Abstract

Rhacophorid frogs, with their unique diversity of reproductive modes, are a perfect model group for studying the evolution of different forms of ontogeny. While phylogenetic studies yield controversial results, the morphological basis of the ontogenetic remodeling remains insufficiently explored. In the present research, we compare embryonic and early-larval development in eight species of Rhacophoridae with four different reproductive modes: jelly nests (Feihyla palpebralis, Theloderma vietnamense), bubble nests (Feihyla hansenae), foam nests (Chirixalus nongkhorensis, Polypedates megacephalus, Rhacophorus annamensis, R. vampyrus), and terrestrial direct-developing egg clutches (Kurixalus gryllus). The main focus is on the degree of embryonization in each species, as well as on the development of three transient early-larval features: external gills, opercular fold, and adhesive organs. The initial signs of gill reduction were observed in biphasic jelly-nesting species with a slightly prolonged embryonic period. In direct-developing species, the reduction of larval structures progressed significantly. Development varied among foam-nesting species: while early ontogeny in Chirixalus showed no signs of significant remodeling, both species of Rhacophorus exhibited a strikingly modified developmental pattern, with a severe reduction in external gills and accelerated formation of tadpole morphology within the foam nest, whereas the studied species of Polypedates displayed transitional developmental features. The morphological changes discovered showed that, in the evolution of ontogeny in Rhacophoridae, direct development and advanced foam-nesting represent two fundamentally divergent and highly derived reproductive modes.

Direct development, embryonization, external gills, foam nests, terrestrialization, tadpole

Among all terrestrial vertebrates, amphibians are unique in having a biphasic life cycle with aquatic larval and terrestrial adult forms, which are especially divergent in anurans with their fish-like tadpoles and quadruped adults. At the same time, anurans display an incomparable diversity of reproductive modes and ontogenetic strategies (Duellman and Trueb 1986; Wells 2007; Elinson and del Pino 2012; Gomez-Mestre et al. 2012; Nunes-de-Almeida et al. 2021). In particular, in some anuran lineages, the evolution of ontogeny followed a path of embryonization (sensu Matsuda 1987), i.e., the incorporation of early larval stages into intraoval development; the extreme embryonization gives rise to species with direct development, completely omitting the free larval phase of ontogeny (Thibaudeau and Altig 1999; Callery et al. 2001; Elinson 2001). This phenomenon is usually considered in relation to the progressive terrestrialization of amphibian reproduction, observed in diverse anuran families (Crump 2015).

This prolongation of the embryonic period at the expense of the larval phase is usually accompanied by some heterochronic shifts in early ontogeny, as well as the loss of certain morphological features (Callery et al. 2001; Naumann et al. 2021). It makes sense that structures, which function temporarily during the early stages of larval life, lose their adaptive value during intraoval development and are reduced. Currently, a significant amount of data has accumulated on how ontogeny is remodeled in direct-developing species in various anuran lineages (Hanken et al. 1992; Bahir et al. 2005; Kerney et al. 2007; Anstis et al. 2011; Goldberg et al. 2020). However, since most studies of the evolution of ontogeny associated with embryonization focus on species with most derived and, therefore, highly modified developmental patterns (which, naturally, are the most intriguing to study), the question arises: when do the rearrangements begin? Are they characteristic only of anurans with some advanced level of embryonization, or can they be found in species making the very first steps on this evolutionary pathway?

Studies aimed at discovering differences in early ontogeny among anurans with biphasic development but different reproductive modes are rather scarce. Nevertheless, they show that, e.g., in the Bufonidae and Leptodactylidae, heterochronies and variability in the formation of early larval transient structures can be found in species with relatively small differences in the duration of the embryonic phase, and that specific conditions of larval development, such as foam nests or micro-waterbodies (soil depressions, phytotelmata, etc.), are also significant (Vera Candioti et al. 2016; Grosso et al. 2017, 2019).

