Land planarians (Platyhelminthes) as a model organism for fine-scale phylogeographic studies: understanding patterns of biodiversity in the Brazilian Atlantic Forest hotspot



    1. Departament de Genètica, Facultat de Biologia, Universitat de Barcelona, Barcelona, Spain
    2. Institut de Recerca de la Biodiversitat, Universitat de Barcelona, Barcelona, Spain
    Search for more papers by this author

    1. Escola de Artes, Ciências e Humanidades, Universidade de São Paulo, São Paulo, Brazil
    Search for more papers by this author
  • J. ROZAS,

    1. Departament de Genètica, Facultat de Biologia, Universitat de Barcelona, Barcelona, Spain
    2. Institut de Recerca de la Biodiversitat, Universitat de Barcelona, Barcelona, Spain
    Search for more papers by this author

    1. Departament de Genètica, Facultat de Biologia, Universitat de Barcelona, Barcelona, Spain
    2. Institut de Recerca de la Biodiversitat, Universitat de Barcelona, Barcelona, Spain
    Search for more papers by this author

Marta Riutort, Departament de Genètica, Facultat de Biologia, Universitat de Barcelona, Diagonal 645, 08028 Barcelona, Spain.
Tel.: (+34) 93 4035432; fax: (+34) 93 4034420; e-mail:


The Brazilian Atlantic Forest is one of the richest biodiversity hotspots of the world. Paleoclimatic models have predicted two large stability regions in its northern and central parts, whereas southern regions might have suffered strong instability during Pleistocene glaciations. Molecular phylogeographic and endemism studies show, nevertheless, contradictory results: although some results validate these predictions, other data suggest that paleoclimatic models fail to predict stable rainforest areas in the south. Most studies, however, have surveyed species with relatively high dispersal rates whereas taxa with lower dispersion capabilities should be better predictors of habitat stability. Here, we have used two land planarian species as model organisms to analyse the patterns and levels of nucleotide diversity on a locality within the Southern Atlantic Forest. We find that both species harbour high levels of genetic variability without exhibiting the molecular footprint of recent colonization or population expansions, suggesting a long-term stability scenario. The results reflect, therefore, that paleoclimatic models may fail to detect refugia in the Southern Atlantic Forest, and that model organisms with low dispersal capability can improve the resolution of these models.


The Brazilian Atlantic Forest (AF) is one of the richest and most threatened biodiversity hotspots of the world (Myers et al., 2000), still harbouring more than 8000 endemic species (Tabarelli et al., 2003). Currently, the remaining 10–16% of its original extent is fragmented in small isolated patches caused by human activities as agriculture and transport networks (Fonseca & Rodrigues, 2000).

Understanding the origin of this biodiversity is critical to address political issues about protection and management. The origin of current biodiversity at the AF is, however, highly controversial. Simpson (1979) proposed that geographic and climatic modifications caused by Tertiary geological events might have promoted species diversification that, in turn, would explain the high levels of biodiversity of the Atlantic Forest. In a different view, Haffer (1969) stated that the origin of this biodiversity would be the Pleistocene glacial cycles promoting vicariance of populations by the regression, fragmentation and expansion of the forest and by sea level changes. Paleoclimatic models, applied to test this latter hypothesis, have predicted two large areas of stability throughout the Late Quaternary in the Central Atlantic Forest (CAF; region extending north of the Doce River, Fig. 1), whereas southern regions (SAF) would presumably have had a lack of refuges (Carnaval & Moritz, 2008). These two scenarios may leave distinctive molecular fingerprint in the genome of organisms; a slow change associated with geological processes would allow populations to remain historically stable and smoothly adapt to the new conditions, thus displaying high nucleotide diversity. Alternatively, a rapid change associated with recent Pleistocene ice advances and retreats would lead to the characteristic footprint of recent bottlenecks or founder events and lower nucleotide diversity. Stability areas predicted by paleoclimatic models, with the subsequent differences in genome patterns, might explain the high species richness (Colinvaux et al., 2000; Ridgely & Tudor, 1996; Bates et al., 1998; Cabanne et al., 2008) as well as the high levels of nucleotide diversity found in some taxa in the CAF and the signs of population expansion in SAF (Pellegrino et al., 2005; Grazziotin et al., 2006; Cabanne et al., 2008; Fitzpatrick et al., 2009; Carnaval et al., 2009).

