True toads of the genus Rhinella are among the most common and diverse group of Neotropical anurans. These toads are widely distributed throughout South America, inhabiting a great diversity of environments and ecoregions. Currently, however, the genus is defined solely on the basis of molecular characters, and it lacks a proper diagnosis. Although some phenetic species groups have traditionally been recognized within Rhinella, the monophyly of some of them have been rejected in previous phylogenetic analyses, and many species remain unassigned to these poorly defined groups. Additionally, the identity and taxonomy of several species are problematic and hinder the specific recognition and description of undescribed taxa. In this work, we first perform phylogenetic analyses of separate mitochondrial and nuclear datasets to test the possible occurrence of hybridization and/or genetic introgression in the genus. The comparative analysis of both datasets revealed unidirectional mitochondrial introgressions of an unknown parental species into R. horribilis (“ghost introgression”) and of R. dorbignyi into R. bernardoi; therefore, the mitochondrial and nuclear datasets of these species were considered separately in subsequent analyses. We performed total-evidence phylogenetic analyses that included revised molecular (four mitochondrial and five nuclear genes) and phenotypic (90 characters) datasets for 83 nominal species of Rhinella, plus several undescribed and problematic species and multiple outgroups. Results demonstrate that Rhinella was nonmonophyletic due to the position of R. ceratophrys, which was recovered as the sister taxon of Rhaebo nasicus with strong support. Among our outgroups, the strongly supported Anaxyrus + Incilius is the sister clade of all other species of Rhinella. Once R. ceratophrys is excluded, the genus Rhinella is monophyletic, well supported, and composed of two major clades. One of these is moderately supported and includes species of the former R. spinulosa Group (including R. gallardoi); the monophyletic R. granulosa, R. crucifer, and R. marina Groups; and a clade composed of the mitochondrial sequences of R. horribilis. The other major clade is strongly supported and composed of all the species from the non-monophyletic R. veraguensis and R. margaritifera Groups, the former R. acrolopha Group, and R. sternosignata. Consistent with these results, we define eight species groups of Rhinella that are mostly diagnosed by phenotypic synapomorphies in addition to a combination of morphological character states. Rhinella sternosignata is the only species that remains unassigned to any group. We also synonymize nine species, treat three former subspecies as full species, and suggest that 15 lineages represent putative undescribed species. Lastly, we discuss the apparently frequent occurrence of hybridization, deep mitochondrial divergence, and “ghost introgression”; the incomplete phenotypic evidence (including putative character systems that could be used for future phylogenetic analyses); and the validity of the known fossil record of Rhinella as a source of calibration points for divergence dating analyses.
INTRODUCTION
General Overview
True toads of the former genus Bufo are a popular group of anurans distributed nearly worldwide, and widely studied by researchers from different disciplines. The classic book “Evolution in the genus Bufo” (Blair, 1972) synthesized knowledge about the morphology, phylogeny, and biology of the group. Despite having integrated evidence from many sources of characters to elucidate the evolutionary relationships among the species groups of true toads, this work largely revealed the difficulties to study their phylogenetic relationships. It was not until the 1990s–2000s that a general picture of these relationships emerged, and the taxonomy of true toads was revised to be consistent with phylogenetic hypotheses (Graybeal, 1997; Pauly et al., 2004; Frost et al., 2006; Pramuk, 2006). Currently, most of the South American true toads of the former genus Bufo are grouped in the large genus Rhinella (Chaparro et al., 2007).
Rhinella includes many of the most conspicuous and ubiquitous species of the anuran fauna in almost all the major biogeographic areas of the Neotropical region (Duellman, 1999; Frost, 2020; IUCN, 2020). With 92 species, Rhinella is the second largest genus of Bufonidae, and its species show considerable morphological and biological diversity, including large variation in size, different levels of cranial ossification, integumentary structure, larval morphology, and ecological and reproductive diversity characteristics (Trueb, 1971; Cei, 1972a; Toledo and Jared, 1993; Pramuk, 2006; Aguayo et al., 2009; van Bocxlaer et al., 2010; Pereyra et al., 2015; Bandeira et al., 2016; Simon et al., 2016; Hudson et al., 2018). Some common species of Rhinella (e.g., R. arenarum, R . horribilis, and R. marina) have been employed extensively as model organisms for various biological disciplines, such as biochemistry (e.g., Abel and Macht, 1912; Cei et al., 1968; Rash et al., 2011), developmental biology (e.g., Markovich and Regeer, 1999; Barisone et al., 2002; Brown et al., 2002), ecotoxicology (e.g., Lajmanovich et al., 2011), molecular biology (e.g., Estoup et al., 2004, 2010; Rollins et al., 2015; Edwards et al., 2018; Ceschin et al., 2020), and especially physiology (e.g., Houssay and Giusti, 1929; Houssay, 1949; Penhos et al., 1967; Sassone et al., 2015). This genus also contains a highly invasive species, R. marina, widely introduced into many countries and islands from different continents (Frost, 2020), where usually it has a highly negative ecological and socioeconomic impact (Jolly et al., 2015; Bacher et al., 2018).
Systematics of Rhinella
For decades, all South American true toads were part of the formerly large and poorly defined genus Bufo, which included a heterogeneous group of toads distributed throughout Africa, America, and Eurasia (e.g., Blair, 1972;; Graybeal, 1997). Frost et al. (2006) partitioned this polyphyletic genus into monophyletic units mostly on the basis of the results of their phylogenetic analysis but also on the results of previous studies (e.g., Graybeal, 1997; Pauly et al., 2004). Frost et al. (2006) resurrected Rhinella for the species of the former Bufo margaritifer Group, which they recovered as distantly related to the other species of South American true toads included in their analysis, including Chaunus and Rhaebo (both also resurrected by Frost et al., 2006)). Frost et al. (2006) noted that Bufo margaritifer was nested within Chaunus in a previous phylogenetic study (Pauly et al., 2004), a finding that was subsequently supported by Pramuk (2006) and Chaparro et al. (2007). Therefore, Rhinella was later redefined to include the species of Chaunus and Rhamphophryne as well (Chaparro et al., 2007).
The species groups of the former Bufo now referred to Rhinella were all recognized primarily on the basis of osteological characters and external morphology that were interpreted without quantitative phylogenetic analyses (Tihen, 1962; Cei, 1972a; R.F. Martin, 1972a, 1972b; Duellman and Schulte, 1992), including the R. crucifer, R . granulosa, R . margaritifera, R . marina, R . spinulosa, and R. veraguensis Groups. Pramuk (2006) studied the phylogenetic relationships of these toads on the basis of a combined analysis of morphological (mostly osteological) and molecular evidence. She rejected the monophyly of some of these species groups (e.g., the R. veraguensis Group is polyphyletic with respect to R. ocellata, the R. margaritifera Group, and Rhamphophryne), but did not modify their composition or diagnosis.
The subsequent increase in the knowledge of relations within Rhinella was limited to the addition of available sequences of some species in extensive phylogenetic analyses of Bufonidae or Anura (e.g., van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Jetz and Pyron, 2018). Figure 1 summarizes the main results of the more inclusive analyses of Rhinella.
For well over a decade, the systematics of Rhinella as a whole has languished, although several efforts focusing on the relationships and taxonomy of parts of the genus have been undertaken. These include phylogenetic analyses of presumptively monophyletic species groups (i.e., the R. crucifer, R . granulosa, and R. marina Groups; Maciel et al., 2006, 2010; Thomé et al., 2010, 2012; Vallinoto et al., 2010; Pereyra et al., 2016a) or fractions of the diversity of certain groups (i.e., the R. festae and R. margaritifera Groups; Fouquet et al., 2007a; Moravec et al., 2014; Santos et al., 2015; Cusi et al., 2017; Avila et al., 2018). Most recent studies on Rhinella aimed primarily to resolve species-level taxonomic problems (e.g., Fouquet et al., 2007a; Narvaes and Rodrigues, 2009; Jansen et al., 2011; Grant and Bolívar-G., 2014; Moravec et al., 2014; Cusi et al., 2017). Consequently, more than a decade after Pramuk's (2006) revision, species groups remain poorly defined, several species cannot be assigned to any of them, and few additional phenotypic synapomorphies have been proposed for Rhinella or its internal clades (Hoogmoed, 1986; 1990; La Marca and Mijares-Urrutia, 1996; Pramuk, 2006; Chaparro et al., 2007; Padial et al., 2009; Blotto et al., 2014; Grant and Bolívar-G., 2014; Pereyra et al., 2016a).
Natural hybridization is common in several groups of Bufonidae, including many species of Rhinella (Blair, 1972;; Feder, 1979; Haddad et al., 1990; Masta et al., 2002; Azevedo et al., 2003; Green and Parent, 2003; Yamazaki et al., 2008; Fontenot et al., 2011; Guerra et al., 2011), and mitochondrial and nuclear introgression have been corroborated in some of these clades (e.g. Green and Parent, 2003; Yamazaki et al., 2008; Fontenot et al., 2011; Dufresnes et al., 2019). Pereyra et al. (2016a) demonstrated the occurrence of hybridization events in the R. granulosa Group and unidirectional mitochondrial introgression of R. dorbignyi into R. bernardoi. A similar situation might exist between R. marina and R. diptycha, although the evidence is not conclusive (Sequeira et al., 2011; Vallinoto et al., 2017). The impact of these phenomena on the inference of phylogenetic relationships (Hennig, 1966; McDade, 1992; Posada and Crandall, 2002) could be mitigated, at least partially, if detected. A detailed evaluation of the discordance between mitochondrial and nuclear genomes together with a critical taxonomic evaluation provide an effective way to detect hybridization/introgression (Pereyra et al., 2016a).
In this paper, we present a densely sampled phylogenetic analysis of Rhinella, including 83 of its 92 species, using molecular (four mitochondrial and five nuclear genes) and phenotypic characters (90 characters from multiple character systems). The goals of this study are to (1) perform a stringent test of the monophyly of Rhinella as well as similar tests on all its species groups, (2) identify phenotypic synapomorphies to diagnose the species groups of Rhinella, and (3) to evaluate the taxonomic status of several taxa.
MATERIAL AND METHODS
Taxonomic Sampling
For the complete dataset (molecular and phenotypic), we sampled 83 described species of Rhinella (including all but nine of the currently recognized species), and 36 exemplar species of other bufonid genera as outgroups (see below). The outgroup species were chosen to provide a severe test of the monophyly of Rhinella, whereas the dense sampling within Rhinella allowed us to rigorously test the monophyly of all its species groups. All specimens scored for phenotypic data were associated with the most morphologically similar and/or geographically closest conspecific terminal of the molecular dataset for the total evidence (TE) analysis.
Collection and locality data of vouchers for sequences used in this study, including the information of the sources of the sequences (this work or previous studies), are detailed in appendix 1, and GenBank accession numbers are listed in appendix 2. A list of the species, specimens, and bibliography analyzed for character scoring of the phenotypic dataset is given in appendix 3, and the collection and locality data of specimens studied for morphology are provided in appendix 4.
Outgroups
For outgroup sampling, we considered the results of the most recent phylogenetic analyses (Frost et al., 2006;; Pramuk, 2006; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Portik and Papenfuss, 2015; Jetz and Pyron, 2018). The previous hypotheses disagree about the phylogenetic placement of Rhinella, recovering it: (1) as closely related to Incilius and Anaxyrus, and deeply nested within an “old world” bufonid clade (Pauly et al., 2004; Frost et al., 2006;; Pramuk, 2006; Chaparro et al., 2007; Pereyra et al., 2016a); (2) as sister taxon of a clade containing all the “old world” bufonid genera (van Bocxlaer et al., 2010); or (3) in a clade together with Anaxyrus + Incilius that is, in turn, sister taxon of the “old world” bufonid clade (Pyron and Wiens, 2011; Pyron, 2014; Portik and Papenfuss, 2015; Ron et al., 2015; Jetz and Pyron, 2018). As all alternative hypotheses have low support values for most relevant nodes around Rhinella, we preferred to sample a broad diversity of bufonid genera representing most of the phylogenetic diversity of the family to rigorously test the relationships and monophyly of the genus. Consequently, we targeted 36 species of 22 bufonid genera as outgroup taxa for the combined molecular dataset and 21 of these species for the phenotypic dataset. Outgroup sequences were obtained exclusively from GenBank (see appendices 1, 2). Thus, in order to increase the number of included genes for outgroup terminals (considering that the number of sampled genes for the ingroup in this work was higher than previous phylogenetic analyses of Bufonidae), we combined sequences from different specimens of the same species to construct several composite outgroup terminals (see justification by Campbell and Lapointe, 2009). These composite terminals (see appendices 1, 2) were constructed only when their uncorrected p-distances (UPDs) in the 16S rRNA gene were less than 0.5%, which is less than the estimated mean divergence observed between sister species of most anurans (Vences et al., 2005a; Fouquet et al., 2007b; Funk et al., 2011). In taxonomy, the exclusive use of pairwise distances and fixed thresholds is questionable (e.g., Will and Rubinoff, 2004; Grant et al., 2006; Meier et al., 2008), but they serve as a useful heuristic for species identification and, in the present context, reduce the risk of constructing composited terminals that could compromise the phylogenetic analysis. Moreover, preliminary analyses including all the sequences of both conspecific specimens recovered them as monophyletic with high support (parsimony jackknife supports >97%, see below).
