The role of Cacao agroforests in the genetic conservation of Cariniana legalis, an emblematic species of the atlantic forest

Abstract

Deforestation in the Atlantic Forest has reached critical levels, threatening multiple levels of biodiversity. In these deforested landscapes, conservation strategies could benefit from preserving agroforestry systems known as Cabruca, a traditional method of cultivating cocoa under a canopy of native trees. In this context, Cariniana legalis (Jequitibá-rosa), an endemic tree species of the Atlantic Forest listed as endangered, was selected to evaluate the role of cocoa agroforests (Cabrucas) and forest remnants in the genetic conservation of this species. The study assessed the genetic diversity, inbreeding levels, and genetic structure of five populations of C. legalis located in forest remnants protected by law and in Cabrucas in southern Bahia, Brazil. Using 11 microsatellite loci, 294 individuals were genotyped for adult and juvenile ontogenetic stages. Despite forest fragmentation, some populations, especially those located in Cabrucas, retained high levels of genetic diversity in both stages. In contrast, a protected area exhibited lower genetic diversity and elevated inbreeding levels (f > 0.43) in both adults and juveniles. Analyses of genetic differentiation (F_ST) and migration rate (Nm) indicated reduced historical gene flow in certain populations, while network analysis and Discriminant Analysis of Principal Components (DAPC) identified Cabrucas as central genetic hubs promoting connectivity across the landscape in both ontogenetic stages. Our results highlight the conservation value of cocoa agroforests for maintaining the genetic diversity and connectivity of this endangered tree species. We therefore recommend the inclusion of Cabrucas in integrated genetic conservation strategies for C. legalis in anthropogenically modified Atlantic Forest landscapes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12862-025-02418-3.

Introduction

The deforested of tropical forests caused by human activities alter environmental conditions and the natural dynamics of ecosystems, compromising the survival of native species and threatening global biodiversity [17, 25, 33]. To illustrate the extent of these anthropogenic impacts on biodiversity, it is estimated that between 2000 and 2018, around 90% of global deforestation occurred in tropical forests [15], which are among the most biodiverse ecosystems on the planet. In Brazil, one of the most biodiverse countries in the world, the Atlantic Forest is the most threatened tropical forest. Only about 28% of its original extent remains, encompassing both well-preserved and degraded fragments. However, more conservative estimates suggest that the cover of truly intact forest is substantially lower [47]. This scenario poses serious challenges for the conservation of this biome, which is globally recognized as a biodiversity hotspot [42].

In this context, the conservation of plant genetic resources, also threatened by illegal logging, becomes even more critical and urgent, as is the case for dozens of species in the Atlantic Forest, including the jequitibá-rosa (Cariniana legalis [Mart.] Kuntze) [13, 35]. Deforestation, combined with selective logging, can cause drastic reductions in population size and affect population structure, especially when individuals are removed before reaching reproductive maturity [50, 72]. These changes can lead to reduced intraspecific genetic diversity, genetic erosion, increased inbreeding, and a loss of evolutionary viability for current and future generations [30, 50, 65, 73]. Thus, empirically assessing how these anthropogenic disturbances impact the genetic level is essential to inform effective conservation strategies [40, 56]. In general, population genetics studies have shown that even in fragmented environments, some tropical tree species with a history of illegal logging can retain substantial levels of genetic variability, particularly in populations located in more forested landscapes or agroforestry systems [34, 53].

Among the anthropogenic systems with the potential to reconcile agricultural production and conservation, the agroforestry system known as Cabruca, predominant in southern Bahia, Brazil. The Cabruca, is a traditional practice of growing cocoa under the shade of native trees of Atlantic Forest. This system has undergone significant transformations in recent decades: during the 1960 s and 1970 s, millions of native trees were cut down to boost productivity [3], and with the cocoa crisis in the 1980 s, illegal logging intensified further [2]. In response, state legislation enacted in 2014 [23] established standards for the sustainable management of Cabruca areas, promoting economic incentives linked to ecosystem services. Furthermore, Cabrucas are recognized as a “biodiversity-friendly” environment capable of retaining part of the region's biological diversity [11, 52]. With studies demonstrating that Cabrucas can function as ecological corridors, contributing to connectivity between forest remnants and facilitating the movement of species [11, 52]. Thus, Cabrucas potentially enhance gene flow between fragmented populations and are relevant for the genetic conservation of native plants. However, empirical studies evaluating the role of Cabrucas in maintaining the genetic diversity of native and timber tree species in the Atlantic Forest remain scarce [34, 69].

