Introduction

Among the most relevant ecosystems for conservation in the Colombian Caribbean is the Tropical Dry Forest (TDF), which provides both economic and regulatory services to local communities, such as the regulation of climate, the fertility of soils, and watershed protection (Gillespie et al. 2013; Quijas et al. 2019; Andrade et al. 2020). Its deciduous vegetation and marked climatic seasonality, coupled with large levels of water stress, contribute to its high turnover of plants and animals, making it highly diverse biologically (Kattan et al. 2019; Romero-Duque et al. 2019). For instance, 3,395 plants, 145 birds, 60 mammals, and 68 dung beetle species have been recorded for the TDF of Colombia (Pizano and García 2014). Unfortunately, it is at risk due to anthropogenic changes in land use and climate vulnerability (Torres-Romero and Acosta-Prado 2022), and currently less than 8% of the TDF’s original extent remains (García et al. 2014). Accordingly, the Colombian government considers the TDF to be a critically endangered ecosystem (Etter et al. 2017).

Dung beetle species are key to ecosystem health due to their dung removal activity that cleans the surface of pastures, favors soil structures, and reduces the presence of parasitic agents (Gómez-Cifuentes et al. 2018). They are sensitive to changes in vegetation cover, variations in microclimate conditions, and the availability of excrement as food sources and for breeding chambers, primarily from primates. Moreover, the density of native vertebrates providing sources of excrement for dung beetles has decreased due to the exploitation of forests (Bicknell et al. 2014). Evidence shows that the species richness of dung beetle species in tropical ecosystems has declined (Scarabaeinae subfamily), resulting in the loss of ecological niches and ecosystem services, especially of the specialist species which are more likely to decrease in abundance (Carrión-Paladines et al. 2021; Noriega et al. 2021).

The Colombian Caribbean region, including the departments of Guajira and Sucre, is experiencing habitat deterioration due to urban, mining, and agricultural activities, which have resulted in the loss of forest cover and soil degradation (Ballesteros-Correa and Pérez-Torres 2022). The TDF in this region provides habitat for Deltochilum guildingii (Westwood 1835), one of the largest roller dung beetle species, which has a limited flight range and a preference for conserved forests. This species plays an active role in recycling nutrients and in removing excrement of native mammals, such as primates of the genus Alouatta whose abundance has decreased locally in the TDF in recent years (e.g., Colosó in Sucre; Martínez-Hernández et al. 2012; De la Ossa et al. 2013; Estrada 2015; Rangel-Acosta and Martínez-Hernández 2017; Noriega et al. 2020). As a large roller, D. guildingii provides a greater burial of organic matter that contributes to soil fertility (Correa-Cuadros et al. 2022), favoring the germination viability of tropical plant species (Andresen 2003; Barragán et al. 2011).

The loss of vegetation cover may restrict D. guildingii dispersal between increasingly isolated habitat fragments (Cox et al. 2020), while the size reduction of habitat fragments coupled with the reduction of excrement sources (Rondón et al. 2017; Medina-Hernández et al. 2020) may have adverse impacts on population size and reproductive success (Leroy et al. 2017). According to the population genetics theory, small and highly fragmented populations will lose genetic variation and adaptive potential because of genetic drift and inbreeding (Hooftman et al. 2016; Dillon and Lozier 2019). Moreover, it has been shown that the largest telecoprids and paracoprids beetles are the most vulnerable to extinction (Tonelli et al. 2020).

Integral conservation plans are required to protect TDF species. Combining ecology and genetics enables a better understanding of the influence of humans on landscapes and the designation of genetically and ecologically representative areas (Etter et al. 2008; Narum et al. 2013; Schierenbeck 2017; Matos-Maraví et al. 2019; Maxwell et al. 2020). Furthermore, we can better understand climate adaptation by examining how the demography of populations has changed historically and how these changes can be attributed to previous climate events (Ho and Shapiro 2011). This understanding is relevant for the Caribbean region, which has three biogeographic regions with heterogeneous climates and topographies: (1) the Sierra Nevada de Santa Marta (SNSM), (2) the mountainous system of Montes de María y Piojó (DMMP), and (3) the Guajira (GUAJ) (Fig. 1) (Arenas 2012; Fremout et al. 2021). For these regions, few genetic studies have been conducted on key species.

