Introduction

Habitat loss, overexploitation of natural resources, invasion by alien species, and global climate change are among the most important factors that threaten biodiversity, and can synergistically act to increase species extinction risk in the future1,2,3. Information used to decide if a species is at risk of extinction, and what threatened category it falls into is generally based on ecological and demographic data such as the number of known individuals, actual or projected declines in population size, and the extent of occurrence (EOO) or area of occupancy (AOO)4. The EOO is an estimate of the total area currently occupied by a taxon and the risks within this area4. The AOO is a metric of current occupation of suitable habitats within the EOO, and measures the risks to a taxon occurring within small patches or in few patches within the EOO4. Demographic information is fundamental in conservation biology, but in several species, it is difficult to directly estimate.

There is a consensus among conservation biologists as to the importance of genetic factors in determining the fates of populations or species, but these factors are not generally used to define the conservation status of a species5,6. The “genetic health” of a population or species is dependent on its genetic diversity and inbreeding level. Rare and endangered species tend to have reduced genetic diversity within populations and low gene flow rates among them, because in general the populations are small and highly disjointed7. Furthermore, habitat loss and fragmentation processes can lead to the disruption of pollination and dispersion, resulting in low reproductive success and making populations more isolated and prone to inbreeding depression, i.e., a reduction in the capacity for survival and reproduction of the offspring of inbreeders5,8.

The genetic diversity and degree of inbreeding in a population are directly dependent on the effective population size (NE), a fundamental parameter for the conservation of threatened species9,10,11. The NE of an actual population can be defined as the size of an ideal population that suffers the same magnitude of genetic drift as the actual population; this genetic drift is generally measured as a loss of heterozygosity, increase of identity by descent, or change in allele frequency over time12. The difficulty and impracticality of directly counting individuals in most species, coupled with the importance of genetic parameters in putatively threatened species, led to the use of NE in place of census population size in species conservation. Furthermore, several population genetic tools based on coalescent theory are currently available for estimating the temporal dynamics of NE and detecting population genetic structure and migration rates among populations13,14,15. Information about these genetic parameters could be used to gain knowledge about how global habitat loss, fragmentation and novel environment will affect a species and thus help to optimize its conservation16,17,18,19,20. These genetic studies are even more important in poorly known species, as they could help to infer a species’ conservation status and aid in making decisions about its conservation.

In this study, we investigated the poorly known legume tree Dimorphandra exaltata Schott (Fabaceae, Caesalpinioideae). Although D. exaltata is the type species of Dimorphandra as described by Schott in 182721, there are only a few herbarium records (Fig. 1) and almost no studies have been performed with it22. The species has a discontinuous distribution in semi-deciduous forests in the Atlantic Forest biome and Atlantic Forest/Cerrado transition areas in southeastern Brazil (Fig. 1). Atlantic Forest and Cerrado (a savanna biome) are hotspots for biodiversity conservation due to the high degree of threats, species richness, and endemism, with each biome showing about 8 000 and 4 400 endemic plant species, respectively23. After five centuries of human exploitation since the Portuguese colonization of Brazil, only 12–15% of the Atlantic Forest remains, in small and isolated fragments24. Deforestation and fragmentation in Cerrado is more recent, but has occurred in high rates; from 1970 to 2005 about 50% of the biome land cover was transformed into pasture and agricultural land25. Dimorphandra exaltata has a distribution partially sympatric with that of the critically endangered species D. wilsonii26. These areas have been highly disturbed by anthropogenic activities, mainly agriculture and urbanization27.

Figure 1
figure 1

Maps showing the Dimorphandra exaltata herbarium records and the populations used for genetic analyses. (A) Map showing the land use of eastern Brazil (ESA/CCI viewer, 2018), MCV = mosaic cropland (50–70%)/vegetation (20–50%), MVC = mosaic vegetation (50–70%)/cropland (20–50%), SDF = evergreen or semi-deciduous forest (>5 m), DEF = deciduous forest (>5 m), SRH = herbaceous vegetation, AUB = Urban areas > 50%. (B) Map showing the Brazilian biomes where Dimorphandra exaltata occurs.

