Systematics of the Rhinella margaritifera complex (Anura, Bufonidae) from western Ecuador and Panama with insights in the biogeography of Rhinella alata

Abstract The Rhinella margaritifera species group consists of 17 species of toads distributed in tropical and subtropical South America and eastern Central America. The identity of some of its species is poorly understood and there are numerous undescribed cryptic species. Among them, the status of Rhinella margaritifera is one of the most problematic. Its range includes lowland rainforests separated by the Andes, the Chocoan rainforest to the west and the Amazonian rainforest to the east. This distribution is puzzling because the Andes are an old and formidable barrier to gene flow and therefore should generate vicariant speciation between disjunct lowland populations. Herein we clarify the taxonomy of populations of the Rhinella margaritifera complex from Central America and the Chocó region of South America. The morphological and genetic variation of Rhinella margaritifera was examined from 39 populations from Chocó, 24 from the upper Amazon region of Ecuador, and 37 from Panama, including the holotype of the Panamanian Rhinella alata. Phylogenetic analyses were performed based on mitochondrial genes 12S rRNA, 16S rRNA, and cytochrome c oxidase I (COI) and the nuclear gene Tyrosinase (Tyr). The genetic and morphological data show that Panamanian and Chocoan populations are conspecific. In the phylogeny, populations from Chocó and Panama form a well-supported clade. The morphology of the holotype of Rhinella alata falls within the variation range of Panamanian and Chocoan populations. Based on all this evidence, we assign the populations from western Ecuador and Panama to Rhinella alata and demonstrate that the unusual distribution pattern of “Rhinella margaritifera” on both sides of the Andes was an artifact of incorrectly defined species boundaries.


Introduction
Rhinella is a genus of bufonid frogs distributed from southern Texas, through southern Sonora (Mexico), south tropical Mexico, Central America, and South America. There are 87 recognized species of Rhinella (Frost, 2014) among which 17 belong to the R. margaritifera species group (Lavilla et al. 2013, Moravec et al. 2014). Thirteen of these species are distributed throughout the Amazon Basin, the Guyanas and Central America, while R. hoogmoedi Caramaschi & Pombal, 2006 occurs in the Brazilian Atlantic Forest, R. scitula (Caramaschi & Niemeyer, 2003) and R. ocellata (Günther, 1858) in the Brazilian Cerrado, and R. paraguayensis Ávila, Pansonato & Strüssmann, 2010 in the Brazilian Pantanal (Caramaschi and Niemeyer 2003, Caramaschi and Pombal 2006, Lima et al. 2007, Fouquet et al. 2007a, Ávila et al. 2010, Frost 2014. They inhabit the forest floor and their cryptic coloration mimics the forest leaflitter. Morphologically they have been characterized by the presence of hypertrophied supra and postorbital crests, especially in females. Putative synapomorphies for the group are the expansion of the posterior ramus of the pterygoid and nasals that articulate laterally with the preorbital process of the maxilla (Pramuk 2006).
The R. margaritifera species group (formerly Bufo typhonius or Bufo margaritifer group) has one of the most complex histories in the systematics of Neotropical anurans (Hoogmoed 1986, 1989, 1990, Hass et al. 1995, Fouquet et al. 2007b). The boundaries among its species member are poorly understood as a result of a highly variable intraspecific morphology and scant morphological differentiation between some species. In addition, some of the type material is unavailable or poorly preserved and several species descriptions lack details. Despite recent progress in the systematics of the group (i.e. Vélez-Rodriguez 2004, Pramuk 2006, Fouquet et al. 2007b, 2012b, Ávila et al. 2010, Lavilla et al. 2013, Moravec et al. 2014) a number of cryptic species still need to be identified, specially among Amazonian populations (Hoogmoed 1990, Hass et al. 1995, Vélez-Rodríguez 2004, Pramuk 2006, Fouquet et al. 2007b, Lavilla et al. 2013, Moravec et al. 2014.
