Introduction

The species-rich family Bufonidae Gray, 1825 is a nearly cosmopolitan clade that originated in South American in the Upper Cretaceous and rapidly expanded all over the major continents except the Australo-Papuan region, Madagascar, and Antarctic (Pramuk et al. 2008; Van Bocxlaer et al. 2010; Frost 2023). Within the more than 600 known species, a relevant diversity has been reported, including a great spectrum of reproductive modes, from oviparity with lentic or lotic free-living exotrophic tadpoles, endotrophic tadpoles that develop in phytotelmata, up to viviparous species with direct development (Baldo et al. 2014; Wake 2015; Chandramouli et al. 2016; Liedtke et al. 2022).

Originally associated with the genus Atelopus Duméril & Bibron, 1841, the bufonid Frostius pernambucensis (Bokermann 1962) was described (as Atelopus pernambucensis) for a remnant of the Atlantic Forest in the municipality of Recife, state of Pernambuco, Brazil (Bokermann 1962). After two decades of the species discovery, Cruz and Peixoto (1982) described its tadpole from the type locality, as well as information on spawning and oviposition site. The species is a phytotelm-breeding frog with endotrophic larvae (Cruz and Peixoto 1982). The reproductive differences observed between F. pernambucensis and the remaining Atelopus suggested that they belong to different lineages (Cruz and Peixoto 1982). Later, Cannatella (1986), based on data on adult morphology, larvae, and spawning, proposes the genus Frostius to accommodate this species. The genus remained monotypic until the recent description of F. erythrophthalmus Pimenta & Caramaschi 2007 for the Atlantic Forest of the municipality of Uruçuca, state of Bahia, Brazil, whose tadpole remains unknown (Pimenta and Caramaschi 2007). Currently, Frostius is endemic to the Atlantic Rainforests of northeastern Brazil (Frost 2023). The genus is one of the early divergent lineages within bufonids, being the sister to all bufonids but Melanophryniscus (Jetz and Pyron 2018).

Larval morphology information proved to be very important and informative in understanding the taxonomy and the phylogenetic relationships of several anurans (e.g., Noble 1929; Orton 1953; Maglia et al. 2001; Haas 2001, 2003; Pugener et al. 2003; Dias et al. 2021, 2023), but data are lacking for several taxa. This is the case of Frostius pernambucensis; even though larval morphology was pivotal for the generic classification of the species, since the original tadpole description no further studies were performed, and aspects of its internal anatomy remain unknown. Moreover, the tadpole of F. pernambucensis was characterized based on seven individuals in early developmental stages and one egg from the same spawning raised in laboratory, and thus, several important characters were not observed or described. Recently, Dubeux et al. (2020a) provided a brief characterization for the tadpoles of F. pernambucensis for state of Alagoas, aiming a generic identification, but not addressing other morphological aspects of the tadpole. The fact that the larvae of F. pernambucensis are endotrophic (Cruz and Peixoto 1982) is contrasting with most of bufonids, making it a very intriguing piece in the puzzled evolutionary history of the family.

Given the above, a more detailed approach is needed to understand the evolution of the unique tadpoles of Frostius pernambucensis. Herein, we present a complete redescription of external morphology and the first description of internal anatomy of tadpole of this species. Additionally, we discuss the importance of larval characters in the systematics of bufonids and address the evolution of endotrophic tadpoles.

Materials and methods

Tadpoles were collected in Atlantic Forest remnants in the state of Pernambuco and Alagoas (Fig. 1). The three first lots (MHNUFAL 14572, 16192, 16196) are from Estação Ecológica de Murici, municipality of Murici, state of Alagoas, Brazil, 167 km south-west from the type locality of species (35° 52′ 41.50" W; 9° 12′ 54.74" S, 462 m a.s.l.; Fig. 1). Tadpoles were collected in July 2015, June, and July 2021, respectively. The other two lots (MHNUFAL 17359, 17360) are from Reserva Particular do Patrimônio Natural Eco Fazenda Morim, municipality of São José da Coroa Grande, state of Pernambuco, Brazil, 99 km south from the type locality of species (8° 52′ 10.46" S, 35° 12′ 28.57" W; 92 m a.s.l.), and were collected in May 2022.

Fig. 1
figure 1

Geographic distribution of the Frostius pernambucensis highlighting the areas in which the tadpoles analyzed in this study were collected (cross and star). A Inset map: South America, highlighting the distribution of the Atlantic Forest (green) and the magnified area showed in A (gray rectangle). B Known localities of the species (diamonds) and its type locality (star). C Panoramic view of Parque Estadual Dois Irmãos, municipality of Recife, state of Pernambuco. D Panoramic view of Reserva Particular do Patrimônio Natural Eco Fazenda Morim, municipality of São José da Coroa Grande, state of Pernambuco. E Panoramic view of Estação Ecológica de Murici, municipality of Murici, state of Alagoas. List of locations and references: 1 = Paraíba: João Pessoa (Pimenta and Caramaschi 2007); 2 = Pernambuco: Paulista (Barbosa et al. 2017); 3 = Pernambuco: Recife (Bokermann 1962); 4 = Pernambuco: São José da Coroa Grande (present study); 5 = Alagoas: Ibateguara (Pimenta and Caramaschi 2007); 6 = Alagoas: Murici (Pimenta and Caramaschi 2007); 7 = Alagoas: Maceió (Dubeux et al. 2020b); 8 = Bahia: Varzedo (Juncá and Borges 2002); 9 = Bahia: São Sebastião do Passé (Freitas et al. 2019)

