INTRODUCTION

Platonia insignis, popularly known as bacurizeiro, is a deciduous tree of the family Clusiaceae native to the Amazon that produces one of the most widely appreciated fruits in the region, the bacuri. It occurs mainly in the states of Pará, Maranhão and Piauí, in northern Brazil (Nascimento et al. 2007). In Pará, it is often found on Marajó Island and in the northeastern part of the state (^{Carvalho 2007}). Its sweet pulp is used in many processed forms, such as desserts, jams, liqueurs and ice cream. Because of its taste and texture, it has been gaining attention in gastronomy beyond the frontiers of the northern and northeastern Brazil.

Platonia insignis trees can reach up to 30 meters. They have very efficient asexual reproduction, generating new buds from the roots. In deforested areas in regions of common occurrence of P. insignis, buds are frequently observed shooting from the ground. Besides asexual reproduction, P. insignis reproduces by seeds, via cross pollination, associated with sporophytic self-incompatibility (^{Maués and Venturieri 1996}).

Since P. insignis has great economic potential in agroforestry, it is necessary to select more adapted and productive genotypes. In this context, the conservation of the species in active germplasm banks is of great importance, to maintain genetic variation and provide material for genetic breeding programs. In Brazil there are germplasm banks of P. insignis in the states of Pará and Piauí. Each one is represented mainly by accessions collected in these states or nearby regions. Some efforts to characterize the conserved germplasm have been carried out by ^{Carvalho et al. (2002}, 2003, 2004) and ^{Souza et al. (2016}). The germplasm bank accessions of Piauí were molecularly characterized with inter simple sequence repeat (ISSR) markers, identifying genetic differentiation among sampling localities in the states of Maranhão and Piauí (Souza et al. 2013). However, studies of the genetic variability of accessions from Pará state are still lacking.

For plant species with no genome sequence available, the use of inter simple sequence repeat (ISSR) molecular markers is a good option (^{Faleiro 2007}). They are dominant and represent genetic variations within microsatellite regions in the genome, detected with random primers (Faleiro 2007). They have been used to estimate genetic diversity and genetic parameters of populations of other Brazilian native fruit species, such as Rollinia mucosa (^{Lorenzoni et al. 2014}) and Genipa americana (^{Silva et al. 2014}), being able to identify considerable genetic variability and divergent genotypes. An additional advantage of ISSR markers is the possibility of obtaining a high number of polymorphic loci without the need for sequence knowledge (Faleiro 2007), as is the case of P. insignis.

The set of accessions evaluated in this study is part of the germplasm bank formed for the selection of superior clones of P. insignis (^{Carvalho et al. 2002}) and identification of morphological variants (Carvalho et al. 2003). Thus, the aim of this study was to estimate the genetic variability and genetic structure of accessions of P. insignis from Marajó Island in Pará, preserved in the germplasm bank of Embrapa Eastern Amazon using ISSR markers.

MATERIALS AND METHODS

To estimate the genetic variability of P. insignis conserved in the germplasm bank of Embrapa Eastern Amazon, we selected 78 accessions belonging to 16 progenies collected on Marajó Island, Pará (^{Guimarães et al. 1992}). These progenies represent open-pollinated families, and were sampled in two localities of Marajó Island: Soure and Salvaterra (Figure 1, Table 1). The accessions were established in the Quatro Bocas Experimental Field of Embrapa Eastern Amazon, in Tomé-Açu, Pará. Four to five plants per progeny were collected (Table 1).

Figure 1 Map of sampling localities of accessions of bacurizeiro (Platonia insignis) on the Marajó Island, Pará, Brazil. 

Total genomic DNA was extracted according to a procedure similar to that of ^{Doyle and Doyle (1990}). Leaves were macerated with liquid nitrogen, and then polyvinylpyrrolidone (PVP) and 3 mL of cetyl trimethylammonium bromide (CTAB) extraction buffer (2% CTAB, 5 M NaCl, 0.5 M EDTA, PVP, 1 M Tris-HCl, and sterile water) were added to the macerate. The macerate was homogenized and incubated in a hot water bath at 65°C for 1 h. Afterwards, chloroform:isoamyl alcohol (24:1) was added followed by homogenization, and the samples were centrifuged for 10 min at 10,000 rpm. Three milliliters of 95% ethyl alcohol were added to the supernatant to precipitate the DNA, and the samples were again centrifuged for 10 min at 10,000 rpm. Next, the precipitate was washed with 70% ethyl alcohol for 10 min and centrifuged at 5,000 rpm. DNA samples were resuspended in 300 μL of TE buffer (10 mM Tris-HCl, 1 mM EDTA, pH 8.0) and RNAse. DNA was quantified on 1% agarose gel using lambda phage DNA as a standard, at different concentrations (50, 100 and 200 ng μL^-1).

