Genetic structure among morphotypes of the endangered Brazilian palm Euterpe edulis Mart (Arecaceae)

Abstract Euterpe edulis (Arecaceae) Mart has high ecological and economic importance providing food resources for more than 58 species of birds and 20 species of mammals, including humans. E. edulis is the second most exploited nontimber product from Brazilian Atlantic Forest. Due to overexploitation and destruction of habitats, E. edulis is threatened by extinction. Euterpe edulis populations have large morphological variations, with individuals having green, red, or yellow leaf sheath. However, no study has related phenotypic distinctions between populations and their levels of genetic structure. Thus, this study aimed to evaluate the diversity and genetic structure of different E. edulis morphotypes. We sampled 250 adult individuals in eight populations with the different morphotypes. Using 14 microsatellite markers, we access genetic diversity through population genetic parameters calculated in the GenAlex program and the diveRsity package in R. We used the Wilcoxon test to verify population bottlenecks and the genetic distance of Nei and Bayesian analysis for genetic clusters. The eight populations showed low allele richness, low observed heterozygosity, and high inbreeding values (f). In addition, six of the eight populations experienced genetic bottlenecks, which would partly explain the low genetic diversity in populations. Cluster analysis identified two clusters (K = 2), with green morphotype genetically distinguishing from yellow and red morphotypes. Thus, we show, for the first time, a strong genetic structure among E. edulis morphotypes even for geographically close populations.

. Studies show that this high percentage of endemism is probably related to environmental characteristics, such as soil fertility and climatic factors (Eiserhardt, Svenning, Kissling, & Balslev, 2011;Salm, De Salles, Alonso, & Schuck-Paim, 2007). For the Brazilian Atlantic Forest, 61 Arecaceae species are found, which are in constant risk of being lost due to the exacerbated loss of their natural habitat (Benchimol et al., 2016;Leitman et al., 2015).
Euterpe edulis Mart (Arecaceae) has a great ecological and economic relevance (Carvalho, Galetti, Colevatti, & Jordano, 2016;Elena et al., 2014;Galetti et al., 2013;Reis et al., 2000). In Brazil, this species is primarily widely distributed in the Atlantic Forest but also occurs in the Gallery Forests of the Cerrado (Leitman et al., 2015). Euterpe edulis seeds and fruits are important food resources for about 58 bird species and 20 mammal species playing an important ecological role in maintaining the diversity of frugivores in the Atlantic Forest (Galetti et al., 2013). In economic terms, the species is the second most exploited nontimber product of the Atlantic Forest Brazilian, through the extraction of its apical meristem called heart-of-palm for human consumption (Pizo & Vieira, 2004;Reis et al., 2000;Silva Matos, Freckleton, & Watkinson, 1999). The illegal harvesting of heart-of-palm causes the death of individuals, which is a serious problem, since the species does not sprout or have tillering. Therefore, illegal harvesting is the primary cause leading to the species listing as endangered flora in Brazil (Leitman et al., 2015). In addition to the negative impacts of predatory harvesting, the reduction and fragmentation of the Atlantic Forest further threaten this species (Carvalho, Ribeiro, Côrtes, Galetti, & Collevatti, 2015;Fleury & Galetti, 2006;Santos, Cazetta, Dodonov, Faria, & Gaiotto, 2016).
Despite of the morphological and demographic differences being are well known and described in literature, E. edulis is the only taxonomically accepted name with the others considered synonyms (Leitman et al., 2015).
There is a shortage of genetic studies that associate the phenotypic distinction between natural populations with genetic structure. Currently, it is only known that populations of E. edulis in different environmental conditions are genetically different (Alves-Pereira et al., 2019;Brancalion et al., 2018). However, it is not known whether populations with morphological differences are also genetically different. The central objective of this study is to evaluate the genetic diversity and to verify whether there is genetic differentiation between the different E. edulis morphotypes.
F I G U R E 1 Phenotypic distinction between the (a) green (b) yellow and (c) red morphotypes found in the populations of E. edulis 2 | MATERIAL AND ME THODS

| Sampling areas
The populations studied were chosen based on their distinct morphotypes ( Figure 1 and Table 1) and geographical distances ( Figure 2). Eight populations were sampled; two in gallery forests from the Cerrado biome (Brazilian savanna in Distrito Federal-DF, central west of Brazil) and six from the Atlantic Forest (Bahia, northeastern of Brazil; Figure 2). We did not observe individuals belonging to different morphotypes within the populations.
The two populations sampled in DF are located in gallery forests, which are a type of perennial forest vegetation immersed in savanna formations, which accompany small streams and usually have nutrient-rich soils (Haridasan, 1998). The DF populations occur within protected areas, namely the Brasilia National Park (PNB) and the Roncador Ecological Reserve (IBGE), which are located 10 km and 20 km from Brasilia-DF, respectively.
The sampled populations in Bahia are located in the region with the largest remnants of the Atlantic Forest of northeastern Brazil (Ribeiro, Metzger, Martensen, Ponzoni, & Hirota, 2009) and are considered a priority for conservation due to high biodiversity of this region (Martini, Fiaschi, Amorim, & Paixão, 2007). Of these popu-