Full size image

The analyzed specimens are deposited in the herpetological collection of the Museu de História Natural of the Universidade Federal de Alagoas (MHNUFAL). Additionally, we also analyzed the original series of larval description of the species (EI 7253; Cruz and Peixoto 1982).

External morphology

Measurements were made in 17 tadpoles in stages 37 to 41 (Gosner 1960). Five of these tadpoles (stage 38) were used for qualitative description (MHNUFAL 16192, 16196). Measurements followed Altig and McDiarmid (1999): body length (BL), maximum tail height (MTH), tail length (TaL), tail muscle height (TMH), tail muscle width (TMW), and total length (TL); Lavilla and Scrocchi (1986): body width at eye level (BWE), body width at nostril level (BWN), extranarial distance (EnD), extraorbital distance (EoD), eye diameter (ED), intranarial distance (InD), intraorbital distance (IoD), maximum body height (MBH), maximum body width (MBW), narial diameter (ND), oral disc width (ODW), snout-spiracle distance (SSD), and spiracle-posterior body distance (SPD); and Grosjean (2005): dorsal fin height (DFH) and ventral fin height (VFH).

All measurements were taken using an ocular micrometer installed on a Leica® MZ6 stereomicroscope, except for TL, which was measured with digital calipers (0.1 mm accuracy). Terminology follows Altig and McDiarmid (1999). For color descriptions, we used the terminology proposed by Köhler (2012) with their corresponding color codes.

Buccopharyngeal cavity

To study buccopharyngeal anatomy, we dissected two tadpoles in Gosner stages 30 and 31 according to Wassersug (1976) and, after inspection using a stereoscopic microscope, we submitted them to the protocol of Alcalde and Blotto (2006) for scanning electron microscopy (SEM). Terminology of buccopharyngeal features follows Wassersug (1976, 1980).

Musculo-skeletal system

To study musculo-skeletal system, six tadpoles in Gosner stages 33 and 41 were processed according to the clearing and double staining protocol of Dingerkus and Uhler (1977); for two individuals, the procedure was interrupted after the alcian blue step, and specimens were manually dissected for inspection of cranial muscles. We employed lugol solution to aid in the identification of muscles. The protocol was carried out until the end for the remaining individuals aiming to study the larval cranium. Terminology for the musculo-skeletal elements is that of Haas (1995, 2001, 2003).

Phylogenetic relationships and character optimization

Besides Frostius pernambucensis, we examined tadpoles of other representatives of several genera of Neotropical bufonids (see complete list of examined material in the Appendix 1) to investigate character evolution within the family; our sampling was complemented with information available in the literature. We based our taxon sampling in the phylogenetic hypothesis of Jetz and Pyron (2018) which has a dense taxonomic sampling, although F. pernambucensis is not included in it. Besides Cannatella`s (1986) hypothesis, F. pernambucensis were never included in another phylogenetic hypothesis and the monophyly of the genus have never been tested. Nevertheless, based on adult morphology and behavior, we assumed the monophyly of the genus and the sister relationship of F. pernambucensis and F. erythrophthalmus—diagnostic characters, such as the iris coloration have been demonstrated polymorphic within F. erythrophthalmus, and all call parameters evaluated so far are overlapping between the two species (Juncá et al. 2012) rendering both species phenotypically very similar—for comparisons and optimizations purposes. We do not claim that we have an exhaustive sampling of bufonids, for that would be beyond of the scope of this paper; we recognize the extensive variation of bufonids, and we claim for further studies with deeper taxon and character sampling. Our analysis represents an initial step toward a better understanding of the evolution of bufonid larvae. Finally, the lack of data for several basal lineages, such as Oreophrynella, prevent an unambiguous optimization of some characters.

We delimited 25 transformation series (Hennig 1960; Grant and Kluge 2004) to describe the variation observed (Appendix 2), focusing on character-states of Frostius pernambucensis. The character matrix was constructed in Mesquite V. 3.51 (Maddison & Maddison 2018). These characters were optimized using parsimony as a criterion (Fitch 1971) into the topology of Jetz and Pyron (2018) using T.N.T.v. 1.5 (Goloboff & Catalano 2016). This tree was pruned to reflect the relationships among taxa for which characters were scored, focusing on the generic-level relationships; notwithstanding, we kept taxa for which no data are available to highlight the parts of the bufonids tree of life that require more studies on larval morphology. Most of the characters in our matrix are invariable within each genus, and most of the unique states are restricted to Frostius pernambucensis. We complemented our observations with information available in the literature. Finally, it is worthy to stress that our analysis is exploratory intended to identify some interesting states and major evolutionary patters in Frostius evolution; future studies including a more detail taxonomic sampling and phylogenetic analysis are required to test our propositions.