Two species of the R. margaritifera group have been reported west of the Andes (Chocó region, humid forests west of the Andes in Colombia and Ecuador) and in eastern Panama: R. alata and R. margaritifera. R. alata was described by Thominot (1884) as Bufo alatus, based on an adult male collected at Obispo, Isthmus of Panama. Boulenger (1885) considered it a junior synonym of "B. typhonius", and Hoogmoed (1986and Hoogmoed ( , 1989 suggested that it was, possibly, a synonym of B. acutirostris (Spix, 1824). La Marca (1997) reported populations of R. alata from northern Venezuela. Gorzula and Señaris (1999) suggested that R. margaritifera only occurs in southern Venezuela and R. alata north of the Orinoco. However, Barrio-Amorós (1999"1998 disagreed with both reports and considered that R. alata was not distributed in Venezuela. Rhinella margaritifera was described by Laurenti in 1768. It occurs in eastern Panama (Frost 2014), the Chocoan lowlands of western Ecuador and western Colombia (e.g. Anderson 1945, Miyata 1982, Ruiz-Carranza et al. 1996, Ortega-Andrade et al. 2010, Ortiz et al. 2013, Ron et al. 2014, Amazonia and vicinities in Bolivia, Brazil, Colombia, Ecuador, French Guiana, Guyana, Peru, Surinam and Venezuela (Lavilla et al. 2013). A genetic study by Fouquet et al. (2007b), using two mitochondrial genes (12S and 16S) and the two nuclear genes (Tyrosinase and 18S), showed that R. margaritifera was paraphyletic and contained up to 11 cryptic species. Populations from the Chocó region have been widely referred as R. margaritifera although Solis et al. (2010) remarked that populations from the Ecuadorian Chocó might belong to a separate species. Unfortunately, they did not provide further details.
The distribution of R. margaritifera in the humid lowlands west and east of the Andes is intriguing because, particularly for amphibians, the Andes represent a formidable barrier to gene flow (e.g. Santos et al. 2009). Despite similar environmental conditions, only four amphibian species are shared between the lowland rainforests of the Amazon basin and the Chocó: R. margaritifera, R. marina, Hypsiboas boans and Trachycephalus typhonius. Moreover, there is genetic and morphological evidence suggesting that populations on each side of the Andes of R. marina and Trachycephalus typhonius represent separate species (Slade andMoritz 1998, Ron andRead 2011). Thus, the distribution of R. margaritifera is suggestive of either an unusual biogeographic history or the existence of cryptic species.
Herein, genetic and morphological information were integrated to clarify the taxonomy of the populations of R. margaritifera from Panama and the Chocoan region. Populations from the western and eastern Andean slopes were compared to test the role of the Andes as a dispersal barrier in shaping the evolution of the R. margaritifera species complex.

Population sampling
Populations from Panama, the Ecuadorian Chocó, and the Amazon basin were sampled (Figs 1 and 2). Specimens examined morphologically are listed in Appendix 1; specimens analyzed genetically are listed in Table 1.
Genetic analyses were based on newly generated sequences of R. margaritifera from 32 individuals and 19 populations: R. margaritifera from the Ecuadorian Chocó (12 individuals, 7 populations); R. margaritifera from Panama (3 individuals, 2 popula-tions) and R. margaritifera from the Amazon basin (17 individuals, 10 populations), and six sequences for the outgroups (see Table 1). Sequences of eight R. dapsilis were generated, including all available homologous sequences for the R. margaritifera species group from GenBank (http://www.ncbi.nlm.nih.gov/genbank; Table 1). R. marina, R. chavin, R. nesiotes and R. festae were included as outgroups. The morphometric and genetic analyses were based on the same individuals, when possible. Several specimens used in the morphological analyses lacked tissues and were not included in the genetic analyses. However, their identification was unambiguous based on geographic distribution and morphological characters.
Examined specimens are deposited at the Museo de Zoología, Pontificia Universidad Católica del Ecuador (QCAZ, Quito, Ecuador), the American Museum of Natural History (AMNH, New York, USA), Círculo Herpetológico de Panama (CH, Panama, Panama), Centro de Ornitología y Biodiversidad (CORBIDI, Lima, Perú) and Museo de Vertebrados de la Universidad de Panama (MVUP). We also examined photographs of the holotypes of R. alata from Musée National d'Historie Naturelle (MNHN, Paris, France). Tissues were obtained from the QCAZ and CH collections. Tissues (liver or thigh muscle) were stored in 95% ethanol.

