Taxonomic revision and phylogenetic position of Osteocephalus festae (Anura, Hylidae) with description of its larva

Abstract Osteocephalus festae is an Amazonian species recently resurrected from a synonymy with Osteocephalus buckleyi. Because few specimens are known, its morphological variation, diagnostic characters, and distribution are poorly understood. Herein we determine its phylogenetic relationships and provide a complete taxonomic account based on recently collected specimens (adults and larvae) from nine localities in Ecuador and Peru. Osteocephalus festae is most similar to Osteocephalus verruciger from which it differs in having less tuberculate dorsal skin on males, smaller tympanum, and more tooth rows in the oral disk of larvae. A phylogeny based on mitochondrial DNA sequences, genes 12S and ND1, shows that Osteocephalus festae is closely related to Osteocephalus buckleyi, Osteocephalus mutabor and Osteocephalus verruciger. A clade consisting of Osteocephalus festae, Osteocephalus verruciger, and Osteocephalus buckleyi is characterized by stream dwelling tadpoles. Surprisingly, we found paraphyly among Ecuadorian populations of Osteocephalus buckleyi and Osteocephalus verruciger. The causes for paraphyly are unknown but in Osteocephalus buckleyi may result from the existence of cryptic species.


Introduction
Osteocephalus is a genus of hylinae frogs (tribe Lophiohylini) distributed in the Amazon Basin and the Guiana Shield (Faivovich et al. 2005). Th ere are 24 recognized species of which half have been resurrected or described since 2000 (Frost 2010). Despite these eff orts, taxonomic problems persist, including undescribed species and binomials of unknown validity or poorly understood boundaries. One such case is O. festae, a species described by Peracca (1904) on the basis of a single specimen.
Th e holotype of O. festae is an adult female collected at "Valle Santiago", Provincia Morona Santiago, Ecuador. After its description, this binomial was largely ignored until Trueb and Duellman (1971) synonymized it under O. buckleyi (Boulenger, 1882) based on comparisons of the holotype of O. festae with series of O. buckleyi from Guyana, Colombia, Ecuador and Peru. Th is synonymy was followed by all systematic accounts until Jungfer (2010) correctly resurrected O. festae on the basis of the distinctiveness between the holotype of O. festae and O. buckleyi. Jungfer (2010) also ascribed to O. festae fi ve specimens from Napo and Sucumbíos provinces, Ecuador.
Recently collected specimens of Osteocephalus from nine populations from southeastern Ecuador and northeastern Peru, one of them at a distance of ~30 km from the type locality ( Fig. 1), closely resemble the holotype of O. festae and are morphologically and genetically distinctive from other species. Th ey also seem to be distinctive from the specimens ascribed to O. festae by Jungfer (2010) which may belong to a diff erent species (see Taxonomic Remarks). Because little is known about O. festae beyond the description of its holotype, below we provide an account of its variation, diagnosis, and distribution, as well as a description of its larvae. In addition, we assess its phylogenetic relationships using mitochondrial DNA sequences.

DNA extraction, amplifi cation, and sequencing
Total DNA was extracted from muscle or liver tissue preserved in 95% ethanol and tissue storage buff er using a guanidine tiocyanate protocol. Polymerase chain reaction (PCR) was used to amplify the mitochondrial genes 12S rRNA and ND1. We amplifi ed one DNA fragment for 12S and one or two overlapping fragments for ND1 using primers listed in Goebel et al. (1999) and Moen and Wiens (2009). PCR amplifi cation was carried out under standard protocols. Amplifi ed products were sequenced by the Macrogen Sequencing Team (Macrogen Inc., Seoul, Korea).