Natural history

We provide comments on the natural history and breeding behavior of Frostius pernambucensis based on field observations. Additionally, we briefly comment on the egg clutch based on field and laboratory observations (MHNUFAL 17359; collected at Reserva Particular do Patrimônio Natural Eco Fazenda Morim).

Results

External morphology

Total length 17.21 ± 1.06 mm (15.65–18.33 mm; n = 5; stage 38; Fig. 2). Body ovoid in dorsal and ventral views (MBW/BL = 0.52–0.58), depressed and globular in lateral view (MBH/MBW = 0.76–0.86). Ventral contour of body sloping, slightly concave in peribranchial region and convex in abdominal region. Body length about 1/4 of total length (BL/TL = 0.26–0.32); maximum body width on middle of body and maximum body height on its end portion. Snout oval in dorsal view and truncate in lateral view, slightly sloping. Eyes located dorsally and directed laterally, representing nearly 32% of intraorbital distance (ED/IoD = 0.30–0.37) and 24% of maximum body width (ED/MBW = 0.23–0.27). Nostrils dorsal and directed frontolaterally, with external opening reniform (Fig. 2F), and a small cutaneous projection (apophysis) on its inner margin; each nostril represents about 14% of intranarial distance (ND/InD = 0.14), 4% of maximum body width (ND/MBW = 0.04) and located closer the eyes than the tip of snout. Oral disc equivalent to 36% of maximum body width (ODW/MBW = 0.34–0.38), ventral, arched-shape, lips modified in a thick dermal fold laterally interrupted (Fig. 2D). Marginal papillae present, short and blump; submarginal papillae absent. Labial tooth row formula (LTRF) 1/1; A1 arched, P1 U-shaped, and shorter in length than A1; large and sparse labial teeth, 24–26 labial teeth in A1 and 8–10 in P1. Labial teeth composed of a basal sheath and convex head with poorly developed cusps (Fig. 3). Jaw sheaths with only the terminal portion pigmented; upper jaw sheath arc-shaped and finely serrated, lower jaw sheath open U-shaped and finely serrated, more robust than the upper jaw. Spiracle sinistral (Fig. 2E), located below the body midline, opening on middle third of body (SPD/SSD = 0.63–0.71), directed posterodorsally, forming an angle of approximately 45° with the longitudinal body axis, visible from ventral view. Inner wall free from body and external wall slightly shorter than inner wall. Vent tube with medial opening (Fig. 2G and H), longer than wide, with both walls attached directly to the ventral fin, except the tip, which is free. Tail length approximately 72% of total length (TaL/TL = 67.92–74.09), higher at the middle of tail; maximum height corresponding to half the height of body (MBH/MTH = 0.74–0.85). Tail tip acute. Maximum tail musculature height less than 1/3 of the maximum height of tail (TMH/MTH = 0.33–0.37), becoming progressively thinner posteriorly and extending to tail tip. Myosepta partially visible from half of its length. Dorsal fin height 32% of tail height (DFH/MTH = 0.30–0.37) and slightly higher than ventral fin (VFH/DFH = 0.30–0.33). Dorsal fin beginning at the end of body, with contour parallel to caudal musculature in almost all its length, slightly arched in its second half, maximum higher at the beginning of the last third. Ventral fin beginning at the base of tail, with origin concealed by vent tube and contour parallel to the caudal musculature; maximum height at first third of tail. Measurements are presented in Table 1.

Fig. 2
figure 2

Tadpole of Frostius pernambucensis, stage 36 (MHNUFAL 16196), in A lateral, B dorsal, and C ventral views. D Detail of the oral disc. E Detail of the spiracle. F Detail of the narial aperture. Detail of the vent tube in G lateral (right side) and H ventral views. Details of the eggs of F. pernambucensis (I). Lateral, dorsal, and ventral views of tadpoles of F. pernambucensis in stages (J) 28, (K) 32 and (L) 42 (MHNUFAL 16196). M Tadpole of F. pernambucensis, stage 37 (MHNUFAL 16196), in life. Scale bars = 3 mm (A–C, I–L), and 0.3 mm (D–H)

Full size image
Fig. 3
figure 3

Labial teeth of Frostius pernambucensis (EI 7253) at stage 30. Scale bar = 10 µm

Full size image
Table 1 Measurements (in mm) of Frostius pernambucensis tadpoles
Full size table

Coloration

In life, body Drab-Gray (256), covered by numerous Fawn Color (258) dots (Fig. 2L). Due to its translucency, in dorsal view, it is possible to observe Tawny Olive (17) color of the calf on lateral margins of the second half of body, as well as Vandyke Brown (281) color in inner part of the eyeballs. In ventral view, Fawn Color (258) points restricted to lateroventral surfaces, becoming less evident in the medial portion of belly. Belly Pale Buff (1), and evident calf in its posterior portion. In lateral view, posterior portion of body darker, Fawn Color (258). Translucent fins color Light Sky Blue (191) and caudal musculature opaque, colors similar to fins. Myosepta partially visible. In preservative, color pattern is maintained, background becomes Pale Horn Color (11) and dots distributed on the body become Dark Brownish Olive (127) (Fig. 2A–C).