Morphological analyses
Morphological terminology and abbreviations follow Vélez-Rodriguez (2004) and Narvaes and Rodrigues (2009). Sexual maturity was determined by the presence of nuptial pads in adult males and convoluted oviducts or mature eggs in gravid females. Specimens from the QCAZ collection were euthanized with the anesthetic spray Roxicaine, fixed in 10% formalin, and preserved in 70% ethanol.
The goal of the morphological analyses was to compare three geographic regions: (1) Chocó (2) Panama, and (3) upper Amazon basin. Because the phylogeny showed that Panama and Chocó populations are conspecific, we also compared Chocó + Panama vs. upper Amazon. Morphometric analyses were based on adult and well-preserved specimens (Simmons 2002). We measured the following variables: (1) SVL (snoutvent length, from the tip of snout to the mid-vent); (2) TL (tibia length, from the outer edge of flexed knee to the heel); (3) FL (femur length, from the mid-venter to the outer edge of flexed knee); (4) HL (head length, from the posterior margin of tympanum to the tip of snout); (5) HW (head width, between knobs at angles of jaws, if present); (6) STCH (supratympanic crest height, the distance between the angle of the jaw and the highest point of the ridge above of the tympanum); (7) SOCH (supraorbital crest height, the distance between the angle of jaw and the highest point of the ridge at the mid-orbit); (8) NSD (nostril-snout distance, from the nostril to the tip of the snout); (9) IND (inter-nostril distance, distance between nostrils); (10) TD (tympanum diameter, from the posterior to the anterior edge of the tympanum); (11) FT (foot length, from the posterior edge of the metatarsal tubercle to the tip of the toe IV). Measurements were taken with digital calipers (to the nearest 0.01 mm). Two qualitative mor-phological characters were also analyzed: (1) vertebral apophyses (present/absent) and (2) bony knob at angle of jaws (present/absent).
Principal Components Analysis (PCA) and Discriminant Function Analysis (DFA) were used to assess morphometric differentiation between Chocó, upper Amazon, and Panama. To remove the effect of body size (SVL), the MANOVA and PCA were applied to the residuals from the linear regressions between the measured variables and SVL, for males and females separately. For the PCA, only components with eigenvalues > 1 were retained. All measurements were first subjected to the Shapiro-Wilk normality to test for normal distribution (Shapiro and Wilk 1965). Data not normally distributed were logtransformed. Levene's test was used to determine if variables were homoscedastic (Levene 1960). Number of analyzed specimens were (1) Chocó: 43 males and 31 females, (2) Panama: 6 males and 8 females, (3) upper Amazon basin: 16 males and 16 females. All analyses were performed using JMP® 9.0.1 (SAS Institute 2010).

DNA extraction, amplification, and sequencing
Total DNA was extracted from muscle or liver tissue preserved in 95% ethanol or tissue storage buffer using standard guanidine thiocyanate protocol (M. Fujita, unpublished) with modifications. Polymerase Chain Reaction (PCR) was used to amplify the mitochondrial genes 12S rRNA, 16S rRNA, cytochrome c oxidase I (COI) and nuclear gene Tyrosinase (Tyr). PCR amplifications were carried out under standard protocols. Using standard primers developed by Bossuyt and Milinkovitch (2000), Goebel et al. (1999), Pauly et al. (2004), and Meyer et al. (2005). Amplicons were sequenced by Macrogen Inc., Seoul, Korea.

Phylogenetic analyses and genetic distances
Preliminary sequence alignment was done with Geneious Pro 5.4.6 (Drummond et al. 2011). The sequence matrix was imported to Mesquite 2.75 (Maddison and Maddison 2011) and the ambiguously aligned regions were adjusted manually to produce a parsimonious alignment. Phylogenetic trees were obtained using Bayesian Inference (BI) in MrBayes 3.1.2 (Ronquist and Huelsenbeck 2003) and Maximum Likelihood (ML) in Garli 2.0 (Zwickl 2006). The best-fit models of sequence evolution were selected under the Akaike information criterion (AIC) and the best partitioning scheme for the combined nucleotide data set and the models of character evolution for the BI and ML were estimated with PartitionFinder 1.0.1 (Lanfear et al. 2012). We ran three analyses: (1) the complete multi-locus data set, (2) only mitochondrial genes, (3) only the nuclear gene.