Given this context, the present study aims to evaluate the levels of inbreeding, diversity and genetic structure, as well as the patterns of genetic connectivity among populations of C. legalis, a species endemic to the Atlantic Forest and threatened with extinction due to deforestation and selective logging [7]. To this end, populations composed of adults and juveniles were evaluated in forest fragments and areas under Cabruca management in the Atlantic Forest of southern Bahia. By incorporating different ontogenetic stages (adults and juveniles) and distinct landscape configurations, we aim to understand the effects of fragmentation, land use, and disturbance history on the species’ genetic dynamics [54, 66, 70]. Our results provide scientific support for in situ conservation actions and sustainable management of C. legalis, highlighting the importance of traditional agroforestry systems for long-term genetic conservation in highly anthropized Atlantic Forest landscapes.

Based on these considerations, we formulated the following hypotheses to guide our investigation: (i) C. legalis populations located in Cabruca agroforestry systems maintain levels of genetic diversity comparable to those found in populations from forest remnants, due to the preservation of native shade trees and the ecological continuity provided by these systems; (ii) C. legalis populations in Cabruca areas play a central role in promoting genetic connectivity, acting as permeable elements in anthropogenic landscapes that facilitate gene flow among forest remnants; and (iii) Genetic patterns differ between adults and juveniles, with juveniles expected to exhibit lower genetic diversity, higher inbreeding, and stronger genetic structure, reflecting more recent effects of selective logging, deforestation, and increased mating among related individuals.

Materials and methods

Target species

The genus Cariniana (Lecythidaceae) widely exploited since the colonial period, includes species of great economic and ecological value. Among them, Cariniana legalis (jequitibá-rosa) stands out as an endemic and iconic species of the Atlantic Forest. This tree species is currently classified as Endangered on the Brazilian Red List and as Vulnerable under criterion A1ac of the IUCN Red List ([7] https://www.iucnredlist.org/). It is a late secondary tree that can reach up to 60 m in height and live for more than 500 years [10]. It has a mixed mating system, with a predominance of allogamy [57], hermaphroditic flowers pollinated by bees of the genera Trigona and Melipona, and seeds dispersed primarily by anemochory, although they are also consumed by local fauna, such as primates [9, 45]. Due to its commercial value, C. legalis is listed in the Tropical Timber Atlas of the International Tropical Timber Organization (ITTO), a key reference for professionals involved in the processing, use, or management of tropical timber (https://www.itto.int/).

Study area and sampling

The research was conducted in the Central Corridor of the Atlantic Forest in southern Bahia, Brazil, a region recognized for its high biodiversity and endemic plant species and designated as a conservation priority [39, 44]. The local landscape combines forest fragments with agroforestry systems known as Cabruca (Fig. 1).

Fig. 1.

Visual differences between the two land use systems sampled in the study. A The photograph of a Cabruca agroforestry system (photographed by Paloma Resende), illustrating its vertical structure, characterized by cacao trees in the understory shaded by a diverse canopy of large native trees; B Forest fragments of the Atlantic Forest (photographed by Alesandro Santos)

Within this land cover mosaic that composes the southern Bahian landscape, five populations of C. legalis were sampled, distributed across protected areas —Reserva Particular do Patrimônio Natural Veracel (VCL) and Reserva Particular do Patrimônio Natural Serra do Teimoso (ST) — and Cabruca agroforestry systems: Agroforestry BC, Agroforestry FR, and Agroforestry JS (Fig. 2). Prior to sampling, authorization for access to Brazilian genetic heritage was obtained from the Ministry of the Environment (registration number A6BAAEC). For the protected areas (VCL and ST) which are privately owned conservation units recognized by the Brazilian government—was conducted with the explicit permission of the landowners responsible for managing these reserves. For the Cabruca agroforestry systems (BC, FR, and JS), sampling was likewise conducted on private properties with the explicit consent of the landowners.

Fig. 2.

