spatial genetic structure and mating system at the hierarchical levels of fruits and individuals of a continuous Theobroma cacao population from the Brazilian Amazon
Understanding the genetic diversity, spatial genetic structure and mating system at the hierarchical levels of fruits and individuals of a continuous Theobroma cacao population from the Brazilian Amazon
C R S Silva1, P S B Albuquerque1, F R Ervedosa1, J W S Mota1, A Figueira2 and A M Sebbenn3
1Comiss Executiva do Plano da Lavoura Cacaueira, ERJOH, Marituba, Par Brazil2Universidade de S Paulo, Centro de Energia Nuclear na Agricultura, Lab. Melhoramento de Plantas, Piracicaba, S Paulo, Brazil3Instituto Florestal de S Paulo, Se de Melhoramento e Conserva Gen Florestal, S Paulo, BrazilCorrespondence: Dr AM Sebbenn, Instituto Florestal de S Paulo, Se de Melhoramento e Conserva Gen Florestal, CP 1322, S Paulo 01059 970, Brazil.
Top of pageAbstractUnderstanding the mating patterns of populations of tree species is a key component of ex situ genetic conservation. In this study, we analysed the genetic diversity, spatial genetic structure (SGS) and mating system at the hierarchical levels of fruits and individuals as well as pollen dispersal patterns in a continuous population of Theobroma cacao in Par State, Brazil. A total of 156 individuals in a 0.56 plot were mapped and genotyped for nine microsatellite loci. For the mating system analyses, 50 seeds were collected from nine seed trees by sampling five fruits per tree (10 seeds per fruit). Among the 156 individuals, 127 had unique multilocus genotypes, and the remaining were clones. The population was spatially aggregated; it demonstrated a significant SGS up to 15 that could be attributed primarily to the presence of clones. However, the short seed dispersal distance also contributed to this pattern. Population matings occurred mainly via outcrossing, but selfing was observed in some seed trees, which indicated the presence of individual variation for self incompatibility. The matings were also correlated, especially within (p(m) rather than among the fruits (p(m) which suggested that a small number of pollen donors fertilised each fruit. The paternity analysis suggested a high proportion of pollen migration (61.3 although within the plot, most of the pollen dispersal encompassed short distances (28 The determination of these novel parameters provides the fundamental information required to establish long term ex situ conservation strategies for this important tropical species.
Keywords: cocoa; effective population size; microsatellite loci; paternity analysis; tropical tree
Top of pageIntroductionKnowledge of the genetic structure, mating system and contemporary gene flow in tropical trees has increased substantially in recent years. However, most studies have focused on timber species, and little emphasis has been given to tree species bearing fruits that are valued by humans (Alves et al., 2007). Various biomes in the tropics, including the Amazon, have suffered from intense colonisation pressure and alarming rates of deforestation. Important genetic resources must be protected against these threats, and therefore, sound in situ and ex situ conservation strategies that are designed specifically for tropical conditions are urgently required.
Understanding the mating system and pollen dispersal patterns of natural populations of tree species is integral for ex situ genetic conservation because these factors determine the kinship within open pollinated families and consequently affect the effective population size in progeny array samples. Mating systems have been shown to be dynamic across a number of tree species; the outcrossing rates and correlated matings have been shown to differ among Embothrium coccineum populations (Mathiasen et al., 2006), among individuals within populations (Platypodium elegans, Hufford and Hamrick, 2003; Magnolia stellata, Tamaki et al., 2009), among different parts of the canopy (Eucalyptus globulus,
cards againts humanity?, Patterson et al., 2004), among reproductive events (P. elegans, Hufford and Hamrick, 2003), and even among and within fruits within individuals (Acacia melanoxylon, Muona et al., 1991; Eucalyptus rameliana, Sampson, 1998; M. stellata, Tamaki et al., 2009). In animal pollinated tree species, mating systems have been shown to be affected by factors such as the reproductive population density (Murawski and Hamrick, 1991) and by anthropogenic processes such as forest fragmentation (Fuchs et al., 2003; Quesada et al., 2004; Eckert et al., 2009) or logging (Obayashi et al., 2002; Lourmas et al., 2007; Lacerda et al., 2008). Because these processes reduce the density of reproductive individuals and may affect the behaviour of pollinators, the outcrossing rate and the number of mating pollen donors may decrease, which results in inbreeding and an increased relatedness within families. Consequently, the increase in inbreeding and kinship within families decreases the effective population size (Ne more than that expected in populations that are characterised by random mating (Ne and therefore a larger sample from a greater number of seed trees is required for ex situ conservation.