The Ingroup: Rhinella
We included 278 terminals representing 83 described species of Rhinella for the combined (molecular + phenotypic) dataset. For practical purposes, the included taxa are presented below in the species groups to which they were assigned by Duellman and Schulte (1992), but considering subsequent modifications to this proposal (details of the assignation of each species to species groups by different authors are given in appendix 5).
For the purposes of our analysis, we recognize the following seven species groups within Rhinella: the R. acrolopha Group, the R. crucifer Group, the R. granulosa Group, the R. margaritifera Group, the R. marina Group, the R. spinulosa Group, and the demonstrably paraphyletic “R. veraguensis Group.” Moravec et al. (2014) also proposed the Rhinella festae Group to include three species of the former Rhamphophryne and four species of the paraphyletic R. veraguensis Group (see Pramuk, 2006; Chaparro et al., 2007; van Bocxlaer et al., 2010; Pyron and Wiens, 2011), which they recovered as a clade in their molecular phylogenetic analysis. Although this resolves the nonmonophyly of the analyzed species of the R. veraguensis Group, the authors did not diagnose either their R. festae Group or their restricted R. veraguensis Group or address the placement of the remaining species of the former Rhamphophryne. Given that recognizing the R. festae Group left many species of the former Rhamphophryne and R. veraguensis Group s.l. unassigned to any group due to the lack of diagnoses, we exclude the R. festae Group below.
Grant and Bolívar-G. (2014) proposed the Rhinella acrolopha Group to include the species previously assigned to Rhamphophryne. Although molecular phylogenetic analyses have consistently supported the monophyly of this group (albeit on the basis of a small fraction of its species; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Jetz and Pyron, 2018), its recognition renders the R. veraguensis Group paraphyletic (see Pramuk, 2006; Chaparro et al., 2007; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Jetz and Pyron, 2018). Thus, as discussed by Grant and Bolívar-G. (2014), the composition and diagnosis of the R. festae Group, the R. acrolopha Group, and the R. veraguensis Group are problematic and will be addressed in the context of our results. For the time being, we employ the putatively monophyletic R. acrolopha Group and demonstrably paraphyletic “R. veraguensis Group” only to characterize and provide background on the ingroup.
The Rhinella acrolopha Group
This group consists of 10 small to medium-sized species of Rhinella characterized by a projecting snout, small and inconspicuous parotoid macroglands, heavily ossified skull with some degree of co-osification, well-defined cranial crests (at least in some species), tympanic membrane and annulus absent (except in R. truebae), m. levator mandibulae externus undivided with trigeminal nerve passing medial (deep) to the muscle, m. adductor longus absent, and large and unpigmented eggs (Trueb, 1971;; Lynch and Renjifo, 1990; Grant and Bolívar-G., 2014). These species are distributed from southern Panama to southern Ecuador, and many of them are critically endangered (Rueda-Almonacid et al., 2004).
We sampled the following species: Rhinella acrolopha, R . festae, R . lindae, R . macrorhina, R . nicefori, R . paraguas, R . ruizi, and R. tenrec. We also included an undescribed species from Colombia (Rhinella sp. C sensu Machado et al., 2016). Sequences of R. macrorhina and R. rostrata available from GenBank (A. G. Gluesenkamp, unpublished) were not included because our preliminary analyses (data not shown) revealed that the sequences of the fragments of 12S and 16S rRNA genes of each specimen appear to be chimeric and/or contaminated with R. festae, and we cannot determine with certainty which sequences correspond to each taxon (see also Cusi et al., 2017). Tissues samples of R. rostrata were not available for this study. This poorly known species (Noble, 1920) was described from “Santa Rita Creek,” 23 km N of Mesopotamia town, in the southern part of the departamento de Antioquia, Colombia. There is great uncertainty about this locality, because it has never been possible to locate or document it in the literature a stream with that name near Mesopotomia (today part of the municipality of La Unión, Antioquia). Additionally, we could not obtain samples of R. truebae, a species known only from the holotype and for which the precise locality is unknown (Lynch and Renjifo, 1990; Vélez-Rodríguez, 2004a).
The Rhinella crucifer Group
This putatively monophyletic species group is currently composed of six medium-sized species whose distribution is mainly associated with the Atlantic Forest of Argentina, Brazil, and Paraguay (Duellman and Schulte, 1992; Baldissera et al., 2004; Thomé et al., 2010, 2012; Roberto et al., 2014). The following characters have been proposed to diagnose this species group: skull heavily ossified with slightly elevated cranial crests, dorsal skin smooth with low, scattered tubercles, lateral row of enlarged tubercles present, pale mid vertebral line well-defined, and parotoid macroglands elongated, moderate in size (Duellman and Schulte, 1992; Baldissera et al., 2004; Pramuk, 2006). This species group was recognized as distinct from the Rhinella marina Group by R.F. Martin (1972b) and Duellman and Schulte (1992) and all its forms were considered as a single species (Bufo crucifer) for a long time (see Lutz, 1934; Cochran, 1955; Cei, 1980; Duellman and Schulte, 1992).
Baldissera et al. (2004) revised the taxonomy of this species group and recognized five species based on morphology and morphometrics: Rhinella abei (Baldissera et al., 2004), R. crucifer (Wied, 1821), R. henseli (Lutz, 1934)), R. ornata (Spix, 1824), and R. pombali (Baldissera et al., 2004). Subsequent to the revision of Baldissera et al. (2004), two additional species, Rhinella inopina and R. casconi, were described from wet forests within the Cerrado and Caatinga habitats of Brazil, respectively (Vaz-Silva et al., 2012; Roberto et al., 2014)). Pramuk (2006) only included one species (R. ornata, as Bufo crucifer) of this group in her phylogenetic analysis, and recovered it as the sister taxon of the R. marina Group. Thomé et al. (2010, 2012) corroborated the monophyly of the R. crucifer Group although the outgroup sampling was limited. They also highlighted problems in the taxonomy proposed by Baldissera et al. (2004), as the recognized species did not fully correspond with genetic structuring in the group. Thomé et al. (2010, 2012) found that samples from specimens identified as R. pombali are nested within R. crucifer and/or R. ornata in the mitochondrial phylogenies and are associated with intermediate nuclear genomes in nonphylogenetic analysis (see factorial correspondence analyses [FCA] in Thomé et al., 2012). In addition to these results, a geographic distribution between that of R. crucifer and R. ornata (Baldissera et al., 2004) is congruent with R. pombali as a hybrid complex between the last two species (Thomé et al., 2010, 2012). Furthermore, samples from R. abei were nested within R. ornata. Thomé et al. (2012) proposed to synonymize R. pombali with both parental species and suggested further reassessment of the taxonomic status of R. abei with additional molecular markers. Their results were congruent with 2D geometric morphometrics of the skull performed by Bandeira et al. (2016), who found R. pombali to be morphologically intermediate between R. crucifer and R. ornata, and R. abei nested within R. ornata in the multivariate space.
Several specimens of the six valid species (Rhinella abei, R . casconi, R . crucifer, R . henseli, R . inopina, and R. ornata) were included in our analyses to test the monophyly of this group and the results of Thomé et al. (2010, 2012). We carried out a preliminary analysis (data not shown) including additional nuclear and mitochondrial sequences of two specimens of “R. pombali” and the results supported their findings (see Hybridization and genetic introgression in Rhinella section), so we did not include specimens of “R. pombali” in our subsequent analyses.
The Rhinella granulosa Group
This monophyletic species group is currently composed of 14 medium- to small-sized species of Rhinella (Pramuk, 2006; Pereyra et al., 2016a; Murphy et al., 2017). The following characters have been proposed to diagnose this species group: skull heavily ossified and exostosed with low, granular or elevated cranial crests, dorsal skin with small, keratinous-tipped tubercles, and lateral row of enlarged tubercles absent (Gallardo, 1957, 1965; R.F. Martin, 1972a, 1972b; Cei, 1980; Duellman and Schulte, 1992; Pramuk, 2006). All species of the R. granulosa Group are mostly distributed in open areas of South America and Panama (Gallardo, 1965; Duellman and Schulte, 1992; Duellman, 1999; Narvaes and Rodrigues, 2009; Sanabria et al., 2010).
The taxonomy of this species group was first revised by Gallardo (1965) and more recently by Narvaes and Rodrigues (2009). The latter authors recognized and diagnosed 12 species on the basis of morphological and morphometrical analyses. Subsequently, Sanabria et al. (2010) described a new species (R. bernardoi) from San Juan, western Argentina. The phylogenetic analyses of Pramuk (2006) and Pereyra et al. (2016a), comprising very different samples of species and characters, recovered this species group as monophyletic and discussed several of its phenotypic synapomorphies. Moreover, Pereyra et al. (2016a) documented the occurrence of hybridization between sympatric species as well as past mitochondrial introgression and proposed several morphological synapomorphies for the group. Vera Candioti et al. (2016) proposed some additional synapomorphies from the embryonic morphology (a very short third pair of gills, type A adhesive glands, the adhesive gland subdivision immediately before the gills reach their maximum development, and a short dorsal line of hatching glands mostly restricted to the cephalic region). More recently, Murphy et al. (2017) found the populations of R. humboldti on both sides of the Andes to be phylogenetically distinct, leading them to restrict R. humboldti to the western Andean populations and resurrect R. bebeei for the eastern ones.
In our phylogenetic analyses, we included most species of this group (Rhinella azarai, R . beebei, R . bergi, R . bernardoi, R . centralis, R . dorbignyi, R . fernandezae, R . granulosa, R . humboldti, R . major, R . merianae, R . mirandaribeiroi, and R. pygmaea) with the exception of R. nattereri, a species known from a restricted area in the border between Brazil, Guyana, and Venezuela (Bokermann, 1967; Narvaes and Rodrigues, 2009).
The Rhinella margaritifera Group
The definition of this species group is controversial, as diagnoses have been largely based on morphological variation of the Rhinella margaritifera species complex (e.g., R.F. Martin, 1972b; Hoogmoed, 1986; Pramuk, 2006) or subjective notions of similarity without consideration of character polarity (e.g., Cei, 1972a; Hoogmoed, 1990; Duellman and Schulte, 1992). The following characters have been used to diagnose this species group: skull relatively lightly ossified with variable amounts of dermal ornamentation and prominent cranial crests, dorsal skin smooth or with small, scattered tubercles, and a lateral row of enlarged tubercles present (Hoogmoed, 1990; Duellman and Schulte, 1992; Vélez-Rodríguez, 2004b; Pramuk, 2006). Nevertheless, this definition does not accomodate the morphology of species recently included in the group (R. ocellata and R. yunga, the putative sister species to the remaining species of the group, see Moravec et al., 2014).
Similarly, the taxonomy of the species of the Rhinella margaritifera Group is also conflicted due to imprecise type localities, extreme sexual dimorphism, and the extensive ontogenetic variation that hinder the specific recognition and description of some putative undescribed species (Hoogmoed, 1977; 1986; 1990; Hass et al., 1995; De la Riva et al., 2000; Vélez-Rodríguez, 2004b; Fouquet et al., 2007a, 2007b, 2007c; Lavilla et al., 2013, 2017). Currently, this group is composed of 20 medium-sized species (see appendix 5) distributed from Panama to southern Brazil, including the Amazonia and Guiana Shield.
We sampled 17 species of this group: Rhinella acutirostris, R . alata, R . castaneotica, R . dapsilis, R . gildae, R . hoogmoedi, R . lescurei, R . magnussoni, R . margaritifera, R . martyi, R . ocellata, R . paraguayensis, R . proboscidea, R . scitula, R . sclerocephala, R . stanlaii, and R. yunga . This sampling also includes numerous specimens of the R. margaritifera species complex throughout its distribution. Additionally, we included two undescribed species of this group, one from Ecuador and another one from Colombia, Peru, and Venezuela. Two species of this group were unsampled: R. roqueana, which occurs along the lowlands east of the Andes in southern Ecuador and adjacent northern Peru (Hoogmoed, 1990), and R. sebbeni, which is known only from a few localities of the riparian and dry seasonal forests in the Cerrado biome (Vaz-Silva et al., 2015).