Geographic distribution of five populations of Cariniana legalis inserted in agroforestry systems (Cabruca) and Atlantic Forest remnants in southern Bahia, Brazil. The state of Bahia is highlighted on the map in dark gray and the red rectangle indicates the study region. A Genetic network analysis for adults using F_ST as a distance measure, with automatic thresholding of 0.3; B Genetic network analysis for juveniles using F_ST as a distance measure, with an automatic threshold of 0.26. The nodes representing the populations are organized according to their geographic origin (white squares). The size and color of each node indicate its betweenness centrality (i.e., the proportion of all shortest paths that pass through the node). BC, FR, and JS represent agroforestry populations, while ST and VCL correspond to protected forest remnants

Within each area (geographic population), an active search was carried out, and cambial scrapings from adults and leaves from juveniles were collected from all located individuals. Trees with a diameter at breast height (DBH) greater than 0.42 cm were considered adults, while those with DBH ≤ 0.42 cm and no evidence of flowering or fruiting — suggesting an age of less than 20 years — were classified as juveniles [46].

DNA extraction and amplification of microsatellite markers

Genomic DNA extraction was performed following Doyle and Doyle [14], using cambial scrapings from adults and leaves from juveniles. Genotyping was conducted using 11 microsatellite loci: nine transferred of Cariniana estrellensis (Ces03, Ces05, Ces07, Ces10, Ces12, Ces13, Ces14, Ces16, Ces18) for C. legalis [24] and two specifics to C. legalis (CL05 and CL11) [34]. Forward primers were modified with an M13 tail to enable fluorescent detection [43]. PCR reactions were prepared in a final volume of 13 µL, containing 7.5 ng of DNA, 1 × (NH₄)₂SO₄ buffer, 3 mM MgCl₂, 0.25 mg/mL BSA, 0.25 mM dNTPs, 0.49 µM of each forward and reverse primer, 0.49 µM of fluorescently labeled M13 primer (HEX, 6-FAM, or NED), and 1 U of Taq DNA polymerase. Amplifications were performed using a GeneAmp 9700 thermocycler (Applied Biosystems) under the following conditions: initial denaturation at 96 °C for 2 min; followed by 35 cycles of 94 °C for 60 s, annealing at a primer-specific temperature for 60 s, and 72 °C for 60 s; with a final extension at 72 °C for 7 min. Amplified fragments were separated by electrophoresis using an ABI377 DNA sequencer, and genotypes were scored using Genotyper software (v. 3.1.2).

Data analysis

Given the heterologous origin of most loci, the presence of null alleles and genotyping errors was evaluated using Micro-Checker 2.2.3 [64], and corrections were applied when necessary. Subsequently, linkage disequilibrium between loci was assessed using FSTAT 2.9.3 [22], and deviations from the Hardy–Weinberg equilibrium were tested with GenAlEx 6.5 [59].

Genetic diversity was estimated separately for adults and juveniles, considering percentage of total alleles observed in the population (A), number of private alleles (AP), allelic richness (Ar), observed (H_O) and expected (H_E) heterozygosity. AP was calculated in GenAlEx 6.5, and the other parameters in the diveRsity package via the divBasic function [31]. The inbreeding coefficient (f) was estimated by the LynchRd method with 95% confidence and 10,000 bootstraps, using the Coancestry program (v. 1.0.1.10; [67]). To evaluate differences in genetic diversity and inbreeding among populations, we performed mean comparisons using estimators that provide per-locus mean values, allowing for replication within each population. Because private alleles are population-specific and not replicated across loci, they were excluded from these comparisons. An ANOVA followed by a Scott-Knott test was used to assess differences among populations within each ontogenetic stage, while a Student’s t-test for independent samples was applied to evaluate differences between stages. As the t-test is parametric, assumptions of normality and homogeneity of variances were tested prior to its application. All statistical analyses were conducted in R (http://www.r-project.org/).

Genetic structure was assessed using four complementary approaches: analysis of genetic differentiation (F_ST) and migrant rate (Nm) between pairs of populations, genetic network analysis, and Discriminant Analysis of Principal Components (DAPC). Nm values were estimated using the equation Nm = (1 – F_ST)/(4F_ST) [71], and in this study, they were used as indirect indicators of historical genetic exchange between populations. The effect of geographic distance (in km) on genetic differentiation was tested using a Mantel test [38] implemented in the ecodist package [21] in R.