The Bayesian search consisted of two parallel runs each with 130 × 10 6 generations with four Markov chains. The convergence of the runs was assessed with Tracer 1.5 (Rambaut and Drummond 2007) evaluating the effective sample sizes and stopping   when all post burn-in values were greater than 200. The first 10% of the sample was discarded as burn-in (Castañeda and Queiroz 2011). For the ML analysis, we carried out 20 replicate searches and increased the setting "genthreshfortopoterm" until all searches resulted in similar likelihood values, indicating an efficient search (Zwickl 2006; final value was 200,000). Ten replicate searches started from stepwise trees and ten from random trees. The setting "limsprrange" was set to 10 (default = 6). Node support was assessed with non-parametric bootstrapping (Felsenstein 1983) with 100 pseudoreplicates with the same settings of the stepwise full search but with a single replicate per search. The 50% majority rule consensus for the bootstrap trees was obtained with Mesquite 2.75 (Maddison and Maddison 2011).
Uncorrected pairwise (p) genetic distances were obtained for gene 16S using software Mesquite 2.75 (Maddison and Maddison 2011). Missing and ambiguous sites were excluded. Genetic distances comparisons were based on gene 16S because it has been widely used as a barcode standard in amphibians (e.g. Vences et al. 2005). We assumed that genetic distances > 3% are suggestive of interspecific differentiation (Fouquet et al. 2007c). Genetic distances thresholds are problematic because they can lead to both false negatives and false positives in species identifications (Collins and Cruickshank 2013). We used the threshold only as a working hypothesis that was tested with morphological comparisons.

Phylogenetic analyses
The complete matrix contained up to four genes and 3045 bp for 92 samples. For the complete data set, PartitionFinder chose seven partitions as the best strategy (best model in parenthesis): 12S (GTR + I + G), 16S (GTR + I + G), COI 1 st position (TIMef + G), COI 2 nd position (TVM + I + G), COI 3 rd position (TrN + G), Tyr 1 st and 2 nd position (TrN + G), Tyr, 3 rd position (TrN + I + G). For the mitochondrial analyses, the same five partitions were chosen, one for each ribosomal RNA gene and each codon position in COI. For the nuclear analysis, two partitions were chosen: Tyr, 1 st and 2 nd position and Tyr, 3 rd position.
The tree topologies for the Maximum likelihood and Bayesian phylogenies were similar except for weakly supported nodes (posterior probability < 0.95 and bootstrap < 75). The Maximum Likelihood tree (Fig. 3) shows a basal divergence of R. castaneotica, which is sister to two clades containing the remaining species of the R. margaritifera species group. One clade is strongly supported in the Bayesian consensus (posterior probability = 1) although it has low bootstrap support (= 63). It contains three groups: Panama (posterior probability = 1.0, bootstrap = 100), Chocó (posterior probability = 1.0, bootstrap = 86) and upper Amazon (posterior probability = 1.0, bootstrap = 68). Chocó and Panama form clade sister to the upper Amazon clade. Both clades, which are on opposite sides of the Andes, are separated by pairwise genetic distances (uncor- rected p for the mitochondrial gene 16S) ranging from 3.01 to 5.5% (average = 4.28, SD = 0.56). The genetic distances and the morphological differences (see next section) between the Chocó-Panama clade and the upper Amazon clade suggest that they are separate species. The 16S genetic distances between the Chocó and Panama clades range from 1.26 to 1.99% (average = 1.63, SD = 0.19). The relatively low genetic distances and the lack of morphological differences between their populations (see next section) indicate that they are conspecific. The Chocó populations further segregate latitudinally in two well-supported clades. One includes the populations in northern Ecuador (e.g. Reserva La Chiquita and Borbón) while the other includes central and southern populations (e.g. Manta Real and Valle Hermoso, Fig. 3). The sister clade to Chocó-Panama + Upper Amazon has weak support and includes other members of the R. margaritifera group (R. dapsilis, R. hoogmoedi, R. lescurei, R. martyi, R. ocellata, R. paraguayensis and "R. margaritifera") from the Guiana region and Amazonian Brazil, Ecuador and Peru. Relationships among them are weakly supported on most branches.