Full size image

Here, we performed population genetic analyses to assess the genetic status of D. exaltata, using nuclear microsatellite markers. We have genotyped a total of 62 individuals from: two populations in transition areas of Cerrado/Atlantic Forest, one population from semi-deciduous Atlantic Forest and seven isolated individuals from three sites in eastern Atlantic Forest (Fig. 1; Table 1). We estimated the genetic diversity and evaluated genetic structure with Bayesian clustering method, as well as the genetic connectivity among extant populations and historical NE using coalescent methods. We looked for genetic evidence of decreases in population size and evaluated the contemporary NE through of molecular coancestry. Further, we compared the genetic status of D. exaltata with its conservation status evaluated using geospatial assessment based on herbarium records, with estimates of EOO and AOO. We also estimated its current and near-future (2050) potential geographical distribution to investigate current habitat suitability and project the impact of climate change on the species. Based on the genetic and ecological data, we suggest conservation measures for the species.

Table 1 Population genetic diversity parameters of Dimorphandra exaltata.
Full size table

Results

Genetic status: genetic diversity and structure, migration, and effective population size

Of the 11 microsatellites loci used, only Dw103 for CON population and Dw28 for all populations showed null alleles, while scoring errors due to stuttering or large allele dropout were not found. However, as the genetic diversity estimates were similar between the original dataset and the corrected for null alleles dataset using the Brookfield 1 method in MicroChecker version 2.2.0.228, we considered the estimates without null allele correction. No significant linkage disequilibrium between loci was observed in any comparison after Bonferroni correction. To estimate the genetic diversity and structure of D. exaltata we also included the seven isolated individuals of the Atlantic Forest in addition to the three sampled populations. These individuals from three sites with distances ranging from 31.5 to 95.7 km were analysed together (AFI group; Table 1) to represent the diversity of eastern Atlantic Forest, although they probably do not represent samples of only one population. For the remaining genetic analyses, the AFI group was not included.

Among the three populations and the AFI group, the mean A and AR ranged from 3.091 to 4.545 and from 2.348 to 3.419, respectively. HO and HE per population ranged from 0.320 to 0.481 and from 0.289 to 0.560, respectively, with mean values of HO and HE of 0.372 and 0.413, respectively (Table 1). Populations CON and SER showed the higher values of private alleles with 20 and 16 alleles, respectively. In addition, nine private alleles were found in the isolated individuals from Atlantic Forest, despite the small sample size (Table 1). The exact test showed that two of the three populations, SER and CON, showed significant positives inbreeding coefficients (FIS) after Bonferroni correction, with estimated values of 0.195 and 0.178 (P < 0.05), indicating deviation from Hardy-Weinberg equilibrium in these populations (Table 1).

AMOVA estimated FST of 0.180 (P = 0.000) and RST of 0.120 (P = 0.000), indicating moderate genetic divergence among populations (Table 2). According to pairwise comparisons, AFI followed by ESM population were the most divergent (Supplementary Table 1). The best number of clusters estimated by Bayesian clustering (Structure), considering both ΔK and mean likelihood methods, was K = 4 with each population been represented mostly by one genetic ancestry and varying admixture levels among them (Supplementary Fig. 1; Fig. 2). The CON followed by SER population showed higher admixture levels in its individuals (68% and 65% of individuals with Q < 0.9 for its most representative cluster, respectively), and AFI the lowest levels of admixture in the relation to other populations (Fig. 2). In addition, the Bayesian clustering showed that SER and CON were more related with each other, followed by ESM. AFI was the most divergent from other populations. These relationships are identified with K = 4 groups, but also when is considered grouping using K = 3, K = 5 and K = 6 genetic ancestries (Supplementary Fig. 2).