The Rhinella marina Group
This species group is currently composed of 11 large species (Duellman and Schulte, 1992; Maciel et al., 2010; Vallinoto et al., 2010; Lavilla and Brusquetti, 2018). The group is distributed from the southern United States to Argentina, and its species inhabit both open and forested areas (Duellman and Schulte, 1992; Frost, 2020). The following characters have been proposed as diagnostic of this species group: extremely ossified and exostosed skulls, elevated (keratinized or not) cranial crests, dorsal skin with small and large tubercles, and lateral row of enlarged tubercles absent (Duellman and Schulte, 1992; Pramuk, 2006; Maciel et al., 2010). Maciel et al. (2010) and Vallinoto et al. (2010) studied the phylogenetic relationships in this species group. Maciel et al. (2010) included phenotypic (morphological and parotoid-macrogland secretions) and molecular (sequences of three mitochondrial and one nuclear genes) characters and found this group as monophyletic, being the sister taxon of the Rhinella crucifer Group. Alternatively, Vallinoto et al. (2010) found the R. crucifer Group nested within the R. marina Group. Sequeira et al. (2011) reported the occurrence of extensive unidirectional introgression between R. diptycha (as R. schneideri) and some populations of R. marina that could contribute to biased inferences in the phylogenetic relationships. More recently, Vallinoto et al. (2017) reevaluated this hypothesis by including additional samples and molecular markers and found a more complex scenario with no evident pattern of unidirectional introgression and a doubtful taxonomic status of some R. marina populations. Finally, based on a phylogenetic analysis using mitochondrial genes and morphometric data, Acevedo et al. (2016) resurrected R. horribilis for the western Andean populations previously considered R. marina . Recently Bessa-Silva et al. (2020) found evidence of interspecific nuclear differentiation between these species and a marked discordance between mitochondrial and nuclear phylogenetic inferences in the R. marina Group.
We included samples of several populations from all the currently recognized species of this group: Rhinella achavali, R . arenarum, R . cerradensis, R . diptycha, R . horribilis, R . icterica, R . jimi, R . marina, R . poeppigii, R . rubescens, and R. veredas . For R. arenarum, we also included samples of the populations historically assigned to the subspecies R. arenarum mendocina (see Laurent, 1969).
The Rhinella spinulosa Group
Nine species are currently assigned to this group, which are distributed in the Andean region from southern Ecuador to southern Argentina and Chile, except for Rhinella achalensis, which is endemic to the Sierras Pampeanas Centrales in central Argentina (Cei, 1972b; Pramuk and Kadivar, 2003). The species of this group are medium sized and have a moderately to lightly ossified skull that lacks dermal sculpturing and exostosis. They also have a marked sexual dimorphism in skin texture and coloration (Vellard, 1959; Cei, 1972a, 1972b; Duellman and Schulte, 1992). This group was recovered as monophyletic in the combined phylogenetic analysis of Pramuk (2006: fig. 4) but paraphyletic in the separate molecular or morphological analyses (Pramuk, 2006: figs. 1–3). Some subspecies have been recognized for the nominal species of this group, which is a putative species complex (Vellard, 1959; Cei, 1972a; Ferraro et al., 2018).
We included all recognized species of this group: Rhinella achalensis, R . amabilis, R . arequipensis, R . arunco, R . atacamensis, R . limensis, R . rubropunctata, R . spinulosa (including populations historically assigned to the subspecies R. s . papillosa, R . s . spinulosa, and R. s . trifolium), and R. vellardi. We were unable to sample populations assigned to two subspecies of R. spinulosa: R. s . altiperuviana and R. s . flavolineata.
The “Rhinella veraguensis Group”
This nonmonophyletic group is composed of 17 small- to medium-sized species, all of which occur in the cloud forest of the Andes from northern Peru to northern Argentina, excepting Rhinella chrysophora, a species from north-central Honduras (Cei, 1972a; Duellman and Schulte, 1992; Chaparro et al., 2007; Cusi et al., 2017; McCranie, 2017). Members of this group are morphologically diverse with terrestrial, semiaquatic, or arboreal habits.
The following characters have been considered diagnostic for the Rhinella veraguensis Group: skull with weak exostosis, cranial crests absent or weak, dorsal skin bearing small elevated tubercles, and a lateral row of enlarged tubercles in some species (Gallardo, 1961; Cei, 1972a; Duellman and Schulte, 1992; Pramuk, 2006). This group has been consistently recovered as nonmonophyletic (Pramuk, 2006; Chaparro et al., 2007; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Moravec et al., 2014; Pyron, 2014; Cusi et al., 2017; Jetz and Pyron, 2018) and its definition and composition are problematic (see Pereyra et al., 2015; and comments regarding the definition of the R. acrolopha and R. festae Groups above). For purposes of description of the ingroup, R. lilyrodriguezae is included in this group (according to the phylogenetic relationships recovered by Cusi et al., 2017), although this species was assigned to the R. festae Group in the original description. We included samples of most species of this group (R. amboroensis, R . arborescandens, R . chavin, R . fissipes, R . inca, R . justinianoi, R . leptoscelis, R . lilyrodriguezae, R . manu, R . multiverrucosa, R . nesiotes, R . quechua, R . rumbolli, R . tacana, R . veraguensis, and R. yanachaga). The only unsampled species was R. chrysophora, which occurs in the Wet Forest in the central and western portions of the Cordillera Nombre de Dios, central-north Honduras. This species has not been observed since 1996 and is thought to be extinct (McCranie, 2017).
Species Unassigned to Any Group
Six species of Rhinella are not currently assigned to any group (for a history of previous group assignments see appendix 5). Among them, we included R. ceratophrys, R . gallardoi, and R. sternosignata in our phylogenetic analyses. The following species were not included in the phylogenetic analyses: (1) R. cristinae (Vélez-Rodríguez and Ruiz-Carranza, 2002), a species known only from the type locality (Vereda Tarquí, km 53–54 on road Altamira-Florencia, Municipio de Florencia, Departamento del Caquetá, Colombia) and collected in 1990 for the last time; (2) R. gnustae (Gallardo, 1967), which is only known from the holotype collected in 1925 from an imprecise locality (Rio Grande, Jujuy, Argentina; see Lavilla et al., 2002); and (3) R. iserni (Jiménez de la Espada, 1875), which is also known with certainty only from the holotype and its type locality is imprecise (Andes de Chanchamayo, Peru).
Molecular Data
Tissue Sampling
The molecular data were the main source of evidence in terms of both number of scored characters and sampled terminals. As one of the main goals of this paper was to test the monophyly of all the species groups of Rhinella, we attempted to obtain tissue samples from as many species as possible, with particular emphasis on putative nonmonophyletic species groups. Additionally, we included specimens from multiple populations of species that include recognized subspecies (e.g., R. arenarum and R. spinulosa), species that might represent species complexes (e.g., R. dapsilis, R . margaritifera, and R. proboscidea), and widely distributed taxa (e.g., R. diptycha, R . marina, and R. veraguensis) to evaluate their taxonomy. We included GenBank sequences only in cases where precise voucher number and locality data are provided, for specimens sequenced for at least the 16S rRNA gene. Besides, we made an effort to corroborate the identity of most relevant vouchers. A detailed list of all the terminals included in our analyses is given in appendices 1 and 2.
Laboratory Protocols
We extracted total genomic DNA from ethanol-preserved tissues (liver, muscle, or fingertips) using the Qiagen DNeasy kit. We carried out PCR amplifications in a total volume of 25 µl reactions using 0.2 µl Taq (Fermentas). The PCR protocol consisted of an initial denaturation step of 3 min. at 94° C followed by 35 (for mitochondrial genes) or 45 (for nuclear genes) cycles consisting of 30 seconds at 94° C for denaturation, 40 seconds at 48°–62° C for annealing, and 30–60 seconds at 72° C for extension, and a final extension step of 10–15 minutes at 72° C. We cleaned PCR-amplified products using 10U of Exonuclease plus 1U of alkaline phosphatase per reaction. We sequenced the products with an automatic sequencer ABI 3730XL (Applied Biosystems) in both directions to check for potential errors and nuclear polymorphisms. We processed the chromatograms using the software Sequencher version 4.5 (Gene Codes, Ann Arbor, MI) and edited the complete sequences with BioEdit (Hall, 1999). Sequences are deposited in GenBank under the accession numbers MW002838–MW003700.
Genotypic character sampling
The mitochondrially encoded loci sampled for the phylogenetic analyses include: (1) the 12S rRNA, the tRNA Valine, and the 16S rRNA genes (12S-tRNAVal-16S; 2469 bp), (2) a fragment comprising the upstream section of the 16S rRNA gene and the tRNA Leucine, NADH dehydrogenase subunit 1, and tRNA Isoleucine genes (16S-tRNALeu-ND1- tRNAIle; 1305 bp), and (3) a fragment of cytochrome b gene (cytb; 700 bp), for a total of up to 4474 bp. The nuclear loci include: (1) the C-X-C motif chemokine receptor 4 gene (cxcr4; 676 bp), (2) the solute carrier family 8 member A1 gene (slc8a1; 715 bp), (3) the proopiomelanocortin gene (pomc; 559 bp), (4) two nonoverlapping fragments of the recombination activating 1 gene (rag1-a and rag1-b; 936 and 429 bp respectively), and (5) the rhodopsin gene (rho; 316 bp), for a total of 3631 bp. Primers and their sources are detailed in table 1.
For the parsimony total evidence and maximum-likelihood (ML) analyses (see below), the amount of sequence data analyzed per terminal ranged from 447 bp (Rhinella gildae URCA 12651 obtained from GenBank) to 8089 bp (R. henseli CFBH 20117), with a mean of 4378 bp per terminal. All the phylogenetic datasets employed in the analyses are available at https://doi.org/10.5531/sd.sp.46).
Phenotypic data
The phenotypic dataset consisted of direct observations on specimens and bibliographic information for 90 characters, scored for 106 terminals (84 from the ingroup, 22 from outgroups). The scoring was recorded using Mesquite version 3.51 (Maddison and Maddison, 2018). The dataset was assembled from the following character systems: 33 from adult osteology, 15 from hand and foot musculature, 3 from the tympanic middle ear, 1 from adult visceral anatomy, 15 from adult external morphology, 9 from larval external morphology, 3 from larval chondrocranium, 4 from embryonic external morphology, 6 from natural history, and 1 from cytogenetics. Phenotypic characters are described below (see List and Description of Characters); the phenotypic matrix is included as supplementary data 1 (available at https://doi.org/10.5531/sd.sp.46).
Cranial and postcranial osteology follows the terminology employed by Trueb (1973, 1993), that of cranial crests follows Mendelson (1997a), and hand and foot myology follows Blotto et al. (2020). Terminology for larval external morphology follows Altig and McDiarmid (1999) and the characterization of embryonic structures follows Nokhbatolfoghahai and Downie (2005, 2008). Osteology was studied in (1) cleared and double-stained specimens prepared following the techniques of Wassersug (1976), (2) dry skeletons, and (3) µ-CT scans (available for download at www.morphosource.org, Duke University). Additional information was obtained from detailed osteological descriptions in the literature (see appendix 3). Visualization and data processing of µ-CT images was done in MeshLab (Cignoni et al., 2008). For the study of myology, dissections of the hand and foot musculature were performed to remove superficial layers and observe successively deeper muscles as outlined by Blotto et al. (2020). Topical applications of the iodine/potassium iodide solution of Bock and Shear (1972) were used when necessary to enhance contrast. The remaining characters were scored from the literature, unless specified (see appendix 3).
We scored multiple states for uncertainty or ambiguity in the condition of a terminal (among some states, but not all the character states) for some characters (see Pol and Apesteguía, 2005). This way of scoring let us incorporate relevant information (mainly from descriptions obtained from the bibliography) even when descriptions were not detailed enough. For 19 series of transformation, we used composite coding (sensu Maddison, 1993), which minimizes the occurrence of inapplicable or missing entries (Pimentel and Riggins, 1987; Maddison, 1993; Strong and Lipscomb, 1999).