Fig. 1
figure 1

Studied fragments of the Tropical Dry Forest (TDF) in the Colombian Caribbean. Three biogeographic regions are denoted with a different grey-coloring background in the map: Sierra Nevada de Santa Marta (SNSM) (fragments PNNT and TG); Montes de María and Piojó (DMMP) (fragments COL, RCM and SFFC); Guajira (GUA) (fragment BRUN). Red dots denote the fragment location

Full size image

In this study, based on double digest Restriction-site Associated DNA sequencing (ddRadSeq) analysis for 60 individuals of D. guildingii sampled in six TDF fragments of these three biogeographic regions in the Colombian Caribbean, each with varying degrees of anthropogenic habitat disturbance, we aimed to: (i) estimate levels of genetic diversity and their relationship with habitat fragmentation, (ii) evaluate patterns of spatial genetic structure, and (iii) infer the demography history of the species and its potential associations with past climatic events. This study will provide valuable information for future conservation decisions for this roller dung beetle species and others that could have similar genetic responses to the changing conditions of the TDF in the Colombian Caribbean.

Materials and methods

Study area and species

In the Colombian Caribbean region, we sampled the following TDF fragments: (1) Bruno (BRUN) in the Guajira region (GUAJ); (2) The Tayrona National Natural Park (PNNT in Magdalena), and (3) Tierra Grata (TG in César) in the Sierra Nevada de Santa Martha (SNSM); (4) the Flora and Fauna Los Colorados Sanctuary (SFFC in Bolívar), (5) the Reserva Campesina “La Montaña” (RCM in Atlántico), and (6) Estación Biológica Primatológica de Colosó (COL in Sucre) in the Montes de María y Piojó region (DMMP) (Fig. 1). All fragments have the typical TDF rainfall regime (900–1200 mm) and vegetation cover (Galván-Guevara et al. 2009). These fragments have different levels of environmental vigilance (Table 1).

Table 1 Ecological attributes of the TDF fragment, fragment type, biogeographic region, index of fragmentation (IF), and conservation status
Full size table

Deltochilum guildingii is a large species, with limited flight range and nocturnal activity. For the genus, it has been observed that the egg balls are buried in the ground in a cup-shaped depression (Howden and Ritcher 1952; Pessôa et al. 2017). It has a wide distribution in Brazil, Colombia, Trinidad and Tobago, Venezuela, and Suriname (González-Alvarado and Vaz-de-Mello 2014).

Sampling and fragmentation index

Deltochilum guildingii individuals were collected during 2018 and 2019 (Colombian sampling permits Resolution 00949, and No. CITES permit 1607). For each fragment, four transects were located 500 m apart and for each transect, five pitfall traps were placed separated by 100 m each following da Silva and Medina-Hernández (2016) for a total of 20 traps per fragment. Each pitfall consisted of a 32-ounce plastic container buried in the ground without lethal liquid and a plastic funnel was placed at the aperture of the container. A rod in the shape of an inverted L was placed above the container, upon which was hung a bait consisting of human excrement and chicken viscera wrapped in gauze (Martínez-Hernández et al. 2022). The traps were revised every 24 h for two days and the material was stored in whirlpack bags with 96% ethanol (on average ~ 52 individuals were collected per fragment). The material was identified taxonomically and with the extraction of genitalia using the keys of González-Alvarado and Vaz-De-Mello (2014). Ten individuals were selected for each fragment to extract the two middle and hind legs, which were stored in cryovials with ethanol 96% until genomic DNA extraction.

For each fragment, we calculated the fragmentation F Index (IF) according to Lozano-Botache et al. (2011), whose value ranges from 0 to 1. The values correspond to the following categories: insularized fragment (IF < 0.5), highly fragmented (IF = 0.5–0.7), moderately fragmented (IF = 0.7–0.99), and no fragmented (IF = 1) (Galván-Guevara et al. 2015). The IF was calculated using the TDF cover map created by the Institute Alexander von Humboldt (González et al. 2014). Moreover, we calculated the area (\(\:{m}^{2}\)) and perimeter (\(\:{m}^{2}\)) of each fragment in ArcGIS pro v.2.9 (Esri Inc 2020).