Figure 1.

 Distribution of the studied Atlantic Forest sampling localities. Parque Nacional da Serra da Bocaina (SB) (yellow diamond); Estação Biológica de Boraceia (BB) (green diamond); Parque Estadual de Intervales (PI) (light blue diamond) and Parque Nacional de Saint-Hilaire/Lange (SL) (dark blue diamond). CAF, Central Atlantic Forest area; SAF, South Atlantic Forest.

Nucleotide diversity patterns observed in some of these studies, however, suggest that paleoclimatic models fail to predict stable rainforest areas in the SAF. Cabanne et al. (2008), for instance, studied different subspecies of the bird Xiphorhynchus fuscus and detected some endemic lineages in the SAF that could be better explained if forested areas have persisted in that region. Carnaval et al. (2009) analysing three Hyla species, predicted a species-specific stable region for H. faber in the São Paulo area (situated within SAF) not found in the other two species. Moreover, other studies (Costa et al., 2000; da Silva et al., 2004; Pinto-Da-Rocha et al., 2005) found centers of endemism in São Paulo (and southern regions) that could be explained by the geological history of the Serra do Mar and Serra da Mantiqueira prior to the Pleistocene glaciations. Moreover, some studies dating diversification processes in the SAF region have also revealed older lineages (Prepleistocenic) than predicted by paleomodelling (Fitzpatrick et al., 2009; Grazziotin et al., 2006); these studies, nevertheless, also found the genetic signature of recent Pleistocene events. As it has been repeatedly proposed, the current data suggest that the processes generating the Atlantic Forest diversity are complex.

Here we have focused the analysis on a small SAF region situated around the border between SP and RJ states (SP-RJ region, Fig. 1). The paleoclimatic model of Carnaval & Moritz (2008) predicts that it had been covered by patches of forest 6000 years ago, but with no-forest at all 21 000 years ago (see Fig. 4 in Carnaval & Moritz, 2008). Moreover, some studies (as those referred above, Cabanne et al., 2008; Carnaval et al., 2009; Grazziotin et al., 2006) also show that this SAF region has been stable, or harbours a high number of endemic lineages. These conflicting results might reflect the difficulty of detecting refugia in this area, perhaps because SAF refugia were smaller than those in northern regions. Indeed, many of the studies surveying the AF region suffer from important limitations, such as a low spatial resolution of the predictions for studied species (e.g. vertebrates) with great dispersal capacity (Carnaval & Moritz, 2008). Specialized and low-dispersal taxa likely would be better models for phylogeographic studies, as their distribution and richness are highly influenced by the historical habitat stability (Cruzan & Templeton, 2000; Hewitt, 2004a,b; Graham et al., 2006; Garrick et al., 2004; Sunnucks et al., 2006). In this context, land planarians (Phylum Platyhelminthes) have adequate biological features as important physiological limitations (reduced retention capacity of body fluids and a high sensitivity to sunlight and heat) that limit their natural dispersal ability and, in consequence, they are likely to prevent the movement of planarians across forest fragments. Actually, they have been shown to be suitable to detect variability at low scale in a study on a system of montane forest in Australia (Sunnucks et al., 2006).

We have applied the state-of-the-art population genetics and phylogeographic tools on two terrestrial planarian species to study a locality within the SAF region, the Serra da Bocaina National Park (SB), to understand the relative role of Pleiostocene glaciations and geological events in shaping current patterns of genetic variability. If the SB habitat had been historically stable during long periods, even presenting relatively small fragments of forest, we should observe a pattern and level of nucleotide diversity different than that expected in areas shaped by recent glaciations, namely (i) a high intrapopulation genetic variability but lower than that expected between populations, (ii) no genetic evidence for bottlenecks or population expansions. Moreover and importantly, this pattern should be shared by the two studied species.