Phylogenetic analyses
A list of the samples included in the phylogenetic analyses is shown in Table 1. For the outgroup, we included sequences of Osteopilus and Hypsiboas obtained from GenBank (http://www.ncbi.nlm.nih.gov/genbank). Outgroup choice was based on Locality data from Trueb and Duellman (1970) and specimens deposited at the Museo de Zoología of Pontifi cia Universidad Católica del Ecuador, the Herpetology Collection, Escuela Politécnica Nacional, and CORBIDI (Appendix 1). Numbers correspond to those on Table 1 and Figure 2. phylogenies showing that Osteocephalus is most closely related to Tepuihyla and Osteopilus (Faivovich et al. 2005 andWiens et al. 2010). Because missing sequence data can result in misleading estimates of topology and branch lengths in phylogenies  (Lemmon et al. 2009), we only included GenBank sequences for which both genes were available. Preliminary sequence alignment was done with CLUSTALW 1.83 (Chenna et al. 2003). Th e sequence matrix was imported to Mesquite (version 2.72; Maddison and Maddison 2009) and the ambiguously aligned regions were adjusted manually to produce a parsimonious alignment (i.e., informative sites minimized). Phylogenetic trees were obtained using Bayesian inference. Th e models of character evolution for the Bayesian analyses were chosen using JModelTest version 0.1.1 (Posada 2008) using the Akaike Information Criterion with sample size correction as optimality measure. We applied independent models to each of four partitions: one for 12S and three for each codon position in ND1. Four Markov chains were utilized in each of two Bayesian analyses, the prior for the rate matrix was a uniform dirichlet and all topologies were equally probable a priori. Each analysis ran for 5 × 10 6 generations. For each analysis, the chain was sampled every 1000 generations.
After 5 x 10 6 generations the average standard deviation of split frequencies was ~ 0.002 indicating that the two analyses have converged into a stationary distribution. Th e fi rst 50% of sampled trees were discarded as the burn-in and the remaining trees were used for estimating the Bayesian tree, posterior probabilities and other model parameters. Phylogenetic analyses were carried out in MrBayes 3.1.2 (Ronquist and Huelsenbeck 2003).

Morphological analyses
For ease of comparison, we generally follow the format of Trueb and Duellman (1971) for diagnosis and description. Morphological terminology and abbreviations follow Lynch and Duellman (1997) for adults and Altig and McDiarmid (1999) for tadpoles. Description of oral disk structure follows Altig and McDiarmid (1999). Notation for hand and foot webbing is based on Myers and Duellman (1982). Sex and reproductive condition was determined by the presence of nuptial pads, vocal sac folds, dorsal skin texture, and/or by gonadal inspection. Tadpoles were staged according to Gosner (1960) and preserved in 10% formalin. Other specimens were fi xed in 10% formalin and preserved in 70% ethanol. To identify the tadpoles and juveniles we grew several tadpoles in captivity until they reached the juvenile stage. Juveniles exhibited a color pattern characteristic of Osteocephalus. Th e only other Osteocephalus known at the Río Napinaza collection site breeds on ponds (O. taurinus) and has a diff erent juvenile morphology (Lima et al. 2006). Examined specimens (listed in the type-series and Appendix I) are housed at Museo de Zoología, Pontifi cia Universidad Católica del Ecuador (QCAZ), the Herpetology Collection, Escuela Politécnica Nacional (EPN-H), and the collection of the División de Herpetología, Centro de Ornitología y Biodiversidad (CORBIDI).
Principal Components Analysis (PCA) and Discriminant Function Analysis (DFA) were used to assess the degree of morphometric diff erentiation between adult O. buckleyi, O. festae, and O. verruciger (Werner 1901). Only well preserved specimens (Simmons 2002) were measured for the following eight morphological variables, following Duellman (1970): (1) Snout-vent length (SVL); (2) head length; (3) head width; (4) tympanum diameter; (5) femur length; (6) tibia length; (7) foot length; and (8) eye diameter. All variables were log-transformed. To remove the eff ect of covariation with SVL, the PCA and DFA were applied to the residuals from the linear regressions between the seven measured variables and SVL. We applied a multivariate analysis of variance (MANOVA) to test for morphometric diff erences between sexes. Because we found signifi cant diff erences in O. buckleyi, the PCA and DFA were applied to each sex separately. For the PCA, only components with eigenvalues > 1 were retained. Sample sizes for O. verruciger were 23 males, 5 females; O. festae 7 males, 18 females; and O. buckleyi 25 males, 3 females. Both PCA and DFA were conducted in JMP® 8.01 (SAS Institute 2008).

Phylogenetic relationships
Th e models with the best fi t and the estimated parameters for each of four partitions for the Bayesian analyses are shown in Table 2
Variation. Variation in dorsal and ventral coloration of preserved specimens is shown in Figures 6 and 7. Dorsal coloration consists of a light brown to dark brown background with irregular marks. Th ere is sexual dimorphism in dorsal tuberculation: in females the dorsum is smooth while in males it varies between having scant and ill-defi ned non-keratinized tubercles (most males from Río Napinaza, e.g., QCAZ 26488) to having abundant keratinized tubercles (two males from Chonza Alta, e.g., CORBIDI 758; Fig. 3B).