DNA extraction and ddRadseq protocol

Genomic DNA was extracted using the NucleoSpinTissue (R) tissue kit (Macherey-Nagel 2017) according to the manufacturer’s instructions. Approximately 200 ng of high-molecular-weight DNA was digested with two restriction endonucleases enzymes (EcoRI and HpyCH4IV) following the dd-RadSeq library preparation of Peterson et al. (2012). DNA fragments were searched in a target range of 280 to 375 bp to obtain the genomic libraries. The DNA extraction and ddRadSeq protocol were carried out by the Australian Genomics Research Facility (AGRF).

Bioinformatic analysis

Raw Illumina reads were demultiplexed, cleaned, and quality checked using the module process_radtags in STACKs v.2.53 (Rochette et al. 2019), setting a Phred quality score of 20 (Mastretta-Yanes et al. 2015; Rochette and Catchen 2017). A reference sequence for the species D. guildingii was constructed from an initial mapping of the available dung beetle genome Onthophagus taurus (NCBI: JHOM00000000.2) using NextGenmap v. 0.6 (Sedlazeck et al. 2013) to inspect 3,200,000 reads per individual against the O. taurus reference. The resulting files were then organized and merged with Samtools v.1.17 (Li et al. 2009) and a final reference fasta file was generated using mpileup and seqtk v.1.3 (Li 2012). All read samples were then aligned against the new reference genome with NextGenMap.

The read depth was evaluated using ref_map.pl in STACKs (Rivera-Colón and Catchen 2022), and populations programs parameters to filter SNPs loci were: minimum percentage of populations of the presence of the locus (3%), minimum percentage of individuals with the presence of a locus (70%) and maximum heterozygosity observed per locus (90%). The average read depth was 37.3x. Potential loci under selection were detected using the R package pcadapt v.4.3 (Luu et al. 2016). For this, the Manhattan plot of SNPs with a min MAF > 0.05 was plotted and a frequency histogram was constructed based on the p-values (Luu et al. 2016). Twelve SNPs potentially under selection were removed to obtain a data set of 2,587 SNPs under neutrality for subsequent analyses.

Genetic diversity and genetic structure

We estimated genetic diversity parameters, such as allelic richness (Ar), observed (Ho) and expected (He) heterozygosity using the R package hierfstat v0.5-11 (Goudet 2005). The number of private alleles (PrA) per fragment was calculated in poppr v.2.9.3 (Kamvar et al. 2014), while FIS per fragment and nucleotide diversity (\(\:\pi\:\)) was obtained in STACKs v.2.53 (Rochette et al. 2019).

Genetic structure was evaluated by implementing a sparse non-negative matrix factorization (snmf) from LEA package v.2.8 (Frichot and François 2015) to identify the best number of genetic clusters, and setting K = 1 to 6, 10 repetitions and 1000 iterations. Moreover, we applied Principal Component Discriminant Analysis (DAPC) and a Spatial Principal Component Analysis (sPCA) in the R package adegenet v.2.1.1 (Jombart et al. 2008). For the DAPC, the fragments were used a priori for population grouping, and the optimal PCs to retain were optimized using xval cross-validation criteria (Jombart 2008). For the sPCA we used the Delaunay triangulation to obtain the connection network between fragments. Furthermore, an analysis of molecular genetic variance (AMOVA) was carried out in poppr v.2.9.3 testing a priori the amount of genetic variation explained by the three biogeographic regions, and by the six TDF fragments.

We performed a Mantel correlation to test the effect of isolation by distance (IBD) using Nei FST pairwise genetic distances and Euclidean distances using the R package ade4 v.1.7 with 999 permutations (Dray and Dufour 2007). Moreover, Pearson correlations between fragment area, perimeter, and fragmentation index with genetic diversity, and FIS estimates were tested using the R package past v.4.02 (Hammer et al. 2001).