Table 2 Analysis of molecular variance of Dimorphandra exaltata based on FST and RST. Significant values (P < 0.05) are in bold.
Full size table
Figure 2
figure 2

Bar plot showing the membership coefficients according to the Bayesian clustering method (STRUCTURE) for individuals of three Dimorphandra exaltata populations and seven isolated individuals (AFI) based on the best number of K = 4 genetic clusters.

Full size image

The contemporary NE estimated for the three populations (SER, CON, ESM) ranged from 13.5 to 22.7, indicating low number of reproductive individuals (Table 3). The bottleneck test showed significant deviation of mutation-drift equilibrium in the three mutation models tested for all three populations (P < 0.05; Table 3), indicating that they suffered recent reductions in population size. The best historical migration model among the three populations was the full matrix migration (Supplementary Table 2). The values of the effective number of migrants (NEm) ranged from 0.086 from SER to ESM to 5.653 from ESM to SER (Supplementary Table 3). The estimated historical NEwere 481, 760, and 11 for SER, CON, and ESM, respectively (Supplementary Table 3).

Table 3 Contemporary effective population sizes (NE) and results of tests for recent bottlenecks in Dimorphandra exaltata populations.
Full size table

Species distribution models

Of the 19 bioclimatic variables from the WorldClim29, the following variables were retained to performing species distribution models after the factor analyses: mean diurnal range (Bio2), isothermality (Bio3), mean temperature of warmest quarter (Bio10), precipitation of wettest quarter (Bio16), and precipitation of driest quarter (Bio17). The values of TSS and AUC for model evaluation ranged from 0.333 to 1.000 and 0.611 to 1.000, respectively (Supplementary Tables 4 and 5). The variables that most contributed to the models were Bio3, Bio10, and Bio17 (Supplementary Table 6). The SDM projections confirmed the pattern of occurrence for D. exaltata, showing higher suitability in the semi-deciduous Atlantic Forest and in the Atlantic Forest/Cerrado transitional zone in southeastern Brazil, mainly in Minas Gerais, with smaller areas in São Paulo and Rio de Janeiro states (Fig. 3). The projections of both models for 2050 showed a loss in suitable area, mainly in the northern and northeastern parts of the range, which was more accentuated in the less-conservative BCC-CSM1-1 model (Fig. 3). In the most conservative scenario of climate change, MIROC-ESM, we observed a displacement of suitability toward the south, the opposite to a decrease in suitability in the southern area in the less-conservative model. There was also a decrease in suitability in the Atlantic Forest/Cerrado transitional zone, the area of most of the registered occurrences of the species (Fig. 3).

Figure 3
figure 3

Ensemble species distribution model showing projections of suitable areas for Dimorphandra exaltata: (A) in the current time and projected into the near future (2050) based on the (B) BCC-CSM1-1 and (C) MIROC-ESM general circulation models of climate change.

Full size image

Geospatial conservation assessment

Using the Geospatial Conservation Assessment Tool software30 (GeoCAT) based on 19 herbarium records to evaluating the conservation status of D. exaltata, we estimated an extent of occurrence (EOO) of 263 905 km2 and a much smaller area of occupancy (AOO) of 76 km2 for the species. Due to the possibility of misidentification in the only record for D. exaltata in speciesLink at Jequié, Bahia state, and because we could not find any other description for the species in Bahia in specialized literature21,31,32, we also performed the EOO and AOO analyses without this record. The estimates of EOO and AOO without Jequié were 174 936 km2 and 72 km2, respectively, similar to estimates with the complete dataset. Based on the B criteria of the IUCN, the species was classified as Low Concern according to the EOO (EOO > 20 000 km2) and as Endangered according to the AOO (AOO < 500 km2) in both analyses, with and without Jequié. In addition, following the B2 criteria, the species should also be considered Endangered because it meets the two following conditions: (a) severely fragmented populations and (b) continuous decline in area as indicated by (bi) projected EOO and (biii) continuing decline in area, extent, and/or quality of habitat.