Old World tree frogs (Rhacophoridae) can also serve as a good model for the comparative study of early ontogeny. This extensive family includes both biphasic (i.e., having a free-living larval stage) and direct-developing species, and the biphasic species exhibit an exceptional diversity of reproductive modes, from Buergeria with aquatic egg clutches (Wells 2007; Kuroshima and Tominaga 2021) to various types of terrestrial clutches in other species. In rhacophorids with terrestrial reproduction, eggs may be included in jelly capsules attached to plants (e.g., in some Feihyla and some Gracixalus – Rowley et al. 2014; Pham et al. 2019; Biju et al. 2020; Vassilieva 2023), foam (e.g., in Chirixalus, Chiromantis, Polypedates, Rhacophorus, Taruga – Seymour and Loveridge 1994; Grosjean et al. 2008; Meegaskumbura et al. 2010) or bubble nests (Feihyla species of the F. vittata group, moved by some authors to a separate genus Rohanixalus – Biju et al. 2020). Many species place their eggs in phytotelmata like tree hollows and bamboo stems (e.g., Theloderma and some Gracixalus species – Orlov et al. 2012; Rowley et al. 2014, 2015, 2020; Poyarkov et al. 2015), where their exotrophic tadpoles pursue their development. Other phytotelm-breeding rhacophorids, such as Philautus nepenthophilus breeding in Nepenthes pitchers, have endotrophic tadpoles completing their metamorphosis at the expense of yolk reserves (Etter et al. 2021) or are obligate oophages relying on maternal provisioning of trophic eggs (e.g., Rhacophorus vampyrus, Kurixalus eiffingeri, Nasutixalus spp. – Kam et al. 1996; Rowley et al. 2012; Vassilieva et al. 2013; Biju et al. 2016). Finally, many species (Raorchestes, Philautus, Pseudophilautus), which lay clusters of large eggs in the moist soil, forest litter or moss, are direct-developers displaying a much more advanced level of embryonization: in them, the intraoval development absorbs not only the larval stage, but also the metamorphic period, their progeny develop in the egg using yolk reserves and hatch as young froglets (Patil and Kanamadi 1997; Biju 2003; Bahir et al. 2005; Biju et al. 2010; Hertwig et al. 2012; Orlov et al. 2012; Yan et al. 2021; Huang et al. 2026).

Despite ongoing intensive phylogenetic research on the Rhacophoridae, taxonomic uncertainties remain numerous, so that the composition and validity of some genera and the relationships among them remain the subject of discussion even in the most recent phylogenetic studies (Wilkinson et al. 2002; Grosjean et al. 2008; Li et al. 2009; Hertwig et al. 2011; Orlov et al. 2012; Meegaskumbura et al. 2015; Chan et al. 2018; Biju et al. 2020; Chen et al. 2020; Boruah et al. 2025). This is one of the main reasons why the understanding of the evolution of reproductive modes within this group still requires the attention of researchers. Morphological studies of the ontogeny of Rhacophoridae are significantly behind; although the development of some species has been studied, the structure of early larval stages has not always been examined in sufficient detail, especially in biphasic species, so the available data are rather fragmentary (Dring 1987; Bahir et al. 2005; Chakravarty et al. 2011; Kishimoto and Hayashi 2017; Abraham et al. 2018; Etter et al. 2021; Kuroshima and Tominaga 2021; Yan et al. 2021).

We studied the embryonic and early larval period of ontogeny in eight Southeast Asian species of rhacophorids with different reproductive modes (jelly, bubble and foam nests, and terrestrial clutches) and varying degrees of embryonization (short and extended embryonic periods, direct development), with the aim of revealing potential differences in the development of transient morphological structures that are typically reduced in species with advanced embryonization. Among them are the external gills, which exist during the early larval stages and then are replaced by internal gills; the opercular fold covering the gill area and forming the opercular cavity; and the adhesive organs, or cement glands, i.e., a paired secreting organ on the head allowing the early larvae to attach themselves to substrates (Pennati et al. 2000; Brunelli et al. 2004).

The results are expected to help identify the morphological changes associated with the origin of various reproductive modes in rhacophorids and contribute to our understanding of general patterns of the evolution of ontogeny in anuran amphibians.

