**4. Discussion**

## *4.1. Phylogenetic Systematics and Biogeography*

The phylogenetic relationships inferred here among *Atelopus* are mostly in agreemen<sup>t</sup> with those found by Lötters et al. [10], a study also based solely on mitochondrial DNA sequences (12S and 16S genes), but which included only three of the Central American species, *A. varius*, *A. zeteki*, and *A. chiriquiensis*. The current study and Lötters et al. [10] found that *Atelopus* is monophyletic and of South American origin. Additionally, both studies recovered a clade of Amazonian-Guianan species which is sister to a clade containing the Andean, Chocoan, and Central American species. In contrast with Lötters et al. [10], however, in this work we included sequences of one species (*A. laetissimus*) from the Sierra Nevada de Santa Marta (SNSM), an isolated massif in north-western Colombia well

known for its endemicity. The phylogenetic position of *A. laetissimus* is noteworthy, as it diverged very early in the history of the genus, being either sister to the rest of *Atelopus* (Figures 2 and 3), or the sister to a clade formed by Andean, Chocoan, and Central American species (Supplementary Figure S1). A similar pattern was found in the frogs of the clade Allocentroleniae, in which the basal split separates the Guianas and Amazonian regions from a clade containing a species from the SNSM that is sister to all other known species [68].

While a single colonization event can explain the presence of *Atelopus* in Central America, our data are also compatible with two dispersal events (Figure 3). Further sampling of other Chocoan *Atelopus*, such as *A. spurrelli*, *A. longibrachius*, *A. balios*, and the mainland populations of *A. elegans*, may be needed to resolve this ambiguity. Savage and Bolaños [16] proposed that the enigmatic *A. chirripoensis* of Costa Rica is more closely related to the *A. ignescens* species complex of Ecuador and southern Colombia than to other Central American species of *Atelopus*, which would imply ye<sup>t</sup> another potential colonization event. However, testing this hypothesis is di fficult given that *A. chirripoensis* has not been seen since 1980, and it is now probably extinct [16,18]. The remarkable external similarity between *A. chirripoensis* and the *A. ignescens* complex might also be explained by morphological convergence, given that both are found in similar *páramo* environments [16].

Our analyses sugges<sup>t</sup> that the ancestral lineage of *Atelopus* reached Central America between 14.64 and 3.29 Ma with point estimates ranging from 10.6 to 4.79 Ma, depending on whether one postulates one or two invasions. In all cases, the estimated time frame is consistent with recent molecular biogeographic studies (e.g., [38,69]) supporting a non-traditional, older closure date for the Panamanian Isthmus by 10 Ma [34]. One could also argue that *Atelopus* dispersed over water prior to the completion of the Isthmus, although amphibians are sensitive to dehydration and to salt water [66,70,71]. Assuming an old date for the formation of the Isthmus, one recent hypothesis predicts that the Isthmus was forested during the Miocene and Pliocene epochs until 2.5 Ma when savanna-like environments appeared [72]. Our proposed time-frame for colonization by *Atelopus* of the Isthmus is temporally consistent with this prediction, as *Atelopus* species are a ffiliated with moist forests. Studies of additional groups of terrestrial organisms with low dispersal capabilities (such as insects, fossorial reptiles, or understory birds) or that are sensitive to salt water (such as amphibians) are needed to confirm whether there may be a tendency of inferring early crossings between continents by wet-forest lineages [31].

## *4.2. Geographic Color Pattern Variation*

Most Central American species of *Atelopus* are strikingly variable in their color patterns, not only between but also within populations (e.g., [63]). In some species, such as *A. chiriquiensis* (Figure 4B1,B2), *A. senex* (Figure 4A3,A4) and populations of *A. varius*, a noticeable sexual dimorphism in coloration is sometimes present. On the other hand, color patterns can be very similar between species, resulting in species misidentification [28,73], especially in species that lack other clear diagnostic morphological characters. Our molecular data revealed the presence of a distinct species in the Monteverde Cloud Forest Reserve, northern Costa Rica, that has until now been confused with *A. varius*, very likely due to their similarity in color pattern.

In general, recently metamorphosed and juvenile individuals of *A. certus*, *A. chiriquiensis*, *A. glyphus*, *A. limosus*, *A. varius*, and *A. zeteki* have a barred dorsal color pattern, consisting of a series of dark and light colored chevrons, which usually changes or fades ontogenetically as they grow older ([74], E.D.L. and R.I. pers. obs.). This barred pattern sometimes remains in adult individuals within certain populations, and may even be the predominant color pattern, as observed in some populations of *A. glyphus* (Figure 4D1), *A. limosus* (Figure 4C4), *A. varius* (Figure 4B3,B4), and *A. zeteki.* Therefore, natural and/or sexual selection pressures might be behind this ontogenetic change in coloration, resulting in color pattern di fferences among geographic areas.

Geographic variation in color pattern has been documented for *A. varius* and *A. zeteki* [63,73], and we found distinct color pattern variants among populations of *A. glyphus* and *A. limosus* along their distribution ranges. In the northern region of the Serranía de Pirre (e.g., Figure 1, localities 5 and 6), adults of *A. glyphus* have a barred dorsal color pattern (Figure 4D1), while in the highlands of the southern portion of Serranía de Pirre above Cana (e.g., Figure 1, locality 7), adults have a predominantly uniform brown dorsum with small, light markings (Figure 4D2). In the western area of the distribution range of *A. limosus* (e.g., Figure 1, localities 8 and 10), adults have a uniform olive green dorsum ([63,75]; Figure 4C3), whereas populations to the East (e.g., Figure 1, locality 9) have a barred color pattern on the dorsum (Figure 4C4). The mtDNA data presented here supports our contention that these color pattern di fferences indeed represent geographic variation between conspecific populations. In contrast, the undescribed *Atelopus* sp. "Puerto Obaldía-Capurganá" from eastern Panama seems to exclusively have a barred dorsal color pattern (Figure 4C5).