Ventral surfaces of preserved specimens (Fig. 7) have a cream (QCAZ 39364) to tan (QCAZ 39806) background with darker brown marks that are more distinct and abundant in females (e.g., QCAZ 39811) than in males (e.g., QCAZ 39799); a male from Río Lejía (CORBIDI 623) has an immaculate venter. Ventrally, limbs vary from brown to cream; in QCAZ 39809 and 39811 cream dots are present on hindlimbs; scant cream tubercles can be present in the external edge of the forearm (e.g., QCAZ 39804). Th e vent region is dark brown to brown bordered by a lighter area (cream to tan). Flanks are areolate in the anterior ha lf and smooth posteriorly. Th e areolate portion is cream with dark brown reticulation; the posterior half is cream (e.g., QCAZ39810) to light brown (e.g., QCAZ 39806) with dark brown blotches.
Head shape is rounded in dorsal view and rounded (e.g., QCAZ 39803) to bluntly rounded (e.g., QCAZ 39800-01) in lateral view. Lateral head coloration varies between dark brown (QCAZ 11625) to light brown (QCAZ 39810). Except for QCAZ    39802 and 41039, there is a lighter (brown to cream) subocular mark. A tan (QCAZ 39805) to cream (QCAZ 39364) labial stripe is always present. Th e tympanic annulus is concealed dorsally and has lighter color than the background. Variation in hand and foot webbing is shown in Table 3. Th e distal subarticular tubercle on Finger IV is single in all specimens. Morphometric data pertain to adults and are summarized in Table 4. In the examined series, the largest male has a SVL of 56.09 mm and the largest female 84.94 mm; mean male SVL = 49.47 mm (n = 12; SD = 5.10), mean fe ma le SVL = 6 7.92 mm (n = 27; SD = 10.08). Females are signifi cantly larger than males (t = 5.52, df = 23, P < 0.001). A MANOVA on the residuals of the regressions between SVL and the other measured variables indicates lack of signifi cant diff erences between sexes in size-free morphometry (F = 1.052, df = 17, P = 0.433).
Color in life. Based on digital photograph of adult female QCAZ 41039 (Fig. 3A): dorsum dark brown with irregular ligh t brown and yellowish green marks; canthal region dark brown with yellowish green subocular mark and labial band; tympanum brown; fl anks greenish brown with dark brown reticulation anteriorly and irregular dark brown marks posteriorly; dorsal surfaces of thighs and shanks dark brown with transversal brown bands bordered with light brown; dorsal surfaces of forelimbs dark brown with irregular brown marks; venter light tan with irregular brown marks; bones green; iris dark brown. Female CORBIDI 761 has a predominantly light brown dorsum with irregular brown marks; clear areas on fl anks and below the eye and tympanum are light yellow.
Th ere is signifi cant change in coloration between juveniles and adults. Th e following description is based on a digital photograph of juvenile QCAZ 38081 (Fig. 3E-F). Th e dorsum beige with black interorbital band and two large medial ovoid black blotches; fl anks dark brown; dorsal surfaces of thighs and shanks brown with cream transversal bars; dorsal surfaces of arms cream, dorsal surfaces of forearms brown with cream transversal bars; knees, elbows, and heels cream; anterior half of the venter cream, posterior half light brown; bones green; iris bright red.
Morphometric comparisons. Th ree components with eigenvalues > 1 .0 were extracted from the PCA for males (Table 5). Th e three components accounted for 76.4% verruciger, (t = 9.03, df = 9, P < 0.001). Th ree components with eigenvalues > 1.0 were extracted from the PCA for females (Table 5). Th e three components accounted for 80.3% of the total variation. Th e highest loadings for the PCA for females were tibia length and femur length for PC I, eye diameter and tympanum diameter for PC II, and head width and head length for PC III (Table 5) Table 6. Measurements of tadpoles of Osteocephalus festae (lot QCAZ 30511). Developmental stages, in parentheses, are defi ned according to Gosner (1960) (1999). Morphometric data are provided in Table 6. In dorsal view, a tadpole in Stage 39 (QCAZ 30511H; Fig. 5A) shows elliptical body, widest between eye and spiracle, with rounded snout. Eyes relatively large (body length about 7.89 times larger than eye diameter), directed and positioned dorsolaterally, not visible in ventral view, and separated by a distance 1.27 times the internarial distance. External nares oval, located dorsolaterally, at about one fourth the distance between anterior margin of snout and anterior margin of eye. In profi le (Fig. 5B) body depressed (body width/body height = 0.18), fl attened ventrally, snout slightly rounded. Oral disc not emarginated. Spiracle sinistral, inner wall free from body, its tip closer to the vent than the eye. Spiracle opening rounded. Tail musculature robust, decreasing in size towards tip of tail. Dorsal fi n not extending onto body, slightly convex and attaining its maximum height at mid length of tail; tail tip rounded; ventral fi n convex, beginning at tail-body junction and tapering gradually towards tail tip. Medial vent tube with both walls attached directly to ventral fi n, opening directed posteroventrally. Limbs with subarticular patches. Dorsal body, middle body, supraorbital, infraorbital, posterior supraorbital, and posterior infraorbital lateral lines evident. No glands.