Discussion

Our study showed that D. exaltata exhibits low genetic diversity, inbreeding, and evidence of a recent decrease in population size. Furthermore, species distribution models project a reduction in range size for the near future (2050). According to the geospatial criteria of the IUCN (2014), the species falls into the Endangered category of extinction risk. Altogether, our data indicate clearly that D. exaltata is a threatened species of the Atlantic Forest. The contribution of genetic data, in addition to ecological data, to establishing the extinction risk of the species and the importance of obtaining both types of data to establish measures of conservation are discussed below.

Dimorphandra exaltata populations have lower genetic diversity (HE = 0.289–0.560) than populations of several other legume tree species from the Atlantic Forest, including Hymenaea courbaril (HE = 0.536–0.804)33, Centrolobium tomentosum (HE = 0.500–0.550)34, Inga vera (HE = 0.655–0.865)35, Copaifera langsdorffii (HE = 0.881–0.893)36,37,38, and the vulnerable Dalbergia nigra (HE = 0.682–0.798)39,40,41, all species evaluated with microsatellite markers. Another appreciation of the level of genetic diversity of D. exaltata could be based on a comparison with closely related species, which discounts the phylogenetic effect42,43. Dimorphandra exaltata populations exhibited considerably lower genetic diversity (mean HE per population of 0.413) than a widespread congeneric species, D. mollis, which shows a mean HE per population of 0.59444, as expected due to its more restricted distribution. Dimorphandra exaltata populations showed even lower diversity than the narrowly endemic and critically endangered congeneric species D. wilsonii, with a mean HE per population of 0.55927. Although our genetic diversity evaluation was limited to 62 individuals, from three populations and isolated individuals from Atlantic Forest due to the rarity of the species, the low genetic diversity found indicates that D. exaltata is genetically depleted.

The contemporary NE of D. exaltata populations ranged from 13.5 to 22.7 and the estimated historical NE ranged from 11 to 700, which points to a low NE for the species. The comparison between the two NE together with the bottleneck tests point to a severe decrease in the population sizes of the species, what is probably associated with habitat loss and fragmentation in the study area. This pattern can explain the low diversity found in D. exaltata, since NE is one of the most relevant parameters to explain the patterns of genetic variation in wild populations11. These results are very important for the management of endangered species since NE of 100 and 1 000 individuals are suggested to prevent loss of genetic diversity by genetic drift for five generations and to maintain long-term evolutionary potential, respectively45.

The population genetic structure of D. exaltata suggests considerable admixture among populations, except for the isolated individuals of the Atlantic Forest. The historical demographic analysis reinforces that the three populations were genetically connected in the past. However, we should take care in extrapolating this connectivity to the current time. In fact, during our fieldwork, we observed a very low number of individuals and they were isolated and separated by large spatial distances, mainly in eastern Atlantic Forest. Also, the land use maps showed great anthropic pressure in the areas of occurrence of the species46 (Fig. 1), which results in fragmented habitats and isolated populations. As consequence of this fragmentation, the gene flow of D. exaltata is likely compromised by a low-quality matrix among fragments. The seed dispersion of the species can be affected by a shortage of dispersers. Two species of the genus, D. mollis and D. wilsonii, exhibit low seed dispersal, which is mainly performed by tapirs (Tapirus terrestris), which are vulnerable and locally extinct in some areas26. Although the seed disperser of D. exaltata is not known, the similarity among species of the genus suggests a similar pattern for it. This is a matter of concern because at least 1.0 migrant per generation is necessary to counteract the genetic drift effects in the loss of diversity of populations47.

