Field Collection and Laboratory Methods.

We sampled a total of 319 adult specimens of P. sextuberculata from 16 localities throughout its known distribution in Brazil. The collecting sites (Fig. 1) included the following: (1) São Paulo de Olivença (SPO; Solimões River, n = 24); (2) ESEC Juami-Japurá (JAP; Japurá River, n = 36); (3) RESEX do Alto Juruá (AJU; Juruá River, n = 10); (4) RESEX do Médio Juruá (MJU; Juruá River, n = 31); (5) RESEX do Baixo Juruá (BJU; Juruá, n = 12); (6) Parque Nacional do Viruá (PNV; Branco River, n = 12); (7) REBIO Abufari (ABU; Purus River, n = 19); (8) RDS Piagaçu-Purus (PUR; Purus River, n = 19); (9) Manicoré (MAD; Madeira River, n = 18); (10) Nhamundá (NHA; Nhamundá River, n = 14); (11) Trombetas Tabuleiro (TRO; Trombetas River, n = 21); (12) Terra Santa (TSA; Urupuanã Creek, n = 12); (13) Oriximiná (ORI; Sapucuá Lake, n = 13); (14) Barreirinha (BAR; Andirá River, n = 16); (15) Parintins (PAR; Amazonas River, n = 22); and (16) Baixo Xingu (XIN; Xingu River, n = 40).

View Figure

• Download as Powerpoint
• Download Figure

Blood was sampled from the femoral artery and preserved in 95% alcohol. The DNA was extracted from blood samples following the phenol-chloroform extraction method of Sambrook et al. (1989). Subsequently, the control region of mitochondrial DNA (mtDNA) was amplified using PCR primers forward primer DLSex 5′-AGTGCTCTTCCCCATATTATG-3′ and the reverse primer 12SR5 5′-GTCAGGACCATGCCTTTGTG-3′ (Hrbek and Farias 2008). Sequencing reactions were performed using both the forward and reverse primers according to the manufacturer's recommendation and resolved on an ABI 3130XL automated sequencer (Thermo Fisher Scientific, Brazil).

Descriptive Statistics.

Sequences were imported and manually verified with the software Geneious (http://www.geneious.com; Kearse et al. 2012). Alignments were carried out in Geneious using the ClustalW tool (Thompson et al. 1996) under default conditions. Subsequently, we estimated the number of segregating sites (S), number of haplotypes (h), haplotype diversity (Hd) (Nei 1987), and nucleotide diversity (Π) (Nei and Li 1979) using DNAsp v5.10.1 (Librado and Rozas 2009). To visualize the relationship among the haplotypes, we estimated a haplotype network in the program Haploviewer (Salzburger et al. 2011) based on a maximum likelihood gene tree resolved with RAxML (Stamatakis 2006).

Population Structure and Gene Flow Estimate.

We performed a Bayesian analysis of population structure to identify the number of distinct genetic clusters within the data set using the mixture model implemented in the software BAPS v. 6.0 (Corander et al. 2004). We set the K cluster interval as 1–16 (the number of localities sampled) and ran 5 independent simulations for each K (1–16). We further assessed population structure using an analysis of molecular variance (AMOVA; Excoffier et al. 1992) implemented in the software Arlequin v3.5 (Excoffier and Lischer 2010).

In order to assess gene flow among sampled localities, we performed a maximum likelihood analysis with the program MIGRATE-N (Beerli and Palczewski 2010). We ran 10 short chains, sampling each chain 10,000 times, and then 6 long chains, sampling each chain 1,100,000 times and discarding the first 100,000 samples as burn-in. In this analysis the population genetic parameter (θ) was used to estimate the female effective population sizes (Nef), and we assumed a mutation rate of 2.48 × 10⁻⁷ mutations per site per year for the control region derived from Chelonia mydas (Chassin-Noria et al. 2004).

We also tested the hypothesis of IBD, which describes a local accumulation of genetic differences when the gene flow among populations is geographically restricted (Wright 1943; Slatkin 1987), such that there is an increase in genetic differentiation as geographic distances increase. To test IBD, we correlated a matrix of Ф_ST/(1 − Ф_ST) values with both a matrix of the Euclidean distances and the river distances (following the river courses) in kilometers between each pair of sampling localities. The river distances were calculated in R (R Development Core Team 2011) using the package gdistance (van Etten 2017). We correlated the Ф_ST with the distance matrices using a Mantel test (Mantel 1967) and tested the significance of the Spearman correlation (Spearman 1904) with 10,000 permutations with the package ade4 (Dray and Dufour 2007) in R (R Development Core Team 2011).

