Hierarchical outcrossing among and within fruits in Bertholletia excelsa Bonpl. (Lecythidaceae) open-pollinated seeds

Abstract

Knowing the mating patterns is important to determine the number of trees necessary for seed collection for conservation ex situ, tree breeding and environmental reforestation purposes. We investigated B. excelsa individuals and fruits, to check mating system index variations in a population by using open-pollinated seeds which were hierarchically sampled within and among fruits from nine trees genotyped for ten microsatellite loci. Outcrossing rate ( )changed between trees (0.49–1.0) and fruits (0.53–1.0), but seeds were predominantly produced by outcrossing (0.92) at mean population level. Mating between related trees ( ) was detected in six trees (0.04–0.08) and in 32 fruits in trees (0.03–0.22), thus it suggests that the population presented some related trees in our intra-population spatial genetic structure. Individual fixation index values of seed trees ( ) were lower than the seedling fixation index values ( ), fact that suggests the selection against inbred individuals between the seed and adult stages. The correlated mating showed that seeds sampled at population level were predominantly composed of half-sibs (66%) and full-sibs (20%). Paternity correlation was significantly higher within ( ) than among ( ) fruits at population level, mostly in individual trees. Results evidenced that mating was not random due to self-fertilization, to mating between related trees and correlated mating, and families comprised different relatedness levels such as half-sibs, full-sibs, self-sibs and self-half-sibs; some cases, mating presented inbreeding. These results were addressed to discuss strategies for seed collection applied to conservation ex situ, tree breeding and environmental reforestation.

Introduction

Non-timber forest products (NTFPs) from native forests contribute to biodiversity conservation (Arnold and Ruiz-Pérez, 1998; Marshall et al., 2005) in the Amazon region, since their exploitation generates income and encourages forest preservation by rural communities in order to improve the quality of life of local producers (Bayama et al., 2014). Thus, it is necessary to reconcile the productivity and ecological sustainability of this tree species to generate NTFP incomes (Wadt et al., 2008).

Bertholletia excelsa Bonpl. (Lecythidaceae), popularly known as Brazil nut, presents seeds with high commercial and nutritional value (Wadt et al., 2015; Baldoni et al., 2017). Brazil nut (NTFP) extraction favors forest preservation; therefore, it is considered a model species for conservation if one also takes into account that wood cutting is prohibited by law (Wadt et al., 2005). The species is distributed in the Amazon region, which is known by its diversity of plant species presenting the potential to be used in agriculture and in tree breeding. However, the lack of knowledge about the species, associated with deforestation, agricultural frontier advancements and wildfire events, contributes to natural population losses, even before they are studied (Maués and Oliveira, 2010).

Understanding species mating systems concerns knowing its composition and the genetic structure of its populations (Richards, 1997; Luna et al., 2005). The mating system is the way individuals, populations or species recombine their genetic variability in each generation in order to generate offspring. Being aware of such system is relevant for population genetic conservation management, for tree breeding and environmental reforestation plans (Sebbenn, 2006). Overall, species can be classified by mating system: i) autogamous, when outcrossing is ≤ 0.2; ii) outcross, when it is ≥ 0.8; iii) mixed mating systems, when outcrossing is between 0.2 and 0.8 (Goodwille et al., 2005). B. excelsa is classified as an outcrossing species and its floral structures preclude self-fertilization, as well as limit the group of animals capable of reaching pollen in their flowers (O'Malley et al., 1988; Maués, 2002; Cavalcante, 2008). Their main pollinating agents are bees belonging to genus Bombus, Xylocopa and Centris (Müller, 1995; Maués, 2002). Flowers house the reproductive organs in a chamber (ula) and these groups of bees are able to reach the pollen, since their physical vigor and robust bodies make the task feasible (Maués, 2002). Despite being an outcrossing species, some studies observed selffertilization in Brazil nut at population and individual level (O'Malley et al., 1988; Wadt et al., 2015; Baldoni et al., 2017).

Molecular markers such as microsatellite loci or SSR (Simple Sequence Repeats) have the advantage of being codominant, heritable and very polymorphic. Moreover, they amplify large numbers of alleles (Hoshino et al., 2002) and represent low cost in investigations about the genetic diversity and mating system of tree species when the primers were already developed, such as in B. excelsa. Reis et al. (2009) developed 12 polymorphic microsatellite markers to the species and Sujii et al. (2013) developed another 12. In addition, genetic studies with molecular markers in fragmented forests using natural seeds or regenerates helped evidencing processes such as genetic drift, gene flow, selection and mating system (Carvalho et al., 2010; Wadt et al., 2015; Baldoni et al., 2017).