Materials and methods

Study site and fieldwork

We conducted the study in the SB Park, a 104 000 hectares protected area mainly covered with Atlantic Forest. The altitudes of the park range from sea level to 2132 m (Behling et al., 2007). Three more southerly parks [Estação Biológica de Boraceia (BB), Parque Estadual de Intervales (PI) and Parque Nacional de Saint-Hilaire/Lange (SL)] within and outside the Serra do Mar corridor were also sampled to be used for the interpopulation analyses (Fig. 1; Table S1).

We have studied two terrestrial planarian species Cephaloflexa bergi (Riester, 1938) and Geoplana goetschi sensu Marcus (1951). They were chosen because of their relatively high abundance and broad distribution across the south Atlantic Forest (see also Carbayo & Froehlich, 2008). We searched for animals under fallen logs and litter for 240 h during the day and directly on the ground during the night using led torches, as worms are active at night. A total of 143 specimens belonging to 27 species were found. Ten and 16 of the specimens are Cephaloflexa bergi (Riester, 1938) and Geoplana goetschi sensu Marcus (1951), respectively, both used in the phylogenetic and population genetics analyses. Specimens were taken within a distance less than four km from the main entrance to the park. Each animal was assigned a code and cut in two pieces. One part was fixed in formalin for histological analysis and the other in absolute ethanol for DNA extraction. The animals were identified by examining their external aspect and internal anatomy on histological sections together with the analysis of the cytochrome oxidase I (COI) sequences.

Morphological analysis

For each animal, we described the external features. Then, we poured boiling water on them to avoid them curling when fixing with 10% formalin, subsequently were stored in 80% ethanol. From ten specimens of each species, we embedded tissue blocks of the cephalic region, the pharynx or the copulatory organs in Paraplast, sectioned them at 7 μm, and stained with Mallory/Cason trichrome stain (Romeis, 1989). As the copulatory apparatus is the main structure for unequivocal identification, we reconstructed the copulatory apparatuses with a camera lucida attached to a light microscope.

DNA extraction and sequencing

We analysed both the mitochondrial COI gene and the nuclear ribosomal internal transcribed spacer (ITS-1). DNA was isolated from the specimens preserved in 100% ethanol using the Wizard® Genomic DNA Purification Kit (Promega, Madison, WI, USA). We amplified a 1 kb fragment covering part of the COI gene by polymerase chain reaction (PCR), using the following primers: BarT (ATGACDGCSCATGGTTTAATAATGAT; specifically designed for this study) and COIR [CCWGTYARMCCHC-CWAYAGTAAA (Lázaro et al., 2009)]. To obtain DNA sequences we used the following internal primers: BarS (GTTATGCCTGTAATGATTG) and BBC (CCAAAAGAAAAATCCTTNCC) specifically designed for this study, and COIF [CCNGGDTTTGGDATDRTWTCWCA (Lázaro et al., 2009)]. For the nuclear fragment ITS-1, we amplified 500 bp using the primers ITS9F (GTAGGTGAACCTGCGGAAGG) and ITSR (TGCGTTCAAATTGTCAATGATC) from Baguñàet al. (1999). The same primers were used to obtain the sequences. The PCR amplification conditions for the COI region were: 30 cycles of 50 s at 94 °C, 45 s at 43 °C and 50 s at 68 °C, with an initial denaturation step of 2 min at 95 °C, and a final extension step of 4 min at 68 °C. The amplification reaction was performed in 25 μL volume, using Go Taq® DNA polymerase (Promega) and the DNA template. Annealing temperature for ITS-1 was 45 °C, the rest of conditions were the same applied to COI. Amplification products were purified with the IllustraTM GFX PCR DNA Purification Kit (GE Healthcare, Buckinghamshire, UK) and directly sequenced using the PCR amplification primers or the internal ones. We determined the DNA sequence from both strands using Big-Dye (3.1, Applied Biosystems, Foster City, CA, USA); the reaction products were separated on the ABI Prism 3730 automated sequencer (Unitat de Genòmica dels Serveis Científico-Tècnics de la UB).

