Estimation of Genetic Parameters in Coffea Canephora Var.Robusta

The objective of this work was to assess the genetic parameters in Coffea Robusta clones using mixed model. The experiments were carried out during four years, in complete block design, and one plant per plot, at Oratorio of Minas Gerais state, Brazil. The clones were evaluated for vigor, reaction to rust, reaction to cercospera, number of or to tropicos branches, number of plegiotropics branches, plant height, diameter of stem, fruit maturity, diameter of canopy, fruit size and production of fruits. The data were analyzed using the mixed model methodology (REML/BLUP) of Selegen software for estimation of genetic parameters in C.canephora breeding. The results showed a low genetic variability among the clones of Robusta for all the evaluated traits. On the other hand, relatively high residual coefficient of variation for most of the traits was recorded implying that these traits seem to be highly influenced by the environmental variation.


I. INTRODUCTION
offee belongs to the family, Rubiaceae and the genus Coffea L. which comprises over 104 species that have been identified so far. Commercial coffee production relies mainly on two species, Coffea arabicaL. (63%) and Coffea canephora Pierre (36%). C. arabicais a natural allotetraploid (2n=4X=44), and is self-fertile (Coste, 1989). Whereas, other species are diploid (2n = 22) and generally selfincompatible. The cup quality made from C. canephorais generally regarded as inferior to that made of C. arabica. However, C. canephoradoes not need to grow at high altitude, requires less care to grow because it is hardier, and it tends to be less susceptible to pests and rough handling (Coste, 1989). C. canephora presents a wide genetic variability, with one of the widest geographic natural distribution within the subgenus Coffea (Maurinet. al, 2007). Hence, genotypic parameters analysis in Coffea canephora has importance, especially for genetic materials from Brazil.
As new coffee varieties are continuously being developed, there is a need to determine the level and sources of genetic variation within and between new and existing coffee varieties (Gichimu and Omondi, diversity and available resources within the genus coffee is an important step in Coffee breeding (Cubryet al, 2008). In view of this, characterization and evaluation of its gene pool is necessary for effective crop improvement programs and for better conservation and management of genetic resources (Prakashet. al, 2005).According to De Vienne et al. (2003), morphological characters are a classical method to distinguish variation based on the observation of the external morphological differences. Morphological and agronomical traits as well as resistance to biotic and abiotic stresses that are known to individual accessions increase the importance of the germplasm. Efficient utilization of indigenous germplasm required knowledge of biodiversity of economic interest (Beer et al., 1993).
Perennialplant speciessuch ascoffee exhibit uniquebiological aspects such abiennialcycle;overlapping generations; expression of characters over several years and differences inearliness and productive longevity (Sera, 2001). Those characteristics lead tosome consequences, such as the use ofselected genetic materialfor several years, reductionin its usefullife survival rateduringthe experiments, a fact that tends to generateunbalanced data for usein the estimationof genetic parametersandprediction of the breeding and genotypic values (Resende,2001). Because of these agronomic peculiarities, genetic improvement of coffeeisdifficult, and recommendedthe use of special methods toestimate genetic parameters and predict thegenetic values (Oliveira et al.,2011;Peteketal, 2008).
Estimations of genetic parameters permit the understanding of the nature of the gene action involved in trait inheritance and lead themselves to assessment of expected progress with selection, besides defining the best selection method to be adopted (Sampaioet al., 2002, Oliveira et al., 2008. Hence, in the Coffea canephora improvement program, the best clones can be selected considering good performance for a C number of breeding target traits, so the use of a selection index is a promising method for simultaneous selection. The use of additive selection indices allows a clear visualization of the performance of the progenies combining most of the agronomic allowing for a better selection of the most promising progenies. The reason is that the correlation may be caused by the action of pleiotropic and/or closely linked genes that affect the traits under study (Falconer and Mackay 1996).
Since the 1930s, several methodologies of genetic evaluation have been proposed, one of them is the least squares for unbalanced data (Yates, 1934). The application of this method is not free of problems, since the variance of the prediction error is minimal, the functions of the prediction are not always estimable and, depending on the degree of data unbalancing, the values of some genotypes may be used overestimated (Henderson, 1974). Whereas, the mixed models equation described by Henderson (1963) introduced changes in the estimation of variance components and breeding values (Searl, 1971). This method consists basically the predication of genetic values considered random to the unequal number of subclasses and to coefficients of relatedness of genotypes (Bernardo, 2002).
Since the prediction of genetic values of superior materials is one of the main problems in the breeding of any species, once it requires the true values of variance components, the use of more sophisticated methods, such as BLUP, allows obtaining better estimates for these parameters (Resende, 2002). This approach takes into account the treatment effects as random, which enables to carry out the genotypic selection instead of the phenotypic one (Resende and Duarte 2007) and their implications in plant selection is presented by several authors in the literature (Duarte andVencovsky 2001, Resende andDuarte 2007). Ramalhoet al.(2013) also emphasizethe advantagesof the application ofBLUPin the improvementof Arabica coffee. This methodologyhas beenusedby other authors invariouscrops such as corn, rice, sugar cane among others. However, there are few reportsin the literatureof theuse of this methodologyin the selectionof individual plantsofCoffea canephoraspecies. Hence, the objective of this study was to estimate genetic parameters and identify the existing variability among C. canephora specie using the mixed model methodology.

