Cupuaçu tree genotype selection for an agroforestry system environment in the Amazon

.

The Amazon has been identified as the only reserve of cupuaçu genetic variability, and it is also the center of the species' diversity.However, this region has undergone significant anthropogenic pressure through predatory exploitation and transformation of its natural landscape, with the consequent genetic erosion of this plant resource (Alves et al., 2007).Therefore, there is a need to understand the remaining genetic variability of the species.Since the late 1980s, Embrapa Amazônia Oriental has been conducting a genetic breeding program of the species in the state of Pará, Brazil, which aims to develop genotypes with superior levels of fruit production and resistance to the fungus Moniliophthora perniciosa (Stahel) Aime & Phillips Mora, an etiological agent of witches' broom disease and the main pest that affects the cupuaçu tree cultivation (Alves et al., 2009).
The most cupuaçu cultivation is carried out in agroforestry systems (AFSs), mainly in the northeastern Pará (Barros et al., 2009).AFSs are productive systems in which land is managed in a consortium, using a range of tree species to promote economic and ecological gains (Silva, 2013).Within AFSs, different plant strata can be created in an attempt to replicate the native forest structure, in which trees and shrubs are considered basic elements, combining floristic diversity with the environmental characteristics of the region in which they are established (Paludo & Costabeber, 2012).
In addition to providing producers with genetic materials that meet their needs, such as high levels of fruit productivity and resistance to phytopathogens, the selection of genotypes adapted to conditions of interspecific competition for light, water, and nutrients that normally occur in AFS environments (Giustina et al., 2017) should be incorporated into the objectives of species breeding programs.Currently, researches trying to develop genotypes for these conditions are very rare, especially those involving perennial species.Alves et al. (2020) report results obtained with Swietenia macrophylla King.and cupuaçu tree, when they identified genotypes with the mentioned traits.
The objective of this work was to select cupuaçu progenies and individuals for their superiority in fruit production and resistance to witches' broom disease, and, indirectly, to select better adapted individuals to the agroforestry system environment.

Materials and Methods
The experiment was carried out over 14 years, from 2005 to 2019, on a rural property in the municipality of Tomé Açu (02°32'53.1"S,48°15'52.8"W,at 45 m altitude), in the state of Pará, Brazil.The soil is classified as Latossolo Amarelo distrófico argissólico according to Brazilian soil classification system (Santos et al., 2018), i.e., Oxisol, and it is deep and well drained, with a medium texture.The climate in the region is classified as Ami, according to the Köppen-Geiger's classification, with 2,716 mm mean annual precipitation, 85% relative air humidity, and 26.4°C mean annual temperature (Bolfe & Batistella, 2011).
The AFS established for the present study consisted of an area of 0.9 ha, composed of four species: black pepper trees (Piper nigrum L.) planted in double rows of 2x2 m and 4 m apart (1,440 plants); free standing cupuaçu seedlings planted at 6x4 m spacing (375 plants), interspersed with banana tree seedlings (Musa spp.) in the planting row, originated from rhizomes, at the same spacing (375 plants); and grafted seedlings of bacuri (Platonia insignis Mart.) planted at 18x10 m spacing (49 plants) (Figura 1).
All species were planted in February 2005, after mechanized clearing of the area and individual planting hole preparation.The experiment was conducted using a randomized complete block design, with 25 cupuaçu progenies, five replicates, and three plants per plot, for a total of 375 plants.The thinning dynamic of the temporary crop (banana) began in the fourth year after planting, and by the end of the sixth year, all banana clusters had been removed.The same occurred with the black pepper trees which remained in the field until the seventh year, at which point they were removed because they were affected by fusarium disease.From that point on, only bacuri and cupuaçu remained in the AFS.
The collection of data related to cupuaçu production began two years after planting, during the 2007/2008 harvest, and continued until the 2018/2019 harvest, for a total of 11 harvests.Data collected at each harvest included the number of fruit produced per plant and the average weight of the fruit for each plant, using a sample of five fruit per plant.Thus, for each harvest, fruit production (kg per plant) was estimated with the formula: production = number of fruit × mean fruit weight.The rate of witches' broom disease was also estimated based on the following ratio: number of plants with symptoms of the disease / number of healthy plants.
For statistical analyses, the phenotypic averages of production and incidence of witches' broom disease, evaluated during 11 harvests, were used.The statistical analyses were performed using the mixed model methodology (REML/BLUP) in the SELEGEN-REML/BLUP software (Resende, 2016), in which the components of variance were estimated by the REML method (Patterson & Thompson, 1971), and the genetic values were predicted by the BLUP method (Henderson, 1975) simultaneously.The following statistical model was used: y Xr Zg Wp e , in which: y is the vector of phenotypic data; r is the vector of replicate effects (assumed to be fixed) added to the general average; g is the vector of individual genotypic effects (assumed to be random) , in which σ g 2 is the genotypic variance among full-sib progenies]; p is the vector of plot effects (assumed to be random , in which σ p 2 is the environmental variance between plots]; and e is the vector of residuals (random , which σ e 2 is the residual variance within a plot].The capital letters represent the incidence matrices for these effects. The significance of the random effects of the statistical model was tested by the likelihood ratio test (LRT) (Wilks, 1938) using the Chi-square statistics, with one degree of freedom, and 0.05 type I probability of error.The individual phenotypic variance ( σ f 2 ), the individual narrow-sense heritability ( h a 2 ), the coefficient of determination of plot effects ( c p 2 ), the average heritability of progenies ( h mp 2 ), the accuracy of progenies selection ( r gg  ), and the additive heritability within a plot ( h ad 2 ) were estimated using the following expressions, respectively (Resende et al., 2014): in which: r is the number of replicates (5); and n is the number of plants per plot (3).The progenies with the highest predicted genetic values for the two fruit production traits that maintained ortets with a low incidence of witches' broom disease (<20%) were prioritized in individual selection.Individuals were ranked based on the predicted genetic values, considering the selection of possible genitors, and the predicted genotypic values, considering the selection of clones for the trait fruit production per plant.This trait was chosen as it is the most economically relevant, since it incorporates both the traits number of fruit and average weight of fruit.Therefore, the selection of individuals was not only based on the previously selected progenies, but also considered the ranking for the genetic and genotypic values and incidence of M. perniciosa.
To avoid a significant reduction in the genetic variability of the breeding population, only two ortets per progeny were selected, to mitigate the risks of resistance breaking and the occurrence of incompatibility that is common in the species and can cause yield diminishing (Alves et al., 2017).The temporal distribution of the productive period of the selected individuals was also evaluated, for which plants with greater production concentrated in the last months of the year (beginning of the harvest) were classified as early, plants with greater production in the first four months of the year (middle of the harvest) were classified as regular, and plants with greater production in the period from May to July (end of the harvest) were classified as late.