Molecular datasets

COI and ITS-1 alignments were obtained using MAFFT version 6 (Katoh & Toh, 2008) with a final editing with Bioedit v. software (Hall, 1999); for COI sequences, we used the translated amino acid sequences to guide the nucleotide alignment. We generated two COI datasets (of 822 bp each). COI-1 dataset includes all presumptive G. goetschi and C. bergi individuals (from the SB and the other three localities) together with sequences of closely related species and Geoplana burmeisteri and Geoplana ladislavii as outgroups (Table S1). COI-2 dataset includes only the sequences of C. bergi and G. goetschi. As the ITS-1 has length variability among species, caused by a high number of insertions and deletions, we performed independent alignments for each species and did not include an outgroup. Cb-ITS dataset contains C. bergi sequences (508 bp), and Gg-ITS dataset G. goetschi (428 bp).

Phylogenetic and population genetics analyses

We estimated the DNA sequence evolution model that best fits the data using jModelTest 0.1.1 (Posada, 2008) and applying the Akaike Information Criterion (AIC) for all datasets. We estimated the phylogenetic relationships by Maximum Likelihood (ML) implemented in RAxML 7.0.0 software (Stamatakis, 2006), and Bayesian Inference (BI) using MrBayes v. 3.1.2 (Ronquist & Huelsenbeck, 2003), and applying the next more complex evolutionary model to the estimated with jModeltest present in the inference program. Bootstrap support (BS) values (Felsenstein, 1985) were obtained from 10 000 replicates in ML analyses. For the BI, we carried out 3 million generations using two independent runs. Markov chains were sampled at intervals of 100 generations to include 30 000 data points. The consensus tree was estimated after removing the first 500 saved trees as ‘burn-in’ (to avoid including samples before reaching stationarity), the “burn-in” value was determined by inspection of the likelihood of the trees obtained in each generation. A 50% majority rule tree was obtained from the remaining data points. We estimated the Median Joining Networks (COI-2 dataset plus two outgroup species) to explore the haplotype relationships within and between populations in the two species using NETWORK software (Bandelt et al., 1999) with the default options. Haplotype networks for the SB population were constructed using the program TCS v1.21 (Clement et al., 2000).

We used the program DnaSP v5.10 (Librado & Rozas, 2009) to conduct most of the population genetic analysis. The levels and patterns of nucleotide diversity were estimated by the haplotype (HD) and nucleotide diversity (π); (Nei, 1987), whereas the levels of DNA divergence among populations from the same species by Dxy parameter (Nei, 1987). We carried out three different approaches to determine if the pattern of polymorphism conforms to that expected under the neutral hypothesis. On the one hand, we applied three neutrality tests that use different pieces of information to try to detect the specific fingerprint of recent population expansions, heavy bottlenecks or other selective and demographic scenarios: Tajima’s D (Tajima, 1989), Fu’s Fs (Fu, 1997) and R2 (Ramos-Onsins & Rozas, 2002) statistics. Their statistical significance was estimated by coalescent computer simulations (10 000 replicates). Secondly, we also searched for atypical patterns of variation along the DNA region by analysing the levels of polymorphism and divergence by the sliding window method (Rozas, 2009). We also estimated the levels of linkage disequilibrium (LD) to get insights on the genealogies pattern and shape, in particular to analyse a putative population structure. LD was computed between pairs of polymorphic sites by the correlation coefficient r2 (Hill & Robertson, 1968), and their statistical significance by the χ2 test. The global levels of LD were estimated by the ZnS statistic (Kelly, 1997), being its statistical significance assessed by computer simulations (10 000 replicates) based on the coalescent process without recombination (Hudson, 1990). Lastly, we conducted a Monte Carlo coalescent simulation-based test to determine whether a structured genealogy with two sets of highly differentiated sequences (a deep genealogy) is compatible with that expected under the equilibrium neutral model, even in the absence of recombination (see Aguadé, 2001 for details). For this analysis, we examined if the number of substitutions (X) between the two deep genealogy lineages (descending from the root) are higher than that expected for a nonrecombining (such as the mtDNA) neutral region. The empirical distribution of X was generated by coalescent simulations (10 000 replicates) without recombination and conditioned on the observed levels of variation -Watterson θ-, sample size and the number of sequences in each subset. The statistical significance of the test was determined by comparing our observed X value with those of the empirical distribution (Rozas, 2009).


