Morphology at tailbud stage.

At tailbud stage (GS17–18), leiuperine embryos vary widely in size, shape, and pigmentation (Figs 1 and 15A–15D). Body length is highly variable, with some species being twice as long as others (S2 Table). Embryos of Pleurodema have an average body length of 2.06 mm (1.65–2.70 mm). Transformation to a larger body is recovered in the clade joining the Pl. thaul and Pl. bibroni groups (mean length = 2.47 mm), which have the largest embryos for leiuperines. The clade Physalaemus + Engystomops has the smallest embryos, with mean body sizes of 1.55 mm (1.36–1.90 mm). Embryos of Pseudopaludicola significantly differ; Ps. falcipes tailbud embryos are only 1.06 mm long whereas embryos of Ps. mystacalis are two times longer, reaching 2.10 mm. Embryos of the confamilial genus Leptodactylus show a comparable range of body length, with lower values in species of the L. latrans group and larger ones in the L. fuscus group [37].

Leiuperinae embryos at tailbud stage differ also in dorsal curvature and yolk shape. Engystomops and Physalaemus embryos are typically curved lying upon a subspherical yolk, while Pleurodema and Pseudopaludicola embryos are straighter and have an oblong yolk (Fig 1). As discussed in a previous work [52], descriptions of dorsal curvature in terms of kyphosis and lordosis are not detailed enough to describe interspecific variation. Furthermore, yolk shape and area that help to characterize embryo shape should be discussed separately from body curvature. In order to obtain a more precise description of the yolk shape and area, we recorded the body region where the curvature occurs, and measured the angle of body curvature and the relative yolk area, independently reconstructing ancestral states for those two characters (Fig 15B and 15C; S1 Table). The ancestral states for Leiuperinae are slightly kyphotic embryos with dorsal curvatures of 112–120° and yolk areas of about 40% of the body area. These are similar to the plesiomorphic states of Leptodactylidae. A slight curvature is maintained in Pseudopaludicola and Pleurodema—in this last case mainly in the caudal region with average angles of 119°—with straighter embryos in Ps. falcipes and Pl. cordobae. Yolk area is slightly larger in the Pl. thaul clade. In Engystomops + Physalaemus the highly pronounced kyphosis along the whole body (64–71°) and the rounded yolk render embryos almost spherical. The yolk area increases significantly only in the derived clade of Ph. cicada + Ph. santafecinus. In contrast, Leptodactylus has large, highly curved, and yolked (≥ 50%) embryos in the L. fuscus group, whereas embryos in the L. latrans group almost lack dorsal curvature and their yolk area represents about 40% of their body area [37].

Regarding coloration, pigmented embryos are the plesiomorphic state of Leiuperinae, with two independent transformations to whitish embryos in Pl. brachyops [32] and the clade joining Engystomops + Physalaemus (Fig 15D). Pigmentation in Pseudopaludicola and Pleurodema embryos range from light brown in Pl. borellii to dark brown in Ps. mystacalis and the Pl. thaul clade. In Engystomops and most Physalaemus embryos are initially whitish and scattered dots appear anterodorsally once the second gill pair develops. Two reversals are observed in the pigmented embryos of Ph. cicada and Ph. riograndensis. The lack of pigmentation has been reported for numerous anuran embryos, and within Leptodactylidae it was described in species of the Leptodactylus fuscus group [37] and suspected from egg features in Edalorhina (the sister genus of Engystomops), species of the Ph. signifer clade, and Adenomera (e.g., [53–55]).

The relationship among size, shape, and pigmentation is well documented in several groups, and highly kyphotic embryos are often also large, yolky, and unpigmented (e.g., [37, 56]). Developmental bases for the correlation between yolk amount and pigmentation extent have been suggested [56], but the relationships between the type of cleavage, yolk uptake, pigment production and mobility, and the vast diversity of reproductive modes in anurans is still under discussion [57]. Some other diverging patterns of association between embryo size, shape, and pigmentation have been described. Kyphotic embryos can also be small, like those of most species of Physalaemus, Hyperolius puncticulatus, Atelopus, and several Melanophryniscus [52, 58], and pigmented, as in Ph. riograndensis, some Melanophryniscus, and species of Batrachyla and Odontophrynus [52, 59–60]. Regarding yolk amount, unlike several kyphotic embryos with relative yolk areas larger than 50%, embryos of Engystomops, Physalaemus, Melanophryniscus, and O. americanus have small yolk provision (30–50%).

Body ciliation.