Phylogenetic Analyses
The final taxon sample for the phylogenetic analyses was defined by means of a series of preliminary analyses that clarified the situation of many problematic terminals. As hybridization and genetic introgression, both nuclear and mitochondrial, seem to be common in some species of Rhinella (Sequeira et al., 2011; Pereyra et al., 2016a; Vallinoto et al., 2017), we first performed exploratory analyses of mitochondrial (MD) and nuclear (ND) datasets independently to detect nuclear-mitochondrial discordance as indicative of putative genetic introgression. Subsequently, we performed a total evidence (TE) analysis (Kluge, 1989, 2004; Nixon and Carpenter, 1996) combining nonintrogressed nuclear and mitochondrial sequences and the phenotypic dataset (see details in appendix 2).
The phylogenetic analyses of each separate molecular dataset (nuclear and mitochondrial, see below) and the total evidence analysis were performed in TNT version 1.5 (Goloboff et al., 2008; Goloboff and Catalano, 2016). Gaps were considered as a fifth state in all parsimony analyses (nuclear, mitochondrial, and total evidence analyses) and all classes of transformation events were equally weighted. In addition, we performed a total evidence analysis considering gaps as missing data for comparisons with the maximum likelihood analysis (see below). Unless otherwise stated, all results shown refer to parsimony analyses in which gaps were treated as a fifth state. We favoured parsimony as optimality criterion because the cladogram that minimizes transformations to explain the observed variation is the simplest one, maximizes evidential congruence, and has the greatest explanatory power (Farris, 1983; Goloboff, 2003; Goloboff and Pol, 2005; Kluge and Grant, 2006; Wheeler et al., 2006). Sequences were aligned using the online software MAFFT v7 (Katoh and Toh, 2008; Katoh et al., 2019) under the strategy E-INS-i (for the 12S-tRNAVal-16S fragment) and L-INS-i or G-INS-i (for remaining fragments), with default parameters for gap opening and extension. These alignments were used for both phylogenetic analyses and clade supports estimations (see details below).
Separate Phylogenetic Analyses of Nuclear and Mitochondrial Sequences
Both nuclear and mitochondrial datasets were analyzed in TNT using “New Technology” searches and performing a combination of sectorial searches, ratchet, and tree fusing (Goloboff, 1999; Nixon, 1999), using the default settings for these strategies. Tree searches were performed until the consensus was stabilized 10 times, with a factor of 75 (see Goloboff, 1999; Giribet, 2005).
The strict consensus tree resulting from the analysis of all sampled taxa of the nuclear dataset (= ND) was poorly resolved (data not shown). A poor resolution of the consensus can be due to the effect of just a small number of wildcard or rogue taxa, which are those that assume varying phylogenetic positions in the most parsimonious trees (MPT) (Nixon and Wheeler, 1992; Wilkinson, 1996; Aberer et al., 2013; Goloboff and Szumik, 2015). To avoid including terminals that act as wildcard taxa due to the lack of evidence, we included only terminals with more than three nuclear sequenced fragments (see appendix 2). Although there is an imperfect relationship between missing data and wildcard behavior, we identified three loci as the critical number to obtain an informative and comparable consensus in preliminary analyses. After excluding terminals with fewer than three nuclear fragments from the dataset, we reanalyzed this restricted nuclear dataset (rND) to estimate the consensus tree and clade supports (see below). The mitochondrial dataset was analyzed using the same terminals as the restricted nuclear dataset (i.e., restricted mitochondrial dataset, rMD) and similar parameters of analysis (see above), to allow the comparison.
Total Evidence Analysis
For the TE analysis, we followed the strategy described above for the separate nuclear and mitochondrial analyses. In this analysis, we included: (1) all the nuclear sequences from the complete nuclear dataset, (2) all the mitochondrial sequences from the complete mitochondrial dataset, and (3) the phenotypic dataset. The following criteria were used to treat putatively conspecific sequences as pertaining to the same or different terminals: (1) sequences from the same individual or conspecific individuals placed in well-supported discordant positions in the separate nuclear and mitochondrial analyses were considered as independent terminals, because discordance suggests mitochondrial introgression between different species (see Pereyra et al., 2016a); and (2) terminals from the phenotypic dataset were combined with the more closely related conspecific terminal of the molecular dataset (mitochondrial + nuclear). When mitochondrial and nuclear sequences of a specimen were included separately, the phenotypic data were combined with the nuclear sequences. Appendix 2 provides a list of all the terminals included and excluded in the TE analysis.
TABLE 1
Primers used to amplify and sequence DNA in this study
See appendix 2 for gene abbreviations.
Resampling Support Measures
Two types of resampling support measures were estimated for the datasets in TNT version 1.5 (Goloboff and Catalano, 2016): (1) parsimony jackknife absolute frequencies (JAF; Farris et al., 1996) and (2) parsimony jackknife frequency differences (JGC; Goloboff et al., 2003). For estimation of both measures, we performed 1,000 replicates using “New Technology” searches consisting of a combination of sectorial searches, ratchet, and tree fusing (Goloboff, 1999; Nixon, 1999), reaching minimum length two times (preliminary analyses showed that minimum lengths are hit with this search strategy). Goloboff et al. (2003) noted that the resampling support for a clade does not necessarily correlate with the absolute frequency itself (i.e., the number of times a group is recovered in the resampled matrices), because groups with positive support (≥ 50%) can have much lower frequencies than groups with no support at all (<50%). To solve this situation, these authors proposed to also consider the value GC (i.e., frequency difference), which indicates the frequency differences between a group and the most frequent contradictory group. Values of this score range between -100% (maximum contradiction) and 100% (maximum support).
Maximum-Likelihood Analysis
Maximum-likelihood analysis was performed with IQ-TREE v1.6.12 (Nguyen et al., 2015) considering the same dataset (DNA sequences + phenotypic characters) as the TE analysis under parsimony. ModelFinder (Kalyaanamoorthy et al., 2017), which is implemented in IQ-TREE, was used to select the optimal partition scheme and substitution models for molecular characters. ModelFinder implements a greedy strategy (Lanfear et al., 2012) that starts with the full partition model and subsequentially merges two genes until the model fit does not increase any further. The best partition scheme included two subsets (see table 2). For morphological data we use the two morphological ML models (see Lewis, 2001) implemented in IQ-TREE (i.e., MK and ORDERED, for unordered and ordered characters respectively) considering the ascertainment bias correction (ASC) method. We consider edge-linked-proportional partition model but separate substitution models and rate evolution between partitions (-spp option). The maximum-likelihood tree was conducted with 1000 ultrafast bootstrap replicates (Minh et al., 2013; Hoang et al., 2018) using the option -bnni that reduces the risk of overestimating branch supports due to severe model violations. The resulting tree was visualized and edited in FigTree 1.4.3 (Rambaut, 2016). Partitions and models selected are detailed in table 2.
Taxonomic Evaluation
We considered the following criteria in assessing the taxonomic status of each lineage: (1) the cladogram topology resulting from the phylogenetic analyses, (2) the uncorrected pairwise distances (UPDs) of a fragment of the 16S rRNA gene (delimited by the primers AR and WILK2; see Vences et al., 2005a, 2005b; Fouquet et al., 2007b) calculated in PAUP* (Swofford, 2002), and (3) the known phenotypic evidence for each taxon. The phenotypic criterion was mainly considered in cases where relationships were unresolved (i.e. occurrence of polytomies) or poorly supported (JGC <50%) within a clade. For estimation of UPDs, datasets containing only sequences of the 16S rRNA gene for each species group (as are redefined in the Results section) were aligned in MAFFT under the strategy G-INS-i.
TABLE 2
Best partition scheme and best-fit models selected by ModelFinder for the molecular data.
For phenotypic data, we used morphological models considering the ascertainment bias correction (ASC) method.
LIST AND DESCRIPTION OF CHARACTERS
Characters modified from previous phylogenetic studies are indicated with an asterisk (*).
Adult Osteology
Most of the osteological characters used here are those of Pramuk (2006), so they are not described in detail except when relevant (e.g., when character states were modified or additional character states were considered). Described characters refer to adult individuals of both sexes unless specified.
Skull
0. Preorbital crest (on the maxillary process of nasal), occurrence: (0) absent or indistinguishable, (1) weak, (2) well developed. Additive. Cranial crests were considered osteological characters, although it could also be scored from whole-preserved specimens. The use of presence/absence of cranial crests has a long history in bufonid taxonomy, and they were used in a phylogenetic context by Pramuk (2006: chars. 63–69). However, unlike Pramuk (2006), we differentiate between weak and well-developed crests. State 1 (weak) refers to cranial crests that are faint or not evident externally in living or intact preserved specimens, but evident in osteological preparations. State 2 (well developed) refers to crests that are evident externally in both intact and osteologically prepared specimens. When osteological preparations were not available to precisely determine the absent or weak state of the crest (since both states are similar in complete specimens) we scored these uncertainties as multiple states (i.e., 0/1, see Phenotypic data scoring in Material and methods section).
Previous usage in phylogenetic studies: Inger (1972: char. 29*), Morrison (1994: char. 13*), Mendelson (1997a: char. 6*), Pramuk (2006: char. 65*), Mendelson et al. (2011: char. 6*).
1. Supraorbital crest (on frontoparietals), occurrence: (0) absent or indistinguishable, (1) weak, (2) well developed. Additive.
Previous usage in phylogenetic studies: Inger (1972: char. 29*), Morrison (1994: char. 14*), Mendelson (1997a: char. 7*), Pramuk (2006: char. 68*), Mendelson et al. (2011: char. 7*).
2. Pretympanic crest (on the zygomatic ramus of squamosal), occurrence: (0) absent or indistinguishable, (1) weak, (2) well developed. Additive.
Previous usage in phylogenetic studies: Morrison (1994: char. 16*), Mendelson (1997a: char. 11*), Pramuk (2006: char. 66*), Mendelson et al. (2011: char. 11*).
3. Supratympanic crest (on the otic ramus of squamosal), occurrence in females: (0) supratympanic crest inconspicuous or developed, but that does not extend beyond the level of the cranial roof dorsally, (1) supratympanic crest hypertrophied extending beyond the level of the cranial roof dorsally. This character was codified separately for males and females since a dimorphic condition was detected. Large supratympanic crest occurs mainly in adult females of many species of the Rhinella margaritifera Group (Hoogmoed, 1990; Duellman and Schulte, 1992). However, males of some of these species also have large supratympanic crest (Hoogmoed, 1990).
Previous usage in phylogenetic studies: Morrison (1994: char. 17*), Mendelson (1997a: char. 10*), Vélez-Rodríguez (2004b: char. 35*), Pramuk (2006: char. 69*), Mendelson et al. (2011: char. 10*).
4. Supratympanic crest (on the otic ramus of squamosal), occurrence in males: (0) supratympanic crest inconspicuous or developed, but that does not extend beyond the level of the cranial roof dorsally, (1) supratympanic crest hypertrophied extending beyond the level of the cranial roof dorsally.
Previous usage in phylogenetic studies: Morrison (1994: char. 17*), Mendelson (1997a: char. 10*), Vélez-Rodríguez (2004b: char. 35*), Pramuk (2006: char. 69*), Mendelson et al. (2011: char. 10*).
5. Parietal crest (on frontoparietal), occurrence: (0) absent or indistinguishable, (1) weak, (2) well developed. Additive.
Previous usage in phylogenetic studies: Inger (1972: char. 29*), Morrison (1994: char. 15*), Mendelson (1997a: char. 8*), Pramuk (2006: char. 64*), Mendelson et al. (2011: char. 8*).
6. Nasals, shape of anterior margins: (0) relatively blunt, (1) acuminate.
Previous usage in phylogenetic studies: Mendelson (1997a: char. 34*), Scott (2005: char. 64*), Pramuk (2006: char. 4), Nussbaum and Wu (2007: char. 52*), Mendelson et al. (2011: char. 25*).
7. Nasals, medial contact: (0) not in contact medially, (1) in contact medially.
Previous usage in phylogenetic studies: Lynch (1978: char. 2*), Clarke (1981: char. 1*), Ford (1990: char. 1), Morrison (1994: char. 1*), Báez and Basso (1996: char. 2), Mendelson et al. (2000: char. 32), Scott (2005: char. 63), Fabrezi (2006: char. 1), Pramuk (2006: char. 3), Nussbaum and Wu (2007: char. 51), Ponssa (2008: char. 52*).
8. Contact between nasal and frontoparietal: (0) anterior margin of frontoparietal does not articulate with posterior margin of nasal (fig. 2A), (1) articulate only laterally (fig. 2B), (2) articulate along most of its margin but not completely (fig. 2C), (3) articulate along the entire margin (fig. 2D). Additive.