Recent phylogenetic studies consistently support that aquatic breeding is plesiomorphic for Rhacophoridae, occurring in basal taxa, and that the more advanced terrestrial jelly-nesting gave rise to all the diversity of reproductive modes in this group, which gradually evolved toward decreasing dependence on breeding in standing water. Thus, foam-nesting, which apparently arose once in rhacophorid evolution (Wilkinson et al. 2002; Grosjean et al. 2008; Li et al. 2009; Chen et al. 2020), and terrestrial direct development, which evolved independently twice (Li et al. 2009; Meegaskumbura et al. 2015; Chen et al. 2020), are derived reproductive modes. Bubble-nesting occurs only in a few species of small rhacophorids that are now allocated to the genus Feihyla; however, some authors found their reproductive mode involving bubbled clutch capsules, green-colored eggs, and parental attendance, to be one of the arguments to classify them in a separate genus (Meegaskumbura et al. 2015; Biju et al. 2020). In their description of the genus Rohanixalus, Biju et al. (2020) also assumed that bubble-nesting may represent an intermediate stage between jelly- and foam-nesting. Bubble-nesting is often not considered a distinct reproductive mode, but based on our observations of Feihyla hansenae, we believe it deserves to be recognized as such. In terms of key reproductive parameters (egg size, clutch size, and timing of development), this species differs significantly from the jelly-nesting F. palpebralis and closely related species, such as F. fuhua (as described in Biju et al. 2020). On the other hand, it is similar in reproductive parameters to Chirixalus nongkhorensis and C. doriae (AV personal observations), but lacks a foam nest, which is undoubtedly an important feature.

The evolution toward terrestrialization in reproduction involves the progressive incorporation of larval stages into the embryonic period, i.e., embryonization. Thus, the complete omitting of the larval stages could represent the extreme point of this evolutionary process, whereas its initial stages may result in the hatching of larvae at just more advanced developmental stages (Smirnov and Vassilieva 2022). Such partial embryonization is widespread among anurans and often accompanies the transition from aquatic to terrestrial breeding, such as the use of phytotelmata, e.g., in Microhylidae (Krügel and Richter 1995; Vassilieva et al. 2017; Vassilieva and Nguyen 2023).

A similar evolutionary pattern is also observed in the jelly-nesting Rhacophoridae. The representative of a basal rhacophorid genus Buergeria, B. japonica, lays egg clutches in open waterbodies, and its larvae hatch at Gosner stage 19 (Kuroshima and Tominaga 2021). This more-or-less corresponds to the hatching stage in many anuran species with aquatic reproduction and without signs of embryonization (Gosner 1960; Bradford 1990). As shown in our study, the jelly-nesting rhacophorids Feihyla palpebralis and Gracixalus supercornutus, which suspend their clutches above open water, hatch at more advanced stages (23 and 25, correspondingly). The phytotelm-breeding jelly-nesters Theloderma spp. hatch at stage 25, and the same was observed for the bamboo-dwelling Kurixalus eiffingeri, as well as for K. naso and Mercurana myristicapalustris, which deposit their jelly clutches in moist soil (Cheng et al. 2013; Kishimoto and Hayashi 2017; Abraham et al. 2018; Shangpliang et al. 2021). Even from these few examples, it is obvious that the progressive terrestrialization of reproduction in the jelly-nesting rhacophorids leads to the early larval stages being gradually absorbed into the embryonic period. Consequently, the development of transient larval features, such as the external gills and opercular fold, gradually shifts to the intraoval development: while in F. palpebralis the external gills start to develop during embryogenesis and are fully functional in hatched larvae, in Theloderma vietnamense the entire developmental cycle of the external gills, from the formation of their primordia to their disappearance beneath the opercular fold, is completed before hatching (Fig. 14).

Figure 14. 

Timing of the development of transient larval features (external gills and opercular fold) in rhacophorids with different reproductive modes: a graphical comparison.

In this case, the fate of the external gills is of particular interest. In hatching larvae that enter an aquatic environment, their functional role seems obvious: like all gills of aquatic animals, they serve for respiration in a liquid medium. But if the existence of external gills in biphasic species is entirely limited to the embryonic period, their adaptive significance becomes questionable.