The Maximum Likelihood tree based on mitochondrial genes (Fig. 4) has similar topology to the Maximum Likelihood tree derived from the analysis of the complete data set (Fig. 3). The Bayesian consensus tree, derived from the Tyrosinase gene, has definitely lower resolution (Appendix 2).
Significant differences were observed in relative crest size between the Chocó-Panama and upper Amazon clades (Fig. 6). In the former, female supratympanic crest height had a range between 51.6 to 63.5% of head length (n = 39); in the later, range was 68.6 to 95.5% (n = 16). Ranges did not overlap and differences were significant (Wilcoxon's Z = -5.77, p < 0.001). Male supratympanic crest height had a range between 49.3 to 59.8% of head length in Chocó-Panama (n = 49); in upper Amazon, range was 50.6 to 78.4% of head length (n = 16). Ranges overlapped but differences were significant (Wilcoxon's Z = 3.11, p = 0.0018).   Figure 6. Box and whisker plots showing relative size of supratympanic crests for adult Rhinella margaritifera (upper Amazon) and R. alata (Chocó-Panama). The central bar indicates the median, the interquartile range is shown by the box length, and the range is shown by the short horizontal lines (whiskers). STCH = supratympanic crest height, HL = head length. The yellow cross is the holotype of R. alata.
Three components with eigenvalues > 1.0 were extracted from the PCA for females ( Table 3). The three components accounted for 67.3% of the total variation. The highest loadings of the PCA for females were supratympanic and supraorbital crest height, and tibia length for PC I, inter-nostril distance and tympanum diameter for PC II, and nostril-snout distance and inter-nostril distance for PC III. Three components with eigenvalues > 1.0 were extracted from the PCA in males (Table 3). The three components accounted for 63.3% of the total variation. The highest loadings for the PCA for males were head length and head width for PC I, inter-nostril distance and tympanum diameter for PC II, and tibia length and foot length PC III. The morphometric space of the Chocoan, upper Amazon, and Panamanian populations broadly overlaps in both males and females (Fig. 7).
In the DFA classification for females, 51 out of 55 females were assigned correctly to their geographic region. The four misclassified females from Ecuadorian Chocó were assigned to Panamanian populations. All specimens from the upper Amazon were correctly classified. In the DFA for males, 56 out of 65 males were correctly classified. The eight misclassified males from Ecuadorian Chocó were assigned to Panamanian populations and only one from upper Amazon to Panamanian populations. All males and females from Panama were correctly classified. The DFA analyses indicate that populations from the Ecuadorian Chocó are morphometrically very similar with those from Panama, both groups being markedly different from R. margaritifera from the upper Amazon.
Finally, evidence of sexual dimorphism was found in relative crest size: females have larger cephalic crests than males (Fig. 6). The ratio supratympanic crest height/

Qualitative morphological characters
The upper Amazon clade differs from the Chocó-Panama clade in having protruding vertebral apophyses in the dorsum and bony knobs at angle of jaws (both absent in the Chocó-Panama clade ; Figs 8-10). The Chocó-Panama clade differs from other species of the R. margaritifera group by a combination of an absence of vertebral apophyses, an absence of bony knob at angle of jaws, low cranial crests, and the tympanum rounded or ovoid (see Systematic account section). A large number of specimens were examined (see Populations sampling section) and all conform to this characterization. Thus, it seems unlikely that there are additional species of the group in the Chocoan and Panamanian regions. The holotype of R. alata (Thominot, 1884) (Fig. 11) is an adult male with an SVL of 39.2 mm. It has poorly developed supratympanic crests and lacks bony knobs at the angle of jaws. The vertebral apophyses are inconspicuous. These characters and the location of its type locality (within 6 km of one of our examined populations) lead us to conclude that it is conspecific with the Panamanian and Chocoan populations examined herein.