Historical demography

To infer historical demographic changes, we estimated the Tajima’s D estimator for each fragment in pegas v.1.0 (Paradis 2010). Tajima’s D with significant negative values suggests rejection from neutrality and can be interpreted as population expansion. Also, we estimated Bayesian skylineplots in BEAST v.2.7.1 (Drummond and Bouckaert 2015) to assess effective population changes (Ne) over time. We used the substitution model GTG + R as selected from the Akaike Information Criterion (AIC) in JModeltest v.2.1.10 (Darriba et al. 2012). A strict molecular clock with 20 million generations, a clock rate of 0.012 mutations per year (Gunter et al. 2016), and trees and parameters sampled every 2000 iterations were chosen. The log file was viewed in TRACER v1.7.1 (Rambaut et al. 2018) after analysis to ensure that effective sample sizes (ESS) for all priors were greater than 100.

Results

Estimates of genetic diversity and patterns of genetic structure

Genetic diversity for the species level was low (Ho = 0.07, He = 0.09). Genetic diversity estimates per fragment were similar and, with the highest values for TG (Ho = 0.083, He = 0.09) and SFFC (Ho = 0.084, He = 0.09), and the lowest for COL (Ho = 0.072, He = 0.086) (Table 2). The fragment with the most exclusive alleles was SFFC with 325 and the lowest with 74 in PNNT. Inbreeding FIS values ranged from 0.076 to 0.122.

Table 2 Genetic diversity estimates of D.guildingii in six TDF fragments
Full size table

Results of genetic structure in LEA showed that the most likely number of genetic clusters was K = 2, of which we observed high admixture (Fig. 2A). The most differentiated fragments were BRUN at the northeast and SFFC to the southwest, each belonging to one of the two groups identified. Results from the DAPC along axis 1 also showed the separation of fragments into two groups: one formed by the northernmost fragments of the Colombian Caribbean coast, BRUN, TG, and PNNT, and the second group by the south fragments, SFFC, COL, and RCM (Fig. 2B). The sPCA retrieved the same spatial pattern as the DAPC along axis 1, with the separation of BRUN, TG, and PNNT in one group, and SFFC, COL, and RCM in a second group (Fig. 2C).

Fig. 2
figure 2

Results of genetic structure in D. guildingii: (A) LEA analysis as piecharts and a barplot, (B) Discriminant Analysis of Principal Components (DAPC), and (C) Spatial Analysis of Principal Components (sPCA). The two different colors denote the two genetic groups identified. Abbreviations: Sierra Nevada de Santa Marta (SNSM) (fragments PNNT and TG); Distrito Montes de María and Piojó (DMMP) (fragments COL, RCM and SFFC); Guajira (GUAJ) (fragment BRUN)

Full size image

The pairwise FST indices showed moderate genetic differentiation between PNNT and COL (FST = 0.070) and the lowest between RCM and SFFC (FST = 0.031) (Supplementary material). The AMOVA showed that the genetic proportion of genetic variance explained between the three biogeographic regions (1.5%) was similar to the proportion between fragments (1.7%). For both analyses, most of the genetic variation was within individuals (71%) (Table 3).

Table 3 Results of AMOVA among six TDF fragments and three biogeographic areas
Full size table

Relationship between habitat fragmentation and genetic diversity

Four fragments (COL, RCM, TG, and BRUN) were insularized (IF < 0.5), one was highly fragmented (PNNT, IF = 0.53) and the other was moderately fragmented (SFFC, IF = 0.7). The most critical situation was for RCM in Atlántico, which was the most insularized (IF = 0.17) (Table 1). None of the Pearson correlations between fragment area, perimeter, and the F index with genetic diversity parameters were significant (P > 0.05). Moreover, we found no significant effect of isolation by distance on genetic differentiation (r = 0.052, P = 0.36).

Demography history

Results from the historical demography analyses using the complete neutral SNPs data set showed that the best nucleotide substitution model was GTG + R (-Ln = 52260.66) and stability was achieved for the reconstruction parameters (posterior-ESS = 380; likelihood-ESS = 180). The Bayesian skyline plot of Ne through time showed a population expansion 2.5 million years ago (Fig. 3). This result was in agreement with the ones of the Tajima’s D estimator, which values were negative and significant suggesting population expansions for all TDF fragments: PNNT (D = -4.16, P < 0.05), COL (D = -4.14, P < 0.05), BRUN (D = -4.07, P < 0.05), RCM (D = -3.56, P < 0.05) and TG together with SFFC (D = -2.37, P = 0.018).