Species identification

Cephaloflexa bergi (Riester, 1938). The mature worms are 5–7 cm in length. The body is subcylindrical, the anterior third narrowing very gradually. The cephalic region is rolled to the dorsal side, with a slightly concave ventral side. The ground colour of the dorsum varies from light brown to green olive; numerous dark-brown pigment spots spread onto the dorsum. Some worms have an additional irregular median line of the same colour at the level of the pharynx and the copulatory complex. The ventral side is ochre to light brown covered with brown spots visible to the naked eye. Internally, the cephalic retractor muscle is lens-shaped in cross section; the pharynx is cylindrical; the copulatory complex (Fig. S1) is long, having an intrabulbar prostatic vesicle with the proximal quarter paired; a penial papilla is absent; the male and female atria are folded; the lining epithelium of the female atrium is ciliated.

Geoplana goetschi sensu Marcus (1951). The mature worms are up to 20 cm in length. The body is flattened, with parallel sides. The dorsal ground colour is dark green olive; a submaginal yellowish stripe contours the entire body. The ventral side is pale greyish. The eyes are located marginally. Additionally to the parenchymatic muscle layers common to virtually most Geoplaninae species (i.e. a dorsal diagonal with deccusated fibres, a supra-intestinal transverse one, and a sub-intestinal transverse one), there is a tubular parenchymatic muscle layer of longitudinal fibres embracing the intestine. This muscle layer had been previously noted by E. M. Froehlich (pers. comm.) and reported (Carbayo, 2003; Froehlich & Leal-Zanchet, 2003). The pharynx is bell-shaped; the oesophagus long and strongly muscularized. The prostatic vesicle is extrabulbar and long, with the proximal third paired. The penial papilla is cylindrical, lined with a ciliated epithelium. The female atrium is a long and dilated cavity becoming narrower close to the gonopore, and with a length equal to three quarters of male atrium. A lacunar, pseudo-stratified, ciliated epithelium clothes throughout almost of the total length of the female atrium (Fig. S1).

The molecular phylogenetic analysis of COI-1 dataset (Fig. S2) clearly confirmed that all G. goetschi, as well as all C. bergi, individuals cluster together in monophyletic clades independently of any other species.

Phylogenetic relationships

We have analysed the COI DNA sequence variation in 38 individuals (COI-2 dataset) and in 36 individuals for ITS-1 (Cb-ITS and Gg-ITS datasets), and inferred their phylogenetic relationships. ITS-1 nucleotide variation within the species is much lower than that for COI (Table 1; Fig. 2). Nonetheless, both markers uncover that the four populations are relatively well differentiated and show similar population clustering patterns in the two species.

Table 1.   Summary of the nucleotide diversity estimates.
LocalitySpeciesnNo. of haplotypesHaplotype diversity (HD)No. of polymorphic sitesNucleotide diversity (π)
 Parque Nacional da Serra da Bocaina (SB)Cephaloflexa bergi1060.889310.0109
 All four localities (SB, BB, PI, SL)C. bergi16110.9501230.0573
 Parque Nacional da Serra da Bocaina (SB)Geoplana goetschi1650.742330.0170
 All four localities (SB, BB, PI, SL)G. goetschi2290.8571280.0463
 Parque Nacional da Serra da Bocaina (SB)C. bergi101000
 All four localities (SB, BB, PI, SL)C. bergi1560.571390.0431
 Parque Nacional da Serra da Bocaina (SB)G. goetschi1620.12520.0008
 All four localities (SB, BB, PI, SL)G. goetschi2170.500340.0157
Figure 2.

