS = moderately susceptible; S = susceptible.

Table 1 .
Analysis of variance for GCA and SCA and reciprocal using Griffing's method for severity.

Table 2 .
Analysis of variance for severity in F1 generation.

Table 3 .
Analysis of variance and AGC, SCA and reciprocal effects by Griffing's method of area under disease progression curve (AUDPC).

Table 4 .
Analysis of variance for AUDPC in F1 generation by Hayman's model.

Table 5 .
Analysis of variance homogeneity test (Wr-Vr) attached to each parent according Hayman.

Table 6 .
Estimated different genetic characters studied components of F1 according to Hayman.

Table 7 .
Narrow sense heritability for severity and AUDPC.

Table 1 .
List of selected rice genotypes used for the study in Kampala, Uganda in the two cropping seasons of 2015.

Table 2 .
Across season analysis of variance of rice genotypes for leaf and panicle blast severity and lesion under field conditions at NaCRRI, Kampala, Uganda during seasons of 2015A and 2015B.

Table 3 .
Disease reaction of rice genotype for blast under field and screen house conditions at NaCRRI Kampala, Uganda during seasons 2015A and B.

Table 4 .
Analysis of variance of rice genotypes for leaf and panicle blast severity and lesion size in the screen house conditions at NaCRRI, Kampala, Uganda in season 2015A.

Table 1 .
The principal reproductive growth stages 5 to 7 of T. cacao var.TSH according to the BBCH (Biologische Bundesantalt, Bundessortenamt and CHemische Industrie, Germany) scale.

Table 4 .
Midge population (%) compared to other insects in cocoa farms over the three locations.

Table 5 .
Cocoa phenological cycle in 6 cocoa farms over 3 Caribbean Islands during a one year period [ 2014/15].
season(May/June, 2015).The mean flower/ cushion in Jamaica did not vary compared to Trinidad (32 ± 5.98), as the trees were of same variety and age, and also displayed two distinct flusher in Sept/Nov, 2014 (29 to 61) and April/June, 2015 (63 to 78).

Table 8 .
Cocoa pod yield in farms with substrate augmentation.

Table 9 .
Cocoa floral phenology and pod yield in 6 cocoa farms over 3 Caribbean Islands during a one year period [ 2014/15].