Results and discussion
Cupuaçu fruit production began two years after planting, and the fourth harvest showed an increase of 666% in the number of fruit over the first harvest (Figure 2).At that point, the production in the area stabilized, with minimal fluctuations in the subsequent seven harvests.Both the rapid evolution in the initial production, and the variations in production in the final harvests can be attributed to the same conditioning factor.Initially, the area was equipped with an irrigation system, which supplemented the need for water during the critical period of the year, providing optimal growing conditions.Later, the irrigation system was removed from the AFS.Thus, fruit production depended exclusively on rainfall, which occurred in the period immediately before harvest (July to December), as well as in the flowering period and in the beginning of fruiting (Venturieri, 2011), leading to annual fluctuations in the production from the fourth harvest onward.When evaluating the behavior of Theobroma cacao in response to variations of water availability in the soil, García Lozano & Fonseca (2016) found a decrease of plant metabolism during water shortage, which was attributed mainly to the stomatal closure that paralyzes photosynthesis and, consequently, the production of photoassimilates.
Based on the likelihood ratio test (LRT), the genetic effects were significant, while the plot effects were not significant (p<0.05) for the average number of fruit and average fruit production (Table 1).Therefore, for both traits there was a genetic variability between fullsib progenies of cupuaçu, and the plot effects did not differ from zero.
The estimated genotypic variance among the fullsib progenies was reasonably high, especially for the variable "fruit production" (Table 2).This is a desirable outcome, as it indicates a high genetic variability and supports the possibility of successful selection (Rosado et al., 2012).Corroborating this result, the selective accuracies were high (0.70 < r gg  < 0.90) for the two traits, based on the classification proposed by Resende & Duarte (2007).According to Resende (2002), the selective accuracy refers to the correlation between the predicted and true genetic values, and it is closely associated with the experimental quality.As verified through the LRT, the coefficient of determination of plot effects was practically nil, which can explain the minor environmental heterogeneity between plots (Alves & Resende, 2008).
The high estimates for average progenies heritability (Table 2) reinforce the potential for genetic selection (Rosado et al., 2012).Moreover, moderate but reasonable values for individual narrowsense heritability were observed (0.15 < h a 2 < 0.50) (Resende, 2015) (Table 2), considering the polygenic control of the traits, which are significantly influenced by the environmental effects.This difference is clear when analyzing the work of Maia et al. (2011), in which the traits fruit length, which is of oligogenic nature, showed narrow-sense heritability of 64%, while the traits fruit weight showed 16% narrow-   crosses using the cultivars 'BRS Coari' (174) and 'BRS Manacapuru' (215) reinforce the importance of these progenies for the breeding program, as crosses based on more productive clones may produce genetic materials that maintain the resistance of these genotypes, while incorporating the productivity of the third parent.The above mentioned progenies showed few plants with symptoms of witches' broom disease over the 14 years of the experiment, thus providing the evidence of good levels of field resistance (Table 3).However, most ortets with symptoms were from progenies 13 and 43, reflecting high susceptibility to M. pernicious.For progeny 13, this behavior was expected, since both parents showed susceptibility to the disease (Alves et al., 2009).The objective of this crossing was to concentrate productivity genes from both parents.Thus, despite its susceptibility, it may be interesting in the future to select at least one ortet with high fruit production within this family to be used in hybridization with resistant genotypes.However, the behavior of progeny 43 was unexpected.Although the male parent (clone 12) was highly susceptible, the other parent (clone 220) showed good resistance in other crossings (progenies 9, 11, 17, and 38).From this, it can be concluded that the susceptibility genes of the second parent (clone 12) prevailed in those progenies.According to Alves et al. (2009), genes related to witches' broom resistance in cupuaçu are in heterozygosis, which is why parents that are resistant may produce progenies that are susceptible to the disease.
Hybridization, the method used to produce the progenies of the present study, seeks to consolidate the favorable alleles present in two genitors into a single individual (Baldissera et al., 2014).In the present work, productive and low-resistance genotypes were crossed with resistant genotypes, but the results were unsatisfactory for productivity.The choice of genetically distinct genitors may increase the efficiency of crosses.
After the identification of the superior progenies, the best individuals were identified and selected.In order to maintain the genetic variability in the population, and to guarantee a broad genetic base, two ortets per progeny were selected.Individuals were selected based on the genetic and genotypic values, excluding individuals whose progenies had high incidence rates of M. pernicious (Table 4).No discrepancies were found between the two lists of selected individuals.That is, the same individuals selected on the basis of genotypic values were also selected on the basis of genetic values.Similar results were obtained by Alves et al. (2010).The production from the ten clones originating from these individuals can provide an mean gain of 7.33 kg of fruit per plant, which corresponds to 40% increase over the original population.If these clones are used to establish a seed orchard, a gain of 6.15 kg of fruit per plant is expected, corresponding to 34% increase in relation to the initial situation.According to Alves et al. (2010), the use of clones as planting material narrows the genetic base, therefore, planted populations should always consist Table 4. Individuals selected for clonal propagation with the genetic values and genetic gains of clones (kg per plant), and genetic values and genetic gains of genitors (kg per plant), based on average fruit production per plant (kg) and productive period of cupuaçu (Theobroma grandiflorum).
Individual (1)  Fruit production per plant (kg) of five or more genotypes, in order to avoid the risk of epidemics caused by the fungus M. perniciosa.This precaution also applies to the selection of clones used as progenitors which, through interbreeding, will produce seed with a broad genetic base.
As to the production period, two selected individuals were considered early, four were regular, and four, late.This balancing of the productive season is important to avoid the concentration of production in a single period of the year, which complicates the logistics of the agroindustry and reduces the price paid to the producer due to excess supply (Hojo et al., 2007).
A good adaptability of the tested progenies to AFS conditions was observed due to the seasonal pattern of fruit production (Table 4).Thus, it was indirectly observed that, along with favorable agronomic traits, the genotypes offer tolerance to the conditions present in AFS environments, such as the competition for light, water, and nutrients.Nevertheless, attention should be paid to the choice of the species that will be used in consortium with cupuaçu.Alves et al. (2018a), when studying the interaction of taperebazeiro (Spondias mombin L.) with cupuaçu in an AFS, reported the viability of the consortium due to the fact that no significant losses were observed in the production of cupuaçu fruit.When choosing the species that will compose the AFS, plants that exploit different strata, both above and below ground, are recommended to avoid competition for light, nutrients, and water (Condé et al., 2013).As to this fact, competition with bacuri was probably mitigated by the fact that grafted seedlings were used, which promotes slower growth than seminal seedlings and produces smaller and more compact plants.As there is economic interest in both species, this consortium is advantageous for producers because it combines two potential products of Amazonian fruit culture (Porte et al., 2010).In addition, both are native species, which increases the likelihood of success due to their adaptability to Amazonian conditions.