Rhinella alata is most closely related to populations of R. margaritifera from the upper Amazon basin in Ecuador and Peru. They can be easily distinguished by differences in body size ( Fig. 5; see morphometric comparisons section) and relative size of cranial crests (Fig. 6).
Holotype. The holotype is an adult male with SVL = 39.2 mm (Fig. 11). Descriptions of the holotype have been provided by Leavitt (1933) and Hoogmoed (1989). The bony knob at angle of jaws and vertebral apophyses are absent. The crests are low and thick. There is a dorsolateral row of conical tubercles from the posterior border of the parotoid gland to the groin. There is a clear mid-dorsal line from the snout to the vent. The tympanum is rounded.
Ventral surfaces of preserved specimens have a cream to yellowish-cream background color with irregular darker marks arranged in diverse patterns; marks can  be light gray (QCAZ 6734, AMNH 88689), light brown (QCAZ 6732, AMNH 104454), dark gray (QCAZ 31606) or dark brown (QCAZ 6733, AMNH 89459), and vary from being restricted to the anterior half of the body (QCAZ 31604, AMNH 89459) to being present over the entire venter (QCAZ 4445, AMNH 88694). A longitudinal mid-ventral cream thin stripe can be present in the gular region (QCAZ 31602, 31606) or from the gular region to the mid-venter (QCAZ 6731, 11598).
Color in life. Based on digital photograph of an adult female QCAZ 50568 (Fig.  8). Dark brown dorsum with irregular light brown and yellowish marks; there is a clear mid-dorsal line. Dorsal surfaces of tights and shanks are dark brown with transversal brown bands. Dorsal surfaces of forelimbs are dark brown with irregular light brown marks. Dark brown tubercles are abundant on the dorsum. Ventral surfaces vary from light brown to dark brown, with some irregularly distributed white and orange spots. The fingertips and the subarticular tubercles on fingers and toes are red-orange. Canthal region and tympanum are dark brown; iris greenish yellow with black reticulation. Based on a digital photography of an adult male QCAZ 37248 (Fig. 8). Light brown dorsum with black spots and light brown and light gray marks. Dorsal surfaces of tights, shanks and forelimbs are light brown with transversal dark brown bands. Brown tubercles are abundant on the dorsum. Ventral surfaces are dark brown with irregularly distributed yellowish marks; the posterior part of the venter is cream. The subarticular tubercles of palms, soles, and fingertips are red-orange. Canthal region and tympanum are dark brown; iris greenish yellow with black reticulation.
In Panamanian populations gravid females were found in January (AMNH 104454), September (AMNH 55461), November (AMNH 88689), and December (AMNH 53699). In central Panama, R. alata breeds in ponds and pools along permanent streams or swamps. Reproduction is explosive and most takes place from the middle of the rainy season to early dry season (Wells 1979, Ibáñez et al. 1999. Choruses last less than 24 hours with males usually calling at night and oviposition occurring by day, especially in the early afternoon (Wells 1979). Otherwise, individuals are primarily diurnal, found active on the leaf litter of the forest floor during daytime, and often found asleep on leaves of low vegetation at night (Ibáñez et al. 1999). Diet is specialized on ants (Toft 1981).
Most of the Ecuadorian specimens are from Reserva Mayronga and Reserva Ecológica Cotacachi-Cayapas. They were found in the leaf litter of secondary forest and in agricultural lands. Some adults were observed at night within the forest in vegetation above the ground and some were found in amplexus (QCAZ 10271, QCAZ 10274, QCAZ 10275 in November 1996, andQCAZ 31604, QCAZ 31605 in February 1996). All the specimens collected in Reserva Ecológica Cotacachi-Cayapas were found in secondary forest. At some collecting sites, the forest has been cleared for cacao plantations (QCAZ specimen database).
