Rediscovery of Osteocephalus vilarsi (Anura: Hylidae): an overlooked but widespread Amazonian spiny-backed treefrog

Osteocephalus vilarsi (Melin, 1941) is an Amazonian treefrog species known for over 75 years from its holotype only. Due to a lack of published data on its morphological diagnostic characters and their variations, as well as the absence of molecular, acoustic and ecological data supporting its identity, a highly dynamic taxonomic history has led this species to be confused and even synonymised with other Osteocephalus species from distinct species groups. The molecular phylogenetic relationships of O. vilarsi were investigated based on recently collected specimens from eight Northwestern Brazilian localities in the state of Amazonas, leading to its removal from the Osteocephalus taurinus species group and placement in the Osteocephalus planiceps species group. Furthermore, detailed data on morphology and colour variation are provided, as well as advertisement call and tadpole descriptions. Finally, the currently known geographic range of O. vilarsi is considerably extended, first data on the natural history of the species are provided, and the possible ecological preference of O. vilarsi for Amazonian white-sand forests is discussed.


INTRODUCTION
The spiny-backed treefrogs belonging to the genus Osteocephalus Steindachner, 1862 are medium to large-sized arboreal frogs inhabiting primary and secondary forests in the vast area of the Orinoco and Amazonian basins (Frost, 2019). They are distributed from Colombia, Venezuela and the Guyanas to Bolivia and Brazil (Jungfer et al., 2013) and their and the specimen of O. planiceps (Ca1_Neblina411) (sensu Jungfer et al., 2013) were collected at a similar elevation. With regard to this, it is interesting that specimens of O. planiceps reported from Jaú National Park , supposed to be O. vilarsi (Jungfer, 2010), were also collected in a white-sand forest.
In the frame of herpetological surveys of white-sand forest areas in the vicinity of the municipality of Novo Airão (west bank of Negro River, in the state of Amazonas, Brazil) carried out from 2015 to 2018, adults, juveniles and tadpoles belonging to a species of Osteocephalus morphologically corresponding to the description of O. vilarsi and genetically clustering with O. planiceps CA1 (sensu Jungfer et al., 2013) were collected. After morphological comparison of these specimens to the holotypes of O. planiceps and O. vilarsi, as well as to museum specimens of other Osteocephalus species, we attributed these individuals to O. vilarsi. Subsequently, we confirmed our determination by comparison of mitochondrial DNA sequences from these specimens with the DNA of individuals collected directly at the type locality of O. vilarsi.
Further revision of available Osteocephalus collections revealed that the specimens of Osteocephalus reported from Jaú National Park by  actually represent O. vilarsi, and that additional voucher specimens of this species from the Park territory are available. Moreover, specimens have already been sampled at several other localities in the west and north regions of the Negro River (Amazonas, Brazil; see Material and Methods and Fig. 1). Recently, the name O. vilarsi was used for specimens collected in Santa Isabel do Rio Negro area (Amazonas, Brazil) by Menin et al. (2017), but without providing any data on the morphology of these frogs or any explanation concerning the determination method. Examination of the respective voucher specimens (see Appendix 1 in Menin et al., 2017) revealed that all seven specimens listed as O. vilarsi represent a species belonging to the O. taurinus group. Interestingly, only one actual specimen of O. vilarsi (CZPB-AA 239) from Santa Isabel do Rio Negro material exists, but it was wrongly determined as O. planiceps (in addition, the second specimen listed by the above authors as O. planiceps, CZPB-AA 240, was determined incorrectly and represents a species belonging to the O. taurinus group).
In this study, we provide a detailed phylogeny of Osteocephalus and report detailed data on morphological variation and colouration of O. vilarsi, also describing its advertisement call and tadpoles. Finally, we extend the known range of O. vilarsi, provide first data on its natural history, and discuss the possible association of this species with Amazonian white-sand forests.