a) Plant materials
A total of 52 Robusta variety were used for this study which are maintained at the Coffee Germ plasm Collection of EPAMIG (Empresa de Pesquisa Federal de Vicosa) at Oratorios, Minas Gerais, Brazil. The trail was established as a randomized complete block design with 5 replications and one plant per plot was used.

b) Data collection
Data on 11 quantitative traits recorded on tree basis include vegetative vigor, reaction to rust, reaction to cercospora, number of ortotropics branches, number of plagiotropics branches, plant height, canopy diameter, stalk diameter, fruit maturity, fruit size and production of fruits.
Vegetative vigor average plant scored by the general appearance of the plant, observing the leafiness, the number of orthotropic and reproductive branches, nutritional status and health of coffee, adopting scores from 1 (completely depleted plant) to 10 (highly vigorous plant); reaction to the coffee rustmeasured in the peak months of the disease in the field (between March and July), considering grades 1-5,

c) Statistical analysis i. Estimation of genetic parameters
In order to estimate genetic parameters among Coffea canephora clones, all the quantitative characters considered in the study were statistically analyzed using REML/BLUP by using Selegen software (Resende, 2008).
Considering the environmental of measurements (m ) as fixed, they can be adjusted together the overall average in a single vector of fixed effects. The following model was adjusted. Where: y, m, g, b, p, gm, e: vectors of data, of measurements (fixed), of genotypic effects (random), of block (random), of permanent environment (random), of genotypic x measurement interaction (random), and random errors, respectively. X, Z, W, S, T: Incidences matrices for the effects in the model. The following mixed model equations were used: : Coefficient of determination of block effects; Economic Weight: estimated from Statistics of the experimental data and the genetic variation coefficient (CVg%) as a reference, which directly proportional to the available genetic variance, which express the proportionality between the characters and it is dimensionless (Cruz, 1993).

Estimation of genetic and phenotypic parameters
The highest coefficients of genetic variation obtained were 14.27%, 17.39%, and 19.49% for fruit size (TFr), number of plagiotropics branches (NRPla) and number of Ortotropicos branches (NRort) respectively indicating the relative importance of these traits for the improvement of these variety (Table 1).
On the other hand, relatively highest residual coefficient of variation (CV%) of 125.32%, 45.688%, 36.578%, 23.50% and 26.269% were observed for yield, number of plagiotropicos branches (NRPla), number of Ortotropicos branches(NRort), Cercospors and Vigor, respectively for Robusta. For Robusta variety, the highest phenotypic variance of 1018.85, 674.9, 623.2146, and 222.7696 were obtained for canopy diameter(Dco), plant height, yield (Sac/ha) and plagiotropicos branches (NRPla), respectively. The genetic correlation across measurements (accuracy) over years (r gmed ) ranged from 0.1938 for yield (Sac/ha) to 0.8682 for Cercospora ( Table 2.) The most important function of heritability in the genetic studies on the metric characters, according to Falconer (1960), is its predictive capacity and the expression of the confidence of the phenotypic values as a guide for the genetic value. However, in this study the estimates of individual heritability in the broad sense (h 2 g) were of low magnitude, but were significant for all traits except yield (Sac/ha). Rodrigues W.P. et al (2013) also found similar results for Coffea arabica. The estimated repeatability for most of the traits was lowest indicating the irregularity of the superiority of the individuals among the measurements for these characters showing high irregularity of the performance across measurement, which demonstrate that genotype selection based on those traits is not reliable strategy. The progress expected with the selection depends on the heritability of the character, intensity of selection and phenotypic standard deviation of the character (Cruz and Carneiro, 2004). Thus, the values of heritability and repeatability achieved in this study allow the prediction of better possibilities of genetic gain (Table 1). This also implies that the selection process also provide satisfactory results for all traits (except yield and cercospore) which are economically important characters for Coffea canephoraspecies.