Fig. 3
figure 3

Bayesian analysis of demographic history in D. guildingii. The solid blue line represents the median population size (Ne) and confidence intervals are shown as blue areas. The dashed line indicates the date of the population expansion

Full size image

Discussion

Genetic diversity and its relationship with habitat fragmentation

Overall, it was observed low genetic diversity for D. guildingii compared to the values observed in other members of the Scarabaeidae family, such as Trypoxylus dichotomus (Ho = 0.097–0.17) (Yang et al. 2021). Inbreeding values were also within the range reported for other Coleoptera using SNPs and microsatellites for the genus Eucryptorrhynchus (Zhang et al. 2021).

The lowest value of expected heterozygosity and the highest level of inbreeding were observed for PNNT, which was the most fragmented site. This observation agrees with another study that observed that anthropogenic pressures, such as ecotourism and other activities affected the quality of the habitat and the abundance of D. guildingii (Noriega et al. 2020). Other fragments, such as BRUN suffer from mining activities with unfavorable implications for the ecosystem (Gómez-Betancur and Vilardy 2022), while RCM is small and where agricultural activities are practiced. Both fragments are insularized (IF = 0.27 and 0.17 respectively) but with similar levels of genetic diversity to other sites. Moreover, livestock, agriculture, and mining are well recognized in Montes de María (Sampedro-Marín et al. 2011), where the COL fragment occurs. The IF in COL, denoting isolation (IF = 0.35) was consistent with those of other localities in Sucre, such as Tolú Viejo and Colosó (IF = 0.36) (Galván-Guevara et al. 2015). COL showed lower genetic diversity and a lower number of private alleles compared to other insularized fragments, which suggests that this fragment is more prone to the erosion of genetic diversity due to land-use changes.

The surrounding areas in TG are highly modified by anthropogenic activities, such as cattle grazing, pig farming, and agriculture, which isolate the population of D. guildingii. Due to the vulnerability of the species to move between areas with low forest cover, these land-use modifications may change matrix permeability, restricting dispersal and gene flow between populations (IAVH 2016). Interestingly, TG has the second highest number of unique alleles, which may be due to the influence of the Serranía de Perijá, a region of high biological diversity and endemism (López-O et al. 2014).

In contrast to the rest of the fragments, the highest degree of genetic diversity and number of private alleles were found in the fragment SFFC, which agreed with its highest conservation status (IF = 0.71). This fragment is within the Serranía de San Jacinto and the primary conservation area for San Juan Nepomuceno, which falls within the Local System of Protected Areas (Galván et al. 2015; Jiménez et al. 2018). Additionally, in this region, socio-ecosystemic connectivity initiatives have been implemented that integrate multiple social and governmental actors to ensure the conservation of vertebrates such as the monkey Alouatta seniculus and Saguinus oedipus, important sources of excrements for dung beetles and the Malibú Biological Corridor for big cats such as Panthera onca (Ange et al. 2020). These initiatives help to maintain viable populations and would indirectly benefit D. guildingii at the genetic level, and for this reason, this protected area in the department of Bolívar is an important natural heritage of the Caribbean region (Bermúdez-Wilches 2012).

Patterns of spatial genetic structure

The reported values of FST for D. guildingii agree with those described for species of the same genus, such as D. speciosissimum (FST = 0.029) (Linck et al. 2020). Interestingly, the absence of isolation by geographic distance suggests that the patterns of genetic differentiation are due to other processes such as genetic drift, geologic history, landscape resistance, or environmental differences (Slatkin 1993; Peterson and Denno 1998; Taylor et al. 2020).

Specifically, patterns of genetic variation were not explained by the three biogeographic regions as initially thought. Instead, from clustering analyses we found consistent patterns in two genetic groups in D. guildingii, but that also can be related to the different relief and geological processes forming the mountain systems in the region. The first group included BRUN, TG, and PNNT, and is defined by the Oca fault in the Guajira, the Santa Marta fault in Magdalena, and the Serranía de Perijá in Cesar, in which the geotectonic history of northern Colombia shows similarities between the Sierra Nevada de Santa Marta (SNSM), the Ranchería river basin, and the Serranía de Perijá (Rodríguez and Londoño 2002; Chicangana et al. 2011). Based on our data, BRUN, situated north of the Colombian Caribbean coast, does not function as an isolated biogeographic region as previously assumed. Instead, it is closely related to other prominent mountain massifs with a shared history, such as PNNT and TG in the SNSM region.