Examined material
The examined material attributed to O. vilarsi consists of 29 adults, two juvenile specimens and 10 tadpoles collected by different researchers from eight Brazilian localities in the state of Amazonas (for details see Fig. 1A; Table S1). This material includes three adult females from the type locality of O. vilarsi (Missão Taracuá) (Appendix I). In addition, we analysed DNA sequences of the only known Venezuelan specimen (AMNH 131254) collected in the Department of Amazonas close to the Brazilian border ( Fig. 1A; Table S2).  Jungfer et al. (2013), GBIF (www.gbif.org) and own unpublished records. Localities in map A (all in the state of Amazonas, Brazil, except for locality 2 in the Amazonas Department, Venezuela): 1, Missão Taracuá; 2, Venezuelan slope of Pico da Neblina; 3, Brazilian foothills of Pico da Neblina; 4, left bank of Japurá River; 5, Ayuanã River, Santa Isabel do Rio Negro; 6, Miratucu Lake, Jaú National Park; 7, Seringalzinho Village, Jaú National Park; 8, Jaú National Park; 9, Rio Negro Sustainable development reserve. A star denotes the type locality of O. vilarsi (Missão Taracuá (Palumbi et al., 1991) were used to amplify a~553 (476-588) bp-long fragment of the 16S rRNA mitochondrial gene. Polymerase chain reaction (PCR) protocols included a reaction mix with final volume of 15 mL, containing 2.0 mL ddH2O, 1.5 mL 25 mM MgCl2, 1.5 mL 10 mM dNTPs (2.5 mM of each dNTP), 3 mL 5X amplification buffer (75 mM Tris HCl, 50 mM KCl, 20 mM (NH4) 2SO4), 1.5 mL 2 mM solution of each primer, 0.3 mL Taq DNA polymerase 5 U/mL (Biotools, Madrid, Spain) and 1.5 mL of genomic DNA (about 30 ng/mL). Fragment amplification involved a pre-heating step at 73 C for 60 s, followed by 35 denaturation cycles at 94 C for 10 s, primer annealing at 50 C for 35 s and an extension at 72 C for 90 s, followed by a final extension step at 72 C for 10 min. Sequencing reactions were carried out after PCR purification using exonuclease and thermosensitive alkaline phosphatase, following the manufacturer's recommendations (Thermo Fisher Scientific, Waltham, MA, USA) and followed by the use of aABI BigDye Terminator Cycle Sequencing Kit following the manufacturer's instructions. Forward and reverse primers were used in the sequencing reactions (annealing temperature of 50 C) and resolved on an ABI 3130xl automatic sequencer. Sequences were manually verified at Geneious (Kearse et al., 2012).
In order to infer the phylogenetic relationships of O. vilarsi, we selected 80 sequences of 16S, 81 sequences of 12S rRNA (12S), 43 sequences of NADH dehydrogenase subunit 1 (ND1), 30 sequences of cytochrome oxidase I (COI) and 27 sequences of cytochrome b (CYTB) from GenBank corresponding to 81 specimens of Osteocephalus and four specimens of Dryaderces. The final matrix was composed by 98 terminals representing all nominal species of Osteceophalus (except O. duellmani) and two candidate species revealed by Jungfer et al. (2013) plus two specimens of Dryaderces pearsoni (Gaige, 1929) and two specimens of Dryaderces sp. CA1 used as outgroups (Table S2). Nuclear markers were not included in the dataset due to the absence of available sequences for approximately 90% of the analysed samples.
Sequences of each marker were aligned separately in Bioedit (Hall, 1999) using the ClustalW algorithm (Thompson, Higgins & Gibson, 1994) and manually checked.
Alignments were concatenated in MESQUITE 3.5 (Maddison & Maddison, 2018) and the final matrix was constituted of 4382 bp. PartitionFinder 2.1.1 (Lanfear et al., 2017) was used to infer the best nucleotide evolution model and partitions schemes through PhyML (Guindon et al., 2010) and Bayesian Information Criterion. The best-fit partition scheme and model evolution are displayed in Table 1.
The phylogenetic relationship was reconstructed using Bayesian Inference (BI). The BI tree was inferred in MrBayes 3.2.6 (Ronquist et al., 2012) using four runs of 10 million generations with a Metropolis-coupled Markov chain Monte Carlo algorithm. Each run comprised four Markov chains, with probabilities sampled every 1,000 generations. Convergence among runs and model stationarity parameters were verified using Tracer 1.7 (Rambaut et al., 2018). Tree files were combined in LogCombiner 1.8.4  after discarding 25% of the trees. The majority rule consensus tree was built using TreeAnnotator 1.8.4 . The average Kimura-2-parameter (K2P) (Kimura, 1980) and uncorrected pairwise genetic distances between O. vilarsi and the remaining dataset of Osteocephalus were estimated using MEGA 6.0 (Tamura et al., 2013).