Variation

Tadpoles show little variation in relation to morphological characteristics analyzed. However, when compared to more initial stages of development, a progressive change in body's coloration in life is observed, ranging from a clear translucent tone in Sulphur Yellow (80) and very evident calf in initial stages to a dark tone in Dark Lavender (203) and calf almost imperceptible in metamorphic stages. The eyes also develop becoming larger as development progresses. It is worth noting that tadpoles at stage 26 already have a total length similar to those in final stages of development (~ 16 mm); all growth is post-metamorphic (Fig. 2I–K; Table 1).

Buccopharyngeal cavity (Fig. 4)

Buccal roof

Buccal roof elliptical (Fig. 4A). Obtuse, medial, short papilla present in the prenarial arena (Fig. 4C). Internal nares elliptical, sagittally oriented in an angle of 60°; posterior valve free (Fig. 4D). Pre and postnarial papillae absent. Median ridge absent. Buccal roof arena absent. Buccal roof devoid of pustulation and papillae. Dorsal velum poorly developed, V-shaped, medially interrupted.

Fig. 4
figure 4

Buccopharyngeal cavity of a tadpole of Frostius pernambucensis (EI 7253) at stage 30. A Buccal floor, B buccal roof, and C details of the prenarial arena papilla, D internal nares, E infralabial papillae, and F lingual papillae. BP buccal pocket, BRP buccal roof papilla, IL infralabial papillae, IN internal nares, LB lingual bud, LJS lower jaw sheath, LP lingual papillae, PNAP prenarial, UJS upper jaw sheath. Scale bars = 100 µc (A, B, E, F), and 20 µm (C, D)

Full size image

Buccal floor

Buccal floor triangular (Fig. 4B). Single pair of short, blunt, infralabial papillae (Fig. 4E). Lingual bud rounded; three short, round, lingual papillae (Fig. 4F). Buccal floor arena absent; single buccal floor papillae present above the buccal pocket line. Prepocket pustulation and papillae present. Buccal pockets deep, perforated, oblique, slit-shaped. Ventral velum present, reduced, inconspicuous; spicular support inconspicuous; secretory pits absent; secretory ridges absent; marginal projections absent; medial notch absent. Glottis fully exposed. Branchial basket reduced, semi-circular; filter cavities reduced; filter plates absent.

Musculature

We observed 22 muscles in the larvae of Frostius pernambucensis (Table 2; Fig. 5). Some muscles typically observed in other tadpoles, such as levator mandibulae lateralis (Haas 2003), were not observed; we stress the lack of this muscle is likely to be an observational problem rather than the loss of the muscle in the species. We suggest that future studies incorporating other techniques, such as histology and immunohistochemistry, could aid in the detection of these small muscles. Nevertheless, it is interesting noting the shift of some muscles, such as the levator arcuum branchialium IV, that inserts on the ceratobranchial III, the subarcualis rectus I that has two slips inserting on the processus branchialis II, and the suspensioangularis originating ventrally in the palatoquadrate.

Table 2 Patterns of origin and insertion of larval muscles of Frostius pernambucensis
Full size table
Fig. 5
figure 5

Larval musculature of Frostius pernambucensis at stage 33 (MHNUFAL 16196). A–B Ventral view and C–D lateral view. GH geniohyoideus, IH interhyoideus, IM intermandibularis, LMLP levator mandibulae longus profundus, OH orbitohyoideus, RA rectus abdominis, RC rectus cervicis, SA suspensorioangularis, SH suspensoriohyoideus, SAR I subarcualis recuts I, SAR II–IV subarcualis rectus II–IV, SO subarcualis obliquus, TP tympanopharyngeus. Scale bars = 1.0 mm

Full size image

Larval cranium

Neurocranium triangular; greatest width at otic capsule level (Fig. 6). Suprarostral cartilage (Fig. 6C) formed by two elements, suprarostral alae and corpus completely fused, although the line of fusion seems to be recognizable; dorsal posterior process of the suprarostral ala present, suprarostral corpora fused distally. Suprarostral corpus fused to the cornua trabeculae. Ethmoidal region short; cornua trabeculae short, parallel to each other and fused terminally with the suprarostral corpus. Basicranial fenestra open. Frontoparietal fenestra large, rectangular. Orbital cartilage (Fig. 6d) poorly developed, not reaching the otic capsule (foramen prooticum open). Capsula auditiva robust, large, rhomboidal in dorsal view, representing about 40% of chondrocranium length. Palatoquadrate thin in lateral view (Fig. 6C), attached to neurocranium through a wide anterior commissura quadratocranialis and an almost perpendicular processus ascendens posteriorly; processus ascendens with a high (sensu Haas 2003) attachment. Processus muscularis quadrati triangular, poorly developed, and curved dorsomedially. Commissura quadratoorbitalis absent. In the lower jaw (Fig. 6E), Meckel’s cartilage sigmoid, transversely oriented, almost perpendicular to the chondrocranium longitudinal axis. Infrarostral cartilages rectangular in frontal view, curved, joined at the symphysis.