Although it is forbidden to cut B. excelsa trees, the strong deforestation in the Amazon region and the illegal logging already extinguished many of its populations; therefore, conservation strategies are required. Knowing the mating system of a tree species is important for genetic conservation, tree breeding and environmental reforestation, since fertilization patterns determine the relatedness and inbreeding of next generations and, consequently, the effective size of seed collection (Sebbenn 2006). The mating system of B. excelsa has been investigated (O'Malley et al. 1988, Wadt et al. 2015), as well as its pollen dispersal patterns (Baldoni et al. 2017). These studies have been investigating the outcrossing rates, the mating between related trees, the paternity correlation among and within fruit variations, and pollen-dispersal distance and patterns. We investigated population and individual outcrossing rate variations, and paternity correlation among and within fruits in order to add information about fruit within tree mating system. We tested the following hypotheses: i) Are seeds in population, individual seed trees and fruit levels produced by outcrossing? As the floral structures of the species preclude self-fertilization, we expected all seeds to be originated from outcrossing; ii) is there mating between relatives at population and individual level, as well as within fruits? There are reports about populations presenting intra-population spatial genetic structure; therefore, we expected to record mating between relatives at all these levels; iii) Is the paternity correlation within fruits higher than among fruits? We expected to find this pattern because of reports in previous studies about higher paternity correlation within than among fruits in the species.

Materials and Methods

Study site and sampling

The study was conducted in a 9ha permanent plot (717971.64 E, 8774887.57 S) located in legal reserve at Santo Ângelo Farm, which belongs to the Dalpai Group. The farm is located approximately 30 km from Itaúba County, Mato Grosso State, Brazil, where seed extraction is performed in a yearly basis. The plot is located in a natural Amazon forest. The Bertholletia excelsa population density was 19.8 trees/ha and presented diameter at breast height > 10 cm. Nine trees were selected in the center of the experimental plot; fruits were hierarchically collected from them. Seeds were germinated in sandbox under 50% shade after their tegument was removed at Embrapa Agrossilvipastoril, Sinop, MT. This procedure generated 300 progenies duly identified based on their origin (seed tree and fruit). The plant material from all adult trees (trunk cambium) and progenies (leaves) was collected for DNA extraction. The leaves were transported in plastic bags containing silica gel and stored in a freezer (-20° C). The trunk cambium was stored in 1 ml of transport buffer (300 μl of 2% CTAB buffer, 700 μl of absolute ethanol, 0.2 g of ascorbic acid) and stored in cold chamber (4° C).

Microsatellite analysis

The vascular cambium and leaf of the trunk were used to extract the total DNA from the samples based on the methodology described by Doyle and Doyle (1990), with modifications (CTAB from 2 to 4%). Polymerase Chain Reaction (PCR) amplification was carried out in ten microsatellite loci: BET12, BET14, BET15, BET16, BEX02, BEX09, BEX22, BEX27, BEX33, and BEX37, as described by Reis et al. (2009), Sujii et al. (2013) and Cabral et al. (2017). These analyses were conducted at the Genetics Laboratory of Embrapa Genetic Resources and Biotechnology, Brasília, DF. Fragment sizes were identified in base pairs in the GeneMapper 4.1^® software (Applied Biosystems). Data extracted from the GeneMapper software were rounded out in the Allelobin software (Idury and Cardon, 1997).

Mating system analysis

The mating system at population, individual seed tree and fruits within tree level was estimated through mixed and correlated mating models that were assessed in the MLTR 3.1 software based on the Expectation Maximization Numerical method (Ritland, 2002). The estimated indices were pollen and ovules gene frequencies, seed trees fixation index multilocus and single-locus outcrossing rates, mating between relatives selfing correlation multilocus paternity correlation within and among fruits (r_p); within and among fruits. Fruits presenting only one seed were excluded from the analyses conducted to estimate and at population and individual seed tree level, only fruits with at least three seeds were used in the analyses. Standard deviation indices at 95% confidence intervals were calculated by using 1,000 bootstraps as re-sampling units between individuals within a single family. The effective number of pollen donors was estimated among and within fruits within fruits and, among fruits based on Ritland (1989). The proportion of pairwise self-sibs, half-sibs full-sibs self-half-sibs and mean coancestry coefficient within progeny wherein s is the selfing rate: followed the methodology by Sebbenn (2006). Seed trees individual fixation index (F_m) was estimated in the SPAGEDI software (Hardy and Vekemans, 2002). The variance effective size (N_e) and fixation index (F_o) within a single family were estimated according to Wadt et al. (2015). Negative F_m and F_o values were assumed to be zero in and and N_e estimates (Wadt et al., 2015). The number of trees required for collecting seeds for conservation purposes was estimated through wherein: is the required effective population size (Sebbenn, 2006), which was herein assumed to be 150 (Lacerda et al., 2008). The 95% CI applied to the indices was described according to Wadt et al. (2015). The spearman correlation coefficient (ρ) was used to investigate whether there was significant association between sample size (n) pairwise variables within families, and within fruits,