Samples were genotyped with 23 ISSR primers (University of British Columbia, Vancouver, Canada). Four randomly selected accessions were used to test and select annealing temperatures for each primer (Table 2). PCR was performed in a final volume of 20 μL, containing 10 ng of genomic DNA, 75 µM of each dNTP, 2.0 µM of primer, 1.0 mg mL^-1 BSA (bovine serum albumin), reaction buffer containing 1.2 mM MgCl₂ and 0.2 U Taq DNA polymerase (Invitrogen, Brazil). Reactions were carried out in 0.2 mL microtubes and amplified in an Amplitherm TX96 thermocycler programmed for 35 cycles. First, there was a denaturation phase at 95 °C for 5 min. Then, each cycle consisted of DNA denaturation at 95 °C for 1 min, primer annealing at temperatures from 50 - 62 °C (depending on the primer, Table 2) for 45 s and elongation at 72 °C for 2 min. After the 35 cycles, there was final extension at 72 °C for 5 min.

Reaction products were run on 1.5% agarose gel (Invitrogen, Brazil) prepared with 1.0X TBE buffer (0.45 M Tris-borate and 0.01 M EDTA). Gels were run in a horizontal electrophoresis unit containing 1.0X TBE at constant voltage of 80 V for 3:30h. Gels were visualized with an ultraviolet light transilluminator and images were digitally captured. Bands with the same run pattern were considered from the same locus, and the presence of a band was scored as (1) and absence as (0), generating a binary matrix. The fragments were compared with the molecular marker 1Kb DNA ladder (Invitrogen). Only polymorphic bands were analyzed.

The matrix of genetic similarity was generated with the PAST program (^{Hammer et al. 2001}) based on Jaccard’s coefficient:

sgij=aa+b+c

Where:

a = number of events where the band occurred in both genotypes;

b = number of events where the band occurred only in genotype i;

c = number of events where the band occurred only in genotype j.

Based on the genetic similarity matrix, a dendrogram was generated using the unweighted pair group method with arithmetic mean (UPGMA). The relation between similarity matrix and dendrogram was estimated by the cophenetic correlation coefficient (CCC), according to ^{Sokal and Rohlf (1962}).

The genetic structure was estimated by analysis of molecular variance - AMOVA (^{Excoffier et al. 1992}) using the GenAlEx 6.501 program (Peakall and Smouse 2012). Two approaches considering two hierarchical levels were used. First the variance within and among the sixteen progenies was analyzed. Then, partition of variance was analyzed based on the sampling localities of the progenies (Salvaterra and Soure).

RESULTS

We amplified 121 products with the 23 primers used, with an average of 5.0 bands per primer (Table 2). Among the 121 amplified products, 54 were polymorphic, which corresponds to a polymorphism rate of 44.62%, with an average of 2.35 polymorphic bands per primer. The most polymorphic primers were UBC 834, UBC 899 and UBC 900 (five polymorphic bands each). All bands amplified by UBC 899 were polymorphic. On the other hand, UBC 810 and UBC 884 did not amplify polymorphic bands. Due to the lower quality data, five polymorphic bands (one each amplified by UBC 808, 825 and 900 and two amplified by UBC 856) were discarded, so further analyses were performed with 49 polymorphic bands.

The genetic similarity based on Jaccard’s coefficient varied from 0.51 to 0.98, with an average of 0.79. The least similar accessions were 213-2 and 101-2 (gs_ij = 0.51) and the most similar pairs were 209-2 and 102-3 and 209-2 and 104-4 (gs_ij = 0.98). Despite the high genetic similarity among 209-2, 102-3 and 104-4, these three belong to different progenies. Also, these accessions were sampled in different localities (Table 1).