Demographic History.

We estimated population demographic histories by mismatch distribution in the program Arlequin v3.5 (Excoffier and Lischer 2010), which estimates the sum of squared deviations (SSD) and Harpending's raggedness index (Hri; Harpending 1994); these estimates were tested for significance through 10,000 permutations under the null model of population expansion. We verified whether the samples were at mutation–migration–drift equilibrium using Tajima's D (Tajima 1989) and Fu's Fs (Fu 1997) neutrality tests. Significant values for these tests would indicate either that the mitochondrial sequences are not evolving according to the hypothesis of selective neutrality or that the populations were previously subdivided or experienced fluctuations (or both) in the past (Hartl and Clark 2006). We implemented the 2 tests in the software Arlequin v3.5 (Excoffier and Lischer 2010), testing their significance with 10,000 random samples.

Genetic Diversity and Demography.

Genetic diversity values obtained for the individuals of P. sextuberculata were fairly heterogeneous throughout the sampled area, ranging from Hd = 0.14 in Nhamundá to Hd = 0.91 in the upper Juruá River, with an average Hd = 0.61. Compared with other species of the genus, the average genotypic diversity of P. sextuberculata was the same as that obtained from data of the mitochondrial DNA control region of P. erythrocephala, Hd = 0.61 (Santos et al. 2016) and similar to that of P. expansa, Hd = 0.65 (Pearse et al. 2006) and P. unifilis, Hd = 0.70 (Agostini 2016). However, these values of genetic diversity are much higher than those reported by Silva et al. (2011) with ND1 data for P. sextuberculata, probably owing to the low level of polymorphism of the ND1 gene compared with the polymorphism of the mitochondrial DNA control region. Across animal mitochondrial genomes in general, the level of genetic variability in the control region is approximately five times greater than that in protein-coding genes (Kocher et al. 1989).

Low values of genetic and nucleotide diversity were obtained for individuals from the localities of NHA, MAD, TSA, and XIN. These low levels of diversity may be explained by 1) historical colonization of these populations by a small number of individuals, or 2) turtles from these sites being under greater hunting pressure than those in the other sampled areas (or both). Colonization by a few individuals is unlikely, considering that there are other geographically close sites and that the observed high levels of gene flow should naturally increase genetic diversity through the introduction of migrants over time. Alternatively, consumption of freshwater turtles as a source of protein by river dwellers has drastically depleted the populations of these animals, particularly of the largest species, P. expansa, leading to the consumption of the smaller species, P. unifilis and P. sextuberculata, in place of P. expansa (Smith 1979; Rebêlo and Pezzuti 2001). Several studies conducted in the Amazon basin have reported declines in the abundance of chelonians in areas close to human settlements (Conway-Gómez 2007; Alcântara et al. 2013). For example, Alcântara et al. (2013) found that the density of P. unifilis tended to increase with increasing distance from urban centers, an effect that could reflect a reduction in local population sizes and a consequent loss of genetic diversity. Among the populations identified here as exhibiting low diversity, those in the lower Amazon basin (in the municipalities of Nhamundá and Terra Santa) are surrounded by a large number of villages and riverine communities, which may explain the relatively lower levels of genetic diversity of these localities. The low levels of genetic diversity found in these localities may also be the result of ecological processes, local adaptations, or historical events not investigated here.

Another factor to be noted is the effect of the seasonal dynamics of the Amazonian wetlands, which can modify the availability of food and nesting sites, causing the local abundance of individuals to fluctuate over time due to seasonal movement of individuals. This has been observed in P. sextuberculata in the Mamirauá Sustainable Development Reserve (SDR; Fachín-Terán et al. 2003) and more recently in P. unifilis in the Xingu River (Alcântara et al. 2013). Our findings indicate a high degree of connectedness and the absence of a clear pattern of genetic structuring, which is consistent with these observations of seasonal movement of individuals.

Generally speaking, genetic diversity at most of the sampled localities indicates that the historical human exploitation of this species has not yet dramatically affected its levels of mtDNA diversity. This was evidenced by the mismatch distribution analysis, which indicated constant population sizes in most of the populations, and by the neutrality test which showed that most populations are in mutation–migration–drift equilibrium. The few exceptions to this pattern do not exhibit a correlation with the localities that presented low levels of genetic diversity. The large effective population size estimated for the species (5.3 million individuals) is a reflection of historical diversity of this species and may not reflect the current census size if the species suffered recent population reduction.