Results in this section emphasize some areas of the embryo body (the ventral aspect, gills, cephalic region); distribution of ciliated cells in the body dorsum and tail is undersampled. Patterns of ciliation in Leiuperinae vary in distribution, cell shape, and persistence (Fig 2). The ciliated cells, rounded or polygonal in most species, are more densely arranged in the cephalic region, especially on those areas surrounding the nares, the oral disc, and the adhesive glands. Body ciliation changes during ontogeny, in general with maximum density at full gill development, and a progressive regression post GS25. Embryos of the Pleurodema thaul clade and of the Physalaemus henselii group have sparse ciliated cells that regress earlier, with gills still exposed (Fig 2B). At the other extreme, embryos of Pl. borellii and the Pl. nebulosum clade have large, densely arranged ciliated cells that regress late after hind limb emergence (Fig 2C). Likewise, body ciliation in Pseudopaludicola falcipes regresses once hind limbs reach GS27 (Fig 2D). A late regression of body ciliation was also described for E. pustulosus, and in the leptodactyline Leptodactylus fuscus [28].

External gills.

The external gills in leiuperines develop between GS18 and GS24, and dispose laterally to the body in a configuration similar to most anuran embryos. Main variations occur in the number and length of pairs and filaments (Figs 3–5; S1 Table). The ambiguity in the reconstructed pair number for the subfamily ancestor (Fig 15E) could be solved with further information on embryos of Edalorhina and species of the Physalaemus signifer clade. Two gill pairs are recovered as a homoplastic feature for Engystomops (see also [23, 61]), Pseudopaludicola (Fig 3), and the Ph. henselii group (Fig 4A). All these species have poorly developed gills, with only 3–6 filaments, primary filaments among the shortest in the sample (0.24–0.30 mm; S1 Table), and ciliation relatively scarce. A comparative analysis of developmental sequences (Fig 14, S2 Table) shows that gill ontogeny was also shorter in these species, i.e., maximum gill development and gill regression occurs earlier than in most leiuperines.

Embryos of Pleurodema and the clade Physalaemus cuvieri + Ph. biligonigerus (Fig 15E) develop three gill pairs (see also [16, 34]. Embryos of Physalaemus (except in the Ph. henselii group) and the clades Pl. brachyops and Pl. nebulosum show the most complex gills at full development. In these taxa, the first pair branches into 6–12 filaments, primary filaments are long (0.55–1.04 mm), third pair always grows beyond the operculum margin (Figs 4B, 4C, 5A and 5B; S1 Table), and gill ciliation is profuse. In contrast, gills in the clade formed by the Pl. thaul + Pl. bibroni groups are scarcely developed and the shortest in Pleurodema (Fig 5C and 5D).

The literature reports variations in leiuperine gill structure. For instance, embryos of Engystomops pustulosus were described with two poorly developed gill pairs [23] or three long and ciliated gill pairs [27]. Besides an explanation related to cryptic diversity in this widely distributed species (as suggested by advertisement call and allozyme evidence [62]), intraspecific variations in gill structure should be explored in this and other leiuperines. In our study, intraspecific variation is low, but three cases are noteworthy. Embryos of a single clutch of Pl. borellii presented four gill pairs instead of three, and we found maximum length of gills widely different in embryos of Physalaemus santafecinus and Ph. riograndensis from different clutches. Wider sampling and experimental works are desirable in order to further characterize these polymorphisms.

Adhesive glands.

The optimization of adhesive gland characters recovers the morphogenetic type C as the plesiomorphic state for Leiuperinae (Fig 15F), but with an ambiguity at the node that relates it to Leptodactylinae. Leptodactylus embryos exhibit two different configurations: species of the L. fuscus group lack adhesive glands, whereas those of the L. latrans and L. melanonotus groups have type D glands defined as a medial heart-shaped structure [37]. Data on the L. pentadactylus group, and Adenomera, Lithodytes, Hydrolaetare and the Paratelmatobiinae are not available.

Type C glands are distinguished by an early division into two individual conical structures, and are known in several neobatrachian clades; e.g., some hylids [25], phyllomedusids [63], basal bufonids [52], batrachylids, and odontophrynids [60, 64]. We found remarkable variations in shape, size, and arrangement in the leiuperines (Figs 6 and 7; S1 Table). Glands of Physalaemus and Pseudopaludicola are usually conical, tall, and the diameter at the base is less than 3% of the body length, except for Ph. albifrons and Ph. fernandezae whose glands are slightly larger (3.3 and 3.7% respectively; Fig 6A–6D; S1 Table). At early stages glands are located lateroventral to the oral disc, later they become lateral, and regress almost at the level of the oral disc commissures. Conversely, in Pleurodema embryos the adhesive glands are rounded and low, with the largest diameters observed in our samples (about 5% of body length; Fig 6E and 6F). In early embryos they are ventral and very close to each other, and regress without changing their position relative to the oral disc. Gland regression in this genus occurs later than in the former two, i.e., when the oral disc and the spiracle are already developed and the hind limbs reach GS27 (Fig 14). Adhesive glands of Pl. thaul (Fig 7) are unusual in that the initial morphogenetic field has a V-shape configuration similar to that of the groove of other gland types (e.g., [37, 52]), and only acquire the typical arrangement for Pleurodema at later stages. This autapomorphy is certainly helpful for embryo recognition.