Adult morphology
Sex and maturity were inferred through the presence or absence of secondary sexual characters (e.g. vocal sac, vocal slits, skin texture, nuptial excrescences on prepollex, presence of eggs). Measurements are given in mm and were taken to the nearest 0.1 mm using a dissecting microscope and a digital calliper. Nine measurements were taken according to Duellman (1970) as follows: snout-vent length (SVL), head length (HL: distance from the posterior edge of the jaw articulation to the tip of the snout), head width (HW at jaw angle), horizontal eye diameter (ED), tibia length (TL), horizontal tympanum diameter (TD), minimal interorbital distance (IOD), upper eyelid width (UEW), eye-nostril distance (EN). One measurement was taken following Heyer et al. (1990): thigh length (THL). In addition, disc width on Finger III (3FD) and Toe IV (4TD) were also measured.
Some Osteocephalus descriptions and redescriptions (Ron & Pramuk, 1999;Jungfer & Lehr, 2001;Jungfer, 2010;Jungfer et al., 2016) have stated that foot length followed Duellman (1970), who stated FL as 'the distance from the proximal edge of the inner metatarsal tubercle (the large tubercle at the base of the first toe) to the tip of the longest (fourth) toe (including the disc)'. However, the FL is always longer than TL in those studies, which is not consistent with Duellman's method. In our study, FL was taken from the tip of Toe IV to the heel, in order to make this measurement comparable to those provided in recent Osteocephalus studies. Webbing formulae follow the standards of Savage & Heyer (1967) as modified by Myers & Duellman (1982). Colour in life was described based on field notes and on digital photographs.

Morphological statistics
Although O. vilarsi and O. planiceps are not sister species according our phylogenetic results and the phylogeny published by Jungfer et al. (2013), these species are morphologically similar. Due to that, we performed a Principal Component Analyses (PCA) with morphometric data in order to investigate if the morphometric multivariate space occupied by these two species overlap. The PCA was performed with 12 morphometric ratios (HL/SVL, HW/SVL, IOD/SVL, EN/SVL, ED/SVL, UEW/SVL, TD/SVL, 3FD/SVL, 4TD/SVL, THL/SVL, TL/SVL, FL/SVL). As the SVL seems an important morphological character to differentiate species of Osteocephalus, we also included the SVL in the PCA. Analyses were conducted separately for adult males and females to avoid biases attributed to sexual dimorphism. PCAs were implemented in R environment (R Core Team, 2018) through the function 'prcomp' of the package stats.
We set the parameters 'scale'. and 'center' as 'TRUE' to scale and centre morphometric variables, respectively. The numbers of retrieved Principal Components (PCs) were determined through Broken Stick Model. PCA graphs were obtained trough the function 'autoplot' of the package ggplot2 (Wickham, 2016). Adult specimens of O. planiceps from Peru and Ecuador housed in the MCZ-A and NMP-P6V collections were used in the PCA and measured as described for O. vilarsi. Morphometric measurements of both species are available in Table S3. Results are presented in the subsection Comparaisons.