In preservative, dorsum brown with darker marks between eyes; dark brown dorsolateral stripes extend from mid-body to base of tail; caudal musculature beige with brown spots (Fig 5); skin of fl anks, spiracle, vent tube, fi ns, and around the eyes transparent; belly and fi ns transparent with white blotches.
Tadpole variation and comparisons with other species. In QCAZ 38074, 26321, 26053, 26498 and 26284 the caudal musculature is cream with brown dots; fi ns can have dark brown spots without white blotches (e.g., QCAZ 38074, 26321). Th e LTRF is the same in all stages but in stage 42, rows A2-A4 and P1-P3 have approximately half the number of teeth.
In preservative, ten tadpoles collected in Chonza Alta, Peru, lot CORBIDI-CL-10 in Stages 37 (A-B), 36 (C), 34 (D), 32 (E), 31 (F), 42 (G, H, I) and 44 (J) have dorsum brown with darker marks between the eyes; dark brown dorsolateral stripes from mid-body to base of tail; caudal musculature cream with abundant melanophores; skin of fl anks, spiracle, vent tube, and fi ns transparent; skin around the eyes brown; belly and fi ns transparent with abundant melanophores. Oral disc with LTRF 5/7; papillae distributed around oral disc; tooth rows complete except for medial gap in row A5.
In life (Fig. 5E-F; QCAZ 38074, stage 33; based on digital photograph), dorsum dark brown with darker marks between the eyes; dark brown dorsolateral lines from mid-body to base of tail; tail musculature light brown with small dark brown melanophores; white dots at tail-body junction; skin is tr ansparent ventrally in anterior half of body and ventrolaterally in the posterior half; tail musculature light brown with dark brown spots; fi ns transparent. Iris bronze. Live tadpoles from Chonza Alta (CORBI-DI-CL-10) have dorsum and caudal musculature olive brown; skin transparent ventrally with bright brown fl ecks, gut visible through the skin; fi ns translucent brown.
Distribution and ecology. Osteocephalus festae has been recorded at nine localities in the Ecuadorian (Loja, Morona Santiago, and Zamora-Chinchipe provinces) and Peruvian Amazon basin (Mariscal Cáceres and Rioja provinces). Localities with known elevation (Río Napinaza, Miasí, San Francisco, Reserva Tapichalaca, Río Lejia, Chonza Alta, Camñopite Bajo, and Naranjillo) range between 1000 and 2200 m of elevation. Th e elevation at San Francisco (2200 m) is the highest known for Osteocephalus. Maximum airline distance between localities is 440 km. Osteocephalus festae and O. verruciger have similar elevational ranges and seem to replace each other latitudinally in Ecuador (Fig. 1). Records of O. verruciger from Peru (e.g., Trueb and Duellman 1970) are likely misidentifi ed O. mimeticus (Jungfer 2010). Th us, the southernmost confi rmed records of O. verruciger are those from Provincia Morona Santiago, Ecuador.
Most of our specimens of O. festae are from Río Napinaza, a river surrounded by secondary forest, pastures and agricultural lands. At the collection site, the river has an average width of 2.85 m and an average depth of 23 cm with fast running water and waterfalls that reach 10 m in height (Salazar-Valenzuela 2007). Tadpoles were found in small ponds in the margins of the river. Adults were observed at night next to the river or within the forest on vegetation 40 to 250 cm above the ground.