Our populations showed significant inbreeding coefficients, indicating a departure from random mating within populations. Although the mating system of D. exaltata is not known, it is probably similar to that of the congeneric D. wilsonii, which has self-compatibility26 but mainly performs outcrossing48. The inbreeding found in D. exaltata may be due to selfing or mating between related individuals (biparental inbreeding). The small population size could favor mating among relatives or selfing in opposition to outcrossing. Increased selfing in preferentially outcrossing species is commonly found after fragmentation5. Furthermore, perennial plants, that have recently undergone habitat loss and fragmentation, such as D. exaltata, have a high probability of suffering inbreeding depression8. Therefore, inbreeding might be an additional threat to D. exaltata.

It is well-established that species with isolated and small populations are subject to loss of genetic diversity, a higher degree of inbreeding, and an elevated extinction risk in the wild45. Thus, the genetic data gathered here for D. exaltata, which showed low genetic diversity, inbreeding, and a recent reduction in population size, indicate that the “genetic health” of the species is compromised, which may decrease its adaptive capacity and threaten its survival in the wild in the near future. Thus, the genetic analyses of the species itself allow us to classify D. exaltata as a threatened species. In addition, the genetic data suggest that the species is more threatened than indicated by geospatial analyses.

The projected distribution for D. exaltata in the near future (2050) suggests a decrease in the total area of occurrence, including in the area in which it is most found currently, along with a displacement toward the south. The climate is considered to be the main driver of the species’ range extent at large scales and of its long-term dynamics44,49,50,51,52,53. In face of climate change, the probability of survival of a species is dependent on its capacity to adapt to the new environmental conditions or to migrate to suitable habitats54. However, the genetic data showed that the adaptive potential of D. exaltata is probably compromised. Thus, the species may depend on its migration capacity to colonize new favorable areas to survive in the face of climate change. However, the current connectivity among populations of the species is likely compromised by both low seed dispersal and pollen flow. Thus, the projected reduction of areas of occurrence and their displacement may be a concern for D. exaltata.

The geospatial assessment analyses showed that D. exaltata should be classified as Endangered, according the B2 criteria4 (AOO less than 500 km2). Furthermore, it could be also considered to be in the Threatened category because the populations are severely fragmented and there is a projected decline of this area, along with the low quality of its habitat (the sub criteria of B2). The Atlantic Forest has a fragmented distribution, with a history of disturbance and low functional connectivity among fragments24. Furthermore, the D. exaltata areas of occurrence in the Atlantic Forest/Cerrado transition zone, where the species is partially sympatric with D. wilsonii, are dominated by anthropogenic landscapes due to conversion of forested areas to agricultural lands, urbanization, and fire. These are the main threat factors for D. wilsonii27 and may also play this role for D. exaltata.

Although D. exaltata occurs in one of the most-inventoried regions of southeastern Brazil, only 19 occurrence records were found, demonstrating the rarity of the species. In several areas of Atlantic Forest/Cerrado transition, D. exaltata occurs together with D. wilsonii, a critically endangered species, which has a National Action Plan for its conservation27 (PAN faveiro de Wilson). It is highly recommended that this Action Plan also include D. exaltata, as the two species share similar ecological requirements and the same threat factors. Furthermore, only three records for D. exaltata were within protected areas. Protecting areas where the two Dimorphandra species co-occur will lead to protection of both threatened trees.

The genetic analyses indicated that the main risk factors to wild persistence of D. exaltata are its low genetic diversity and its low effective population size. Considering the ecological and genetic risk factors, conservation actions are urgently needed to decrease the probability of species extinction. Firstly, we recommend more search to find new individuals/populations of the species, mainly in easternmost areas Atlantic Forest. In addition, we recommend reintroduction, ex situ conservation, and genetic rescue to increase genetic diversity and maintain genetic connectivity among its populations. Genetic rescue is a highly beneficial strategy for management of genetically depleted populations55. For this, the creation and permanent maintenance of ex situ production of saplings from seeds of trees from different populations is a priority. As the analysed populations are not very genetically divergent, saplings from different populations could be introduced into the same locality, which could increase the genetic diversity and decrease the rate of inbreeding. However, considering the high genetic divergence of individuals from the eastern Atlantic Forest in relation to other areas, we suggested reintroduction in the eastern Atlantic Forest preferentially with saplings produced by seeds from these locations. The saplings should be used to increase the sizes of current populations, which must be protected, and to reintroduce the species into suitable areas according to SDM projections. It should be noted that the current suitable areas will suffer significant reduction in even the most optimistic projection of climate changes. Thus, it is recommended to take in account this projection when choosing localities for reintroduction.