Tadpoles
The developmental stages of 10 tadpoles were determined according to Gosner (1960). Eight morphometric measurements were taken according to Altig & McDiarmid (1999) as follow: total length (TL); body length (BL); tail length (TAL); maximum tail height (MTH); tail muscle height (TMH); tail muscle width (TMW); internarial distance (IND); interorbital distance (IOD). Three measurements were taken following Randrianiaina et al. (2011): maximum body width (BW); maximum oral disc width (ODW) and size of the dorsal gap of marginal papillae (DG). Morphometric measurements were taken to the nearest 0.1 mm with a micrometer coupled to a dissecting microscope. Tadpole descriptions follow Schulze, Jansen & Köhler (2015) and were based on five tadpoles in Gosner stage 36. Colour was described based on one tadpole raised in laboratory until metamorphosis. Although the advertisement call of O. planiceps has been described by Ron & Pramuk (1999), we analysed 19 advertisement call from three males of this species from Ecuador in order to obtain the same acoustic parameters measured for O. vilarsi in the present study. Advertisment calls of O. planiceps were downloaded from Anfibios del Ecuador (https://bioweb.bio/faunaweb/amphibiaweb/: Ron, Merino-Viteri & Ortiz, 2019). Recordings were made by Morley Read in: (1) Reserva Biológica Jatun Sacha, Provincia Napo; (2) Parque Nacional Yasuní and (3) Pompeya Sur, cerca del río Napo, both in Provincia Orellana.

Molecular phylogenetic analyses
The Bayesian phylogenetic reconstruction based on five mitochondrial markers recovered the genus Osteocephalus as monophyletic (Fig. 2). The five species groups previously determined within Osteocephalus were strongly supported, as well as their evolutionary relationships. The O. taurinus species group is the sister clade to a large clade containing the remaining species groups. Within this large clade, the O. alboguttatus species group is the sister clade to the O. planiceps species group. The O. alboguttatus + O. planiceps species group is sister to the clade containing the O. leprieurii and O. buckleyi species groups (Fig. 2). With few exceptions, the inter-species relationships were highly supported (Fig. 2). Our phylogenetic analyses clearly confirmed the status of O. vilarsi as a valid species. All samples related to this species, including those from the type locality, form a strongly supported and monophyletic clade (Fig. 2). In addition, the individual from the Venezuelan slope of Pico da Neblina, previously considered a candidate species related to O. planiceps (Ca1_Neblina411 sensu Jungfer et al., 2013), nested within the same clade as O. vilarsi. Despite the extensive geographic distances between the known localities of O. vilarsi, the average intraspecific K2P distance within this clade was surprisingly low (ca. 0.4%; Table 2  Osteocephalus vilarsi (Melin, 1941) Hyla (Trachycephalus) vilarsi Melin, 1941: p. 40, Fig. 21.