Fig. 6
figure 6

Larval cranial skeleton of Frostius pernambucensis at stage 41 (MHNUFAL 16196). A Dorsal view; B ventral view; C detail of the suprarostral cartilage; D lateral view; E detail of the lower jaw; F hyobranchial apparatus. CA capsula auditiva, CB ceratobranchial, CI cartilago infrarostralis, CM cartilago Meckeli, CT cornua trabeculae, FO foramen opticum, FOC foramen oculomotorium, FPO foramen prooticum, PAH processus anterior hyalis, PAQ pars articularis quadrati, PAS processus ascendens, PHB planum hypobranchiale, PM processus muscularis, PRU pars reuniens, SA suprarostral ala, SC suprarostral corpus. Scale bars = 0.5 mm

Full size image

Ceratohyals (Fig. 6F) short, flat, and subtriangular; anterior margin with poorly developed anterior and anterolateral processes, posterior processes triangular. Ceratohyals confluently joined by a chondrified pars reuniens. Basibranchial rectangular, with rounded processus urobranchialis present; basihyalis absent. Hypobranchial plates long, rectangular, free from each other. Branchial basket with four curved ceratobranchials, lacking lateral projections; ceratobranchials I, II, and III attached to the hypobranchial plate; ceratobranchial IV reduced, free. Ceratobranchial I with a triangular anterior processus branchialis. Spicules absent.

Natural history notes

Spawns and tadpoles of Frostius pernambucensis were observed both in hollow trees (n = 7) and in bromeliads (n = 2) containing accumulated rainwater. All reproductive sites were inside the forest, at least 200 m from the nearest edge, varying between 60 and 150 cm in height (Fig. 7). Males of F. pernambucensis were observed close to or inside the reproductive sites in almost all visits, many times in vocalization activity (during the night or late afternoon and on rainy days) and showed parental care behaviors during the collection of larvae.

Fig. 7
figure 7

Breeding sites of Frostius pernambucensis in Estação Ecológica de Murici, municipality of Murici, Alagoas state. A–C Different spawning developing in the same tree hollow with water accumulated at about 150 cm in height. D–E Spawn developing in the central cup of an epiphytic bromeliad with water accumulated at about 100 cm in height (see a video in Supplementary Material 1). F–G Spawn developing in tree hollow with water accumulated at about 60 cm in height

Full size image

All clutches found had a large number of eggs and/or tadpoles (~ 20 to 50 individuals), even in reproductive sites with reduced space, such as in bromeliads (Fig. 7D and E; see a video of the density of tadpoles in the same bromeliad in Supplementary Material 1). The clutches (Fig. 2I) consist of gelatinous string of yellowed eggs; the entire string measured 4.4 cm on average (4.3–4.5 ± 0.08; n = 5), accommodating eggs of 2.0 mm (1.8–2.3 ± 0.13; n = 10).

Some clutches had tadpoles in early (newly hatched) and advanced (above stage 38) stages of development, indicating that the same reproductive site was used simultaneously more than once. The reuse of the same reproductive site consecutive times during the reproductive period was observed, with different spawns observed in June, July, and August 2021. During a visit to a tree hollow whose tadpoles were in an advanced stage of development, the behavior of cannibalism was observed, where some individuals feeding on two dead tadpoles, which already had most of the tail devoured.

Discussion

Larval morphology and the systematics of bufonids

More than 65% of known genera of bufonids have information on larval morphology, showing high phenotypic and ecological diversity (e.g., Duellman and Lynch 1969; Lamotte and Xavier 1972; Müller et al. 2005; Haas et al. 2009; Meegaskumbura et al. 2015; Viertel and Channing 2017; Müller 2019; Romero-Carvajal et al. 2023), with many larval characters presenting interesting evolutionary history and phylogenetic significance (Channing 1978; Haas 2003; Frost et al. 2006; Hirschfeld et al. 2012). Very few larval external features of Frostius pernambucensis support the notion that the species is part of the family Bufonidae, such as the eyes placed dorsally, the medial vent tube, and the sinistral spiracle (located in general below the medial line). On the other hand, many features are unique to F. pernambucensis tadpoles, probably related to its phytotelmata habits and endotrophic development (discussed below), suggesting putative synapomorphies—although the lack of information for its putative sister species, F. erythrophthalmus, and some other lineages of bufonids, precludes unambiguous optimizations.