The dendrogram formed by UPGMA and Jaccard’s genetic similarities among the 78 accessions showed no clustering according to progenies or sampling localities (Figure 2). Accessions 101-1, 101-2, 103-5, 216-1 and 103-4 were the most divergent and clustered separately from the other accessions.

Figure 2 Cluster analysis of 78 accessions of bacurizeiro (Platonia insignis) based on 49 ISSR markers. The dendrogram was generated using UPGMA based on the similarity coefficient of Jaccard. Cophenetic correlation coefficient = 0.80.  

Based on AMOVA, there was a significant genetic differentiation among progenies (Φ_PT = 0.064, P<0.001). We found that 6% of total variation was among progenies and 94% was within progenies of P. insignis (Table 3). When AMOVA was performed considering partition of variance between sampling localities, there was low but significant genetic variation between Soure and Salvaterra (Φ_PT = 0.02, P<0.011). Genetic variation among sampling localities was 2% and within 98% (Table 3).

DISCUSSION

Despite the strong potential of P. insignis for commercial fruit production, there are few studies about the genetic variation of this species, and fewer considering molecular aspects (^{Almeida et al. 2007}; ^{Souza et al. 2013}). In this study, we characterized 78 accessions of P. insignis with ISSR markers, which employ universal primers and can be used to study species with no genomic information.

The mean genetic similarity among accessions could be considered high (x̅ = 0.79) compared to other studies of genetic diversity of Brazilian fruit species (^{Santana et al. 2011}; ^{Lorenzoni et al. 2014}). Besides this, the percentage of polymorphic loci was 42%, even though a high number of ISSR primers were used (23). In a previous study to evaluate the genetic diversity of 28 accessions of P. insignis from the Pará germplasm bank, high similarity levels among some accessions were detected with RAPD markers (^{Almeida et al. 2007}). ^{Souza et al. (2013}) analyzed 72 accessions of P. insignis from a germplasm collection formed by accessions from northeastern Brazil using 18 ISSR primers and obtained 221 polymorphic loci and mean genetic similarity of 0.52. Those authors analyzed samples from ten different localities in the states of Maranhão and Piauí, which may have contributed to the higher detection of genetic variation.

The analysis of molecular variance showed that the highest amount of genetic variation was contained within progenies or sampling localities (Table 3), which was expected for allogamous species and explained by the higher efforts to collect samples within places. However, the work of ^{Souza et al. (2013}) showed that higher variation can be obtained with samples in a wider geographical range, which can enrich genetic breeding programs. Perhaps the low genetic differentiation between sampling localities was an effect of the geographical proximity between Soure and Salvaterra, since higher values of genetic differentiation were detected by Souza et al. (2013). The effect of sampling of P. insignis in different localities was observed in morphological and chemical variation of fruits (^{Silva et al. 2009}; ^{Carvalho-Saraiva et al. 2014}).

We observed high genetic similarity among accessions from different progenies and sampling localities. On the other hand, the least similar accessions were from different sampling localities, besides belonging to different progenies. This can be an effect of pollen dispersion from different origins, since these progenies are half-sib families. The clustering of accessions in the dendrogram was not associated with the progeny or sampling localities, which was reflected in the AMOVA analyses. Cophenetic correlation of the dendrogram with the similarity matrix was r = 0.88, which is high and confirms the reliability of the results. Considering genetic partition within and among sampling units, ^{Souza et al. (2013}) also identified higher genetic variation within populations from Maranhão and Piauí (71.82%), but genetic differentiation among populations was higher (28.18%) than in this study. This might be an effect of collecting samples from more distant areas. Since P. insignis is an allogamous species with sporophytic self-incompatibility (^{Maués and Venturieri 1996}) the higher portion of genetic variation within populations or progenies was expected. Again, the geographical proximity likely favoured a more frequent gene flow among trees of our two sampling localities.

CONCLUSIONS

The genetic variation of Platonia insignis from two localities on the Marajó Island (Pará state, northeastern Amazon) was higher within than among progenies and sampling localities, which means that sampling efforts for germplasm enrichment should consider a higher sampling effort within localities. The low genetic differentiation between geographically close sampling places probably was a result of the allogamous behavior of P. insignis trees.