 Maximum likelihood trees (ML) inferred from the two genes. Left, COI-2 dataset, the tree has been rooted on the midpoint between the two species. Right-top, Cb-ITS dataset; Right-bottom, Gg-ITS dataset. Bootstrap values are shown only at nodes over population level.

COI phylogenetic trees inferred from the COI-2 dataset by ML (Fig. 2) and BI (not shown) have similar topologies, and are congruent with those obtained by the haplotype network (Fig. 3). The two methodologies show that the SB individuals in both species cluster together independently from the rest of populations, constituting a monophyletic group. Also in both species, COI data in the SB locality show two relatively high divergent clades, whereas all ITS-1 sequences are practically identical (in fact, there is nearly no nucleotide variability in this population, most variability is due to a few indels). The comparative phylogenetic analysis of two unlinked markers indicates, therefore, that the two deep mitochondrial clades of the SB population represent intraspecific genetic lineages.

Figure 3.

 Median joining network based on COI sequences. Code colour as in Figs1 and 2 for Cephaloflexa bergi and Geoplana goetschi, orange and black circles indicate that the species used as outgroups (E. pseudorhynchodemus and C. iheringi). Each circle represents an haplotype and the circle’s area the haplotype’s relative frequency. In the insets are displayed the haplotype networks for each particular species (left, C. bergi; right, G. goetschi) calculated with TCS software.

Intraspecific DNA sequence variation

Table 1 summarizes the nucleotide variation for the COI-2, Cb-ITS and Gg-ITS datasets. The two species show high levels of nucleotide diversity in the SB population for COI, π = 0.011 and 0.017 for C. bergi and G. goetschi, respectively. The global levels of nucleotide diversity including the individuals from all four parks are much higher: π = 0.057 and 0.046 for C. bergi and G. goetschi, respectively. The levels of intrapopulation nucleotide diversity for ITS-1 are very low, precluding the use of this marker for the population genetic structure analyses.

Population structure in Serra da Bocaina using COI data

Despite the very high COI intraspecific nucleotide variability levels, the pattern of the Tajima’s D neutrality test conforms to that expected by an equilibrium stationary population (Table 2). Geoplana goetschi, however, exhibits a positive Tajima’s D value (1.689), and although the value does not differ significantly from 0, it is very close to the critical point (right tail). This species also has a significant Fu’s Fs statistic (right tail). The R2-based neutrality test (Ramos-Onsins & Rozas, 2002) which is more powerful for small sample sizes, also yields a significant result (at the right tail). Although the Fs and R2 parameters are not significant in C. bergi, they also lay at the right portion of the distribution. Therefore, there is no evidence that the SB population has been recently colonized or has suffered a population expansion. Indeed, in the latter case, the Fs and R2 neutrality tests should exhibit statistics situated at the left tail of its distribution. Lastly, the polymorphism to divergence analyses using the sliding window approach (a useful approach for detecting some forms of selection acting along the DNA region) did not reveal any clear nonneutral signature (decoupling of polymorphism from divergence) in any of the two species (Fig. 4).

Table 2.   Summary of the neutrality tests and LD values in the SB population.
PopulationNeutrality testLinkage desequilibrium
nTajima’s DFu’s FSRamos-Onsins & Rozas R2LD
D95% CIFs95% CIR295% CIZnS95% CI
  1. *< 0.05, **< 0.01.

Cephaloflexa bergi10−0.9933(−1.729, 1.655)1.842(−3.636, 4.782)0.205(0.104, 0.235)0.4811(0.145, 0.709)
Geoplana goetschi161.6886(−1.774, 1.692)8.801**(−4.519, 5.167)0.204*(0.087, 0.204)0.6507*(0.117, 0.589)
Figure 4.