According to the classification of Sierra et al. (1999) (2004), these populations differ from the holotype of R. alata and populations of R. alata in Ecuador and Panama (in parentheses) in having: (1) a canthus rostralis protruding in females and ill-defined in males (inconspicuous in males and females), (2) parietal crests well defined in females, ill-defined in males (ill-defined in males and females), (3) vertebral apophyses slightly visible externally (absent). The differences suggest that those specimens are not R. alata and may belong to a different species. Alternatively, differences between R. alata described by Vélez-Rodriguez (2004) and our study could be an artifact resulting from the use of distinct terminology for similar character states.
In contrast, Mueses-Cisneros and Moreno-Quintero (2012) reported two species of the R. margaritifera group form Barbacoas, Nariño, Colombia (Rhinella sp. 9 and Rhinella sp. 10). Two photographs of live individuals (pp. 45) show morphological features that fall within the observed variation of R. alata. We tentatively assign them to R. alata but direct specimen examination is required to confirm this identification.

Discussion
The taxonomic status and phylogenetic position of populations traditionally ascribed to R. margaritifera (= Bufo typhonius; e.g. Anderson 1945, Miyata 1982, Ortega-An-drade et al. 2010) from western Ecuador and Central America were reviewed. The examination of the holotype of R. alata in combination with the morphological and genetic information from 72 populations from the Chocó region and Panama indicate that those populations should be referred to R. alata. The similarity between Chocoan and Panamanian populations was previously noted by Hoogmoed (1990). Hoogmoed (1990), Lescure and Marty (2000) and Fouquet et al. (2007b) considered that R. margaritifera from French Guyana, with hypertrophied crests, corresponds to R. margaritifera sensu stricto. In a recent review, however, Lavilla et al. (2013) assigned a neotype with the type locality in "Humaitá, State of Amazonas, Brazil". In our phylogeny (Fig. 3), the sister clade of R. ocellata include the closest localities to the new type locality for R. margaritifera and are likely to contain populations of R. margaritifera sensu stricto. Our phylogeny and previous reviews (e.g. Fouquet et al. 2007b) indicate that species diversity in the R. margaritifera group is greatly underestimated. In our phylogeny, two R. margaritifera from the southern Amazon in Ecuador (QCAZ 18241 and QCAZ 23917) are more closely related to R. margaritifera from French Guyana and R. dapsilis than to other R. margaritifera from Amazonian Ecuador. They probably represent an undescribed species, characterized by the presence of vertebral apophyses, bony knobs at the angle of jaws, and poorly developed crests. More studies are needed to define the status of these populations, as well as that of R. cf. paraguayensis from Bolivian and Brazilian Amazon and R. cf. hoogmoedi from Brazilian Atlantic Forest.

Systematics and morphology
The identity of the upper Amazon clade (Ecuador-Peru) remains unresolved. It was not possible to ascribe it unequivocally to any described species of the R. margaritifera species group and it is unlikely to be R. margaritifera sensu stricto (as defined by Lavilla et al. 2013). Thus, these populations may belong to an undescribed species characterized by having prominent supratympanic crests, conspicuous vertebral apophyses in the dorsum and bony knobs at angle of jaws (Fig. 10). We refrain from describing this species until genetic samples of R. margaritifera sensu stricto are available and a comprehensive review of the group is carried out. For now, we suggest that these populations are referred as R. margaritifera sensu lato.
These results raise some rather interesting questions. For instance, the complete distribution range of R. alata is yet to be determined. Extensive and explicit studies are necessary to reveal whether the species is continuously distributed from Ecuador to Panama or if it consists one, two (or more) disjoint population nuclei. This would be an indispensable step before planning further studies on the evolutionary history or conservation status of the species. Moreover, future studies including a larger number of samples, more representative of the geographic range of each species within the R. margaritifera group, from Colombia, Venezuela and Suriname, will help to clarify their evolutionary identity. It will also be necessary to re-evaluate, using molecular, mor-phological, ecological, behavioral, and phylogenetic analyses, the taxonomic status of species that have been previously described only morphologically such as R. acutirostris, R. magnussoni, R. proboscidea, R. roqueana, R. sclerocephala, R. scitula and R. stanlaii. Integrative approaches like the one we pursued in this study will help to disentangle the complex evolutionary history, systematics, and taxonomy of this species group.