The high levels of genetic diversity observed in most P. sextuberculata populations in the Brazilian Amazon are relevant in evolutionary, but not ecological, time and should be viewed with caution because the mtDNA locus is a better indicator of demographic history over evolutionary time frames (White et al. 2008). Similarly high genetic diversity has also been observed in other turtles species that have been extensively exploited in the past and whose populations were significantly reduced (Kuo and Janzen 2003; Pearse et al. 2006; Willoughby et al. 2013). An earlier study documented the demographic impacts of overharvesting in P. sextuberculata (Fachín-Terán et al. 2003), so we interpret the high mtDNA haplotype diversity of P. sextuberculata as a genetic “signature” of recent evolutionary history (Pleistocene/Holocene) and life-history traits (longevity, generation overlap, late maturation, and multiple paternity; Erickson et al. 2015; Fantin et al. 2015; Freda et al. 2016). Future studies based on highly variable nuclear markers with high allelic diversity (e.g., microsatellites) will provide a better signal of recent demographic history.

Connectivity and Population Structure.

Podocnemis sextuberculata shows a panmictic pattern of population structure across a large part of its range, as confirmed by the results of the pairwise Ф_ST, haplotype networks, and Bayesian Analysis of Population Structure. However, samples from the NHA and XIN localities had genetically structured signatures in all of these analyses. Silva et al. (2011) reported a lack of population structuring among P. sextuberculata localities in central regions of the Amazon basin, and our findings now extend this pattern of panmixia to a larger part of Amazonia, spanning distances from 47 km (between TSA and ORI) to 3641 km (between AJU and XIN). This finding corroborates the high migration levels reported in previous studies (Fachín-Terán et al. 2006; Andrade 2012, 2015) and, relative to other Amazonian Podocnemis species, P. sextuberculata shows the most panmictic pattern.

The patterns of genetic structure in these Amazonian chelonians have proven to be varied. Pearse et al. (2006) found a genetic structuring pattern in P. expansa, consistent with drainage basins, indicating low structuring within the basins and strong genetic structuring between drainage basins with a pattern of absence of genetic differentiation in the channel of the Solimões-Amazonas river system. Escalona et al. (2009) observed the same pattern for P. unifilis, but found some cases of structuring within the same drainage basin, particularly within the Orinoco River basin, which they attributed to IBD and to the environmental heterogeneity of the drainage system, including the presence of physical barriers such as waterfalls and rapids. Agostini (2016) found a pattern similar to that observed in P. expansa in P. unifilis populations in Brazilian Amazonia, but showing population groups structured by the presence of waterfalls and rapids, which are present in several Amazonian rivers. Santos et al. (2016) found panmixia in much of the area of occurrence of P. erythrocephala but found genetic structuring among populations that are separated by waterfalls and rapids, even suggesting that the Amazon River could be a barrier to gene flow of the species because the occurrence of P. erythrocephala is limited to blackwater and clearwater rivers. In all these cases, large stretches of drainage have confirmed the panmictic pattern of the species in the mainstream of the Solimões-Amazonas River system; this region is characterized by varzea (seasonally flooded forests). Similar panmictic population structures have also been found in other aquatic Amazon basin species (e.g., Hrbek et al. 2005; Vasconcelos et al. 2006, 2008; Farias et al. 2010; Frederico et al. 2012; Willis et al. 2012; Ochoa et al. 2015; Machado et al. 2017). This pattern supports the hypothesis that the populations of freshwater vertebrates in the mainstream of the Solimões-Amazonas River system are not genetically structured because their life cycles are strongly influenced by the flood pulse of the wetland system, which favors the long-distance movement of animals for breeding and feeding.

The genetic structuring observed in the individuals from the localities of NHA and XIN and the indexes of genetic differentiation among all the population pairs were not correlated to a pattern of IBD, suggesting that for this species geographic river distances are not causally related to accumulation of genetic differences among populations. Although IBD is a simple pattern of genetic structuring often found in populations (Jenkins et al. 2010), more-complex models that also consider the role of the environmental heterogeneity of the landscape, such as isolation-by-environment (IBE), may be responsible for part of the structuring (Wang and Bradburd 2014). An IBE pattern may arise when environmental conditions affect the synchronization of reproduction and the dispersal patterns among populations or may be the result of strong natural selection (Sexton et al. 2014), acting against genotypes poorly adapted to the local environment (Nosil et al. 2005). The context of flood-pulse dynamics in the Amazon basin seems to be conducive to maintaining genetic diversity and the development of more-complex population genetic patterns (Flowers et al. 2002; Machado et al. 2017) and deserves further investigation in the future.