| Microsatellite marker analysis
In total, samples were collected from the root tissue of 250 adult individuals through active search (Table 1), in order to cover as much of each area as possible, avoiding the sampling of individuals geographically close to each other, aiming to sample the genetic pool of each collection area. Posteriorly, the DNA was obtained according to Doyle and Doyle (1987), and the genetic material was amplified with a total of 14 nuclear microsatellite fluorescent primer pairs (EE2/ EE5/EE23/EE32/EE9/EE25/EE43/EE45/EE47/EE48/EE52/EE54/EE59/ EE63) developed by Gaiotto, Brondani, and Grattapaglia (2001). with a final elongation performed at 72°C for 7 min. The analysis of amplicons in a denaturing gel (7 mol/L urea) with 4% polyacrylamide (Long Ranger 50%-Cambrex) was performed in a multiplex system, in a semiautomatic ABI 377 sequencer (Applied Biosystems) by using virtual filter D.

| Data analysis
To estimate the frequency of null alleles and to identify possible  (Goudet, 2001) and to calculate the Hardy-Weinberg equilibrium at each locus within each population in the GenAlex 6.5 program (Peakall & Smouse, 2012 (Cornuet & Luikart, 1996) was used to test whether populations have excess heterozygosity (Piry, Luikart, & Cornuet, 1999). The Wilcoxon test is recommended because of its power to detect population bottleneck when few molecular markers (<20 loci) are used, as is the case in our study.
As microsatellites can present different mutational models, we performed the Wilcoxon test considering the infinite allele model (IAM), stepwise mutation model (SMM), and two-phase model (TPM; Piry et al., 1999). For the TPM model, the proportion of SMM in TPM = 0.000, and variance of the geometric distribution for TPM = 0.36, which generally represent the most sensitive values for microsatellite markers (Piry et al., 1999).
To estimate the level of genetic structure among the populations, we calculated the observed F ST values and the 95% confidence interval with 10,000 bootstraps, using the diffCalc function of the diveRsity package in software R (Keenan et al., 2013; http://www.r-proje ct.org/). Subsequently, to assess whether geographical dis- we performed a Mantel test (Mantel, 1967) using the ecodist package (Goslee & Urban, 2007) in the software R (http://www.r-proje ct.org/).

| Genetic diversity and population bottlenecks
Our results did not show evidence of genotypic linkage disequilibrium between loci (results not shown). However, only 36% in average of the 14 loci used present expected genotypic frequencies at Hardy-Weinberg equilibrium (Table S1)

| Genetic structure
The analysis of the genetic structure with F ST revealed that the populations have moderate to high genetic differentiation, though this is lessor for BN-RE populations (Table 2). Although all F ST values are moderate or high, they are within the confidence interval and do not differ statistically from zero ( Table 2). The Mantel test showed that the geographical distance does not correlate significantly with the pattern of genetic differentiation found between populations (r = .13, p = .33, Figure S1).

| Population clustering
The heatmap with clustering using the Nei's genetic distance showed that the eight sampled populations form two clusters ( Figure 3 and Figure S2)   (Figure 1).

| D ISCUSS I ON
In the present study, we recorded low genetic variability and genetic bottleneck effect in Euterpe edulis populations, which also have unusual genetic structure for populations of this species (Gaiotto et al., 2003;Santos et al., 2015). In addition, the evaluated popu-