Conclusions
1.The agroforestry system allows of the discrimination of the most adaptable cupuaçu (Theobroma grandiflorum) tree genotypes.
3. Ten matrices show agronomic potential and indirect adaptation to the agroforestry system, and they should be used as clonal cupuaçu cultivars in this environment.

Figure 1 .
Figure 1.Spatial distribution of species in the agroforestry system in the implantation (A) and in the maturation period (B).

Figure 2 .
Figure 2. Average number of fruit per plant and average fruit production per plant (kg), for eleven harvests (from 2007/2008 to 2018/2019), in an experimental trial of full-sib cupuaçu (Theobroma grandiflorum) progenies.

Table 1 .
Analyses of deviance for the traits average number of fruit per plant and average fruit production per plant, evaluated in an experimental trial of full-sib cupuaçu (Theobroma grandiflorum) families.
LRT: likelihood ratio test.*Significant at 5% type I probability of error by the Chi-square test.ns Nonsignificant.The null hypothesis was that the complete and reduced models do not differ from each other.

Table 2 .
Estimates of variance components and genetic parameters for the traits average number of fruit per plant and average fruit production per plant, evaluated in an experimental trial of full-sib cupuaçu (Theobroma grandiflorum) families(1).
2 , average progeny heritability; r gg  , progeny selection accuracy; and h ad 2 , additive heritability within a plot.

Table 3 .
Predicted genetic values for the traits average number of fruit per plant, average fruit production per plant, and rate of Moniliophthora perniciosa infection evaluated in an experimental trial of full-sib cupuaçu (Theobroma grandiflorum) families.