The genetic structuring observed in the municipality of NHA seems to be explained by the low number of haplotypes; of the 2 haplotypes observed at this locality, one is shared with the other localities and the other is exclusive (see Fig. 2). The presence of a “private” haplotype restricted to this location was probably the determining factor for the result observed here because Ф_ST is based on the variance of the frequency of haplotypes between population pairs. The results obtained in BAPS (Fig. 3) indicate that the NHA locality is part of the same biological group occurring in most of the sampled sites, i.e., the individuals at this location are part of the larger panmictic population. Conversely, turtles from the Xingu River locality are characterized by several segregating sites that separate their locality from all others (Fig. 2). This sample is recovered as a structured biological group in the BAPS analysis, and it differs significantly from all other samples in pairwise Ф_ST values. Genetically, the individuals of the Xingu River site should be considered a distinct management unit (MU; Moritz 1994). This population should be studied further to determine if evolutionary (e.g., local adaptation), ecological, or historical events are the causative agents of this differentiation; however, regardless of the cause, they will remain a distinct management unit.

Conservation and Management Units

Within a Threatened Watershed. — Based on the results presented here and following the criteria of Moritz (1994), who defined MUs as “… populations with significant divergence of allele frequencies at nuclear or mitochondrial loci, regardless of the phylogenetic distinctiveness of the alleles,” we suggest that the populations of P. sextuberculata of the Brazilian Amazon basin are composed of 2 MUs, one represented by all populations except those of the Xingu River and the other represented by the populations restricted to the Xingu River. Conservation strategies should be designed to treat these 2 MUs separately, but we emphasize the need for follow-up studies based on sampling of the nuclear genome, including both selectively “neutral” Mendelian loci (e.g., microsatellites), loci possibly under the influence of natural selection (Funk et al. 2012), or both; either of which may show patterns of population structure that differ from the mitochondrial genome structure (Palsbøll et al. 2007).

In the Amazon basin currently, the greatest threats to aquatic turtles include both direct exploitation and modification of the watersheds, mainly from the construction of hydroelectric dams. It is important to identify MUs throughout the Amazon basin, for which massive systems of hydroelectric dams are likely to isolate populations of these turtles and other aquatic species. The large Belo Monte dam on the Xingu River deserves considerable attention because of its severe impacts in isolating this entire river from all others (Lees et al. 2016). Gene flow was almost entirely in the direction from the Xingu population, above the Volta Grande waterfall, to the Amazon. Directionality of this gene flow is most likely facilitated by the flow of the river itself. In this way individuals above the waterfall are important contributors of the genetic diversity present in the drainage of the Amazon River; thus this locality behaves as a source population. With the construction of the hydroelectric dam in the Volta Grande region of the Xingu, this dynamic will inevitably change, potentially resulting in the complete isolation of Xingu populations upstream of the Volta Grande from those downstream.

It is also important to consider the modifications of downstream sediment loads deposited below dam sites including the timing, location, and extent of below-dam sediment deposition and sand-bar formation (crucial nesting sites for all turtles). Finally, flood pulses in the Amazon basin drive reproduction of many if not most floodplain-associated fish species (Junk 1997). Therefore, disruption of these cycles will also influence availability of food for P. sextuberculata and other piscivorous species.

Although P. sextuberculata is protected by the Federal Wildlife Protection Law, Law No. 5197 (http://www.planalto.gov.br/ccivil_03/leis/L5197.htm), the species is still subject to predatory and clandestine commercial and subsistence hunting of adults and collection of eggs. Currently, P. sextuberculata is listed by the International Union for Conservation of Nature (IUCN; http://www.iucnredlist.org/) as Vulnerable. For this reason, it is necessary to invest in actions which promote the region's riverine communities' understanding of the need for turtle conservation and management and to involve these communities in preserving the turtle nesting areas. The environmental management agencies must work together with the units of conservation (or any other type of protected areas) to create strategic conservation areas for adult females as well as for their nesting beaches. Moreover, it is essential to study additional populations to better understand population dynamics, sex ratios, and abundance, as was done by Fachín-Terán et al. (2003) in the Mamirauá SDR. This will enable us to expand the body of knowledge about the ecological characteristics of not only P. sextuberculata but of all Amazonian Podocnemis species and to better identify and ameliorate anthropogenic threats to these species.