Correlation among traits
The study of correlation provide the information that how strong traits are genetically associated with one other. Thus through the estimate of genotypic and phenotypic correlation among yield components, it paved the basis for selection of superior genotypes from the diverse breeding populations. Therefore, the present study was undertaken to find association of different characters of Coffea canephora, Robusta and Conilon varieties. Correlation among 11 traits was studied using 52 genotypes of Robusta varieties ( Table 2) It is also observed that correlation between NRort and DCA was significant but negative suggesting that there is a negative genetic influence involved in the relationships of these variables. Thus, selection for higher NRort will directly select against DCA. On the other hand, Plant vigor was positively and significantly associated with NRPla, APL, DCO and yield indicating that vigorous of plants expressed in these traits. Montagnonet al. (2001) noted that for young tree Robusta coffee, vigor was best correlated to competition effects,ie, vigorous clones were more aggressive than others. Aggressiveness of clones is reflected either in completion for their neighbor and or stimulating or promoting yield (Montagnonet al; 2001). Similarly, Leroy et al. (1997) reported that more vigorous young plant reflected high yields of Robusta coffee. Generally, fruit maturity showed negative association with most of the traits for both varieties indicating earliness of the varieties. Simultaneous selection of characters and mean performances Using REML/BLUP, the selection gains (SG) can be obtained directly from the BLUP predictions of the progenies, since these reflect the estimated genotypic c)

III.
values already adjusted to the fixed environmental effects. In percentages, the SG ranged from 15.39% to 6.60% among agronomic traits based on the selection intensity of the best 20% (Table 2 and 3). The response due to index selection was largest for 1018 and 4021 and lowest for 401 and 4022 genotypes. Moreover, 50% of genotypes accounted for more than 40% of total advance.
When expected gains were computed as percentage of the means, genotypes responded less to selection using index. It was observed that the SGs was positive and high gains from selection indicating the great potential of improvement expected for this variety. The lower gains in the desired direction for clones 306 and 4022 was explained by the negative genetic association between the agronomic traits. However, the negative genetic correlation does not necessarily imply that progenies combining good agronomic performances cannot be obtained. Since this correlation is determined by the action of linked genes, success in simultaneous selection could be achieved by evaluating a larger number of traits to raise the chances of finding promising recombinant genotypes within the population (Falconer and Mackay, 1996). IV.

Discussion and Conclusion
Generally, the low the genetic coefficient of variation, characterizes the existence of low genetic variability between genotypes, showing that the greater part of the total variation is due to non-genetic parameters. Resendeet al. (2001) reported similar results for Coffea arabica. The high and very high values of CV% observed here are partly due to the ample variation for these traits in the treatments, above all these traits also seems to be very influenced by the environment which may have contributed to high CV%. According to these results, generally there was low interaction with year, as observed by the genotypic correlation across measurement (r gmed ) for most of the characters evaluated, demonstrated that selection can be performed at any of the development stages used for measurement.
Since the environmental variance obtained in this study was low in relation to phenotypic variance, it can be concluded that genotypes with better yield will have the same response at each season while maintaining predictability in the face of environmental variations. However, the genotypic correlation across the measurement was low for yield (sac/ha). As the environment has strong influence on productivity, the highest coefficient of variation was obtained implying that experimental precision in different harvests was very poor. Similarly, low selective accuracy (r = 0.1130) showed that there was weak relationship between predicted and actual values. In contrary to this, Resende and Duarte (2007) obtained high selective accuracy for Coffea arabica which results in surety in the selection of agronomically superior genotypes. This statistic was preferred to the experimental variation coefficient, because it does not only relies on the magnitude of the residual variance and number of replications, but also on the proportion between the genetic variations and residual nature associated with the character under question (Resende, 2002).
The repeatability represents the maximum value that heritability may achieve in the broad sense, since the genotypic variance used to estimate the repeatability is not only of genetic origin, but still masked by the variance components of the permanent environment and among individuals (Cruz et al, 2004). The study of correlation among yield and yield contributing traits suggests that Vigor, DCO, DCA and TFR were the most important characters which possessed positive association with yield. Therefore, these characters could be utilized in breeding program to improve varieties for higher yield. Positive correlations were also observed between most vegetative characteristics and productivity. This shows that plants that have good initial development can provide good yields. Silvarolla et al. (1997) found a correlation between productivity and vegetative characters for Timor Hybrid. On the average of four harvests, the authors obtained a high phenotypic correlation of productivity gained by plant height and canopy diameter.