## *4.3. Taxonomic and Conservation Implications*

The phylogenetic results obtained here are broadly concordant with the current taxonomic arrangemen<sup>t</sup> of the species of *Atelopus*. Our data reveal two Central American lineages whose taxonomic status needs further attention, however. First, the specimens from Puerto Obaldía, Panama, and Capurganá, Colombia, were already of uncertain taxonomy. The names *A. varius glyphus* [76,77], *A. varius* [78] and *A.* <sup>a</sup>ff. *limosus* [23,62] have been applied to this binational population, but our molecular phylogenetic results are not compatible with these proposed names. We cannot, however, reject the hypothesis that these Caribbean samples might correspond to a previously unreported population of *A. spurrelli* from the Pacific coast of Colombia, since we were unable to sample this latter taxon. More likely, however, these specimens represent an undescribed species.

Second, a specimen from the Monteverde Cloud Forest in Costa Rica (MVZ 164818) is most likely a member of an undescribed (and probably extinct) species, given its large genetic distance from all other specimens, including several sympatric individuals of its sister lineage, *A. senex*. The specimens from the Monteverde, Chompipe, and Tapantí areas in Costa Rica highlight a taxonomic problem that requires further study. Specimens from populations at Chompipe (the slopes of Volcán Barba) and Tapantí are traditionally considered to be *A. senex* (Figure 4A3–A5), while the ones from around Monteverde are assigned to *A. varius* ([15,63,79]; Figure 4A1,A2). The specimen (MVZ 164818) from the Monteverde Cloud Forest Reserve stood as a member of an undescribed (and probably extinct) species, although its phylogenetic position had no significant statistical support. This specimen was inferred to be the sister clade to either *A. senex*, *A. chiriquiensis*, or to a larger clade containing both. Richards and Knowles [28] also reported on the taxonomic confusion around Monteverde specimens and noted the low genetic divergence (i.e., 0.5–1.2%) between samples from Monteverde and *A. senex* specimens. However, we confirm that the specimen MVZ 164818 from Monteverde is certainly not *A. varius*, and it should likely be assigned to an undescribed species. Furthermore, the specimens from Monteverde MVZ 164816 and MVZ 164818 share the same morphology and color pattern based on our examination, but here we show they are genetically distinct at the level of species, as *A. senex* and *A.* sp. "Monteverde", respectively. Assuming no errors on the original source of the samples and DNA sequencing, possible explanations for the di fference between morphological features and our mtDNA analysis are that genetic introgression between species has occurred or both species and/or their hybrids were present at Monteverde. To resolve this taxonomic problem, we sugges<sup>t</sup> including more DNA data in future analyses, since further sampling of individuals would be di fficult given the catastrophic population declines. Finally, the poorly known *A. spurrelli*, *A. certus* (Figure 4D3,D4), and *A. glyphus* (Figure 4D1,D2) are in need of further taxonomic study.

We confirm the results of Richards and Knowles [28] regarding the distinctiveness of *A. zeteki* (Figure 4B5) from *A. varius*, and we uncovered the locality (Juan Lana, Panama) in which these species are found in sympatry, as was hinted by Zippel et al. [73]. As previously suggested [28,73], these two species could have been hybridizing in sympatry at this locality, a hypothesis that deserves further exploration given their morphological similarity and recent divergence (2.12 Ma, CI: 1.41 to 2.86 Ma). Additionally, we established the identity of individuals from a population in Cerro Azul (Cerro La

Victoria), Panama (Figure 4C2). Savage [63] suggested this to be an introduced population of *A. varius* that resembled those from El Valle de Antón and Cerro Campana. According to our phylogenetic analyses, the individuals from this population are *A. zeteki*.

Finding support for the traditional taxonomy of Central American *Atelopus* brings hope to the success of the *ex situ* conservation programs being undertaken with the described species, as it indicates that resources are being used efficiently in frog conservation, given that all surviving species are being included by these programs. In addition, since each named species appears to be a distinct genetic entity, species ranges are not being underestimated, and no species is being protected unnecessarily. However, this is not the case for the two candidate species uncovered in this study, as they are not currently part of any *ex situ* or *in situ* conservation program. Furthermore, finding that the original identification of most specimens (except some specimens of *A. varius* and *A. zeteki*, Supplementary Table S1) is concordant with the results of the phylogenetic analyses, suggests that cryptic diversity is low among Central American *Atelopus*, and that hybridization resulting from misidentifications of specimens in captive programs should be rare. At least in the case of *A. limosus* this has been shown to be the case [80]. Further study is needed using multilocus nuclear markers or genomic data, especially on the likely case of hybridization between *A. varius* and *A. zeteki*. Additional genotyping should be conducted on captive specimens currently housed in *ex situ* programs. Finally, recognizing the taxonomic distinctiveness of the candidate species found herein is urgently needed in order to initiate the corresponding conservation measures to guarantee their long-term survival.