Hatching gland.

The hatching gland is placed in the cephalic region and along a long dorsal line in most leiuperine species (Figs 8 and 9). The cephalic portion has a T-shaped configuration, progressing into two circular and lateral regions with dense, secretory cells with long microvilli, similarly to what has been described for Leptodactylus [37]. The hatching gland is conspicuously visible in embryos of Pleurodema, Pseudopaludicola, and in the pigmented embryos of Physalaemus (Ph. riograndensis and Ph. cicada). Conversely, in the remaining species of Physalaemus it is only visible with scanning electron microscopy. Although our sampling of this structure was not exhaustive (some stages were not imaged), we recorded a remarkable diversity in gland arrangements, dorsal extensions, shapes of secretory cells, and density and length of microvilli (Fig 8). In Pseudopaludicola, fully developed glands are represented by secretory cells with very short microvilli, disposed in a rostral transverse patch without a dorsal extension (Fig 8A and 8B). In Physalaemus, the hatching gland is arranged in a discontinuous patch of secretory cells with small microvilli (Fig 8C–8E). Rounded areas are lateral in the cephalic region and show a high density of secretory cells, and the dorsal line reaches midbody at advanced stages. In Ph. fernandezae, secretory cells are very small and mostly scattered, and only present close to the nares (Fig 9E). In Pleurodema, secretory cells are arranged in a large cephalic patch and in a dorsal line of variable length; some interspecific variations also involve the density of cells and the length of microvilli (Fig 8F and 8G). Pleurodema borellii presents the longest microvilli among leiuperines (Fig 8F), and in Pl. bibroni the dorsal line only reaches the anterior portion of the cephalic region. In the Pl. thaul clade the cells are sparsely distributed and not clustered in large groups. They have minute microvilli, and the dorsal line does not extend pass the gills level (Fig 9).

Hatching glands reach full development shortly before hatching, when cells become denser and microvilli length generally increases. As indicated by pigmentation, gland regression occurs in a postero-anterior direction, thus the frontal area is the last region to lose hatching cells. Pleurodema cordobae and species of the Pl. nebulosum group exhibit the latest regression of leiuperines, approximately at GS26–27.

Oral disc.

Descriptions of the larval oral discs are available for the majority of leiuperine species, but data on their ontogenetic trajectories are scarce ([35]). The plesiomorphic and most generalized labial tooth formula includes two upper and three lower labial ridges. While the morphology of the upper lip is highly conserved, the number of ridges of the lower lip varies, and a reduced formula with two lower ridges evolved in the Physalaemus biligonigerus group, Pleurodema tucumanum and the Pl. nebulosum clade, and in several species of Pseudopaludicola including Ps. mystacalis in which it is polymorphic (Fig 16A). Marginal papillae of the lower lip are also variable, and may appear complete or interrupted by ventral and/or ventrolateral gaps.

In a research that included 12 leiuperine species, Vera Candioti et al. [35] identified varied developmental pathways related to the diversity of oral configurations within the group. According to their results, Physalaemus and Pseudopaludicola are characterized by the occurrence (transient or maintained in larval stages) of ventrolateral gaps, whereas in Pleurodema marginal papillae develop only from commissures and thus ventrolateral gaps never define. Additionally, the five types of oral disc known for Physalaemus result from common ontogenetic trajectories ending at different states, or from trajectories that differ initially in the formation of a ventral gap (see also Fig 10). Labial ridges are formed in slightly different ways among genera (Fig 14). In Physalaemus and Pseudopaludicola, the first ridge to differentiate is row P1 as an indented margin, and then row P2 forms as a transverse division along the ridge. Conversely, in Pleurodema row P2 develops first, and soon afterwards row P1 forms in a proximal direction. This is followed later by tooth emergence, which occurs first in older ridges.

Among the species we analyzed, the oral discs of Physalaemus aff. albonotatus and Ph. albifrons develop identically to those of other species of the Ph. cuvieri group ([35]; Fig 10). Conversely, oral ontogeny of Ph. carrizorum and Ph. cicada differs from the known trajectories. In Ph. carrizorum, the larval oral configuration shows three lower labial ridges and continuous lower papillae as described in other species of the Ph. gracilis group. However, the oral ontogenetic trajectory is similar to that of the Ph. henselii group, since it involves the initial presence of a ventral gap that, along with ventrolateral gaps, later complete with papillae (Fig 10A). Additionally, the P3 ridge develops before the mental papillae are outlined. Event sequence analysis showed that the larval oral disc acquires its definitive configuration earlier than in other species of the genus (Fig 14). In turn, Ph. cicada exhibits two larval oral configurations that develop following two different trajectories. A configuration with a ventral gap was previously described [65], and in half of our samples we observe comparable transient morphologies (e.g., Fig 10B); thus in these embryos we interpret an ontogenetic trajectory similar to that of the Ph. henselii group. In the other half, specimens show a different trajectory that ends in complete marginal papillae; in this case from the configuration with row P3 and both ventral and ventrolateral gaps, mental region completes first (Fig 10C). A transient oral morphology including the P3 row and only ventrolateral gaps was not reported for Physalaemus, but interestingly, it corresponds to the larval configuration of the recently described Pseudopaludicola jaredi [66].