 Sliding window plot of the COI nucleotide diversity (Β) within the SB population of Cephaloflexa bergi (red-dotted line) and Geoplana goetschi (purple), and nucleotide divergence (K) between C. bergi and G. goetschi (green, bold). Step size: 50 nucleotides; window size: 10 nucleotides.

We also studied the intrapopulation linkage disequilibrium (LD) to analyse the pattern of the population genetic structure (Table 2). Although ZnS values are high (0.4811 and 0.6507 for C. bergi and G. goetschi, respectively), these values are significant only in G. goetschi (P-values obtained by coalescent simulations with no recombination). Moreover, G. goetschi has significant LD polymorphic site pairs even after applying the conservative Bonferroni correction (335 significant pairs out of 528). C. bergi has 210 significant pairs (out of 435), but none of them remains significant after applying the Bonferroni correction; nevertheless, the relation between the high number of pairs of polymorphic sites analysed (435) to the relative small sample size (10), precludes detecting any Bonferroni significant pair. These features, shown by the two species, indicate that the SB population has a clear molecular signature of population structure. These significant LD results are explained by the presence in both species of two very deep lineages, deeper than that expected for a nonrecombining region (Slatkin & Hudson, 1991). To assess if such deep genealogies are expected in a single panmictic population, we studied the distribution of the number of substitutions expected between the two most divergent lineages in a neutral genealogy (see Material and methods). This analysis was carried out separately for each species of the SB population. We found that the number of substitutions observed between the two most divergent lineages of G. goetschi (which generates two sets of 10 and 6 sequences) is under the neutral expectations (= 0.213). The same happens for C. bergi: the genealogy exhibits two divergent lineages (of 1 and 9 sequences) which are also compatible with the neutral expectations (= 0.171). Therefore, although the observed genetic structure is compatible with that expected of a nonrecombining region in a panmictic population, it might also reflect the presence of old lineages of a stable and structured population.


Species identification

Cephaloflexa bergi is known to show several colour patterns; that of specimens from SB match the colour observed by Marcus (1951) for some of the specimens he sampled in São Paulo and surrounding areas. The internal morphology agrees with that described by Riester (1938), Marcus (1951) and Carbayo & Leal-Zanchet (2003). The identification of the second species deserves some major comments. The worms we identified as Geoplana goetschi are morphologically similar to those from São Paulo that Marcus (1951) studied and identified as Geoplana goetschiRiester, 1938. However, the morphology of Marcus’s specimens is different in key diagnostic features from that of Riester’s (1938). Unlike Marcus’s specimens, Riester’s specimens show a female atrium as long as the male one, or slightly longer; the pseudostratified epithelium of the female atrium occupies only the posterior half of it; and, more importantly, there is no parenchymatic tubular muscle layer of longitudinal fibres, as we have observed in specimens we have sampled. Furthermore, the reconstructed phylogenetic tree (Fig. S2) confirmed that all individuals of G. goetschi sensu Marcus (1951) used in our study form a different clade than that constituted by G. goetschiRiester, 1938;. Thus molecular data confirmed the morphological results indicating that G. goetschi sensu Marcus (1951) is a different and undescribed species. Still, throughout the manuscript we denote Geoplana goetschi sensu Marcus (1951) as Geoplana goetschi.

Genetic variation in Serra da Bocaina

Our phylogenetic analyses show that all individuals of the SB population cluster together (Fig. 2). Nonetheless, current COI data reveals that both species have high levels of genetic variation in this locality (Table 1), clearly over the range observed in two Australian terrestrial flatworm species (π = 0.0008–0.0073; Sunnucks et al., 2006). This population also shows neutrality test statistic values at the right tail of the distribution (some of them significant), and a good correlation between DNA polymorphism and divergence levels (Fig. 4). The results suggest, therefore, that the SB population has been historically stable, without any evidence for a recent population expansion.