Results

The mean population fixation index of seed trees was low, but significantly higher than zero, thus it indicated an inbreeding process; however, it was negative at individual level and ranged from -0.71 to -0.37 (Table 1). The mean population fixation index of seedlings was similar to zero but ranged from - 0.42 to 0.12 between families (Table 1) and from -0.43 to 0.34 (Table 2) between tree fruits. All F_o values were higher than F_m (Table 1), and it suggested the selection against inbred individuals between the seed and adult stages. The mean outcrossing rate at population level (0.92) was not significantly different from the unity (1.0), but, at individual level, it was significantly lower than 1.0 in five seed trees (it ranged from 0.49 to 0.98) (Table 1) and in fruits within trees in three fruits (it ranged from 0.53 to 0.79) (Table 2). Self-fertilization correlation was not significantly different from zero. Mating between related trees was significantly higher than zero at individual level in six trees (it ranged from 0.04 to 0.08) and it ranged from 0.03 to 0.22 in 32 fruits within trees (Tables 1 and 2).

Table 1. Inbreeding and mating system indices at individual and mean population level for Bertholletia excelsa

Table 2. Inbreeding and mating system indices at individual fruit level for fruits with at least three seeds

Paternity correlation indicates that 23.0% of outcrossing seedlings at population level were full-sibs; it ranged from 0.02 to 0.62 between seed trees, and from 0 to 1.0 between fruits within trees. Thus, 4.3 pollen donors effectively fertilized seed trees, on average; fertilization ranged from 1.6 to 52.6 between trees and from 1.0 to 28.6 within fruits (Tables 1 and 2). On average, 0% (95% CI: 0-2%) of the seedlings were selfsibs 66% (95% CI: 44-97%) were half-sibs were full-sibs and 14% (95% CI: 0-27%) were self-half-sibs at population level. Accordingly, the coefficient of coancestry at population and individual level, and within fruits, was higher; the variance effective size within a single family was lower than the expected in panmictic populations The number of seed trees for seed collection (m) was estimated in 64 (95% CI: 52-79). Paternity correlation was significantly higher within fruits than among fruits at population level, and higher within fruits than among fruits in eight of the nine seed trees.

The sample size was not significantly correlated to the mating system indices at family level, but it was significantly associated with at fruit within trees level (Table 3). Therefore, we excluded the fruits with less than three seeds in order to decrease sample size variation between fruits within trees. Subsequently, only n and N_e was significantly associated The indices and N_e had significant positive association at family and fruit level, and the index had significant positive association at fruit level.

Table 3. Results of Spearman correlation coefficient () and statistical probability (P) between sample size (n) pairwise variables, t_m , t_m - t_s , N_ep [], F_o and N_e

Discussion

Outcrossing rate

Our results indicate individual and fruit outcrossing rate variation, and t_m lower than the unity (1.0) in five seed trees (0.49 - 0.98), as well as in three fruits within trees (0.53 - 0.79). Therefore, B. excelsa overall produces seeds through outcrossing; however, there are some variations between tree and fruits within trees, due to individual variations in self-incompatibility or due inbreeding depression; thus, same trees may produce seeds from self-fertilization. The species presents floral structure capable of precluding self-fertilization (O'Malley et al., 1988; Maués, 2002; Cavalcante, 2008), although self-fertilization at population level (at range 0.85 to 0.98) and at individual level (at range 0.90 to 0.98) have been reported in other studies (O'Malley et al., 1988; Wadt et al., 2015; Baldoni et al., 2017). These results suggest that the species presents latent self-incompatibility; some self-fertilization events may result in seed production. An alternative explanation for t_m variation between trees and fruits lies on genetic load population and on individual variations that result in the survival of self seeds originated from populations, or from trees, with low genetic load (Hufford and Hamrick, 2003; Tambarussi et al., 2016). Our results also evidenced that outcrossing can vary between fruits within trees; thus, some openpollinated seeds are inbred due to self-fertilization.