Previous usage in phylogenetic studies: Heyer and Liem (1976: char. 2*), Morrison (1994: char. 6*), Mendelson (1997a: char. 2*), Pugener et al. (2003: char. 12* [adult morphological characters]), Pramuk (2006: char. 8*), Mendelson et al. (2011: char. 2*).
9. Dermal roofing bones, sculpturing: (0) dermal bones of the skull completely smooth, (1) lightly exostosed, (2) heavily ornamented with pits, striations, and rugosities. Additive. Hyperossification in anurans involves the level of sculpturing and the number and identity of exostosed bones (see revision by Blotto et al., 2021). Although species of Rhinella display a relatively high diversity of hyperossification, for the time being, we scored the variation only in the dermal roofing bones (nasals and frontoparietal), until more detailed analyses of the skull morphology are carried out.
Previous usage in phylogenetic studies: Clarke (1981: char. 2*), Ford (1990: char. 4*), Morrison (1994: char. 11), Mendelson et al. (2000: char. 28*), Pugener et al. (2003: char. 10* [adult morphological characters]), Scott (2005: char. 61*), Fabrezi (2006: char. 2*), Pramuk (2006: char. 2), Nussbaum and Wu (2007: char. 62*).
10. Occipital artery pathway, coverage with bone: (0) occipital canal not covered by bone, (1) partially covered, (2) completely covered with bone. Additive.
Previous usage in phylogenetic studies: McDiarmid (1971: char. 7*), Inger (1972: char. 10*), Heyer and Liem (1976: char. 3*), Lynch (1978: char. 4*), Clarke (1981: char. 4*), Morrison (1994: char. 10*), Mendelson (1997a: char. 4*), Mendelson et al. (2000: char. 38*), Pugener et al. (2003: char. 15* [adult morphological characters]), Wiens et al. (2005: char. 16*), Pramuk (2006: char. 9), Mendelson et al. (2011: char. 4*).
11. Squamosal, medial extension of otic ramus: (0) otic ramus of squamosal present, but not enlarged, (1) otic ramus of squamosal slightly enlarged, overlapping with the dorsal surface of the crista parotica, (2) otic ramus enlarged, in contact with posterolateral margin of frontoparietal, forming a continuous temporal arcade. Additive.
Previous usage in phylogenetic studies: Inger (1972: char. 12*), Lynch (1978: char. 5*), Clarke (1981: char. 5*), Ford (1990: char. 29*), Báez and Basso (1996: char. 16*), Mendelson (1997a: char. 33*), Faivovich (2002: char. 4*), Scott (2005: char. 65*), Wiens et al. (2005: char. 15*), Fabrezi (2006: char. 10*), Pramuk (2006: char. 15), Nussbaum and Wu (2007: char. 85*), Araujo-Vieira et al. (2019: char. 28*).
12. Nasals, extension of anterior margin: (0) anterior margins extend beyond the dorsal margins of the alary processes of the premaxillae (fig. 3A), (1) anterior margins are flush with the dorsal margins of the alary processes (fig. 3B), (2) anterior margins lie posterior to the dorsal margins of the alary processes (fig. 3C). Additive.
Previous usage in phylogenetic studies: Pramuk (2006: char. 21), Ponssa (2008: char. 57*).
13. Premaxilla, orientation of alary process: (0) angled posteriorly to the anterior margin of the premaxillae (fig. 4A), (1) dorsally projected to the anterior margin of the premaxillae (fig. 4B), (2) angled anteriorly to the anterior margin of the premaxillae (fig. 4C). Additive.
Previous usage in phylogenetic studies: Ford (1990: char. 12), Morrison (1994: char. 42), Mendelson (1997a: char. 23), Scott (2005: char. 78), Pramuk (2006: char. 26), Nussbaum and Wu (2007: char. 68), Ponssa (2008: char. 33), Barrionuevo (2017: char. 6*), Araujo-Vieira et al. (2019: char. 19*).
14. Septomaxilla, level of development of the anterior end: (0) not developed, (1) very developed and exposed anteriorly to the alary process of the premaxilla. Alcalde (2017) showed that bones previously described as “rostrals” (Pregill, 1981) or “prenasals” (Pramuk, 2000, 2006) in some bufonids are actually part of the enlarged and exposed anterior ends of the septomaxillae (and thus char. 42 of Pramuk [2006] refers to this structure instead to prenasals bones). Alcalde (2017) also pointed out the presence of an unpaired bone in the anterior end of the snout in Rhinella dorbignyi (as R. fernandezae, from the R. granulosa Group). He stated that it is homolog to the prenasal bone in some Lophyohylini (Hylidae; Trueb, 1970); even if primary homologs, they clearly represent independent instances of evolution. We observed this element in R. beebei (USNM 566017–8), but we could not determine its occurrence in other species of the group for which we do not consider this bone as a different character (see comments on the preservation and identification of this structure in Alcalde, 2017).
Previous usage in phylogenetic studies: Pramuk (2006: char. 42*).
15. Squamosal, articulation of zygomatic and ventral rami: (0) the zygomatic ramus of the squamosal is free from the ventral ramus, (1) the zygomatic ramus of the squamosal articulates with the ventral ramus of the squamosal.
Previous usage in phylogenetic studies: Mendelson (1997a: char. 32*), Vélez-Rodriguez (2004b: char. 32*), Pramuk (2006: char. 14*).
16. Jaw articulation: (0) posterior to the fenestra ovalis, (1) opposite to the fenestra ovalis, (2) anterior to the fenestra ovalis. Additive.
Previous usage in phylogenetic studies: Pramuk (2006: char. 25), Báez et al. (2012: char. 39).
17. Supraorbital flange on the frontoparietals: (0) frontoparietal does not extend laterally beyond the lateral margin of the sphenethmoid, (1) frontoparietal extends laterally beyond the lateral margin of the sphenethmoid.
Previous usage in phylogenetic studies: Morrison (1994: char. 4), Mendelson (1997a: char.1), Mendelson et al. (2000: char. 36), Pugener et al. (2003: char. 13 [adult morphological characters]), Wiens et al. (2005: char. 13), Pramuk (2006: char. 72), Mendelson et al. (2011: char. 1).
18. Sphenethmoid, extent of anterior ossification: (0) bony sphenethmoid reaches the level of palatines, but not beyond, (1) bony sphenethmoid beyond palatines, but does not reach the level of the premaxillae, (2) bony sphenethmoid reaches the level of the premaxillae anteriorly. Additive.
Previous usage in phylogenetic studies: Morrison (1994: char. 20), Mendelson (1997a: char. 13*), Vélez-Rodriguez (2004b: char. 21*), Pramuk (2006: char. 34*), Araujo-Vieira et al. (2019: char. 9*).
19. Pterygoid, articulation of the anterior ramus with maxilla: (0) anterior ramus of pterygoid articulates along the margin of maxilla, but does not contact with the palatine, (1) anterior ramus of pterygoid articulates along the margin of maxilla and contacts the palatine.
Previous usage in phylogenetic studies: Ford (1990: char. 32*), Morrison (1994: char. 52), Clarke (1981: char. 13*), Mendelson (1997a: char. 28*), Pugener et al. (2003: char. 40* [adult morphological characters]), Vélez-Rodríguez (2004b: char. 17*), Ponssa (2008: char. 67), Barrionuevo (2017: char. 29*).
20. Palatine, ventral ridge: (0) absent or indistinguishable, (1) present.
Previous usage in phylogenetic studies: Inger (1972: char. 18*), Morrison (1994: char. 33*), Mendelson (1997a: char. 15*), Mendelson et al. (2000: char. 10*), Pramuk (2006: char. 38), Mendelson et al. (2011: char. 14*).
21. Pterygoid, contact of medial ramus with ala of parasphenoid: (0) the medial ramus of the pterygoid is not in contact nor fused with the anterolateral margin of the ala of the parasphenoid, (1) the medial ramus of the pterygoid is fused with the anterolateral margin of the parasphenoid, (2) the medial ramus of the pterygoid is fused and extends medially along approximately half the length of the parasphenoid ala. Additive.
Previous usage in phylogenetic studies: Lynch (1978: char. 9*), Clarke (1981: char. 14*), Ford (1990: char. 34*), Morrison (1994: char. 54), Báez and Basso (1996: char. 28*), Mendelson (1997a: ch 29*), Vélez-Rodríguez (2004b: char. 19*), Pramuk (2006: char. 19).
22. Pterygoid, suture between the medial ramus and parasphenoid alae: (0) the surface of contact is smooth, (1) jagged or scalloped. This character is not applicable for specimens where the medial ramus of the pterygoid is not in contact or not fused with the anterolateral margin of the ala of the parasphenoid (char. 21.0).
Previous usage in phylogenetic studies: Grandison (1981: char. 13*), Pramuk (2006: char. 31).
23. Parasphenoid, shape of anterior margin of cultriform process: (0) acute and narrow (fig. 5A), (1) broadly rounded anteriorly (fig. 5B), (2) truncated (fig. 5C), (3) jagged or scalloped (fig. 5D). Nonadditive.
Previous usage in phylogenetic studies: Clarke (1981: char. 12*), Ford (1990: char. 45*), Morrison (1994: char. 36*), Mendelson (1997a: char. 20*), Scott (2005: char. 54*), Pramuk (2006: char. 29*), Nussbaum and Wu (2007: char. 98*), Araujo-Vieira et al. (2019: char. 42*).
24. Bony protrusion at the angle of jaws: (0) absent or indistinguishable, (1) weak, (2) developed into a processus. Additive. A bony protrusion (“or bony knob”) is caused by a variable level of thickening of the ventrolateral margin of the quadratojugal. The level of development of the bony protrusion could also be determined both in living or intact specimens as in osteological preparations.
Previous usage in phylogenetic studies: Vélez-Rodriguez (2004b: char. 36*).
25. Hyoid, posterior lobe of the anterolateral process: (0) absent or indistinguishable (fig. 6A), (1) present (fig. 6B).
Previous usage in phylogenetic studies: Vélez-Rodriguez (2004: char. 42).
Vertebral Column
26. Presacral vertebrae, level of development of neural spine: (0) neural spine flat or slightly elevated, (1) neural spine notably elevated, protruding externally. The level of development of the neural spines can be determined both in intact-preserved specimens and in osteological preparations.
Previous usage in phylogenetic studies: Vélez-Rodriguez (2004b: char. 40*).
27. Presacral vertebrae, number: (0) eight, (1) seven. This number refers to the number of vertebrae even if there is some level of fusion between them. The number can be traceable even when there is fusion of centra due to the persistence of the intervertebral foramina (see Trueb, 1973; Cannatella, 1986).
Previous usage in phylogenetic studies: McDiarmid (1971: char. 23*), Lynch (1973: char. 1*), Grandison (1981: char. 15*), Cannatella (1986: char. 3*), Morrison (1994: char. 65*), Báez and Basso (1996: char. 30*), Wiens et al. (2005: char. 51*), Fabrezi (2006: char. 34), Pramuk (2006: char. 44*), Nussbaum and Wu (2007: char. 139), Mendelson et al. (2011: char. 43).
28. Presacral vertebrae I and II, fusion: (0) absent, (1) present. The fusion of the centra of both vertebrae into a single element may be identified for the occurrence of transverse processes and two foramina for vertebral nerves in the anterior presacral element.
Previous usage in phylogenetic studies: McDiarmid (1971: char. 24*), Lynch (1973: char. 2), Heyer and Liem (1976: char. 9), Cannatella (1986: char. 4*), Ford (1990: char. 66), Morrison (1994: char. 66), Wiens et al. (2005: char. 50), Grant et al. (2006: char. 145*), Nussbaum and Wu (2007: char. 137), Báez et al. (2012: char. 49*), Barrionuevo (2017: char. 43).
29. Sacrum, shape of sacral diapophyses: (0) the maximum width of the sacral diapophysis is smaller than its maximum length, (1) the maximum width of the sacral diapophysis is equal to, or greater than, its maximum length.
Previous usage in phylogenetic studies: Heyer (1975: char. 34*), Heyer and Liem (1976: char. 12*), Ford (1990: char. 75*), Morrison (1994: char. 70*), Báez and Basso (1996: char. 36*), Faivovich (2002: char. 21*), Pugener et al. (2003: char. 57* [adult morphological characters]), Fabrezi (2006: char. 42*), Grant et al. (2006: char. 143*), Pramuk (2006: char. 51), Araujo-Vieira et al. (2019: char. 95*).