Studies of the morphological diversity of external gills in anuran embryos demonstrate that these organs can be highly diverse in size and structural complexity. However, no universal patterns in their development were identified. Some comparative data on frogs, including species with terrestrial or semi-terrestrial clutches, such as Phyllomedusidae, Hylidae, and Leptodactylidae, show that species with delayed hatching tend to have the most highly developed gills (Warkentin 2000; Nokhbatolfoghahai and Downie 2008). Apparently, this is also true for Colostethus spp. (Dendrobatidae), which exhibit varying degrees of embryonization (Caldwell and Lima 2003; del Pino et al. 2004). However, other research demonstrates that in Cycloramphidae hatching at Gosner stages 25–26 and using similar breeding sites (rocks above fast-flowing streams), the length of the external gills differs greatly (Colaço and da Silva 2022). At the same time, larvae of Boophis spp. (Mantellidae), also hatching with already fused opercular folds, apparently totally omit external gills in their development (Blommers-Schlösser 1979), as do Leiopelma spp. (Leiopelmatidae) hatching at the stage of rather advanced tadpoles (Stephenson 1951; Bell and Wassersug 2003).

It appears that the morphology of the external gills is strongly influenced by the conditions affecting embryonic development (Seymour and Bradford 1995; Warkentin 2000, 2002; Nokhbatolfoghahai and Downie 2008). In a hypoxic environment (with limited oxygen access in the microhabitat, high temperatures, thickened egg capsules, or high clutch density), the gills may be hyper-developed; in such cases, they often adhere to the egg membrane to facilitate gas exchange. Conversely, in the absence of oxygen deficiency, e.g., in species with air-exposed terrestrial clutches, the gills may reduce.

Presumably, the evolutionary restructuring of these organs may involve the same heterochronic mechanism that leads to the loss of larval traits in direct-developing species: a change in the timing of gene expression and organogenesis. This results in a shortening of the lifespan of these structures and their morphological reduction, as they have lost their original functional significance (Callery et al. 2001).

The external morphology and developmental timing of the external gills in T. vietnamense show that even a relatively slight prolongation of embryonic period is sufficient for adaptive early-larval traits, whose functional significance has diminished, to display signs of reduction. Compared to F. palpebralis, T. vietnamense develops only two pairs of external gills (instead of three); these gills are noticeably shorter and less branched, and their developmental cycle from initial buds (stage 19) to complete disappearance (stage 25) lasts just over two days (about 54 hours) versus more than eight days (about 195 hours) in F. palpebralis or more than four days (about 100 hours) in the bubble-nesting F. hansenae, which develops under the same temperature conditions as T. vietnamense. The time required for the opercular cavity to form (stages 23–25) also decreases: in T. vietnamense, the opercular fold takes only 24 hours to completely close off the gill area, whereas in F. palpebralis this takes 144 hours, and in F. hansenae, 63 hours (see Table 3). Although, given the presence of blood circulation, the external gills in T. vietnamense retain the ability to perform gas exchange, their short existence, small size, and simplified morphology suggest that their functional significance is hardly substantial. Regarding Gracixalus, Rowley et al. (2015) confirmed the presence of “long, branched gills” in the embryos of G. quangi before their hatching at stage 25, but we lack details of their morphology to compare them with the embryonic external gills of Theloderma and reveal some potential signs of reduction. In K. eiffingeri and K. naso, the morphology of the external gills is not described in detail, but from illustrations, they appear to be significantly reduced, although they still retain blood circulation (Kishimoto and Hayashi 2017; Shangpliang et al. 2021).