Conclusion

Our integrative genetic and ecological framework to assess the degree to which a poorly known tree species is threatened and to guide conservation measures for that tree species, which occurs in the Atlantic Forest, a large and highly threatened biome in eastern South America, was demonstrated to be highly effective. This strategy of investigation, which includes genetic data, could be used for species in other conservation hotspots worldwide, where generally there is a lack of demographic information to establish the conservation status of species. From the results, our integrative framework supplies essential information for conservation policies.

Materials and Methods

Study species

Dimorphandra exaltata is a medium to tall tree (15–30 meters), with bipinnate leaves, opposite leaflets pairs, and densely hairy petioles. Its inflorescences have a corymbe-type panicle consisting of multiple spikes. The flowers are cream or yellow with a glabrous chalice. Its fruits are flat legumes, which are almost straight, woody, and dark, indehiscent with sub-cylindrical seeds21. It flowers during November–December and fruit maturation occurs in September–October21. The pollinator and seed disperser of D. exaltata are not known. However, it has been proposed21 that the pollination is performed by insects because of the similarity of its inflorescences and flowers with those of the congeneric species D. vernicosa, which is pollinated by bees of the Apidae family.

Herbarium records and study area

Our study initially evaluated the occurrence of D. exaltata through analyses of herbarium records. Georeferenced herbarium records of D. exaltata were obtained mainly from the speciesLink database56, but to check for duplicates and to add missing records, other databases were used as well21,31,32. Due to the similarities among several species of the genus, the records of D. exaltata were compared with those of other Dimorphandra species to avoid misidentification. We retained only the most recent records by locality to avoid bias due to differential sampling efforts in certain areas and aggregated surveys. The localities were defined as the exact coordinates of the sites available in the record or alternatively, when the coordinate of the site was not available, by the municipalities coordinates. After the filtering process, we obtained only 19 georeferenced records in the speciesLink database (Supplementary Table 7 and Fig. 1), mostly in the semi-deciduous Atlantic Forest and in the Atlantic Forest/Cerrado transition zone. In the transition zone, the occurrences were mainly in cropland or urbanized areas46 (Fig. 1). In addition, most of the peripheral records at the southern and eastern edges of occurrence were from near highly urbanized areas46 (Fig. 1). For species distribution models (SDM) and geospatial conservation assessment analyses, we used the 19 records.

We also made a great effort to collect D. exaltata on Minas Gerais and Rio de Janeiro states, which concentrate the highest number of herbarium records and the most potential distribution areas for the species (Figs 1 and 3). In our collect expeditions in the eastern of Minas Gerais and the Rio de Janeiro, we found only four juveniles in one area and three isolated adults in two distant sites with distances ranging from 31.5 to 95.7 km apart from each other evidencing that the species is rare.

For the genetic analyses, we sampled 55 individuals of D. exaltata from three populations and seven isolated individuals from the three sites in eastern Minas Gerais and Rio de Janeiro, totaling 62 individuals (Fig. 1, Table 1). Two populations are located in Atlantic Forest/Cerrado transition areas, Contagem (CON) and Esmeraldas (ESM), where the species occurs in higher density and the third population is located in Serro (SER) in the semi-deciduous Atlantic Forest (Fig. 1). The isolated individuals of the species were in two sites already recorded in the speciesLink database for the species, Rio de Janeiro (RJA) and Valença (VAL) and in a municipality that do not had register for D. exaltata occurence, Santa Bárbara do Monte Verde (SBM) in the semi-deciduous Atlantic Forest. These areas are highly fragmented and disturbed due to urban pressure and soil conversion to croplands46 (Fig. 1). The seven isolated individuals from RJA, VAL and SBM were used as representatives of the eastern Atlantic Forest, and were analysed together (AFI group), although they probably do not represent samples of only one population. AFI group was used only for estimates of genetic diversity and evaluation of genetic structure of species through of the analysis of molecular variance (AMOVA) and by the Bayesian clustering method implemented in STRUCTURE software57,58. The further genetic investigation was performed using only the populations SER, CON and ESM.