Genetic network analyses were conducted using the EDENetworks program. In this approach, nodes represent populations connected by edges reflecting their genetic relationships based on pairwise F_ST values [32]. The networks were spatially organized according to the geographic origin of each population. To identify populations central to gene flow, node size and color were scaled by betweenness centrality, which quantifies the proportion of shortest paths passing through each node. Networks were constructed using the automatic threshold detection feature in EDENetworks, set at the percolation threshold to ensure network connectivity. The thresholds were 0.30 for adults and 0.26 for juveniles, both below the percolation limit.

The DAPC was performed separately for adults and juveniles using the adegenet package v.2.0.0 [28, 29]. First, the number of clusters was determined using the find.clusters function, retaining all principal components and the optimal number of clusters was selected using the Bayesian Information Criterion (BIC), with the lowest BIC indicating the best model (Figure S1). Then, the relationships between the clusters were identified through the DAPC function, using the cross-validation procedure to choose the optimal number of principal components to be retained in each DAPC (Figure S1). For adults, 60 PCs and 4 discriminant functions were retained, explaining 90.5% of the variance. For juveniles, 70 PCs and 5 discriminant functions were retained, explaining 91.6% of the variance. The scatter function was used for the graphical representation of individual assignments to genetic clusters based on their gene pool.

Results

Genotypic data quality and Hardy–Weinberg equilibrium

Our results showed that C. legalis populations had a low overall average frequency of null alleles in adults (0.12) and juveniles (0.07) (Table S1), and no significant genotypic linkage disequilibrium was identified between the 11 loci used (data not shown). However, of the 11 loci evaluated, only three loci in adult populations and 4.4 loci in juvenile populations exhibited genotypic frequencies consistent with Hardy–Weinberg equilibrium (Table S2). These results ensure the quality and reliability of the genotypic data used in subsequent analyses.

Patterns of genetic diversity and inbreeding coefficient among populations and ontogenetic stages

Genetic diversity parameters varied markedly among populations and between ontogenetic stages (Table 1). In general, adult populations from ST, JS, and FR exhibited the highest values of allelic richness (Ar = 3.41, 3.32, and 3.29, respectively), observed heterozygosity (H_O = 0.594, 0.581, and 0.566), and expected heterozygosity (H_E = 0.639, 0.636, and 0.630), along with the lowest inbreeding coefficients (f = 0.071, 0.092, and 0.103). These values suggest a relatively efficient maintenance of genetic variability in these areas.

Table 1.

Genetic diversity and inbreeding (f) of five Cariniana legalis populations in agroforestry systems (Cabruca) and remnants of the Atlantic Forest in southern Bahia, Brazil

Population	N	A (%)	AP	Ar	HO	HE	f (CI 95%)
Adults
ST	39	52.6a	32	6.77a	0.56a	0.78a	0.30 (0.02; 0.68)b
Agroforestry BC	13	27.9b	10	4.42b	0.46b	0.64b	0.56 (0.21; 0.87)a
VCL	25	26.4b	11	3.81b	0.34b	0.52b	0.47 (0.15; 0.88)a
Agroforestry JS	37	44.8a	22	5.87a	0.55a	0.75a	0.31 (−0.01; 0.75)b
Agroforestry FR	22	43.5a	25	6.45a	0.66a	0.78a	0.19 (−0.02; 0.53)b
Mean	27.2	39.1	20	5.46	0.51	0.69	0.34 (0.05; 0.73)*
SE	10.8	11.4	9.4	1.29	0.12	0.11	-
Juveniles
ST	22	36.4	9	5.23a	0.59a	0.70a	0.23 (0.02; 0.60)b
Agroforestry BC	38	48.9	30	5.36a	0.49b	0.71a	0.44 (0.06; 0.84)a
VCL	11	18.7	5	3.26b	0.44b	0.55b	0.43 (0.08; 0.88)a
Agroforestry JS	31	47.8	19	6.17a	0.63a	0.76a	0.25 (−0.01; 0.63)b
Agroforestry FR	56	61.0	32	6.46a	0.65a	0.79a	0.19 (−0.03; 0.52)b
Mean	31.6	42.6	19	5.3	0.56	0.70	0.28 (0.01; 0.65)*
SE	17.0	15.9	12.1	1.25	0.09	0.09

N Sample size, A Percentage of total alleles observed in the population, AP Number of private alleles, Ar Allelic richness, H_O Observed heterozygosity, H_E Expected heterozygosity, SE Standard error, CI 95% 95% confidence interval. The subscript letters represent the results of the Scott-Knott tests for each genetic parameter