Theobroma cacao L. (cacao; Malvaceae sensu lato; Alverson et al., 1999) is an important tropical tree species that is cultivated for its valued seeds, which comprise the sole source of cocoa butter and solids for the chocolate and confectionary industries. This diploid species (2n is monoecious with bisexual flowers, but it demonstrates high outcrossing rates that range from 30 to 100 (Voelcker, 1938; Benton, 1986; Efombagn et al., 2009a). Cacao is typically pollinated by midges, mainly Forcipomyia species, or by other small insects, such as ants and aphids. Two morphological and adaptive flower traits favour outcrossing in cacao: (i) the presence of a crown of staminodes (modified sterile stamens) around the stigma, which represents a physical barrier against self pollination; and (ii) the anatomical structure of the petals, which contain a distal ligule (blade of the petal) and a proximal cowl, a shell like modification that completely surrounds each anther (Cuatrecasas, 1964). In addition, cacao possesses a unique gameto sporophytic self incompatibility system (Knight and Rogers, 1955; Cope, 1976).
Incompatible mating is characterised by a failure in gametic nuclei fusion at the embryo sac that results in flower abscission (Knight and Rogers, 1955; Cope, 1962), which is considered a late acting mechanism (Gibbs and Bianchi, 1999). The genetic control of self incompatibility in cacao appears to be determined by a nuclear multiallelic S locus (Knight and Rogers, 1955; Cope, 1976), which is affected by other independent loci (Cope, 1958,
crimes against humanity game, 1962). However, the incompatibility mechanism in cacao is not strict but quantitative because it depends on the ratio of fused to non fused ovules and individual differences in incompatibility (Cope, 1962; Warren and Kalai, 1995). The self incompatibility mechanism in cacao can be overcome intentionally by employing a mixture of compatible and incompatible pollen with successful self fertilisation (Glendinning, 1960) or naturally via pollination with a similar blend of pollen under field conditions. The latter conditions have been shown to result in self pollination rates that range from 0 to 89 (Lanaud et al., 1987).
Under natural conditions, cacao seeds appear to be dispersed by animals, which mostly include small primates, rodents or birds; however, humans might be the most efficient dispersal agents. Cacao trees display a natural propensity to develop many orthotropic stems ( that are capable of bending in response to environmental factors. This process initiates the development of additional orthotropic shoots (Bartley,
cards of humanity, 2005). Over time, this vegetative propagation may represent a relevant dispersal mechanism, yet it still demonstrates greater restrictions in comparison to animal seed dispersal. Cacao seeds are recalcitrant, and ex situ conservation requires the maintenance of living clonal plant repositories.
The putative centre of T. cacao diversity was originally hypothesised to be located in the region between Ecuador, Colombia and Peru (Cheesman, 1944) and was later confirmed by a microsatellite marker analysis (Motamayor et al., 2002, 2008). Despite the importance of this plant, little is known about the natural genetic structure of the T. cacao population because most studies have utilised accessions that were originally collected in the wild, but maintained in ex situ germplasm repositories. The accessions have been analysed as a group according to the approximate collection location, the river basin (Sereno et al., 2006; Motamayor et al., 2008; Zhang et al., 2008). Moreover, to our knowledge, no report has estimated the clonality, intrapopulation spatial genetic structure (SGS), effective population size, mating system at the hierarchical level of fruits and individuals, and pollen flow in natural cacao populations, which are fundamental parameters that are required to establish long term ex situ or in situ conservation strategies. Collection expeditions conducted in Amazonia have employed empirical approaches to sample budwood, seeds, or both from trees without previous definitions of the ideal number and distance between individuals for ex situ conservation sampling. The high rate of deforestation in Amazonia threatens the conservation of the untapped genetic diversity of natural cacao. In Brazil, a systematic collection project designed to obtain a representation of the genetic diversity of natural or cultivated cacao in the whole Brazilian Amazon region was conducted by the Brazilian government from 1976 to 1991 (Almeida et al., 1995), but only a small fraction of the Brazilian Amazon river basins have been sampled and maintained in ex situ collections (Sereno et al., 2006).