DNA isolation and microsatellite genotyping

Frozen leaf or vascular tissue was used for DNA extraction according to the cetyltrimethylammonium bromide (CTAB) protocol59, slightly modified as suggested by60. Individuals were genotyped for 11 microsatellite loci, five isolated in D. mollis61 (Dmo5, Dmo7, Dmo13, Dmo20, Dmo21) and six isolated in D. wilsonii62 (Dw21, Dw28, Dw33, Dw105, Dw52 and Dw103) Amplifications of all markers were performed according to61. The amplicons were submitted to capillary electrophoresis in an ABI prism 3500xl automated sequencer with GeneScan TM ROX500TM as a standard fragment (Applied Biosystems, Foster City, CA, USA) and genotyped in GeneMapper software Version 5.0 (Applied Biosystems). Scoring errors due to the presence of null alleles, stuttering, or large allele dropout were evaluated in MicroChecker version 2.2.0.228. Linkage disequilibrium for each pair of loci for each population was tested using a randomization-based test in FSTAT version 2.9.3.263.

Genetic status: genetic diversity and structure, migration, and effective population size

The genetic diversity of D. exaltata populations was estimated by the following parameters: mean number of alleles per locus (A), expected and observed heterozygosity (HE and HO respectively) using ARLEQUIN 3.564 and the allelic richness (AR) using the rarefaction method of65 in FSTAT version 2.9.3.263. The number of private alleles were investigated in GenALEx 6.50366,67. The inbreeding coefficient (FIS) was calculated and departures from Hardy-Weinberg expectations were tested with exact tests in FSTAT version 2.9.3.263, with significance assessed after sequential Bonferroni correction for multiple comparisons. The FIS was not estimated for the AFI group because they can represent samples of more than one population. Analyses of molecular variance (AMOVA) implemented in ARLEQUIN 3.564 were used to estimate the hierarchical genetic structure and the pairwise FST and RST among populations.

The number of genetic clusters was evaluated using a Bayesian clustering method implemented in STRUCTURE version 2.3.457,58. Parameter sets assumed an admixture model with correlated allele frequencies among patches. Ten independent simulations were run with 100 000 chains as burn-in and 1 000 000 Markov chain Monte Carlo (MCMC) replicates after burn-in, testing for K = 1–8. To evaluate the best K number, the mean log likelihood of the model through all runs and ΔK statistics were used68. To visualize the individual ancestry proportion, the average of ten runs for K was calculated in CLUMPP version 1.1.269 and the bar plot membership coefficient of each individual was built in DISTRUCT 1.170.

We estimated the contemporary effective populations sizes (NE) using the relation NE = 0.5/Θ71; Θ is the group molecular coancestry for the total number of sampled trees in each population72. The calculation of Θ was made using the equation (1), \(\Theta =\frac{{\sum }_{i=1}^{n}0.5(1+Fi)+{\sum }_{i=1}^{n}{\sum }_{j\ne i}^{n}\theta ij}{{n}^{2}}\)72; Fi is the inbreeding coefficient of the i-th individual, θij is the pairwise kinship between the individual i-th and j-th38 and n is the sample size. Fi and θij were estimated in SPAGeDi 1.1 software73. We evaluated for recent reductions in population size using a test for heterozygosity excess implemented in BOTTLENECK version 1.2.0274. Populations that have suffered a recent size decrease exhibit reductions of allele numbers and heterozygosity. However, reduction in allele diversity is faster than that in heterozygosity, which leads to an excess of heterozygosity under Hardy-Weinberg equilibrium in comparison with the expectations of mutation-drift equilibrium75. Bayesian simulations were carried out using three mutation models for microsatellites, infinite allele models (IAM), stepwise mutation model (SMM), and two-phase mutation model (TPM), with 30% multi-step mutations to avoid biases due to wrong assumptions about mutation models. We ran 1 000 000 MCMC and used the Wilcoxon sign-rank test to analyse the significance of differences in heterozygosities.