^* significant difference for the Student's t-test between means, comparing the adult and juvenile stages. BC, FR, and JS represent agroforestry populations, while ST and VCL correspond to protected forest remnants

In contrast, BC and VCL populations showed reduced genetic diversity and elevated inbreeding. In adults, BC had Ar = 2.20, H_O = 0.413, H_E = 0.599, and f = 0.433; VCL had Ar = 2.25, H_O = 0.355, H_E = 0.640, and f = 0.506. Similar patterns were observed in juveniles, although in BC juveniles the values of Ar (2.85) and H_E (0.631) were statistically comparable to those of more genetically diverse populations. Nonetheless, inbreeding remained high in both BC and VCL juveniles (f = 0.460 and 0.443, respectively).

Overall, only the inbreeding coefficient differed significantly between ontogenetic stages, being higher in adults (mean f = 0.340) than in juveniles (mean f = 0.280; p < 0.05). No significant differences were detected between stages for the other diversity parameters.

Genetic structure and historical gene flow (Nm) between populations

Genetic differentiation analyses (F_ST) and estimates of the number of migrants (Nm) between pairs of populations revealed substantial variation in the genetic structure in the two ontogenetic stages analyzed (Fig. 3). The highest F_ST and lowest Nm values were observed in pairs involving the VCL population, with VCL_BC standing out (F_ST = 0.26 and Nm = 0.71 in adults; F_ST = 0.21 and Nm = 0.94 in juveniles). On the other hand, F_ST values ≤ 0.12 and Nm ≥ 1.85 were found for the JS_FR, ST_FR, and ST_JS pairings, in both stages (Fig. 3). The Mantel test indicated that geographic distance does not significantly explain the genetic differentiation observed between the adult (r = 0.66; p = 0.13) or juvenile (r = 0.64; p = 0.17) populations, despite the considerable geographic distances between some populations (up to 198 km).

Fig. 3.

Visualization of genetic differentiation (F_ST) and migration rate (Nm) between pairs of Cariniana legalis populations within agroforestry systems (Cabruca) and Atlantic Forest remnants in southern Bahia, Brazil. Fst_A = Genetic differentiation between pairs of populations with adult individuals; Fst_J = Genetic differentiation between pairs of populations with juvenile individuals; Nm_A = Migrant rate (Nm) between pairs of populations with adult individuals; Nm_J = Migrant rate (Nm) between pairs of populations with juvenile individuals. BC, FR, and JS represent agroforestry populations, while ST and VCL correspond to protected forest remnants

Genetic connectivity and population clusters

Network analysis indicated gene pool sharing between populations, with networks remaining connected at both ontogenetic stages, reflecting a certain degree of genetic connectivity between populations (Fig. 2). Populations JS and FR were identified as the most genetically connected, occupying central positions within the network and playing a crucial role in the maintenance of genetic connectivity between adult and juvenile populations, respectively. In contrast, population VCL presented the weakest connections at both ontogenetic stages.

DAPC analysis revealed that the five adult populations formed five distinct genetic clusters (Figure S1A-B; Fig. 4A), with JS and FR in the same quadrant, ST in the same quadrant as BC, and VCL genetically more isolated. Among the juveniles, six genetic clusters were identified (Figure S1C-D; Fig. 4B), with JS subdividing into two overlapping groups that partially overlapped with each other and with FR, while BC and VCL occupied the same quadrant, and ST appeared isolated from the other clusters.

Fig. 4.

Discriminant Analysis of Principal Components (DAPC) for Cariniana legalis populations within agroforestry systems (Cabruca) and Atlantic Forest remnants in southern Bahia, Brazil. The dots represent individuals within the population, while the ellipses denote the genetic clusters defined by DAPC. A DAPC for adult individuals; B DAPC for juvenile individuals. Note: The numbers represent the number of clusters identified in the DAPC, and the colors represent the populations of origin. Orange = ST; Firebrick = BC; Dark blue = VCL; Black and Light blue = JS; Purple = FR. BC, FR, and JS represent agroforestry populations, while ST and VCL correspond to protected forest remnants

Discussion

Our study on the genetic diversity and population structure C. legalis, an endemic and endangered tree from the Brazilian Atlantic Forest, shows that, despite intense illegal logging and deforestation of this ecosystem, the species still exhibits a degree of genetic resilience. In particular, the Cabruca agroforestry system has proven to be extremely important both for retaining genetic diversity and for connecting populations by gene flow. However, some populations evaluated already show signs of genetic erosion, most likely in response to the anthropogenic disturbances experienced by the species over the years [65].