30. Sacrum, orientation of anterior edge of sacral diapophyses: (0) posterior to the midline axis of the vertebral column, (1) perpendicular to the midline axis of the vertebral column, (2) anterior to the midline axis of the vertebral column. Additive.
Previous usage in phylogenetic studies: Scott (2005: char. 16), Pramuk (2006: char. 52), Nussbaum and Wu (2007: char. 142).
31. Sacrum and urostyle, fusion: (0) absent, (1) present.
Previous usage in phylogenetic studies: McDiarmid (1971: char. 25*), Lynch (1973: char. 3), Ford (1990: char. 76), Pugener et al. (2003: char. 58* [adult morphological characters]), Wiens et al. (2005: char. 60*), Nussbaum and Wu (2007: char. 138), Báez et al. (2012: char. 51*).
32. Ilium, dorsal protuberance, level of development: (0) large and slightly anteriorly or more dorsally directed, (1) small, low, and laterally projected. Gómez and Turazzini (2016) comment on the morphological variation and taxonomic distribution of this structure in anurans.
Previous usage in phylogenetic studies: Clarke (1981: char. 21*), Morrison (1994: char. 87*), Scott (2005: char. 12*), Pramuk (2006: char. 54), Báez et al. (2012: char. 65*).
Adult Musculature
Foot (ventral surface)
33. Discrete superficial cutaneous tendons, occurrence: (0) absent, (1) present. Burton (2004: 212, 220) described briefly this group of superficial tendons and Blotto et al. (2020) formalized this name. We scored if the superficial tendons are discrete or if they are absent or transformed into a sheet of connective tissue or fascia over the plantar side of the foot. Additional studies are needed to determine whether this group of tendons must be considered as a whole (as here) or individual superficial cutaneous tendons of each digit should be treated as independent characters. See further comments in Blotto et al. (2020).
34. M. interphalangeus proximalis digiti V, medial slip, occurrence: (0) absent, (1) present. See Dunlap (1960), Burton (2001, 2004), and Blotto et al. (2020) for descriptions of the mm. interphalangei of the foot and comments on its taxonomic distribution in Anura.
35. M. interphalangeus proximalis digiti V, lateral slip, occurrence: (0) absent, (1) present.
36. M. abductor brevis plantaris hallucis, occurrence: (0) absent, (1) present. See Burton (2001, 2004) and Blotto et al. (2017) for characterization of this muscle and taxonomic distribution in nonbufonid taxa.
Previous usage in phylogenetic studies: Burton (2004: char. 30*), Faivovich et al. (2005: char. 7), Hoyos et al. (2014: char. 44), Blotto et al. (2017: char. 1).
37. M. flexor digiti II (FDM II), position of the origin with respect to the m. intermetatarsalis 1 (IMT 1): (0) FDM II ventral to the IMT 1, (1) FDM II dorsal to the IMT 1, (2) FDM II ventral and dorsal to the IMT 1. Nonadditive. See Dunlap (1960: 42) for an account under the name of m. flexor teres (for the FDM II) and transversus metatarsus (for the m. intermetatarsalis).
38. M. interosseus cruris, presence of an additional origin from the tibiale: (0) absent, (1) present. Most species have both an origin from the tibiale and from the fibulare (Gaupp, 1896; Dunlap, 1960; Burton, 2004). Among bufonid taxa, state 0 was reported for Atelopus (see Dunlap, 1960: 30), under the name of m. intertarsalis. The only species from our sampling that has state 0 is Rhinella paraguas .
Foot (dorsal surface)
39. M. extensor digitorum longus (EDL), insertion on metatarsophalangeal joint of digiti IV: (0) absent, (1) present. We scored the insertion of the EDL in each digit as an independent character, contra Burton (2004: char. 48), as discussed by Faivovich et al. (2005: 201). We found informative variation for the insertions on digits IV and V (next char.). The insertions on the metatarsophalangeal joint of the digits IV and V may be by an independent tendon or through a common tendon with the m. extensor brevis superficialis, m. extensor brevis medius, and/or the m. dorsometatarsalis proximalis, a source of variation not considered in the present study. See Dunlap (1960) and Burton (2004) for descriptions and variation of the insertion of this muscle, under the name m. extensor digitorum communis longus. The intraspecific variation reported by Inger (1972: 103) for the absence/presence of the insertion on each digit should be further tested; only Nannophryne variegata from our sampling was studied from more than one specimen to test this potential intraspecific variation.
40. M. extensor digitorum longus, insertion on metatarsophalangeal joint of digit V: (0) absent, (1) present.
Previous usage in phylogenetic studies: Inger (1972: char. 26*).
41. M. extensor brevis medius hallucis, occurrence: (0) absent, (1) present. See Dunlap (1960: 52–53) for description and variation across Anura.
Previous usage in phylogenetic studies: Hoyos et al. (2014: char. 37).
42. Lateral m. dorsometatarsalis proximalis digiti IV, discrete and independent tendon inserting on the proximal interphalangeal joint of digit IV: (0) absent, (1) present. Dunlap (1960: 57) considered the muscles dorsometatarsales proximales and the dorsometatarsales distales (both as mm. extensores breves profundi) as the same muscle (see discussion in Blotto et al., 2020). This fact partially precludes the understanding of the variation and taxonomic distribution described by Dunlap (1960). On the other hand, the extensive study of Hylidae by Burton (2004: char. H) suggests a great intraspecific variation when considering the number of tendons of insertion of the mm. dorsometatarsales proximales III–V (as extensores breves profundi). In our sampling, all species have a tendon of the lateral m. dorsometatarsalis proximalis digiti IV inserting on the distal interphalangeal joint of digit IV, while Rhinella crucifer and R. henseli have an additional independent tendon of insertion on the proximal interphalangeal joint. In the light of the variation found in Bufonidae, as well as in other clades of Anura (B.L.B., personal obs.), we decided to tentatively consider each tendon to each interphalangeal joint as independent transformation series.
Hand (ventral surface)
43. Medial m. lumbricalis brevis digiti V, slip from distal carpal 3-4-5: (0) absent, (1) present. The medial m. lumbricalis brevis digiti V may have two slips, one from the distal carpals and the other one from the flexor plate/adjacent tendo superficialis digiti V; both with a common or independent insertions (Burton, 1998: 59; this study). Nevertheless, Burton (1998: char. 18) discarded further discussion and comparison of the nature of this muscle given the extreme degree of variation found within his sampling (“Leptodactylidae” s.l.).
Previous usage in phylogenetic studies: Burton (1998: char.18).
Hand (dorsal surface)
44. M. extensor digitorum, insertion on metacarpophalangeal joint of digiti III: (0) absent, (1) present.
Some species lack the insertion on the metacarpophalangeal joint of the digit III. This insertion may be through a common tendon after inserting on the dorsal fascia of other muscles (usually mm. extensores breves superficiales) or by an independent tendon (Burton, 1998; this study).
Previous usage in phylogenetic studies: Burton (1998: char. 22*).
45. M. extensor digitorum, insertion on metacarpophalangeal joint of digiti V: (0) absent, (1) present. The slip of the m. extensor digitorum to the digit V may have two insertions, one on the metacarpophalangeal joint and a second insertion on the lateral side of the metacarpal V. The presence of both insertions varies independently across Anura (B.L.B., personal obs.), for which we scored their presence as independent transformation series. Within the current sampling of Bufonidae, the lateral insertion on metacarpal V is invariably present, and thus variation is restricted to the presence of the insertion on the metacarpophalangeal joint. This insertion may be through a common tendon after insertion on the dorsal fascia of other muscles or by an independent tendon (Burton, 1998; Araujo-Vieira et al., 2019; this study).
Previous usage in phylogenetic studies: Araujo-Vieira et al. (2019: char. 171).
46. M. extensor carpi ulnaris, occurrence of a head from the radioulna: (0) absent, (1) present. This head was not previously reported in the literature. It originates from the distal half or quarter of the radioulna, laterally to the origin of the m. abductor pollicis longus. The head converges with the head from the humerus, which attaches to the ulnare and distal carpal 3–4–5 (fig. 7).
47. M. extensor carpi ulnaris, nature of the origin of the head from the radioulna: (0) fleshy (fig. 7B), (1) via a flat tendon (fig. 7D). This character is not applicable for specimens that lack a supplementary head from the radioulna (char. 46.0).
Tympanic Middle Ear Complex
Pereyra et al. (2016b) reported the range of variation in structures of the tympanic middle ear (i.e., columella, annulus tympanicus, and tympanic membrane) in Bufonidae and demonstrated its unique evolutionary pattern within Anura.
48. Columella, occurrence: (0) absent, (1) present.
Previous usage in phylogenetic studies: Grandison (1981: char. 1*), Cannatella (1986: char. 6*), Ford (1990: char. 11), Morrison (1994: char. 27), Mendelson (1997a: char. 38*), Pugener et al. (2003: char. 47* [adult morphological characters]), Scott (2005: char. 81), Pramuk (2006: char. 17), Nussbaum and Wu (2007: char. 67*), Mendelson et al. (2011: char. 27).
49. Annulus tympanicus, occurrence: (0) absent, (1) present.
Previous usage in phylogenetic studies: Inger (1972: char. 28*), Drewes (1984: char. 23*), Cannatella (1986: char. 8*), Scott (2005: char. 80*), Wiens et al. (2005: char. 35*), Nussbaum and Wu (2007: char. 66).
50. Tympanic membrane: (0) absent, (1) present.
Previous usage in phylogenetic studies: Inger (1972: char. 28*), Heyer (1975: char. 2*), Drewes (1984: char. 23*), Cannatella (1986: char. 8*), Morrison (1994: char. 96*), Scott (2005: char. 144*), Wiens et al. (2005: char. 108*), Ohler and Dubois (2006: char. 4*), Nussbaum and Wu (2007: char. 4), Barrionuevo (2017: char. 60).
Adult Visceral Anatomy
51. Inguinal fat bodies, occurrence: (0) absent, (1) present. Boulenger (1910) first reported the occurrence of elongated bodies associated to the muscles of the inguinal region in several species of Bufonidae. Later, Plytycz and Szarski (1987) and da Silva and Mendelson (1999) corroborated the occurrence of these inguinal fat bodies in many other species of several bufonid genera.
Previous usage in phylogenetic studies: Mendelson (1997a: char. 45), Pramuk (2006: char. 79), Mendelson et al. (2011: char. 34).
Adult External Morphology
All described characters of external morphology refer to adult individuals of both sexes, except when sexually dimorphic characters are considered (i.e., chars. 53–54 and 58–59).
52. Dorsal skin, macroscopic glandular structures in females: (0) indistinct; (1) small and smooth; (2) greatly enlarged and smooth; (3) with small tubercles, without cornified tip; (4) conical with a single cornified tip; (5) hemispherical with multiple cornified tips. Nonadditive. The skin of bufonids varies from completely smooth to highly tuberculated and warty due to modifications of dermal and/or epidermal components (Elias and Shapiro, 1959). It differs between sexes and its structure is affected (at least in males) by the reproductive condition of the specimens (see Cei, 1980; Duellman and Trueb, 1986). Therefore, when scoring skin diversity we considered: (1) as independent character structures present in both sexes (chars. 53, 54); (2) the maximum level of development reported for the dorsal structures of skin within the studied specimens; and (3) the most common structures present in the dorsal skin. Although some character states seem to be composed of a progressive series of transformation of glandular structures, detailed histological studies are required to understand the various components that are differentiated in each structure.
Previous usage in phylogenetic studies: McDiarmid (1971: char. 39*), Morrison (1994: char. 99*), Grant et al. (2006: char. 0*).
53. Dorsal skin, macroscopic glandular structures in males: (0) indistinct; (1) small and smooth; (2) greatly enlarged and smooth; (3) with small tubercles, without a cornified point; (4) conical with a single cornified point; (5) hemispherical with multiple cornified points. Nonadditive.
Previous usage in phylogenetic studies: McDiarmid (1971: char. 39*), Morrison (1994: char. 99*), Grant et al. (2006: char. 0*).
54. Vertebral line, occurrence: (0) absent, (1) present. Boulenger (1897) first noted the distinctiveness and independence between a filiform line or raphe and the vertebral line. The former is a line of thin skin extending along the middle of the back from the snout to the vent. This raphe is very common in bufonids and gives rise to a light vertebral line. As pointed out by Boulenger (1897), the independence of these structures is evident in cases of deviation of the former (see Boulenger, 1897: fig. 9). We consider the occurrence of a distinctive light vertebral line only.