Therefore, we examined the genetic diversity, inbreeding, SGS and mating system at hierarchical levels of individuals and fruits within individuals and pollen dispersal patterns and distance in a continuous T. cacao population in the Brazilian Amazon. Specifically, we aimed to answer the following questions: (i) Is there a SGS in this population? (ii) Is there selfing and inbreeding in this population as observed by Efombagn et al. (2009b) for some accessions of cacao? (iii) What is the level of correlated paternity, coancestry and the effective population size within and among the fruits? (iv) What is the rate of pollen migration, and the distance and patterns of pollen dispersal in the plot? (v) What is the minimum number of seed trees necessary to collect representative seeds for conservation programmes?
Top of pageMaterials and methodsStudy site and samplingThis study was conducted to evaluate a continuous cacao population in the Amazonian forest near Mocajuba, Par State, Brazil (02 S; 49 W). The population was located in a floodplain that displayed an insular ecotype; the area was exposed to daily floods from tidewater from the Tocantins river basin caused by high tides from the ocean. This site has been used to collect cacao fruits for over 200 years, and it had been selectively logged for economic timber trees. For the study, an 80 70 plot (Figure 1) was established in 2008. This plot contained 156 individuals, which were all sampled, mapped and genotyped. The cacao trees were not randomly distributed in the population, and a clear grouping was apparent in parts of the plot. No growing seedlings were found in the area, but some clonal individuals were visually identified and later confirmed by DNA analysis. Seeds from open pollinated fruits were sampled from nine individual trees that were located close to the centre of the plot. Five fruits were sampled from each tree, and ten seeds from each fruit were used for the genetic analysis (50 seeds per seed tree for a total of 450 seeds for nine seed trees). The numbers in boxes represent the seed trees.
Full figure and legend (46K)
DNA extractionDNA extractions and microsatellite analyses were conducted at the Molecular Biology laboratory of the Executiva do Plano da Lavoura Cacaueira (CEPLAC), Marituba, Brazil. DNA was extracted using a protocol adapted from Doyle and Doyle (1990), as described by Sereno et al. (2006). The DNA was quantified using a spectrophotometer (Biomate 3; Thermo Electron Co., Madison, WI, USA), and the DNA quality was evaluated by gel electrophoresis. If RCE RCE or RCE then the distribution of the individuals is considered random, aggregate or uniform, respectively. To determine whether repeated multilocus genotypes were clones (ramets) of the same genotype (genet), we calculated the probability of observing at least the given number of samples with the same multilocus genotypes using the following equation:
where g is the Hardy probability of the multilocus genotype under random mating; m is the observed number of clones (identical multilocus genotypes) and n is the total number of sampled individuals. The gene frequencies in this analysis were estimated using only unique multilocus genotypes that contained one sample of each repeated multilocus genotype (clones). If Psex individuals with identical multilocus genotypes were ramets of the same genet. This analysis was run using GenClone 2.0 (Arnaud Haond and Belkhir, 2007). All parameters and permutations were estimated using Fstat (Goudet, 1995). Coancestry coefficients were estimated using the method of J Nason (described in Loiselle et al., 1995). To visualise SGS, values were averaged over a set of distance classes and plotted against the distances (classes of 5 up to 80 To test whether there was a significant SGS, the 95 CI was calculated for each observed value, and each distance class was calculated from 1000 permutations of individuals among the locations. The CI was used to construct a coancestry graph. Coancestry coefficients and CIs were calculated using SPAGeDi version 1.3a.
Estimation of historical gene dispersal from SGSSPAGeDi version 1.3a was used to estimate the historical gene dispersal for adults from SGS with the assumption that the observed SGS represented the equilibrium between isolation by distance and genetic drift (Hardy et al., 2006). According to previous plant studies (Hardy et al., 2006), D and D were adopted as a minimum and a maximum estimate of De, respectively. of bk which was calculated by jackknifing the data across each loci (Hardy et al., 2006).