| Genetic diversity and population bottlenecks
In general, there is a low genetic diversity in the eight populations of E. edulis when considering the richness of alleles compared to other populations of the species, using microsatellite markers (Carvalho et al., 2015(Carvalho et al., , 2017Conte, Sedrez Dos Reis, Mantovani, & Vencovsky, 2008). In theory, several ecological or anthropogenic factors can negatively impact genetic diversity, such as the reduction or inefficiency of dispersers and pollinators, reduction in effective population size, causes related to habitat loss and fragmentation, or even illegal extraction of species (Browne, Ottewell, & Karubian, 2015;Carvalho et al., 2016;Chung et al., 2014;da Carvalho et al., 2017;Ellegren & Galtier, 2016;Young, Boyle, & Brown, 1996). Even in this context, all populations had a considerable proportion of private alleles, demonstrating the importance of these areas for the conservation of this endangered species.
Considering the observed heterogeneity (H O ), six populations have low or moderate values and lower than those reported in other populations of the species (Carvalho et al., 2015). In addition, expected heterozygosity (H E ) values are moderate to high and always higher than H O values in all populations, indicating deviation from that expected under drift-mutation equilibrium (see Table S1) in an idealized, panmictic population (in the absence of evolutionary forces). As a consequence, all populations have positive and high values of inbreeding (f), indicating nonrandom mating occurring, which may cause reduction of genetic diversity (Xue et al., 2015). In this scenario, the genetic erosion may arise and even increase over generations, as individuals would become more related to each other as result of nonrandom mating within populations (Xue et al., 2015).
On the other hand, the populations RE and BN have the highest genetic diversity (H O , H E ) values and the lowest inbreeding measures, indicating the best genetic conservation status among the studied populations. These values are probably influenced by the occurrence of gene flow between RE and BN, which may favor the maintenance of genetic variability over time (Gaiotto et al., 2003;Santos et al., 2016).
The Wilcoxon test, with the infinite allele mutational model (IAM), revealed that six of the eight populations studied have recently experienced genetic bottlenecks. However, the two-phase model (TPM) and stepwise mutation model (SMM) had divergent results, not detecting genetic bottleneck in any of the evaluated populations. This result was expected to some extent because, although microsatellite markers may adhere to different mutational models (AIM, TPM, or SMM), genetic bottlenecks have been reported mainly for the AIM model (Piry et al., 1999;Santos et al., 2019). The genetic bottleneck or excess of H E occurs when the population has experienced a recent reduction in effective population size, causing a reduction in the number of alleles faster than in H E (Cornuet & Luikart, 1996;Piry et al., 1999). The genetic bottleneck identified in most of the populations evaluated in this study would explain the low allelic richness. Even if some population did not go through a genetic bottleneck (e.g., BR and EV), there is evidence of genetic structure (Figures 3 and 4), which could potentially lead to nonrandom crossing within these areas due to isolation (Mosca, González-Martíınez, & Neale, 2014;Sexton, Hangartner, & Hoffmann, 2014).
Thus, we believe that both the genetic bottleneck and the degree of  (Conte et al., 2008;Santos et al., 2015). Thus, we believe that the differences found between the populations of this palm tree should not be related to habitat fragmentation, but to naturally occurring evolutionary events, such as local adaptation (Brancalion et al., 2018 Wright, 1943). Thus, we emphasize that the pattern of genetic differentiation found between populations is probably influenced mainly by morphological differences, as demonstrated in the cluster analysis.

| Population clustering
Although F ST values were moderate to high between the sampled populations, clustering using Nei genetic distance and Bayesian analysis identified two genetic groups. In the Bayesian approach, F ST value greater than 0.20 can allow for more accurate cluster identification and correctly attribute individuals to a true cluster (Latch, Dharmarajan, Glaubitz, & OER, 2006). Thus, the reported F ST values probably contributed to a greater inference power of the genetic clusters.
The grouping into two clusters (Figures 3 and 4) indicates a genetic ancestry between populations with green morphotypes and that it is genetically distinguished from yellow and red morphotypes. Thus, our results are the first to show that there is genetic divergence within morphotypes of E. edulis rather than phenotypic plasticity, possibly due to a response in environmental or ecological differences (Funk & Murphy, 2010). The grouping of the yellow morphotype with the red indicates that this morphological differentiation is probably more recent compared to the green morphotype.
In this context, considering that there are proposals in the literature to taxonomically separate the morphotypes: green sheath as

| CON CLUS ION
The genetic grouping of yellow and red morphotypes that are only restricted to the state of Bahia and Espírito Santo in Brazil is genetically different from the green morphotype of E. edulis that have occurrence areas in across the entire distribution of the species. The relationship between phenotypic and genetic variation is an important advance in scientific knowledge for future conservation measures for this endangered species. In consideration of the low genetic variability reported in this study, the development of conservation plans is highly recommended, particularly for populations with yellow or red morphotypes that have limited geographical occurrence.

ACK N OWLED G M ENTS
This study was financed in part by the Coordenação de

CO N FLI C T O F I NTE R E S T S
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
Relationship between pairwise F ST and geographical distances ( Figure S1), estimated number of groups using the average method ( Figure S2), variation of the second order of the average values of maximum likelihood ( Figure S3) uploaded as online Supplementary material. In addition, the raw data were deposited in the Dryad: https://doi.org/10.5061/dryad.547d7 wm5g.