Previous usage in phylogenetic studies: Heyer (1978: char. 1*), Ohler and Dubois (2006: char. 13*), Ponssa (2008: char. 1*).
55. Parotoid gland, occurrence: (0) absent, (1) present.
Previous usage in phylogenetic studies: Cannatella (1986: char. 11), Morrison (1994: char. 95*), Mendelson (1997a: char. 43*), Wiens et al. (2005: char. 109), Fabrezi (2006: char. 72), Pramuk (2006: char. 73*).
56. Parotoid gland, shape: (0) approximately ellipsoid, longer than wide (fig. 8A); (1) subtriangular (fig. 8B); (2) round to ovoid mostly symmetrical (fig. 8C); (3) triangular and bulky (fig. 8D). Nonadditive. This character is not applicable for specimens that lack a parotoid gland (char. 55.0).
Previous usage in phylogenetic studies: Morrison (1994: char. 95*), Pramuk (2006: char. 73*), Mendelson et al. (2011: char. 32*).
57. Skin, occurrence of row of dorsolateral tubercles: (0) absent, (1) present.
Previous usage in phylogenetic studies: Morrison (1994: char. 97), Mendelson (1997a: char. 44*), Pramuk (2006: char. 81), Mendelson et al. (2011: char. 33).
58. Vocal sac, occurrence in adult males: (0) absent, (1) present. The vocal sac develops as ventral diverticula of the mouth floor into spaces among submandibular muscles (Noble, 1931; Tyler, 1971). This second cavity communicates with the buccal cavity via single or paired apertures, the vocal slits. In this way, the presence of a vocal sac automatically implies the presence of at least one vocal slit (and vice versa). Modifications of the gular skin (i.e. “internal” or “external” vocal sacs sensu Günther, 1858a) can be absent or present without affecting the codification of this character. Vocal sacs are either absent or present in adult males of the majority of species, with few exceptions where both states cooccur in different specimens (Liu, 1935; Inger and Greenberg, 1956, Hayes and Krempels, 1986; Mendelson, 1997b).
Previous usage in phylogenetic studies: Liem (1970: char. 36*), Drewes (1984: char. 20*), Cannatella (1986: char. 10*), Hillis and de Sá (1988: char. 6), Mendelson et al. (2000: char. 51*), Grant et al. (2006: char. 76*), Pramuk (2006: char. 75*), Ohler and Dubois (2006: char. 14), Mendelson et al. (2011: char. 31*).
59. Vocal slits, number: (0) unilateral, (1) bilateral. Several authors (e.g., Boulenger, 1897; Liu, 1935; Inger and Greenberg, 1956) reported the occurrence of specimens with a single vocal slit. This condition was observed in some species of Bufonidae and has not been reported in other anuran families. The single vocal slit can either be on the left or the right side of the tongue in different specimens of the same species. Furthermore, there are species where one (on either side) or two vocal slits can occur. This character is scored as not applicable for taxa lacking vocal sacs (see char. 58.0).
Previous usage in phylogenetic studies: Drewes (1984: char. 20*), Cannatella (1986: char. 10*), Mendelson (1997a: char. 42*), Mendelson et al. (2000: char. 51*), Pramuk (2006: char. 75*), Mendelson et al. (2011: char. 31*).
60. Vocal sac, shape when fully inflated: (0) spherical or subspherical, (1) projected anteriorly. Simple subgular vocal sacs are often spherical or subspherical. Nevertheless, in a few species, they project anteriorly deviating from a spherical shape. The degree of projection ranges from a slight deformation to a large, vertically oriented lobe. McAllister (1961) reported on this variation in North American bufonids and their putative relationship with vocalization, but this character has not been used in phylogenetic studies. This character is not applicable for specimens that lack a vocal sac (char. 58.0)
61. Nuptial pads, occurrence in males: (0) absent, (1) present. Nuptial pads are sexually dimorphic structures that can be present in the fingers of males; their structure and diversity were recently studied (Luna et al., 2018).
Previous usage in phylogenetic studies: Liem (1970: char. 35), Heyer (1975: char. 3*), Scott (2005: char. 132*), Wiens et al. (2005: char. 100), Grant et al. (2006: char. 23), Ohler and Dubois (2006: chars. 16–18*), Ponssa (2008: char. 24*), Barrionuevo (2017: char. 69).
62. Nuptial pads, coloration: (0) light colored, (1) dark colored. Following Luna et al. (2018) we distinguished between dark- and light-colored nuptial pads, where “dark-colored” includes all tones of brown and black and “light-colored” includes beige/uncolored pads. These differences in coloration result from minor changes in the stratum corneum of the epidermis and are independent of the number of layers of this stratum (Luna et al., 2008). This character is not applicable for specimens that lack nuptial pads (char. 61.0).
Previous usage in phylogenetic studies: Ohler and Dubois (2006: char. 24*).
63. Manus, occurrence of webbing between fingers: (0) absent or poorly developed, (1) present, well developed.
Previous usage in phylogenetic studies: Wiens et al. (2005: char. 99), Pramuk (2006: char. 77*), Nussbaum and Wu (2007: char. 12*).
64. Pes, edge of foot webbing: (0) smooth, (1) serrated.
Previous usage in phylogenetic studies: Vélez-Rodriguez (2004b: char. 12*).
65. Tarsus, occurrence of tarsal fold: (0) absent, (1) present. A tarsal fold is a dermal fold on the medial-ventral surface of the foot, extending proximally from the inner metatarsal tubercle.
Previous usage in phylogenetic studies: Inger (1972: char. 31*), Heyer (1975: char. 6*), Scott (2005: char. 156*), Grant et al. (2006: char. 28), Ohler and Dubois (2006: char. 11*), Ponssa (2008: char. 19), Barrionuevo (2017: char. 77).
66. Relative size of adult females and males: (0) adult females similar in size or larger than adult males, (1) adult males much larger than adult females. As a first approximation, we consider only two states due to the occurrence of a more evident gap in size according to published data. However, a more detailed study of sexual dimorphism in Rhinella could help to partition these into more additional states.
Previous usage in phylogenetic studies: Scott (2005: char. 139), Fabrezi (2006: char. 78*), Ponssa (2008: char. 110*).
Larval External Morphology
67. Body, morphology of the peribranchial and abdominal regions: (0) absence of external modifications, (1) presence of bulging regions lateral to the oral disc, (2) occurrence of an abdominal sucker. Additive. Most species of Rhinella have lentic larvae that lack external modifications in the peribranchial and abdominal regions (state 0). Modifications in these regions are typical of some lotic forms (McDiarmid and Altig, 1999; Hoff et al., 1999) and two different states occur within Rhinella. Larvae of Rhinella rumbolli have a central depression delimited by bulbous lateral regions in the peribranchial zone (state 1). Moreover, some other species of the R. veraguensis Group have a well-developed abdominal sucker that is bounded anteriorly by the oral disc, and the lateral and posterior edges are free from the body (state 2). We consider the character states to represent an ordered series of transformation for which the states are considered as additive.
68. Body, dorsal coloration: (0) light brown, (1) dark brown, (2) sharply defined dark markings on pale ground. Nonadditive.
69. Caudal musculature, ocurrence of an unpigmented longitudinal stripe along the inferior edge in the caudal musculature: (0) absent, (1) present. An unpigmented longitudinal stripe along the inferior edge of the caudal musculature sometimes occur in the caudal musculature of larvae having a dark coloration of the tail.
Previous usage in phylogenetic studies: Mendelson et al. (2011: char. 38*).
70. Caudal musculature, occurrence of irregular transverse whitish stripes: (0) absent, (1) present. In some species of the Rhinella granulosa and R. veraguensis Groups there are irregular transverse whitish stripes of variable extension due to the absence of melanocytes contrasting with the general dark coloration of the dorsal musculature (see Blotto et al., 2014, for taxonomic distribution in Rhinella).
71. Oral disc, occurrence of submarginal papillae: (0) absent, (1) present.
Previous usage in phylogenetic studies: Grant et al. (2006: char. 91*), Barrionuevo (2017: char. 86*), Araujo-Vieira et al., (2019: char. 135).
72. Oral disc, number of posterior labial tooth rows: (0) two, (1) three.
Previous usage in phylogenetic studies: Hillis and de Sá (1988: char. 2*), Wiens et al. (2005: char. 122*), Grant et al. (2006: char. 94*), Ohler and Dubois (2006: char. 31*), Barrionuevo (2017: char. 90*), Araujo-Vieira et al. (2019: char. 141*).
73. Oral disc, condition of the labial tooth row A2: (0) complete, (1) divided.
Previous usage in phylogenetic studies: Mendelson et al. (2011: char. 37*).
74. Oral disc, condition of the labial tooth row P1: (0) complete, (1) divided.
Previous usage in phylogenetic studies: Wiens et al. (2005: char. 124), Araujo-Vieira et al. (2019: char. 142)
75. Vent tube, opening: (0) medial, (1) dextral.
Previous usage in phylogenetic studies: Grant et al. (2006: char. 96*), Barrionuevo (2017: char. 93), Araujo-Vieira et al., 2019 (char. 145).
Larval Chondrocranium
Oliveira et al. (2014) studied the chondrocranium of some species of Rhinella and reviewed the information available for other bufonids.
76. Otic capsule, larval crista parotica, occurrence of processus anterolateralis: (0) absent or indistinguishable, (1) poorly developed with a rounded aspect, (2) well developed with an acute appearance. Additive.
Previous usage in phylogenetic studies: Larson and de Sá (1998: char. j*), Haas (2003: char. 66*); Miranda et al. (2015: char. 61*).
77. Procesus ascendens, angle of attaching to the braincase: (0) obliquely attached, (1) perpendicularly attached.
Previous usage in phylogenetic studies: Larson and de Sá (1998: char. o*), Miranda et al. (2015: char. 65*).
78. Copula anterior, occurrence: (0) absent, (1) present.
Previous usage in phylogenetic studies: Haas (2003: char. 105), Pugener et al. (2003: char. 35 [larval characters]), Miranda et al. (2015: char. 73).
Embryonic Morphology
Vera Candioti et al. (2016) studied the early ontogeny and described the informative variation found in several species of Bufonidae. All the characters considered on embryonic morphology were described in detail in that publication.
79. Third pair of external gills, condition: (0) absent or indistinguishable, (1) short, (2) long. Additive.
80. Dorsal line of hatching glands: (0) short (cephalic region only), (1) long (beyond cephalic region).
81. Type of adhesive gland: (0) A, (1) B.
82. Time of division of adhesive gland: (0) slightly after the second-gill pair branches off before operculum at the gill base, (1) immediately before the gills reach their maximum development, (2) immediately after opercular fusion. Additive.
Natural History
83. Diel activity of adults: (0) diurnal, (1) nocturnal.
Previous usage in phylogenetic studies: Grant et al. (2006: char. 115*).
84. Habits: (0) terrestrial, (1) arboreal, (2) aquatic. Nonadditive. Some species of the Rhinella veraguensis Group are completely arboreal. We do not consider as arboreal the mostly terrestrial species that have the ability to climb up the vegetation to rest during the night (de Noronha et al., 2013).
Previous usage in phylogenetic studies: Grant et al. (2006: char. 114*).
85. Oviposition site: (0) aquatic, (1) terrestrial, (2) phytotelmata. Nonadditive. Following van Bocxlaer et al. (2010), terrestrial oviposition refers to eggs that are placed on the ground, in leaf litter, or under stones, and are exposed to little or no free water at the time of oviposition. Phytotelmata refers to any chambers in a plant that is used as oviposition site (e.g., water-filled nut, tree holes, leaf axils; see Lehtinen et al., 2004; Grant et al., 2006)).
Previous usage in phylogenetic studies: Faivovich (2002: char. 83*), Grant et al. (2006: char. 107*), Araujo-Vieira et al. (2019: char. 191*).
86. Structure of the spawn: (0) strings, (1) open clump, (2) mass, (3) strands. Nonadditive. Altig and McDiarmid (2007) reviewed in detail the terminology and diversity of arrangement of deposited eggs in Amphibia.
Previous usage in phylogenetic studies: Haas (2003: char. 141*).
87. Egg disposition in strings: (0) uniserial, (1) biserial, (2) multiserial. Nonadditive. Mature oocytes are surrounded by jelly layers as they are displaced through the different regions of the oviduct (Salthe, 1963; Altig and McDiarmid, 2007). The number and type of jelly layers are not well characterized in Rhinella (Pereyra et al., 2015), and there is no information about a direct relation between the diversity of strings and the eggs disposition within the string. Thus, we cannot infer a series of transformation and we consider this character as nonadditive.