There is, in addition, evidence of intrapopulation genetic structure in this sampling locality: the two species have two deep mtDNA lineages and that in turn generate high LD values. Despite the fact that this feature might indicate the co-existence of two reproductive isolated genetic groups (in each species) likely it is not the case. First, the deep lineages feature is not observed using information of the nuclear marker ITS-1; nevertheless, as mtDNA and nuclear markers differ on their mean coalescent times (four times lower for the mtDNA), these markers might uncover evolutionary histories at different times. Second, and more notably, our coalescent simulation analyses indicate that the genetic structure formed by the two deep lineages, although important, is compatible with the neutral equilibrium model, especially in the absence of recombination (such as for the mtDNA). Moreover, the values of the neutrality statistics tend to fall within the right part of the distribution; that is, a number of SNPs segregate at intermediate frequencies. Despite the fact that a number of evolutionary processes, such as balancing selection, population subdivision or population decline, might generate this pattern, it is compatible with a long-term habitat stability scenario, and actually is highly inconsistent with a recent population expansion or recolonization scenario. Remarkably, the two species share these features. It is important to note that the high levels of nucleotide variation observed within species do not correspond to any methodological species-identification problem.

Taken together, our results clearly suggest that the habitat of the SB population has been stable for a relatively long period of time. The observed population structure might be caused by some micro-spatial genetic structure or by an ancient population fragmentation with a subsequent admixture of the populations associated with ice advances and retreats during the Pleistocene. In the latter case, the ancient populations should be well differentiated. Results of the ITS-1 data, which do not show this intrapopulation structure, would not seem to support the admixture scenario. The ITS-1 results, however, are not completely conclusive as (i) this nuclear region harbours little nucleotide variation and (ii) the effective population size of nuclear regions is four times higher than that of the mitochondrial ones. Therefore it is possible that this admixture scenario really exists, and that it was only detected by the COI genetic data. Nonetheless, the interpopulation nucleotide divergence values (Table 3) of both species are much higher than within population estimates (Table 1; Fig. 2). All these features, intrapopulation nucleotide diversity values and interpopulation coalescent times older than intrapopulation ones, are the genetic pattern expected for a low mobility species in a long-term stable habitat, and therefore with reduced gene flow among parks.

Table 3.   COI sequence divergence between populations.
  1. Dxy values among the four Cephaloflexa bergi (lower left) and Geoplana goetschi (upper right) populations.

SB 0.0660.0800.093
BB0.101 0.0770.094
PI0.0850.065 0.078

Origin and maintenance of the Atlantic Forest biodiversity

Our analyses of the patterns of genetic diversity do not support a recent colonization of the SB region, neither that they have suffered recent population growth. Taking all the results together, the patterns of genetic diversity found at the SB park fits well with that expected for a long-term stability region which, although reduced in area, would have maintained its functional ecological properties at least for small invertebrates. Indeed, the high levels of nucleotide variation and the strong genetic structure cannot easily be explained by recent climatic events, such as the last glacial periods (only tens of thousands years ago).

In conclusion, our analysis uncovers that paleoclimatic models are likely failing in the detection of refugia areas in SAF and that an invertebrate model organism as planarians, with low dispersal capability and high dependence on forest, can improve the resolution of these models. Further phylogeographic studies of land planarians would likely yield new and informative data on the origin and current maintenance status of the Atlantic Forest biodiversity. Undoubtedly, this information will be very useful to provide scientific-based guidelines for conservation-policy makers.


We thank the Instituto Chico Mendes de Conservação da Biodiversidade (ICMBio) for sampling licence, Francisco Livino and co-workers for their kind help in logistic and field transportation in the SB Park. We also thank Júlio Pedroni, Débora Redivo, Cláudia Olivares, Marília Jucá and Welton Araújo for sampling help and Italo D’Elia, Ana Cavalcanti and Lígia Domingos for the histological work. This work was supported by Fundación BBVA grant BIOCON 06 – 112 to M.R. and Grup de Recerca Consolidat of the Generalitat de Catalunya: 2009SGR1462. We also thank two anonymous referees who helped to improve the original ms.