In Pseudopaludicola, the oral disc ontogeny of Ps. mystacalis differs from what was described for Ps. falcipes (i.e., a trajectory almost identical to that of species of the Physalaemus cuvieri group; [35]). In most specimens, only two lower tooth ridges form and ventrolateral gaps persist in larval stages (Fig 11A–11D) [19]. A single specimen showed labial teeth on a mental papilla/ridge defining an incipient P3 (Fig 11E), a polymorphism also reported in Ps. falcipes. Additionally, a significant number of specimens (30%) have a wide ventral gap resembling bufonid tadpoles (Fig 11F). This oral configuration has been previously described for Ps. boliviana and Ps. pusilla [67–68].

The oral disc ontogeny in Pleurodema species was synthesized by Vera Candioti et al. [35], and it is simpler than those of Physalaemus and Pseudopaludicola. Data on five additional species in our study allow to discuss further variations among main clades. Embryos of Pl. nebulosum develop only two lower tooth ridges in a sequence similar to that of its sister species Pl. guayapae. In all other Pleurodema, three lower labial ridges appear. In Pl. cordobae, as in Pl. bibroni of the same intrageneric group, row P3 is very short and coexists briefly with a short ventral gap (Fig 12). As already stated [35], the marginal papillae in the genus develop from commissures and progresses in a medial direction, being the mental region the latest to complete. Alignment of marginal papillae is also variable. In Pl. nebulosum and Pl. guayapae, papillae are disposed in a single row, whereas in most other species they first appear widely spaced, and small gaps between them fill progressively resulting in alternate papillae (Fig 13). Finally, marginal papillae in the Pl. bibroni clade have the longest development in the genus, with an early differentiation of the first papillae and delayed acquisition of the larval configuration.

The ancestral state reconstruction of ventral and ventrolateral gaps allows for an interpretation of oral disc evolution that slightly differ from previous discussions by Vera Candioti et al. [35] and Lourenco et al. [46]. First, our sampling of Physalaemus does not include species of the Ph. signifer clade; consequently what we recover as synapomorphies of the genus may be so at the level of Ph. cuvieri clade, as already suggested [46]. Second, intraspecific variations (e.g., in Ph. cicada and Pseudopaludicola species) turn the reconstruction ambiguous at corresponding nodes. Third, developmental trajectories of newly sampled species (e.g., Ph. carrizorum) reveal unknown apomorphic features. Having said this, in our study we found that from a plesiomorphic absence, transient ventrolateral gaps evolve in the ancestor of Physalaemus (Fig 16B). These gaps persist in Ph. riograndensis and (as already stated in [46]) as a synapomorphy of the Ph. cuvieri group. In Pseudopaludicola, the sampling is still incomplete and includes polymorphic species, which makes the scenario uncertain. Nevertheless, given that ventrolateral gaps are never present in Ps. boliviana, Ps. pusilla, and some specimens of Ps. mystacalis, this feature does not characterize the genus as previously suggested [35]. A transient ventral gap occurs in most species of Leptodactylidae because of the way marginal papillae develop (only from commissures; Fig 16C). A larval ventral gap appears in the ancestor of Physalaemus, but it is soon lost at the most derived clade joining the Ph. gracilis and Ph. biligonigerus groups. A further change within that clade occurs in Ph. carrizorum embryos that show a reversal to the plesiomorphic, transient state. Knowledge of the ontogenetic trajectories for species belonging to the Ph. signifer clade, and the sister genera Edalorhina and Engystomops are crucial to resolve the ambiguities in the internal nodes of Leiuperinae, and establish the level at which these characters are apomorphic.

Hatching.

We found substantial interspecific variation in the hatching time of leiuperine embryos (Fig 14; S1 Table). Pseudopaludicola experiences the earliest hatch, which occurs when gill arches are still differentiating (GS17–18). In Pleurodema and Physalaemus, most embryos hatch with long gills and operculum at their bases (GS21–22), but Pl. nebulosum, Pl. diplolister, species of the Pl. bibroni clade, Ph. fernandezae, Ph. riograndensis and Ph. carrizorum hatch relatively early. We observed some intraspecific variation as well in Pl. borellii and five species of Physalaemus. Such variation is likely underestimated since in nine of the species studied we only examined development from a single clutch.