Amended diagnosis
Osteocephalus vilarsi is a medium-sized species (as defined by Jungfer, 2010) of the O. planiceps species group according to the phylogeny presented in Jungfer et al. (2013) and in the present study. The species can be diagnosed by the combination of the following characters: (1) SVL 47.5-58.4 mm in adult males and 54.6-65.3 mm in adult females; (2) frontoparietal ridges on the head; (3) truncate snout in dorsal view and rounded in lateral view; (4) skin on dorsum of adult males conspicuously tuberculated; (5) vocal slits in males; (6) vocal sac distinct, subgular, moderately expanded laterally to the area between the tympanum and forearm insertion; (7) absence of subdigital nuptial excrescences in breeding males; (8) distal subarticular tubercle on Finger IV bifid; (9) toe length I < II < III < V < IV; (10) adults exhibit white tibiofibular bones, bicoloured iris (upper part bright golden with dark brown veins and fine incomplete dark brown radiation, lower part silver to bronze with dense bold dark brown veins and/or incomplete radiation) and light subocular spot; (11) metamorphs present iris entirely bright red without black reticulation, dorsum and flanks grey with dark grey blotches and spots and absence of reddish orange blotches on hand, elbow, knees, discs, and heels; (12) tadpoles at Gosner stage 36 TL = 33.0-34.5 mm, rounded snout in dorsal view, and LKRF = 2(2)/5-6(1); (13) advertisement call composed by two (169 ± 9 ms (144-180 ms)) or three notes (276 ± 48 ms (162-337 ms)), first note always formed by two pulses, and note duration of single notes lasts 51 ± 9 ms (36-65 ms).   (Fig. 3D) has an entirely bright red iris without black reticulation (black-reticulated iris, upper portion with red pigmentation on yellow ground and lower iris tan with red pigmentation near the pupil in metamorph of O. planiceps; Fig. 3H). The PCAs showed that the multivariate morphometric space occupied by females of O. vilarsi and O. planieps does not overlap and poorly overlaps in males (Fig. 4). The first two PCs retrieved by the Broken Stick Model in the PCA with females explained approximately 59% of the morphometric variation, while the first three PCs in the male analyses explained 63% of the variation (Fig. 4). The most important morphometric variables contributing to PC1 in female PCA were SVL, UEW, HW and ED. In the male PCA, the most important morphometric variables to PC1 were HW, T4W, TD and FL. See Table 4 for the contribution of all morphometric variables for other PCs. Additionally, the advertisement call of O. vilarsi is different from that of O. planiceps in structural and temporal parameters. The call duration of two-note (169 ± 9 ms Osteocephalus vilarsi can be differentiated from all members of the O. taurinus species group by a bicoloured iris, with the upper part bright golden with dark brown veins and fine incomplete dark brown radiation and lower part silver to bronze with dense dark brown veins and/or incomplete radiation and by white tibiofibular bones (iris greenish gold with bold dark brown regular radiation, green tibiofibular bones in O. taurinus and O. oophagus at their type localities; Jungfer & Schiesari, 1995;Lima et al., 2006;Jungfer, 2010). Additionally, males of O. vilarsi differ from those of O. taurinus by having vocal sac distinct, subgular and moderately distensible (vocal sac subgular, paired and strongly distensible). Tadpoles of O. vilarsi at Gosner stage 36 differ from those of O. oophagus at same stage by having TL = 33.0-34.5 mm, rounded snout in dorsal view, and LKRF = 2(2)/5-6(1) (TL = 28.9 mm, nearly truncate snout in dorsal view and LKRF = 2(2)/3 in O. oophagus at its type locality; Jungfer & Schiesari, 1995). Moreover, recently metamorphosed O. vilarsi individuals differ from members of the O. taurinus species group by the absence of reddish orange blotches on hand, elbow, knees, discs and heels (reddish orange blotches present; Lima et al., 2006;Jungfer et al., 2013).
In life (Figs. 7 and 8), adult specimens pale brown, dorsally yellowish brown to reddish brown, with or without a pattern of dark brown to black irregular markings; interorbital stripe and dark brown canthal stripe present (interorbital stripe narrower than the diameter of the eye in all specimens except for female INPA-H 40462, in which a triangular spot is formed); a narrow pale supralabial line expanding in a subocular spot; creamy to yellowish white flanks, with or without irregular brown markings or small spots; light brown hidden thigh surfaces; creamy or yellowish white throat and belly with or without diffuse small pale brown markings or spots; a narrow dark line present along the mandible; fleshy pink to slightly orange ventral thigh surfaces; bicoloured iris with a dark brown horizontal stripe, bright golden above with dark brown veins and fine incomplete dark brown radiation, silver to bronze below with dense bold dark brown veins and/or incomplete radiation (Figs. 7 and 8); white tibiofibular bones (Figs. 7E and 8F).

Sexual dimorphism
In addition to the presence of vocal slits, vocal sac and dark keratinous nuptial excrescences in males, O. vilarsi also exhibits sexual dimorphism of body size and dorsal skin texture . Adult males possess numerous protuberant spinous tubercles
The dominant frequency is located in the fundamental harmonic. The overall dominant  Table 5.

Tadpole and metamorph colouration
In life, tadpoles at Gosner stage 36 display a light brown dorsum with irregularly distributed dark brown blotches (Fig. 11A); a dark brown band from the nostril to the anterior portion of the orbits; a black iris with a red ring around pupil. Translucent venter. Dorsal portion of the tail muscle bronze; ventral portion pinkish cream; fins translucent with light brow blotches. Tadpoles at Gosner stage 39-40 are similar in colour pattern (Fig. 11B), except for the dorsum colour and fin blotches (that become darker) and by the presence of a large cream blotch on the knee and heel. Tadpoles at Gosner stage 45 display a silvery cream dorsum with dark and light grey blotches and spots (Fig. 11C); a dark grey band covering the lateral of head, from the snout to the posterior portion of the tympanum and on the upper lips; black iris with a red ring around pupil; grey fingers, hands, forearms, toes and feet; white upper arm; white blotch on the knee and heel; light grey groyne; dark grey tail; white venter. In life, metamorphs present a grey dorsum with dark grey blotches (Fig. 11D); a large dark grey band on the interorbital region. White subocular region, extending to the anterior infratympanic area; a dark stripe covering supratympanic region, from the posterior corner of the eyes to the arm insertion to the body. Bright red iris and black oval pupil. Limbs grey, except for the upper arm (white); four transversal dark grey stripes on the thigh; three horizontal dark grey stripes on the forearm and tarsus; light grey blotch on knee; white blotch on the heel. Venter white.