In this context, we believe that, because this is a species with a long generation time and high life expectancy, there may be a delayed genetic response (genetic extinction debt), masking the real magnitude of the harmful effects of anthropogenic disturbances [20]. Therefore, we emphasize caution when interpreting the results, due to a possible genetic time lag, and highlight that, in addition to the forest remnants of the Atlantic Forest, conservation efforts must consider the Cabruca agroforestry system as a key piece for the genetic conservation of C. legalis.

Patterns of genetic diversity and inbreeding coefficient among populations and ontogenetic stages

The low average proportion of loci in Hardy–Weinberg equilibrium at both ontogenetic stages may indicate the action of microevolutionary forces, such as genetic drift, migration, selection, or non-random reproduction in the populations [68].

Moreover, the patterns of genetic diversity observed in C. legalis show marked variation among the evaluated populations. This may reflect both human activities, such as different intensities of selective logging between areas, and historical processes (e.g., pollen and seed dispersal) related to the establishment and natural dynamics of populations across different locations [34, 50, 60, 65]. For instance, the ST, JS, and FR populations with adult individuals exhibited the highest values for the genetic diversity estimators evaluated here (A, AP, Ar, H_O, and H_E) and the lowest inbreeding rates, suggesting a relatively efficient maintenance of genetic variability in these areas. The values observed for these genetic variability estimators are within the upper range observed in neotropical tree species that maintain high levels of genetic diversity, even in fragmented landscapes [36, 53]. Thus, although a myriad of factors of natural or anthropogenic origin may influence this pattern, we believe that the greater genetic connectivity between these populations (see F_ST and Nm) is favoring the maintenance of genetic diversity in these populations [49].

In contrast, BC and VCL populations showed reduced genetic diversity and high levels of inbreeding, which may be indicative of isolation effects, genetic drift, and crosses between related individuals [30, 49]. This pattern is consistent with studies in other plant species subjected to fragmentation, in which reduced diversity and increased inbreeding are associated with habitat loss and decreased effective population size [30, 49]. Although the general pattern observed in adult populations is maintained in the juveniles of the VCL population, it is worth noting that the pattern is only partially retained in BC. In the latter case, the values of A, AP, Ar, and H_E in BC are statistically equal to those reported in genetically more diverse populations. These findings reveal that within the same population, there may be genetic differences between individuals of different ontogenetic stages depending on the sensitivity of the genetic variability estimator used in responding to environmental changes [36, 53, 66].

For example, the adult population of BC is made up of only 13 individuals, which would partially explain the low H_O value and high inbreeding found in the following generation of young individuals [61]. In this context, even if new alleles potentially arrived from neighboring populations and increased the number of individuals, which would explain the increase in estimators calculated from the number and/or allele frequency, BC is still at risk of genetic erosion via genetic drift and crosses between related individuals [61]. In this sense, the inbreeding coefficients observed in adult and juvenile populations, particularly in BC and VCL (f > 0.43) are considered high and suggest potential risks of inbreeding depression, reduced reproductive vigor and impairment of natural regeneration in the long term [61].

Furthermore, the absence of significant differences in the average of genetic diversity parameters between adults and juveniles, except for the inbreeding coefficient, may indicate either genetic resilience to anthropogenic disturbances or a time lag in the genetic response (genetic time lag), as already reported for C. legalis and other tropical plant species [4, 53, 54, 61]. On the other hand, the lower inbreeding found in juveniles may suggest a slight recovery in outcrossing patterns. However, considering the still high levels of overall inbreeding (adults f = 0.34 and juveniles f = 0.28), there is clear evidence of crosses between close relatives or limitations in the activity of dispersers is affecting the populations [1, 4]. We hypothesize that the low number of reproductive individuals in populations favors an increase in genetic relatedness among individuals, generating the observed pattern of inbreeding, as already described for other populations of C. legalis [34, 62].