To test for models of historical gene flow among populations, we used a coalescent model implemented in MIGRATE-N 3.6.1113. Mutation-scaled effective population size (Θ = 4ΝEμ) and migration rates (M = m/μ), where μ is the mutation rate, were estimated using the Bayesian inference method. We considered μ = 4.76 × 10−3 mutations per locus per generation for the microsatellite loci76. The Brownian motion model for microsatellite data was defined as the mutation process of the markers and the prior distribution was set as uniform with Θ (10−5 to 100) and M (0 to 1 000). We ran 3 iterations with 2 000 000 MCMC after a burn-in of 50 000 generations, with sampling every 100 generations using a static heating scheme with four chains, swamping at every 10 chains based on an exponential temperature scheme. The independent runs were summarized to estimate the parameter values. The following models were tested: panmictic model, full migration model, zero-migration, a model with CON and ESM as one population, and a model with SER and CON as one population. We obtained the effective numbers of immigrants (NEm) using the equation (2) M * Θ = 4NEm, with the number 4 representing a scalar due to ploidy or inheritance of diploid nuclear data. The historical effective population size was estimated through the expression (3) (Θ/4μ) = ΝE.

Species distribution models

To assess the current potential distribution of D. exaltata, species distribution models (SDM) were performed using 19 bioclimatic variables available in the WorldClim database, with resolution of ~1 km29. To obtain a subset of uncorrelated explanatory variables, an exploratory factor analysis (EFA) was used in the psych 1.7.5 package implemented in the R environment77,78. The algorithms Maxent, Generalized Linear Models (GLM), Generalized Boosted Models (GBM), Artificial Neural Network (ANN), Classification tree analysis (CTA), Surface Range Envelopes (SRE), Flexible Discriminant Analysis (FDA), Multiple Adaptive Regression Splines (MARS), and Random Forest (RF) were implemented in the Biomod2 3.3-7 package in the R environment79. To evaluate model performance, we randomly subset points for training (75%) and points to test the predictive power of the models (25%). Then, we used the receiver–operating characteristic (ROC) curve and its area under curve (AUC) and the true skill statistic (TSS) to select the best models. Consensus maps were built based on five independent runs of each algorithm using the TSS as a criterion to discard low-predictive-power models (TSS > 0.7). To evaluate the impacts of future climate change on the distribution of the species, we projected its potential distribution in 2050 using the same set of variables for MIROC-ESM and BCC-CSM1-1 projections in an arc resolution of 30 s representing an intermediary and a maximum representative concentration pathway for this period, respectively.

Geospatial conservation assessment

The geospatial conservation status of D. exaltata was evaluated using the extent of occurrence (EOO) and area of occupancy (AOO), as estimated using the Geospatial Conservation Assessment Tool software30 (GeoCAT). Using this framework, we evaluated the conservation status of D. exaltata based on the International Union for the Conservation of Nature (IUCN) criteria B (EOO and AOO) to identify the threat category of the species4. The estimation of EOO was based on the minimum convex hull that contains all sites of known occurrence30. The estimation of AOO was based on the sum of squared grids of known occurrence based on grids of 2.0 km2 as suggested by IUCN guidelines4,30. The same dataset used in the SDM was used for geospatial analysis.