The majority of bufonids tadpoles are characterized by a tail of moderated length that ends in a rounded tip. Moreover, the oral disc is in general emarginated, with marginal papillae absents dorsally and ventrally (putative synapomorphy of the family; Haas 2003) and a standard LTRF 2/3 (e.g., McDiarmid and Altig 1990; Inchaustegui et al. 2014; Vera Candioti et al. 2020). The long tail (representing ¾ of the total length) and the acute tail tip are two features that seems singular to F. pernambucensis, standing out from most other bufonids. However, the simplified aspect of the oral disc is the one that deserve the most attention. The marginal papillae are reduced, the lips are modified in a thick dermal fold interrupted laterally, the labial teeth are reduced to two rows (one anterior and one posterior) with few and sparce teeth, and jaw sheaths are almost inconspicuous with only the terminal portion pigmented. Among bufonids, endotrophic larvae are rare, also being reported for Blythophryne, Pelophryne, and Altiphrynoides (Grandison 1978; Müller 2019). Endotrophy evolved independently in those taxa, and some character-states associated with lack of feeding, such as the reduction of mouthparts, are convergent in those lineages.

The oral discs of these taxa present unique character-states; Blythophryne beryet has marginal papillae (with dorsal gap) and a well-developed jaw sheath, but no labial teeth (LTRF 0/0) (Chandramouli et al. 2016); while Pelophryne spp. have the oral disc with no observable marginal papillae and a LTRF 1/0 (Inger 1960; Malkmus et al. 2002; Leong and Teo 2009). The LTRF 1/1 and the presence of a modified lip in a thick dermal fold are autapomorphic conditions of F. pernambucensis and could represent a synapomorphy of Frostius, pending data on F. erythrophthalmus.

Data on buccopharyngeal cavity are available for several lineages of bufonids: Anaxyrus (Wassersug and Rosenberg 1979); Altiphrynoides (Müller 2019); Ansonia (Inger 1985); Bufo (Viertel 1982); Ingerophrynus (Inger 1985); Melanophryniscus (Baldo et al. 2014); Osornophryne (Romero-Carvajal et al. 2023); Poyntonophrynus (Lambris 1994); Rentapia (Inger 1985); Rhinella (e.g., Vera Candioti 2007; Oliveira et al. 2013); Schismaderma (Lambris 1994; Viertel and Channing 2017); Sclerophrys (Lambris 1994); and Vandijkophrynus (Lambris 1994). Although some features may be missing in some taxa, the almost complete absence of elements in the buccal roof and buccal floor is unique to the tadpoles of F. penambucensis. However, the lack of information for several lineages of bufonids (e.g., Oreophrynella) prevents an unambiguous optimization of these characters. We address, though, some interesting patterns that must be elucidated in the future.

Frostius pernambucensis has a single pair of infralabial papillae, the same as other bufonids for which the buccopharyngeal cavity is known (e.g., Oliveira et al. 2013; Müller 2019; Vera Candioti et al. 2020; this study), including the most basal genus of the family, Melanophryniscus (Baldo et al. 2014), contrasting with the four papillae (two pairs) observed in Bufonidae`s sister taxa Odontophrynidae (e.g., Nascimento et al. 2013; Dias et al. 2014, 2019). Other closely related lineages, such as centrolenids or leptodactylids (mostly), also present two pairs of infralabial papilla (e.g., Wassersug and Heyer 1988; Vera Candioti et al. 2007; Rada et al. 2019; Dias et al. 2020; Nascimento et al. 2020). Thus, we suggest that the presence of a single pair of infralabial papillae is a putative synapomorphy of Bufonidae.

We observed three, rounded, poorly developed lingual papillae in F. pernambucensis. Most bufonids present four lingual papillae with few exceptions: lingual papillae are absent in members of the Rinella veraguensis species group (Cadle and Altig 1991; Aguayo et al. 2009), and in Ansonia tadpoles (Inger 1985), R. rumboli may have 3 or 4 lingual papillae (Haad et al. 2014), and Ingerophrynus divergens present a single pair (Inger 1985). Three lingual papillae seem to be a unique feature of Frostius, although the occurrence of polymorphic condition in other bufonids, such as R. rumboli, cast a shadow in that interpretation; further specimens of Frostius need to be examined regarding their buccopharyngeal cavity to determine the occurrence or not of polymorphism in this character. It is worthy to note that, despite the number of papillae, the reduction in its size has never been reported in any other bufonid as far as we know.

The reduction in the number of buccal floor and roof papillae and pustulation is another characteristic of Frostius pernambucensis—single pair of buccal floor arena papillae and no pustulation at all—in comparison with other bufonids. Buccal roof papillae are absent in the larvae of Ansonia (Inger 1985), but all other bufonids with their buccopharyngeal cavity described present some papillae and pustulations on the buccal roof and floor. Also, postnarial papillae are missing in Frostius. We thus suggest that the absence of pustulations on the roof and floor, as well as the absence of papillae in the buccal roof and postnarial papillae could represent synapomorphies for the genus Frostius. Other elements are missing the buccopharyngeal cavity of Frostius tadpoles: median ridge, lateral ridge papillae, secretory ridges and secretory pores, and filter plates, that should be better explored in other bufonid genera, but could represent putative synapomorphies for the genus.