Reproductive behaviour
Two O. vilarsi calling males were observed at RDS Rio Negro (municipality of Novo Airão, Negro-Solimões interfluve) during 24-26 October 2017. The first male (not collected) called from the same place for three consecutive nights. It occupied a calling place on a horizontal leaf of herbaceous plant ca. 60 cm above the ground. The plant was growing close to a small phytotelma formed by trunk ridges of a large fallen tree (Fig. 12B). The phytotelma was approximately at the same height above ground as the calling place. The second male was found calling on vegetation ca. 40 cm above the ground at a small temporary forest puddle. Both breeding places were also occupied by calling males of Rhinella sp. (Rhinella margaritifera species group). Tadpoles from RDS Rio Negro were found in a shallow puddle not connected to a stream in a secondary white-sand forest. Specimens from the Jaú National Park were collected and observed in reproduction (May 2000) on the ground or up to 1 m high in shrub branches in open white-sand forests known as campina (canopy below 10 m), near or above temporary puddles. Some calling males presented territorial behaviour by altering their vocalisation in response to a playback (not recorded). Based on those notes, it appears that O. vilarsi can breed in small water body sources in closed forests (e.g. small temporary puddles, phytotelmata), as well as in open forests.

Distribution and habitat
Until now, O. vilarsi was recorded in localities inside the interfluve between the Negro and Solimões rivers, as well as at the left bank of the upper Negro River (Figs. 1A and 1B). Such a wide geographic range covering approximately 212,000 km 2 indicates that O. vilarsi is widely distributed across the extreme North-western Brazil (state of Amazonas) and the adjacent Southern Venezuela. The western most known locality is ca. 40 km from the Colombian border. Therefore, its occurrence can be expected in Colombian territory. During field surveys in the vicinity of the municipality of Novo Airão, O. vilarsi was recorded both in relatively undisturbed forests and in heavily altered habitats adjacent to communities and farmlands. It was found both in the semi-open white-sand forest known as campinarana (canopy below 20 m; Fig. 12A) and in high closed forests (canopy above ca. 20 m; Fig. 12B). No individuals were found in the open white-sand vegetation known as campina (canopy below 10 m). After dusk, individuals were found perched on vegetation (mostly on narrow vertical trunks of smaller trees) ca. 50-200 cm above the ground. In two cases, adult individuals were found hidden in open vertical plastic tubes, used for the  Several non-reproductive specimens of O. vilarsi at Jaú National Park were observed in both flooded and unflooded rainforests (known as igapó and terra firme forests, respectively; canopy above ca. 20 m), campinarana (canopy below 20 m), and campina (comprising shrubs and exposed sand). Specimens were perched on horizontal branches of small trees or on vertical trunks. Other hylid species found in sympatry with O. vilarsi at the Jaú National Park are listed in Neckel-Oliveira & .
Similarly, individuals recorded during the field survey at the Brazilian foothills of the Pico da Neblina, close to the Venezuelan border, were found perched on horizontal branches of small trees, in an ecotonal white-sand forest (campinarana) with a low canopy and several small temporary puddles. In this region, O. vilarsi occurs in sympatry with the hylids Boana aff. cinerascens, B. lanciformis, Dendropsophus minutus, D. tintinnabulum (Melin, 1941), Osteocephalus aff. taurinus, Scinax ruber and S. cruentommus (Duellman, 1972).
In fact, all individuals of O. vilarsi analysed in this study were recorded in localities covered by white-sand forest, or in close vicinity. Such an association indicates an ecological preference of this species to this particular and threatened type of Amazonian forest. In fact, the estimated distribution of O. vilarsi presented herein mostly coincides with the occurance of this forest type inside the Amazon in North-western Brazil and adjacent Colombia and Venezuela (Fig. 1A).