Biogeographic implications
Because all species in the R. margaritifera species group are distributed in South America, it is reasonable to assume that the presence of R. alata in Central America is the result of a single dispersal event from South America. The genetic distances between Chocoan and Panamanian populations are low (range 1.2-1.9%) and suggest that their divergence was recent and occurred after the closure of the Panamanian isthmus during the late Pliocene. Assuming a rate of evolution of the gene 16S of 0.00249-0.00277 substitutions per site per lineage per Myr (Evans et al. 2004;Lemmon et al. 2007), the divergence between these populations occurred ~ 2.16 to 3.42 Myr ago (under the 0.00277 rate) or ~ 2.41 to 3.81 Myr ago (under the 0.00249 rate). Thus, it is likely that the divergence between Panama and Chocó took place after the completion of the Panamanian Isthmus (~ 3.5 Myr ago; Coates et al. 1992, Coates andObando 1996). These estimates of time of divergence, however, should be considered with extreme caution because they assume a molecular clock at a rate estimated for species in different families. Further explicit studies will be necessary to estimate divergence times with more confidence.
Rhinella alata is sister to populations of R. margaritifera from the Ecuadorian and Peruvian Amazon and the eastern Andean slopes, up to 2000 m of elevation, forming altogether a robust clade. The two lineages are highly divergent from each other (uncorrected p distances 3.0-5.5%, mitochondrial gene 16S) and are morphologically distinctive. Therefore, both clades clearly represent separate species. Previously, R. margaritifera was considered to occur on lowland rainforests east and west of the Andes of Ecuador. This distribution was atypical because out of 174 amphibian species inhabiting the Amazonian rainforests of Ecuador below 600 m of elevation, only three also occur in the rainforests of the Chocó region west of the Andes: Hypsiboas boans, Rhinella marina and Trachycephalus typhonius (Ron et al. 2014). Despite having similar environmental conditions and being geographically close (as low as 100 km of airline distance), rainforests on both sides of the Andes share few amphibian species, a result of the barrier effect of the Andes. Our results showing that R. margaritifera only occurs on the eastern side demonstrate that their unusual distribution was an artifact of the incorrect delimitation of species boundaries. We suspect that the same problems could explain the disjunct distributions of Rhinella marina, Trachycephalus typhonius and Hypsiboas boans. Therefore, tropical rain forests of the Amazon and the Chocó may not share amphibian species.

Acknowledgments
This work was funded by Secretaría Nacional de Educación Superior, Ciencia, Tecnología e Innovación (Arca de Noé initiative) and Pontificia Universidad Católica del Ecuador. RI was supported by the Sistema Nacional de Investigación de Panama. We are particularly thankful to Annemarie Ohler, the Curator of Muséum National d'Histoire naturelle, who provided data and photos of the holotype of R. alata. David Buckley for valuable comments on the text. Ángel Sosa and Mario Yánez-Muñoz shared photographs of R. alata. Darrel Frost, Curator of AMNH, loaned specimens of R. alata from Panama. Pablo Venegas, curator of CORBIDI, provided a tissue sample from Peru. Nadia Páez for help in taking measurements of specimens. We are thankful to Andrea Manzano and María Ordoñez for carrying laboratory work. For specimen collection and locality data, we are indebted to Néstor Acosta, Silvia Aldás, Alejandro Arteaga, Fernando Ayala, Alvaro Barragán, Paola Buitrón, David Cannatella, Edwin Carrillo, Luis Coloma, Rafael de Sa, Alexandra Endara, Jorge García, Stella de la Torre, Mireya Dimas, Sehoya Harris, César Jaramillo, Fidel Jaramillo, Diego Lombeida, Alfredo Lopez, Luis E. Lopez, Fernando Nogales, Giovanni Onore, Aida Ortiz, Alexandra Quiguango, Fabián Sáenz, Frank Solís, Samuel Sucre, Italo Tapia, Queti Tapia, Eduardo Toral, J. M. Touzet, Ana M. Velasco, and Tania Villegas.