As might be expected, progressive embryonization in direct-developing rhacophorids leads to a more pronounced reduction in early-larval traits that have lost their functional value. As our data show, in K. gryllus, the existence of external gills is even more ephemeral than in Theloderma; they appear as a single pair of small, papilla-like rudiments with the faintest hint of ramification and disappear quickly, playing no obvious functional role. The opercular fold, however, develops fully; it covers the external gill rudiments and the forelimbs develop inside the opercular cavity, as in typical tadpoles. Other larval traits, such as oral disc, keratinous beak and keratodonts, do not appear at any stage. Apparently, the tail may play a significant role in facilitating cutaneous respiration during the late embryogenesis of K. gryllus; it possesses broad vascularized fins that persist until the prehatching stages. A similar adaptation is observed in other direct-developing species, those in the rhacophorid genus Pseudophilautus (Bahir et al. 2005) and E. coqui (Townsend and Stewart 1985); the extreme development of tail fins as effective respiratory organs occurs in some arthroleptids (Schweiger et al. 2017) and craugastorids (Nokhbatolfoghahai et al. 2010; Goldberg et al. 2012; Goldberg and Vera Candioti 2015; Salica et al. 2023). The advanced embryos of Leiopelma lack developed tail fins, but the skin of their muscular tails is vascularized and presumably participates in the respiratory function (Stephenson 1951).

To understand the extent of ontogenetic remodeling in K. gryllus associated with the evolution of direct development, it is necessary to compare it with other direct-developing species. Apart from K. gryllus, direct development among the Rhacophoridae is known only for the genera Philautus, Pseudophilautus, and Raorchestes. The systematic placement of K. gryllus is debatable; different authors have mostly attributed it to these three genera: Philautus (Smith 1924; Bossuyt and Dubois 2001), Pseudophilautus (Li et al. 2009), Raorchestes (Orlov et al. 2012; Meegaskumbura et al. 2015; Biju et al. 2020; Chen et al. 2020). Yet recently, this species was moved to Kurixalus (Poyarkov et al. 2021; Du et al. 2024), making it the only known species in this genus to date with direct-development. This move was suggested by Poyarkov et al. (2021) based on unpublished molecular data of specimens from the type locality and unpublished morphological data from type material according to Du et al. (2024). This data has yet to be published. Therefore, the placement of this species in Kurixalus should be considered premature as it cannot be verified. Given that the composition of the genera and the phylogenetic relationships among them are still subjects of intense research, the understanding of how direct development has evolved across the various clades of the family requires further elaboration. Furthermore, descriptions of the development of different species are often not detailed enough to allow for direct, qualitative, and quantitative morphological comparisons at different stages of development. At this point, it is more accurate to conclude that the degree of ontogenetic remodeling in different direct-developing rhacophorids may vary in regard to the extent of reduction in certain morphological structures or the heterochronic shifts in their appearance. For example, in the Sri Lankan direct-developers Pseudophilautus viridis and P. silus (previously referred to as Philautus), the early ontogeny is generally very similar to that of K. gryllus: embryogenesis closely coincides in stages, the embryos develop a tail with tall fins, the forelimbs develop under the opercular fold, and there are rudiments of the adhesive organs. However, unlike K. gryllus, they do not develop external gills even in a rudimentary form. On the other hand, data on skeletogenesis in rhacophorids with different developmental modes indicate that, compared to biphasic rhacophorids (Chirixalus, Polypedates, and Rhacophorus), changes in the ossification sequence (in particular, the accelerated formation of jaw and suspensorium elements) are more pronounced in K. gryllus than in P. silus (Kerney et al. 2007; Vassilieva 2017). Since direct development appears to have evolved independently in the Rhacophoridae of East/Southeast Asia and the Indian subcontinent (Chen et al. 2020), it is reasonable to expect that the morphological reorganization associated with progressive embryonization may differ somewhat between species belonging to these different lineages.