The musculo-skeletal system of Frostius pernambucensis is also highly modified, with interesting evolutionary implications. Haas (2003) suggested some larval synapomorphies derived from the muscles and cartilages for the family Bufonidae: (1) the absence of m. interhyoideus posterior; (2) the absence of the diaphragmatopraecordialis; (3) a slip of the subarcualis rectus II–IV reaches far laterally into interbranchial septum IV; and 4) the absence of processus anterolateralis. Frostius pernambucensis is in accordance with three of these synapomorphies, but differs regarding the subarcualis rectus II–IV, which originates in the ceratobranchial III and reaches the ceratobranchial I, both conditions are putative synapomorphies for the genus.

Many bufonids have the m. subarcualis rectus I divided into two slips inserting on the ceratobranchials II and III (e.g., Haas 2003; Vera Candioti 2007). In Frostius pernambucensis, the two slips are also present, but both are inserted in the ceratobranchial III; the changing in the pattern of insertion of the dorsal slip is a putative synapomorphy of Frostius. Also, the insertion of the levator arcuum branchialium IV and that of the tympanopharyngeus are modified in F. pernambucensis; in this species, all the mentioned muscles are inserting on the ceratobranchial III, in contrast with all other bufonids, in which these muscles insert on the ceratobranchial IV.

Regarding the larval cranium, Frostius pernambucensis also has several interesting transformations. Suprarostral corpus and alae completely fused it is not commonly observed among bufonids, but it was reported for the gastromyzophorous larvae of Atelopus (Lavilla and de Sá 2001). Amazophrynella, Dendrophryniscus, and Melanophryniscus do not present such fusion (Baldo et al. 2014; P.H.D. pers.obs.), but the lack of data for other basal lineages creates a difficulty in optimizing this character state. Notwithstanding, the fusion on the suprarostral with the cornua trabeculae had never been reported in bufonids—Gastrophryne carolinensis larvae have these cartilages fused (Trueb et al. 2011), but this is not closely related species—which represents an additional putative synapomorphy of Frostius.

The commissura quadratoorbitalis is present in the majority of bufonids (Haas 2003), but gives its presence in other closely related taxa, such as Odontophrynidae (Dias et al. 2019); it could represent a synapomorphy of a more inclusive clade. Such commissura was lost in F. pernambucensis, representing another putative synapomorphy for the genus.

Finally, the ceratobranchials II and III attached to the hypobranchial plates, the absence of commissura terminalis uniting the ceratobranchial, and the free ceratobranchial IV are also putative synapomorphies of the genus, given that these conditions are not observed in other members of the family (e.g., Haas 2003; Aguayo et al. 2009; Haad et al. 2014).

Living in phytotelmata and the evolution of endotrophy

Phytotelmata, such as bromeliads tanks, bamboo stumps, tree holes, nut shells, and palm leaves, can be defined as water filled plants or parts of plants (Thienemann 1934). This habitat can be drastically different from other water bodies in its physicochemical properties, presenting low dissolved oxygen, low pH, and elevated viscosity (Laessle 1961; Maguire. 1971; Richardson 1999). Another characteristic of phytotelmata is that the food chain and the primary productivity are very particular (Richardson et al. 2000; Kitching 2001) with reduced availability of food resources, and thus, organisms inhabiting those habitats may face distinguished selective pressures. Frogs have colonized phytotelmata several times (Lehtinen et al. 2004) and tadpoles developing in that environment have often evolved differentiated trophic strategies; although many phytotelm dwellers are detritivorous and filterers (e.g., Caldwell 1993; Dias and Pie 2021), several species are predators (carnivory), oophagous, or endotrophic (Lannoo et al. 1987; Lehtinen et al. 2004). Within bufonids, phytotelm usage for development has been reported in several taxa (e.g., Caldwell 1993; Langone et al. 2008; Malkmus and Dehling 2008), and the tadpoles of some species inhabiting these plants are endotrophic (e.g., Leong and Teo 2009; Chandramouli et al. 2016).

Most of unique character-states in Frostius pernambucensis are related to the reduction or loss of elements in oral disc, buccopharyngeal cavity, musculature, and larval cranium. Such pattern can be directly related to the endotrophic development of these tadpoles. Endotrophic tadpoles usually lack mouthparts, pigmentation, tail fins, and spiracle (e.g., Kaiser and Altig 1994; Caldwell and Lima 2003; Randrianiaina et al. 2011; Etter et al. 2021), although a variety of combination of traits losses can be observed in different lineages (Altig and Johnston 1989).

The buccopharyngeal cavity of several endotrophic larvae follow similar pattern, with the reduction or loss of features. Wassersug and Duellman (1984) investigated the phenotypic variation in buccopharyngeal cavity of several hemiphractid tadpoles. Their findings regarding the endotrophic forms are similar to our results with Frostius pernambucensis, including the loss or reduction of several papillae and other buccal features.