DISCUSSION
The results of the phylogenetic analysis based on five mitochondrial markers are in good agreement with the previous broader phylogenetic hypotheses proposed for Osteocephalus by Ron et al. (2012) and Jungfer et al. (2013). Monophyly of the five species groups was strongly supported. The main difference between the present study and previous phylogenetic analyses lies in the topology of the species groups. Whereas Ron et al. (2012) and Jungfer et al. (2013) recovered the O. planiceps species group as sister to the O. leprieurii species group, our analyses placed this group in a sister position to the O. alboguttatus species group. Different topologies concerning these three phylogenetic hypotheses may reflect differences in specimen sampling, analysed genes and amount of missing data from each dataset.
Our phylogenetic and morphological analyses revealed O. vilarsi to be a valid species belonging to the O. planiceps species group. It is interesting however, that one individual of O. vilarsi (AMNH 1312546, Venezuelan slopes of the Pico da Neblina mountain range) was already recognised at the specific level in most of the previous phylogenetic hypotheses concerning the genus. Due to the absence of other voucher specimens and lack of comparative morphological data, the individual from the Venezuelan slope of the Pico da Neblina was wrongly associated with other Osteocephalus (e.g. O. leprieurii, O. 'leprieurii', or O. planiceps CA1; Wiens et al., 2006;Moravec et al., 2009;Wiens et al., 2010;Ron et al., 2012;Salerno et al., 2012;Jungfer et al., 2013).
The apparent strong ecological association of O. vilarsi to white-sand ecosystems (e.g. primary and secondary white-sand forests, as well as ecotonal zones between white-sand forests, igapó and terra firme forests) represents a rarely reported example of such an association in Amazonian anurans. To the best of our knowledge, a similar ecological association is known only in the case of the casque-headed frog Aparasphenodon venezolanus (Mertens, 1950), whose distribution partially overlaps with that of O. vilarsi at Jaú National Park (De Carvalho et al., 2018). Despite rarely reported in anurans, such a specific association is widely known and studied in Amazonian birds, which present highly endemic white-sand assemblages (Adeney et al., 2016;Borges et al., 2016).
As the white-sand forests cover large but patchily distributed portions of the upper and middle Negro River basin (Adeney et al., 2016), we expect O. vilarsi to present a similar distribution pattern across this area. White sand ecosystems belong to the most sensitive vegetation types in Amazonia (Adeney et al., 2016), and the effects of their exploration and alteration on the dynamics of O. vilarsi populations in different parts of the Amazon are still unknown.
Osteocephalus vilarsi is a remarkable case of a widely distributed Amazonian anuran species, which has been overlooked for decades. Its rediscovery illustrates how an integrative approach to the study of Amazonian amphibians is vital. Our extensive examination of morphological, molecular, geographical, ecological, and bioacoustical data, combined with the revision of museum collections has updated and extended the geographic range of the species (previously known only from its type locality) to an area covering over 200,000 km 2 at the Negro-Solimões Interfluve and surrounding Brazilian and adjacent Venezuelan areas (the species is also expected to occur in adjacent Colombia). The scientific knowledge concerning O. vilarsi should rapidly increase, since the data reported herein (including morphological variation in adults, juveniles and tadpoles and advertisement call characteristics) should serve as a basis for further identification of individuals belonging to this species.

CONCLUSION
Osteocephalus vilarsi (Melin, 1941) has been rediscovered 75 years after its description. Based on morphological and molecular analyses, O. vilarsi is replaced from the O. taurinus species group to the O. planiceps species group. Morphological variation of adult, subadult and juvenile specimens, morphology of tadpoles, advertisement call, and natural history of O. vilarsi are described for the first time. Biogeographic data demonstrate that O. vilarsi, previously known only from its type locality, is widely distributed across the interfluve between the Negro and Solimoês rivers in North-western Brazil and adjacent southern Venezuela. The species displays strong ecological association to white-sand ecosystems (primary and secondary white-sand forests).