Mating among related trees

Our results also showed mating between related individuals in six trees (from 0.04 to 0.08) and within fruits (from 0.03 to 0.22). According to these results, the population presents some related trees, due to the possible occurrence of intra-population spatial genetic structures (SGS). Related individuals were often closer than the unrelated ones, due to short seed dispersal and closeness to the mother. SGS was detected in two populations living 175 m from each other, thus it indicated that near-neighbor trees living at this distance from each other may be related (Baldoni et al., 2017). Other study also detected mating between relatives (from 0.02 to 0.08), and it is explained by the fact that the mean pollen dispersal distance (159 m) lies within the distance SGS occurs in (Wadt et al., 2015); it may also explain our results. Furthermore, it also reinforces that the species presented self-incompatibility variation between trees in our genetic load and that inbreeding may also due to mating between related trees.

Inbreeding

Our results showed that has positive association at fruit level. Thus, the increased selfing or mating between related trees increased inbreeding within fruits. The individual fixation index values of seed trees were lower than the seedling fixation index values and it suggested the selection against inbred individuals between the seed and adult stages. Thus, inbreed seeds originated from selfing and mating between related trees will probably die before reaching the adult stage due to inbreeding depression. Inbreeding between the seeds in adult stage is apparently a common pattern in tropical trees, as it has been reported in many studies (Hufford and Hamrick, 2003; Degen and Sebbenn, 2014; Wadt et al., 2015; Tambarussi et al., 2016).

Correlated mating

The sampled seeds were predominantly composed of half-sibs (66%) and full-sibs (20%) at population level. The mean population paternity correlation was 0.23, but it varied between seed trees (0.02-0.62), and between fruits within trees (0 to 1.0). Thus, in mean population level, a low number of pollen donors effectively (N_ep = 4.3) fertilized the seed trees; it ranged from 1.6 to 52.6 between trees, and from 1.0 to 28.6 within fruits. The maximum estimated N_ep of 28.6 is an obvious overestimate, because a single fruit produces from 8 to 24 seeds. Our N_ep results at population level are similar to those recorded for other populations (4.5) distributed in natural forests (Wadt et al., 2015). The paternity correlation within fruits was higher than between fruits at population level as well as at individual level (Table 1). Thus, the number of effective pollen donors was often lower within than between fruits Wadt et al. (2015) detected similar pattern in other populations of this species higher than is also reported in many studies involving insect-pollinated tree species (Muona et al., 1991; Sampson, 1998; Quesada et al., 2001; Tamaki et al., 2009; Silva et al., 2011; Manoel et al., 2015; Tambarussi et al., 2015).

CONCLUSION AND IMPLICATIONS FOR SEED COLLECTION

Our results evidenced that mating was not random due to self-fertilization, through mating between related trees and correlated mating. Consequently, families comprised different relatedness levels such as self-sibs, half-sibs, full-sibs, and self-half-sibs; eight of them presented inbreeding from selfing, from or from both. Such results are the key to determine the strategies to collect seeds for conservation ex situ, tree genetic breeding and environmental reforestation purposes. The coancestry coefficient at population and individual level, as well as within fruits, was higher due to the mix of relatedness and inbreeding within families; the variance effective size (N_e) within a single family was lower than the expected for panmictic populations fact that resulted in at least 64 seed trees for seed collection. However, as there were variations in the outcrossing rate, in mating between related trees and in correlated mating within fruits. We suggest that seed collection must involve many fruits from each seed tree. Seeds from different seed trees must be mixed, first. Subsequently, they should be mixed with seeds from other trees at the same proportion to maternal gametic controls; therefore, all mothers will contribute with the same number of genes.

Conflicts of interest

The authors declare no conflict of interest.

Acknowledgment

The authors are grateful for the research support granted by Embrapa (Empresa Brasileira de Pesquisa Agropecuária), CNPq (Conselho Nacional de Desenvolvimento Científico e Tecnológico), and FAPEMAT (Fundação de Amparo à Pesquisa do Estado de Mato Grosso). Alexandre M. Sebbenn and Flávio D. Tardin were granted with the research fellowship from CNPq.