Similarly, the musculo-skeletal system is quite modified in endotrophic forms. We reported some differences regarding the pattern of origin and insertion of some muscles (e.g., levator arcuum branchialium), although the general pattern of origin and insertion of muscles in quite similar to that of exotrophic tadpoles. This is also true for other endotrophic tadpoles for which larval muscles are known—Eupsophus emiliopugini (Vera Candioti et al. 2011a, b) and Fritziana goeldi (Haas 1996).

Larval cranium morphology data are available for few endotrophic tadpoles: Cycloramphus stejnegeri (Lavilla 1991), Eupsophus calcaratus (Vera Candioti et al. 2005), E. emiliopugini (Vera Candioti et al. 2011a, b), E. nahuelbutensis (Nuñez and Úbeda 2009), E. queulensis (Cárdenas-Rojas et al. 2007), Rhinoderma darwinii (Lavilla 1987), and Fritziana goeldi (Haas 1996). Some character-states present in most of these larvae, as well as in Frostius pernambucensis, are the short processus muscularis quadrati, large capsula auditiva, short cornua trabecula, tall orbital cartilage, and reduced hyobranchial apparatus. With exception of the enlargement of capsula auditiva, these character-states can be directly correlated with the lack of feeding activity in these larvae.

Two no exclusive mechanisms have been evoked to explain the evolution of endotrophy: (1) the loss of some specific tadpole characters and (2) the acceleration in the development of some adult characters (Elinson 2001). Empirical evidence seems to support these ideas; data suggest that endotrophy is associated with the remodeling and/or loss of cartilaginous elements typical of tadpoles, as the suprarostral cartilages and palatoquadrate, and with changes in the onset of some skeletal elements (Hanken et al. 1992; Yeh 2002; Kerney et al. 2007). The fusion between the suprarostral and cornua trabeculae provides additional evidence to the idea of truncation of development. The suprarostral cartilage is likely evolved by the addition of an articulation in the existing trabecular cartilages (Sevensson and Haas 2005; Lukas and Olsson 2018) during development; the fact that these cartilages are not completely separated could indicate that tadpole`s development was truncated.

Changes in development is also valid for buccopharyngeal cavity; Wassersug and Duellman (1984) examined the ontogeny of buccopharyngeal cavity in the exotrophic larvae of Gastrotheca riobambae and observed that early stage embryos (Gosner 23) lacked most of buccal features and resembled the cavity of endotrophic forms and concluded (Wassersug and Duellman 1984:35) that endotrophic forms may have arisen by a simple truncation or acceleration of the “tadpole developmental program”. As far as we know, there is no information on the buccopharyngeal cavity of bufonids in early developmental stages (24-), but the external morphology of F. pernambucensis resembles that of the embryos of other bufonids, with poorly developed oral discs and mouthparts and poorly pigmented bodies (Vera Candioti et al. 2016).

The real processes involved in the evolution of endotrophy are still poorly understood. Altig and Crother (2006) hypothesized that endotrophic forms evolved from developmental reppartening via changes in some gene family responsible for the development of tadpole-specific traits (see also Altig 2006). Although there are no molecular and developmental studies addressing this issue, the fact that the loss of exotrophic tadpoles and the evolution of endotrophic forms occurred independently and repeatedly in some lineages (besides bufonids), such as aromobatids (e.g., Vacher et al. 2017) and hemiphractids (e.g., Castroviejo Fischer et al. 2015), suggests that some regulatory program can be altered relatively easily. Further studies are necessary to test these hypotheses.

Finally, endotrophy also had an impact in the ecological interactions and space use in Frostius pernambucenis larvae. We observed several individuals in the same phytotelm (up to 50 tadpoles at different developmental stages). Usually, phytotelm breeders tend to lay few eggs per site; for example, in dart-poison frogs (Dendrobatidae) the number of eggs in the clutches and tadpoles deposited per site tend to be small (Caldwell and Araújo 1998; Lehtinen et al 2004; Duarte-Marin et al. 2020). Large number of tadpoles inhabiting the same phytotelm seems to be correlated with oophagy (Schiesari et al. 2003), and we suggest that the large number of F. pernambucenis tadpoles found in a single site could be related to endotrophy, giving that feeding resources would not be a limiting factor for special usage. Further studies are necessary to test this hypothesis.

Remarks and conclusions

The endotrophic tadpoles of Frostius pernambucenis are highly modified in comparison with exotrophic larvae of other bufonids, leading us to propose several novel synapomorphies for the genus. Notwithstanding, data are scarce for many lineages of bufonids (especially for early divergent clades), and a more comprehensive study will be necessary to test our hypotheses.

The colonization of phytotelma, an environment relatively poor of food resources, could have created a selective pression on the development of F. pernambucenis, favoring the evolution of endotrophy, although further evidence is necessary to test this hypothesis. The evolution of non-feeding larvae is associated with the loss and/or remodeling of several elements of the anatomy of these tadpoles. This fact suggests that F. pernambucensis might be an excellent model for future evo-devo studies.