Very similar morphological features related to the reduction of larval structures as a result of embryonization are also observed in direct-developing anurans from other families, although the degree of this reduction may vary. For example, among Eleutherodactylidae, Eleutherodactylus coqui retains the temporal presence of the external gills, which are small, but display signs of initial ramification (forming 2–4 “terminal convolutions”) and functional potentials (blood circulation is visible in these gill rudiments); the opercular fold also develops in this species, but is incomplete, covering only the bases of the forelimb primordia (Townsend and Stewart 1985; Callery and Elinson 2000). In the direct-developing Adelophryne even the ephemeral development of the external gills and opercular fold is completely eliminated from the ontogeny (de Lima et al. 2016). In Platymantis vitiana (Ceratobatrachidae), neither external gills nor an opercular fold are observed at any stage of embryonic development (Narayan et al. 2011). In the brachycephalid Ischnocnema henselii buds of external gills are present at early embryonic stages, but the opercular fold is partially reduced (Goldberg et al. 2020). In Oreobates (Craugastoridae), external gills are absent, but the forelimbs may develop concealed under the opercular fold in O. berdemenos or fully exposed, with only small skinfold covering their bases in O. barituensis (Goldberg et al. 2012; Salica et al. 2023). Among Myobatrachidae, in Arenophryne, Myobatrachus, and Metacrinia, the external gills are absent at all stages, but forelimbs develop beneath the operculum (Anstis et al. 2007; Anstis 2008); this is also the case of Arthroleptis spp. (Arthroleptidae) and Arthroleptella villiersi (Pyxicephalidae; Schweiger et al. 2017, 2026). Among direct-developing Australopapuan microhylids, external gills may be present in a rudimentary form, but the opercular fold is always absent and the forelimbs develop externally (Anstis et al. 2011). In contrast, in the neotropic microhylid Myersiella the opercular fold is fully developed (Izecksohn et al. 1971). Thus, most non-rhacophorid direct-developing anurans display a dramatic evolutionary transformation of early ontogeny with full or partial loss of transitory larval features. Additionally, the presence of the egg tooth (a keratinized protuberance on the upper jaw helping to disrupt egg membranes at hatching) in the direct-developing Brachycephaloidea represents an evolutionary novelty, also indicating the derived state of ontogenetic remodeling in some direct-developing taxa. Compared to them, the Rhacophoridae are not apparently among the most advanced direct-developers, since they retain rudimentary external gills and adhesive organs, as well as the fully developed opercular fold. Moreover, the remodeling of the skeletogenesis in E. coqui, compared to the basal ossification sequence for anurans, goes further than in the direct-developing rhacophorids studied in this respect (Kerney et al. 2007; Vassilieva 2017). Additionally, P. silus was found to retain the transient larval cartilages (suprarostrals and trabecular horns), whereas these skeletal elements fail to appear in E. coqui, although short-term expression of the genes responsible for their formation was detected (Hanken et al. 1992; Kerney et al. 2007, 2010).

Another evolutionary pathway in the reproduction of rhacophorids led to the origin of terrestrial foam nests. This mode of reproduction evolved independently in several anuran families and is considered a key innovation that enabled anurans to colonize new ecological niches (Faivovich et al. 2012; Meegaskumbura et al. 2015; Pereira et al. 2017). Among the adaptive benefits of foam nesting, studied mostly in leptodactylids, the most commonly mentioned are the protection of embryos from desiccation, insolation, and predation, as well as microbial defense, temperature acclimation, and post-hatching refuging (Gorzula 1977; Dobkin and Gettinger 1985; Méndez-Narváez et al. 2015; Gould 2021).

Our results reveal that, despite superficial similarities in this reproductive mode in foam-nesting rhacophorids, remarkable differences exist in the early ontogeny between Polypedates and Rhacophorus, on the one hand, and Chirixalus, on the other.

In C. nongkhorensis, early ontogeny differs little from the typical development of anurans with a biphasic life cycle, although it exhibits early signs of embryonization: hatching occurs at stage 23, when many early-larval features (pigmentation, oral disc, coiled gut) are already partially formed. The external gills, which begin to form during embryogenesis, reach their maximum development soon after hatching, function during the first few days of the larva’s life, and then degenerate, disappearing under the opercular fold. This development is very similar to that of the bubble-nesting Feihyla hansenae, except that the latter hatches slightly earlier, at stage 22 (Fig. 14). It is noteworthy that, with eggs and clutches of similar sizes, development in C. nongkhorensis proceeds slightly faster than in F. hansenae, and progresses more uniformly, without significant variation in stages, which argues in favor of the beneficial effect of the foam nest on the developing embryos. The hatched larvae of both species do not linger in the foam or in the bubble capsule and fall directly into the water immediately after hatching.