Mating system analysisThe mating system was analysed using the mixed mating model and the correlated mating model in the MLTR program, version 3.4 (Ritland, 2002). The parameters that were calculated at the population level were the multilocus outcrossing rate ™, the single locus outcrossing rate (ts),
cards against humanit, the selfing correlation (rs), the multilocus paternity correlation (rp(m)) and the single locus paternity correlation (rp(s)). The difference between the single locus and multilocus outcrossing rates (tm was used to determine whether any mating had occurred among the relatives in the population. Positive and significant differences between tm and ts were attributed to mating among relatives because ts represents the rate of mating among non relatives, and the complement 1 includes apparent selfing due to mating among relatives and true selfing, and tm excludes all apparent selfing due to biparental inbreeding (Shaw et al., 1981). The parameters tm, ts, tm and rp(m) were also estimated at the levels of families and fruits within families, and the multilocus paternity correlation (rp(m)) was estimated at different hierarchical levels among and within fruits. The analysis was performed at the population level using the Newton numerical method and at the individual family and fruit levels using the Expectation numerical method (Ritland, 2002). The 95 CI of the parameters was calculated from 1000 bootstrap permutations. was calculated using individuals within the families as the baseline for re sampling. The average coancestry coefficient ( within families was calculated from the estimator of the coefficient of relatedness within families (rxy), as derived by Ritland (1989). As the studied population was not inbred (see results), rxy was estimated according to rxy and the coancestry coefficient within families was estimated by dividing Ritland’s estimators by two or directly from the following expression: . We also estimated the average effective population size within families (Ne) using the Cockerham (1969) estimator: , where n is the number of analysed seeds within families (n or fruits (n and Fo is the coefficient of inbreeding in the offspring (estimated by the fixation index). In this estimate, negative Fo values were assumed to be zero. From the effective population size within families, we estimated the number of seed trees (m) necessary for seed collection to retain the reference effective population size (Ne(reference)) of 150 (3 50; Lacerda et al., 2008). This expression is based on two suppositions: (i) the seed trees are not relatives; and (ii) the seed trees do not overlap in the pollen pool. Therefore, in this case, we assumed that the seed trees were not mating with one another and were not receiving pollen from the same fathers. Thus, if related individuals are only present within families, but not among families, the Ne values of the families can be added, and a total effective population size of the progeny array can be estimated.
Paternity analysisThe theoretical power to exclude the second parent (when the first is known) assuming random mating was calculated using Cervus 3.0 (Marshall et al., 1998; Kalinowski et al., 2007). The cryptic gene flow (CGF), or the probability of assigning a candidate father inside the population when the true father was outside the plot, was calculated as described by Dow and Ashley (1996): CGF where N is the number of candidate fathers and P2 is the combined non exclusion probability of the second parent when the first (here the mother) is known. The parentage analysis was conducted based on the multilocus genotypes of the 450 seeds and all 127 unique multilocus genotypes present in the plot via a maximum likelihood paternity assignment (Meagher, 1986) using Cervus 3.0. The most likely parental pair was determined with the statistic (Marshall et al., 1998) using the reference allele frequencies that were calculated for the adult trees, as indicated by Meagher and Thompson (1987). To determine the putative father of the seeds, all 127 genets were tested as putative paternal parent candidates. The significance of was determined using paternity tests that were simulated by the software (critical using a confidence level of 80 a genotyping error ratio of 0.01 and 50 repetitions. The calculation of critical values was based on the assumption that 90 of the sampled candidates were located within the plot. We also evaluated self fertilisation. If a father candidate had a value higher than the critical value calculated by the simulations, it was considered the true parent. If the same individual was found to be the paternal parent, this seedling was considered selfed. Thus, the selfing rate (s) was estimated as the number of selfed seeds divided by the total number of analysed seeds. The pollen immigration rate (m) was calculated as the ratio of seeds that had no parents inside the plot relative to the total number of sampled seeds. Because all of the sampled individuals had known spatial positions, the effective pollen dispersal distance was calculated from the position of the seed tree relative to its putative father. Because of the presence of clones in the population, the distance between the seed tree and each identical clone that was determined to be the father of a seed was averaged. To investigate whether mating success was a function of the distance between trees, we compared the frequency distribution of the realised pollination with the frequency distribution of the distances among all trees using the Kolmogorov test (Sokal and Rohlf, 1995).
Top of pageResultsClonal diversityThe