All the specimens collected in Las Cataratas de Paraiso (Chonza Alta) and Camñopite were found at night on vegetation 50 to 300 cm above the ground, next to fast running streams. Tadpoles (CORBIDI-CL-10) were found in a rocky stream with average width of 4 m and an average depth of 30 to 40 cm with fast running water, close to the base of a waterfall. At both sites the streams are surrounded by secondary forest, pastures and agricultural lands. Th e specimens from Bajo Naranjillo and Río Lejia were found at night on branches 150 to 200 cm above the ground (in primary forest at Río Lejia and secondary forest surrounded by pastures at Naranjillo).
At Las Cataratas de Paraiso (Chonza Alta), on 1 December 2007, we found twelve males calling from the low vegetation and six amplectant pairs (Fig. 3B). Recently metamorphosed individuals were perching on leaves and rocks at the shore. On 4 November 2008, at the same stream, we only found one adult non-amplectant female, two adult males, several tadpoles, and 12 freshly metamorphosed juveniles on leaves and rocks (e.g., CORBIDI 1962-64). Th e rainy season in this region generally starts in December but during our surveys heavy rains fell since the fi rst week of November.  (2010), both females diff er from the holotype and 18 adult females analyzed here (in parentheses) in having a venter uniform tan without marks (brown marks present), webbing almost reaching the ultimate subarticular tubercle in the inner edge of third fi nger (web reaching half the distance between ultimate and penultimate subarticular tubercles), and larger tympanum size with TD/HL = 0.23-0.25 (TD/HL = 0.14-0.22 among 39 adult males and females). In addition, we did not fi nd the sexual dimorphism in relative tympanum size reported by Jungfer (Student's t = 1.227, df = 23, P = 0.232). Males assigned to O. festae also seem to differ from our series in the extent of the axillary membrane (covering half of the upper arm vs. one third to one fourth in our series). Th e discrepancies suggest that at least some of the specimens assigned to O. festae by Jungfer (2010) may belong to a diff erent species.

Discussion
Th e phylogenetic relationships recovered by this study are consistent with the phylogenies reported by Wiens et al. (2010) and Moen and Wiens (2009) -O. verruciger))) is composed by species that mainly reproduce along streams or slow fl owing ditches. At several sites we have found tadpoles of O. verruciger in ponds on stream banks and also on slow fl owing ditches confi rming the riparian habits reported by Trueb and Duellman (1970). Osteocephalus buckleyi breed along streams (Jungfer 2010;Lima et al. 2006) and O. mutabor has been found breeding along ditches (SRR pers. obs.;Jungfer and Hödl 2002) and temporary ponds (M. Read, pers. comm.) Th e predominance of stream breeding habits among these species suggests that reproductive mode may be phylogenetically conserved in Osteocephalus. Th e same pattern is suggested by the close relationship between O. planiceps, O. deridens, O. fuscifacies implied by our phylogeny and that of Moravec et al. (2009) because these species share phytotelmata breeding (SRR pers. obs.; Jungfer et al. 2000). Th is reproductive mode, however, may have an additional independent origin in O. oophagus (Moravec et al. 2009).
An unexpected result in our phylogeny is the fi nding of paraphyly among populations of O. buckleyi and O. verruciger from Ecuador. Plausible explanations include incomplete lineage sorting, mitochondrial gene capture, and the existence of cryptic species hidden within each taxa. In the case of O. verruciger, the position in the phylogeny of the population that generates paraphyly, Pacto Sumaco, is weakly supported (Fig. 2) and the genetic distance between Pacto Sumaco and the other O. verruciger populations is lower (1.5-1.8% sequence divergence) than the distances between O. verruciger and O. buckleyi or between any other species pair in the phylogeny (> 2.1%). In addition, we could not fi nd conspicuous morphological diff erences between Pacto Sumaco and the other O. verruciger populations. Th e observed pattern suggest that our mitochondrial gene tree may not correctly refl ect the history of divergence among O. verruciger and the morphologically distinctive populations of O. buckleyi from Hola Vida and Bobonaza. Conversely, the paraphyly among populations of O. buckleyi has a strong support with higher genetic distances between both clades (4.7%-6.0%). Neither of these clades represent O. cabrerai, a species morphologically similar to O. buckleyi for which there are not confi rmed records from Ecuador (Jungfer 2010). Examination of additional characters (morphological and molecular) is underway by SRR to determine the taxonomic status of the populations of O. buckleyi from western Amazonia.