The differentiation of COL, SFFC, and RCM between PNNT, TG, and BRUN can be explained by different biogeographic influences (Jiménez et al. 2018). Both biogeographic regions experienced different paleoclimatic events, causing changes in the distribution of genetic variation and other biotic responses (Riddle et al. 2008). Likewise, the lower differentiation between the SFFC and the RCM can be associated with the shared geomorphological history between the Luruaco and San Jacinto anticlines (Banco de Occidente 1999), and the fact that there is still a forest belt between Bolívar and Atlántico which would promote gene flow.

Historical demography and climatic events

The negative Tajima D values for all fragments suggest demographic expansions after bottleneck events (Dogan and Dogan 2017). With the complete data set for the species level, the estimated population expansion event of 2.5 million years ago corresponds to the transition between the Pliocene and Pleistocene, a period of orogenic and paleoclimatic relevance that marked the evolutionary history of many biological groups, such as vertebrates and insects. For instance, it has been documented that the Canthonini tribe, where Deltochilum was formerly included, appeared in the Pliocene. During this period the reduction in rainfall and temperature up to 5 °C (Hooghiemstra and Ran 1994) were important for the development of seasonally dry areas that marked the trajectory of the TDF and allowed the diversification of Alouatta primates with the expansion of families such as Fabaceae. Thus, the increase in effective population size of D. guildingii coincides with the phylogeny of Alouatta in South America, during which the primate clades expanded and diversified between 3.3 and 2.4 million years ago (Cuervo-Díaz et al. 1986; Cortés-Ortiz et al. 2003; Padilla-Gil and Halffter 2007; Gattepaille et al. 2013; Mercado-Gómez et al. 2019).

Conservation implications

For the first time, a population genetic study for D. guildingii was conducted, contributing to our understanding of the genetic status of dung beetles in the Colombian Caribbean. Large telecoprids such as D. guildingii, which are normally associated with primate excrement and dense forest cover, and with low dispersal capacity may be affected by the reduction in excrement availability and limited connectivity under the current TDF situation with accentuated differences between conserved and not protected areas (Díaz et al. 2010). According to our results, D. guildingii populations are vulnerable to the effects of genetic drift and inbreeding as populations become smaller under increasingly isolated fragments and land-use changes.

Specific conservation efforts to protect viable populations of D. guildingii are necessary given the dynamics of each TDF fragment and their human threats. For example, increasing environmental vigilance and genetic monitoring in areas that have a high number of exclusive alleles but that are not included as protected areas and that have many anthropogenic pressures such as RCM and TG are critical, which, without proper intervention can be detrimental to the gene pool.

COL and PNNT have lower genetic diversity than other areas without permanent environmental monitoring and demonstrate that even in protected areas, dung beetles can be affected by human disturbance at different levels. Specifically, it would be important to increase the genetic monitoring in COL and surrounding reserves because the high frequency of local extractive pressures in Sucre may affect the population densities of vertebrates such as A. seniculus on which D. guildingii relies as sources of excrement (Ochoa et al. 2011; Martínez-Hernández et al. 2012; Fajardo-Patiño and De la Ossa 2014).

Anthropogenic disturbances in the TDF of the Colombian Caribbean cause changes in the ecological requirements and microhabitat conditions for dung beetles due to the generation of relict areas with variable excrement conditions (Martínez-Hernández et al. 2012). These modifications in habitat connectivity increase the occurrence of small and isolated patches and affect the composition and abundance of dung beetles (Pryke et al. 2013; Pablo-Cea et al. 2020). Future studies are needed to expand the number of fragments sampled for genetic analysis and to evaluate if the landscape matrix or the environment influences patterns of gene flow in D. guildingii, which can inform the implementation of dispersal corridors among isolated fragments and the effect of climate adaptation for the long term viability of populations (Ho and Shapiro 2011).