Paradoxically, although fragmentation reduces C. legalis populations over time, it can also increase gene flow distances, which would help explain the slight reduction in inbreeding in juveniles [4, 36]. This hypothesis is because C. legalis flowers are pollinated by bees of the genera Melipona and Trigona [45], while its seeds are dispersed by gravity and wind [10]. Thus, it would be expected that bees would expand their foraging range in environments with low density of C. legalis, and that the wind would favor the dispersal of seeds over greater distances in more fragmented environments, promoting greater gene flow and, consequently, reducing inbreeding [26, 48, 62]. Despite this apparent attenuation of inbreeding in juveniles, values still range from moderate to high, which suggests source limitation (reproductive individuals), reflecting a restricted parental base [1, 4].

In this context, our findings emphasize the importance of sustainable management initiatives aimed at promoting gene flow between genetically distinct populations and implementing strategic reforestation efforts in the region. In this way, it would be possible to carry out a genetic rescue of compromised populations, favoring intraspecific genetic variability and maximizing the adaptive resilience of C. legalis in the face of environmental changes and anthropogenic pressures in the Atlantic Forest [18, 63].

Genetic structure and historical gene flow (Nm) between populations

The results obtained for genetic differentiation and historical gene flow from the F_ST and Nm estimators indicate significant variations in the degree of genetic structure of the populations analyzed, both for adults and juveniles [5]. The high F_ST values and low Nm values between VCL and the other populations suggest a clear barrier to historical gene flow [5, 37]. Certainly, a myriad of factors potentially influences this observed pattern. For example, isolation by geographic distance and forest fragmentation can create obstacles to gene flow, resulting in a significant genetic differentiation between populations [8, 37, 58]. On the other hand, different colonization histories or anthropogenic disturbances such as selective logging can also lead to genetic differentiation between populations [27, 41].

In contrast, the F_ST and Nm values between the JS and FR, ST and FR, and ST and JS population pairs indicate a moderate genetic structure, suggesting a more efficient historical gene flow between these populations at both ontogenetic stages [5, 37]. Taken together, the results of our study suggests that the pattern of genetic differentiation observed among C. legalis populations in fragmented landscapes of the Atlantic Forest may be more complex than the simple spatial separation of populations [12, 58].

In this context, although the Mantel test did not detect a statistically significant correlation between geographic distance and genetic differentiation for adults and juveniles, the correlation coefficients suggest a moderate to strong relationship, which may have biological relevance. The lack of statistical significance may be partly related to the low number of population pairs (n = 10), which reduces the statistical power of the test and limits the detection of significant patterns, especially in contexts with high spatial variability. Therefore, although it is not possible to state that isolation by distance is the main factor of genetic structuring in C. legalis, the observed coefficients suggest that spatial structure may still contribute to the detected genetic patterns, together with other ecological and historical factors [12, 58].

Finally, although the Nm results are convergent with other analyses (e.g. DAPC and genetic network analysis), it is important to highlight that the use of Nm values derived from the classical equation Nm = (1 – F_ST)/(4F_ST) is based on highly simplified assumptions, such as equal and constant population sizes, symmetrical migration, and equilibrium between drift and gene flow [71]. The violating these assumptions can lead to substantial biases, making such estimates unreliable indicators of recent gene flow. Therefore, in this study, Nm values are presented as indirect indicators of historical gene exchange but should not be interpreted as direct measures of contemporary gene flow. To overcome these limitations, we also employed complementary approaches, such as DAPC and genetic network analysis (EDENetworks), that allowed a more nuanced interpretation of connectivity among populations beyond what F_ST alone can reveal.

Genetic connectivity and population clusters

Genetic network analysis also revealed an interesting pattern of connectivity between populations. The JS and FR populations from both ontogenetic stages appeared as the most central in the network, maintaining stronger connections with the other populations, which may be indicative of these populations acting as genetic"hubs,"facilitating the exchange of genetic material between different locations [51, 55]. This pattern of centrality of these populations inserted into Cabruca agroforestry highlights the crucial role of this production system in maintaining the genetic diversity of C. legalis in fragmented landscapes [34]. Thus, our study highlights the relevance of Cabruca for maintaining biodiversity at the genetic level, corroborating other studies reporting the importance of this production system in conserving biodiversity at the species level [11, 16, 52].

In contrast, the VCL population, which is an environmentally protected area, demonstrated the lowest genetic connection for both ontogenetic stages, reflecting its genetic isolation in relation to the other populations, as already evidenced in the F_ST and Nm analyses.