In contrast, the larvae of Polypedates megacephalus and Rhacophorus spp. show no signs of embryonization and hatch as early as stage 19, as does, e.g., the aquatic-breeding Buergeria. However, unlike C. nongkhorensis, their larvae remain in the foam nest and spend one and a half to two days there. During this time, they undergo significant changes and emerge into the external environment at a much more advanced stage. At the same time, in P. megacephalus, the external gills do not reach the same level of development as in C. nongkhorensis: they are significantly smaller and have a shorter lifespan, less than four days (about 81 hours) versus nearly five days (about 117 hours) in C. nongkhorensis. Nevertheless, they are functional and serve the larva during the first day of its free aquatic life.

Unlike P. megacephalus, in which gill development continues after emergence into the water medium, the larvae of Rhacophorus annamensis leave the foam nest already as fully formed tadpoles, having completed the development cycle of the external gills and opercular fold (Fig. 14). At the same time, their external gills are extremely reduced to tiny rudiments and are non-functional, while the opercular fold completely covers them in less than two days (about 45 hours). In the phytotelm-breeding R. vampyrus, we did not find even rudimentary external gills; hence, in this species, their reduction goes even further, possibly due to the extensive morphological remodeling related to the obligatory larval oophagy in this species (Rowley et al. 2012; Vassilieva et al. 2013; Vera Candioti et al. 2021). Thus, unexpectedly, the reduction of the external gills in some biphasic rhacophorids with foam nests turns out to be more profound than even in certain direct-developing species, including K. gryllus.

The observed contrast in the development of the external gills among different species of foam-nesters is striking. When comparing P. megacephalus and Rhacophorus spp. with C. nongkhorensis, the key difference between them is likely the prolonged larval stay in the foam. The function of the foam nest in C. nongkhorensis appears to be purely ecological; it allows the clutch to be placed out of the water and provides favorable conditions for embryonic development, but has no noticeable effect on the larvae’s life after hatching. In the African rhacophorid Chiromantis xerampelina, a representative of a sister genus to Asian Chirixalus (according to Wilkinson et al. 2002 and Chen et al. 2020), the reproductive mode is rather similar; mating frogs construct a spherical nest consisting of dense foam, and the larvae hatch at stage 22, but then spend about two days inside the nest. At emergence from the foam to water, their stage is the same, but larvae are some more advanced in their growth, pigmentation, and external gill development (Seymour and Loveridge 1994). Thus, they derive certain adaptive benefits from the delay in the nest before entering the aquatic environment, but their ontogenetic pattern remains fundamentally unchanged compared to that of C. nongkhorensis. On the other hand, in Rhacophorus spp., the hatched larvae develop rapidly through several stages (19–25) while protected within the nest, which gives them an obvious advantage when they enter the water as markedly more developed individuals. Possibly, P. megacephalus represents an intermediate stage in this trend, at which the loss of larval stages with developed, functional external gills has not yet progressed very far.

The reduction of external gills in larvae living in a foam nest may be due to various possible causes. On the one hand, it is plausible that the function of the external gills in aerated mucus is not as effective as it is in water, and therefore they may gradually disappear as structures that have lost their adaptive significance. However, this requires further study: in C. xerampelina external gills are well developed and perform a respiratory function while the larvae live in the nest (Seymour and Loveridge 1994); the example of various leptodactylids with foam nests shows that in some species, the external gills are small in size, although they are rather well differentiated and functional (Romero-Carvajal 2009; Grosso et al. 2017, 2019). On the other hand, it can be hypothesized that the ability to accelerate development under the protection of the nest leads to heterochronic shifts in ontogeny, causing some structures to fail to develop fully. In any case, such a drastic reorganization of ontogeny apparently imposes certain limitations; while P. megacephalus larvae are capable of surviving by leaving the nest prematurely at any post-hatching stage, Rhacophorus annamensis larvae often die if they fall into the water before reaching stages 24–25.