Furthermore, Discriminant Principal Component Analysis (DAPC) revealed a clear differentiation between adult populations, with the formation of five distinct genetic clusters, corroborating the results of F_ST, Nm, and networks. The JS and FR populations clustered closely, suggesting a significant degree of genetic sharing between them. In contrast, VCL appeared as the most isolated population for adults, which reinforces the idea that this area experiences a greater genetic barrier, potentially reflecting the genetic isolation and low genetic diversity observed [4, 19].

Regarding juveniles, the formation of six genetic clusters, with the subdivision of JS into two groups, indicates the complexity of genetic interactions between populations. The overlap of the JS and FR clusters may suggest that, although there is some differentiation, there is also considerable genetic flow between them. The BC and VCL populations are in the same quadrant, while ST remained more isolated. These patterns indicate that gene flow between juvenile populations was not homogeneous, which may reveal different rates of seed dispersal and pollination between populations, reflecting the observed genetic structure [4]. In summary, our results point to a complex genetic structure in C. legalis, with a large variation in connectivity among populations. The centrality of some populations, such as JS and FR, and the isolation of VCL suggest that maintaining genetic connectivity and implementing conservation strategies that promote gene flow are essential for the long-term conservation of this endangered species [61, 62].

Conclusion and recommendations for the conservation

The relatively high genetic diversity observed in most C. legalis populations, despite habitat fragmentation, highlights the potential for effective, evidence-based conservation strategies. The presence of high inbreeding rates and the signs of genetic erosion observed in some populations, especially BC and VCL, indicate the urgency of mitigating the negative effects of population isolation.

In this sense, we recommend that conservation programs include functional connectivity actions, such as the implementation of ecological corridors, and active management, such as the controlled introduction of genetic material from more diverse populations into areas with less variability, aiming at genetic rescue and reducing the effects of inbreeding. This recommendation should also consider and respect the local gene pool, avoiding exogamic depression.

Additionally, the inclusion of Cabruca areas as relevant components in in situ conservation strategies is a promising measure, especially in highly anthropized landscapes. Since these systems harbor functionally connected populations of C. legalis, they should be integrated into public conservation policies and territorial management plans. In addition, continuous genetic monitoring of populations, including different ontogenetic stages should be incorporated as an essential tool to detect early signs of genetic erosion and guide corrective actions.

Finally, we emphasize that the conservation of C. legalis should not be limited to the protection of forest remnants, but should also include the strengthening of partnerships with rural producers, traditional communities, and owners of agroforestry systems. Effective conservation of the species requires an integrated approach that combines science, adaptive management, and public policies aimed at promoting genetic resilience and the socioeconomic value of agricultural practices compatible with biodiversity conservation.

In line with recent guidelines for the management of tropical timber species [6, 12], we also highlight the importance of basing restoration and translocation efforts on genetic data. The selection of seed or seedling sources should consider the genetic structure of populations, avoiding the indiscriminate mixing of individuals from distinct evolutionary lineages. This precaution reduces the risk of outbreeding depression and ensures that local adaptations are preserved, thereby increasing the success and long-term resilience of conservation interventions.

Authors’ contributions

The authors Leal JB, Corrêia RX and Gaiotto FA contributed to the study conception and design. Leal JB, Corrêia RX, and Gaiotto FA, contributed to the field collection. Material preparation and laboratory analysis were performed by Leal JB, and Gramacho, KP. The genetic analysis were performed by Santos AS. The first draft of the manuscript was written by Santos AS and all authors commented on previous versions of the manuscript. All authors have read and approved the final manuscript. The authors Santos AS and Leal JB contributed equally to this work and shared first authorship.

Funding

The research was supported by grants from the Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq) (#421179/20180) and its productivity fellowship provided to Corrêa RX (#315913/2023–1) and Gaiotto FA (#312065/2021–3).

Data availability

The files containing sampling sites and genotyping data with microsatellite markers will be available under request to gaiotto@uesc.br.

Declarations

Ethics approval and consent to participate

Not applicable. This study did not involve human participants. However, all necessary permissions for field sampling of Cariniana legalis were obtained. In the case of the Reserva Particular do Patrimônio Natural Serra do Teimoso (ST) and Reserva Particular do Patrimônio Natural Veracel (VCL), sampling was conducted with the explicit authorization of the landowners. For Cabruca agroforestry systems (BC, FR, JS) sampling was conducted with the consent of the respective private landowner.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.