The revealed differences in early ontogeny may be of interest from the perspective of genus differentiation. As Grosjean et al. (2008) pointed out, the genera Polypedates and Rhacophorus are close phylogenetically, and generic attribution based on morphology of adult frogs is rather difficult, but larval forms provide more divergent traits for reliable species allocation. It is possible that the differences we have identified in early ontogeny may also prove useful to distinguish the representatives of both genera. Although detailed descriptions of the early ontogeny of members of these genera are few, they are consistent with our findings: Chakravarty et al. (2011) and Tesia et al. (2017) report functional external gills in early larvae in Polypedates spp., whereas Sandeep et al. (2022) describe the development of opercular fold, but do not mention developed external gills in Rhacophorus malabaricus; the former Rhacophorus arboreus with short external gills (Iwasawa and Kawasaki 1979) is now moved to the genus Zhangixalus (Frost 2026). The accumulation of new data on the early ontogeny of foam-nesting rhacophorids will help clarify whether the observed developmental features are genus-specific and whether there are transitional forms between them.

The evolution of reproductive modes in Rhacophoridae is regarded as a transition from basal aquatic breeding to various forms of terrestrial reproduction. The first comparative study of early ontogeny in rhacophorids with biphasic and direct development, as well as with different types of terrestrial clutches (jelly, bubble, and foam nests), contributes to the understanding of the morphological transformations and heterochronies underlying the origin of the derived reproductive modes in this group. Accepting jelly-nesting as a transitional form of reproduction, it is possible to trace the stages of the evolution of rhacophorid ontogeny toward direct development through progressive embryonization. This evolution is accompanied by morphological changes leading to the loss of certain developmental stages and the reduction of transient larval traits. Even a relatively slight prolongation of the embryonic period due to partial embryonization, as seen in Theloderma, already results in some reduction in the size and lifespan of the external gills compared to the jelly-nesting Feihyla palpebralis, hatching at an earlier stage. As expected, this reduction goes further in the direct-developing Kurixalus gryllus, which retains only tiny ephemeral rudiments of the external gills.

Unexpectedly, even stronger reduction of these structures is observed in Rhacophorus spp., in which larvae spend the first posthatching days inside a foam nest. These rhacophorids show no signs of embryonization, but due to the accelerated passage through the early larval stages in the foam, their tadpoles emerge into the external environment already at an advanced stage of development. Members of the closely related genus Polypedates apparently display the intermediate stage of this ontogenetic remodeling. In contrast, in another foam-nester, Chirixalus nongkhorensis, the ontogeny is more similar to jelly-nesting and bubble-nesting species with typical traits of early-larval development. This suggests, first, that in some respects, advanced foam nesting is no less derived than direct development, and second, that developmental heterochronies evolved in different rhacophorid lineages leading to a diversity of ontogenetic trajectories.

All of the fieldwork was conducted with the permission of the Department of Forestry of the Ministry of Agriculture and Rural Development of Vietnam and Forest Protection Departments of the Peoples’ Committees of Gia Lai, Lam Dong and Khanh Hoa provinces (permit numbers 432/TCLN-BTTN, 3369/UBND–NV4, 175/BQL–HTQT, 382/vPUBND-Nv3, 304/BQL-HTQT, 2783/SNN-NVTH). The authors thank the Administration of Cat Tien National Park for the opportunity to perform the long-term research in accordance with Agreement № 37/ HD on the scientific cooperation between Cat Tien National Park and the Joint Vietnam-Russia Tropical Science and Technology Research Center (JVRTSTRC). We thank the administrators of Loc Bao Forest Enterprise for facilitating the research allowed by the Administration of Lam Dong Province. We are deeply indebted to Nguyen Van Thinh, Vu Manh, and Igor Palko for the organization of the fieldwork. We are very grateful to Vitaly Trounov for his extensive help with the field and photographic work and to Fedor Shkil for his kind assistance with microphotography. The study was performed in frame of IPEE RAS scientific program 1022040700412-9-1.6.12 and as part of the research project of the JVRTSTRC “Biodiversity, structural and functional organization, conservation, restoration, and sustainable use of tropical forest ecosystems” (subsection “Research on Amphibians and Reptiles”). The work was supported by the Russian Science Foundation, grant no. 25-24-00301.