88. Ovum pigmentation: (0) unpigmented, (1) animal pole pigmented.
Previous usage in phylogenetic studies: McDiarmid (1971: char. 42*), Grandison (1981: char. 4*), Cannatella (1986: char. 14*), Grant et al. (2006: char. 68*), Ohler and Dubois (2006: char. 29), Mendelson et al. (2011: char. 40*).
Cytogenetics
89. Nucleolar Organizer Regions, location: (0) terminal position of the short arms of the chromosome pair 1, (1) pericentromeric position of the long arms of the chromosome pair 1, (2) terminal position of the long arms of the chromosome pair 5, (3) terminal position of the long arms of the chromosome pair 6, (4) interstitial position of the short arms of the chromosome pair 7, (5) interstitial position of the long arms of the chromosome pair 10, (6) terminal position of the long arms of the chromosome pair 10, (7) interstitial position of the short arms of the chromosome pair 11. Nonadditive.
Previous usage in phylogenetic studies: Faivovich (2002: char. 82*).
RESULTS
Separate Analyses of Restricted Nuclear (rND) and Mitochondrial (rMD) Datasets
The parsimony analyses, reaching a stable consensus 10 times, retained 706 unique MPTs of length 1757 for the rND and one MPTs of length 11,436 for the rMD. Within the ingroup (i.e., Rhinella), the main incongruence between the rND and rMD analyses involved the position of the specimens of R. horribilis, which are deeply nested within the R. marina Group in the rND analysis, but were recovered as the sister clade of the R. marina + R. crucifer Groups in the rMD analysis (fig. 9). Based on these observations and previous published results (Pereyra et al., 2016a), we included the mitochondrial and nuclear genomes of R. bernardoi and R. horribilis as independent terminals in the TE analysis (see Discussion section for comments on the putative mitochondrial or nuclear introgression in these terminals and the rationale for the considerations of both genomes as independent terminals). Mitochondrial introgression and hybridization between R. diptycha and R. marina might have occured in the area south of the Amazon River (see Sequeira et al., 2011), but the evidence is not conclusive (see Vallinoto et al., 2017). For this reason, we did not include sequences of these species from this complex area. In appendix 2, we list the terminals considered in the TE analysis.
Total Evidence Analysis
Molecular data were included for all 320 terminals of 124 species, whereas phenotypic data were restricted to 106 specimens of 102 species (90 characters; ∼50 scores/terminal). The TE analysis using parsimony, reaching a stable consensus 10 times, retained 657 unique MPTs (length 25,399). One of the optimal topologies is shown in figures 10–14 (fig. 10 for outgroup relationships, figs. 11–14 for Rhinella relationships). A summary tree of Rhinella relationships to species level is shown in the supplementary data 2 (available at https://doi.org/10.5531/sd.sp.46). In depicting all unrefuted clades, we employ the strict consensus of the optimal phylogenetic hypotheses resulting from this TE analysis treating gaps as fifth state as the basis of our discussion of taxonomy. The results of the TE analysis considering gaps as missing data (see supplementary data 3.1–3.5, available at https://doi.org/10.5531/sd.sp.46) and the ML analysis (see supplementary data 4.1–4.5, available at https://doi.org/10.5531/sd.sp.46) were highly congruent with the TE analysis considering gaps as fifth state. The few differences between these hypotheses are discussed when relevant.
The MPTs resulting from the TE analysis recovered Rhinella as nonmonophyletic due to the position of R. ceratophrys that is the sister taxon of Rhaebo nasicus with strong support (JGC and JAF = 100%; see fig. 10). Among outgroups the strongly supported Anaxyrus + Incilius (JGC = 96%, JAF = 97%) is the sister clade of all the other species of Rhinella. The monophyly of the clade composed of these three genera is poorly supported (JGC = 63%, JAF = 73%). The species of Rhinella (excluding R. ceratophrys) are monophyletic, well supported (JGC and JAF = 98%), and grouped in two major clades. One of these is moderately supported (JGC = 88%, JAF = 92%) and includes the species of the former R. spinulosa Group (including R. gallardoi; see Discussion) and those of the R. granulosa, R . crucifer, and R. marina Groups (figs. 11, 12). The other is strongly supported (JGC and JAF = 99%) and composed of all the species from the nonmonophyletic R. veraguensis and R. margaritifera Groups, the former R. acrolopha Group (see Discussion section), and R. sternosignata (figs. 13, 14).
Uncorrected P-Distances
The patterns of UPDs found within each species group vary largely (see below), so we did not consider a single value as a threshold to delimit species, but each particular situation was considered in the context of the genetic distances found within each species group. Interspecific distances among all the species addressed by the taxonomic revision are presented in the Discussion section of each species group. Throughout the text the UPDs are expressed as percentage.
DISCUSSION
Systematics and Taxonomy
Relationships among Outgroups and Rhinella
Our outgroup sample was designed exclusively to provide a rigorous test of the monophyly of Rhinella and does not constitute a critical test of previously hypothesized relationships among other clades of Bufonidae (e.g., Frost et al., 2006;; Pramuk, 2006; Pramuk et al., 2008; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Portik and Papenfuss, 2015; Jetz and Pyron, 2018). Indeed, most of the basal relationships of Bufonidae are unresolved or poorly supported in the TE analysis (fig. 10). Nevertheless, we found Anaxyrus + Incilius to be the sister clade of Rhinella with low support (JGC = 63%, JAF = 73%). The clade composed of these three genera, in turn, is the sister taxon of a large and poorly supported clade (JGC <25%, JAF <50%) of African and Eurasian bufonids. An almost identical relationship was recovered in the ML analysis (see supplementary data 4.1). The sister-group relationship between Rhinella and Anaxyrus + Incilius is consistent with the results of most previous phylogenetic analyses (e.g., Pramuk, 2006: fig. 4; Pramuk et al., 2008; Pyron and Wiens, 2011; Pyron, 2014: suppl. information “amph_shl. tre”; Portik and Papenfuss, 2015; Jetz and Pyron, 2018: suppl. information “amph_shl_new.tre”). Alternatively, van Bocxlaer et al. (2010: fig. S1) recovered Rhinella as the sister taxon of a clade comprising all African and Eurasian bufonids.
Although the vast majority of species of Rhinella form an exclusive clade, it is polyphyletic because R. ceratophrys was recovered as the sister taxon of Rhaebo nasicus with strong support (JGC and JAF = 100%). This relationship is not surprising, given that the morphological resemblance between both species was pointed out previously (e.g., Hoogmoed, 1977; Fenolio et al., 2012). Although Rhaebo was paraphyletic in our TE analysis (fig. 10; but see results of the ML analysis in supplementary data 4.1), our taxon sampling was not designed to test its monophyly. Thus, we transfer Rhinella ceratophrys to Rhaebo as Rhaebo ceratophrys (Boulenger, 1882), new combination.
Rhinella and Its Internal Relationships
In the parsimony total evidence analysis, Rhinella was recovered as monophyletic (after transferring R. ceratophrys to Rhaebo) and well supported (JGC and JAF = 98%). The monophyly of Rhinella was previously recovered by several phylogenetic studies that used fewer taxa (e.g., Pauly et al., 2004: fig. 4; Pramuk, 2006; Pyron and Wiens, 2011; Pyron, 2014: suppl. information “amph_shl.tre”; Portik and Papenfuss, 2015; Jetz and Pyron, 2018: suppl. information “amph_ shl_new.tre”). In contrast to all previous studies, we found that Rhinella is composed of two major, well-supported clades (figs. 11–14; see below). Our results support the R. crucifer, R . granulosa, and R. marina Groups as monophyletic. Otherwise, the R. spinulosa Group is recovered paraphyletic due to the nested position of R. gallardoi (a species unassigned to any group). The R. margaritifera Group is polyphyletic due to the position of the former R. ceratophrys nested in Rhaebo. The R. veraguensis Group is polyphyletic due to the position of several taxa (i.e., R. arborescandens, R . chavin, R . lilyrodriguezae, R . manu, R . multiverrucosa, R . nesiotes, R . tacana, and R. yanachaga) more closely related to the R. margaritifera Group, and with the monophyletic R. acrolopha Group nested within them. The ML analysis of the molecular + phenotypic datasets supported most of these results (supplementary data 4.2–4.5), and we only discuss the relevant differences between analyses. Below, we provide a revised account and comments for Rhinella and its main clades and species groups on the basis of these results.
Rhinella
Diagnosis: The long third pair of external gills (char. 79.2) optimizes as the only phenotypic synapomorphy of Rhinella in all the MPTs, which reverts to short third pair of external gills, the plesiomorphic bufonid condition, in the R. granulosa Group. An unequivocal diagnosis of this genus is obscured by the large phenotypic variation within Rhinella that overlaps with the diversity of many of the related bufonid genera. Nevertheless, this genus can be diagnosed from most of the related bufonids by the combination of the following phenotypic characters: (1) nasals and frontoparietal heavily ornamented with pits, striations, and rugosities (char. 9.2); (2) presence of a row of dorsolateral tubercles on skin (char. 57.1); and (3) nucleolar organizer regions (NORs) located on interstitial position of the short arms of the chromosome pair 7 (char. 89.4).
Sister clade: The well-supported clade composed of Anaxyrus + Incilius (JGC = 96%, JAF = 97%).
Distribution: Mostly Neotropical, ranging from the southern United States to southern South America. Rhinella marina is a highly invasive species introduced in many countries and islands outside its native distribution (e.g., Antilles, Australia, Hawaii, Philippines, Taiwan, etc.; see Frost, 2020; IUCN, 2020)).
Comments: The exclusion of the former Bufo ceratophrys renders Rhinella monophyletic. The two large clades of Rhinella were not recovered in previous phylogenetic analyses (e.g., Pramuk, 2006: fig. 4; Chaparro et al., 2007; Pramuk et al., 2008; van Bocxlaer et al., 2010: suppl. information S1; Pyron and Wiens, 2011; Pyron, 2014: supp. information “amph_shl.tre”; Portik and Papenfuss, 2015; Pereyra et al., 2016a; Jetz and Pyron, 2018: suppl. information “amph_shl_new. tre”). Vera Candioti et al. (2016) proposed the long third pair of external gills as a putative synapomorphy of Rhinella in the context of a review of embryonic morphology of Bufonidae. Our TE analysis supports this character state as synapomorphy of the genus, although the embryonic morphology of many genera of Bufonidae and species of the R. margaritifera Clade (see below) is unknown. This synapomorphy of Rhinella reverts to the plesiomorphic state (short third pair of external gills) in the R. granulosa Group.
As a result of our TE analysis (also see ML result), we define two major clades, the Rhinella marina Clade and the R. margaritifera Clade, composed of eight species groups within Rhinella. The R. marina Clade includes (1) the R. arunco Group (new species group); (2) the R. crucifer Group; (3) the R. granulosa Group; (4) the R. marina Group; and (5) the R. spinulosa Group as redefined here. The second clade, the R. margaritifera Clade, is composed of (1) R. sternosignata, a species unassigned to any group; (2) the R. festae Group as redefined here; (3) the R. margaritifera Group as redefined here; and (4) the R. veraguensis Group as redefined here. Below, we provide diagnoses, content, and comments on the distribution and systematics of each of the newly defined major clades and all species groups of Rhinella. The clades and species group are presented in the order described above and correspond to the sequence in which they appear in the TE tree (figs. 10–14) from base to tip and top to bottom.
The Rhinella marina Clade (figs. 11, 12)
Diagnosis: This clade is moderately supported (JGC = 88%, JAF = 92%) and diagnosed by a phenotypic synapomorphy: larval otic capsule with poorly developed processus anterolateralis with a rounded aspect (char. 76.1), with one instance of homoplasy in Sclerophrys regularis.
Sister clade: The Rhinella margaritifera Clade (figs. 13, 14).
Contents: The Rhinella marina Clade is composed of the R. crucifer, R . granulosa, and R. marina Groups, the R. spinulosa Group as redefined here, and the R. arunco Group, a new group defined here (see below). Moreover, we found a divergent mitochondrial lineage introgressed into R. horribilis (hereafter referred to as GIM [ghost introgressed mitochondrion], see below and discussion) that does not seem to belong to any recognized extant species of Rhinella and was recovered as sister clade of the R. marina + R. crucifer Groups (see fig. 12), although with poor support (see below).
Distribution: The species of this clade naturally occur in all main biogeographic regions of the Neotropics.
Comments: The Rhinella marina Clade is composed of two subclades. One is poorly supported (JGC = 68%, JAF = 82%) and includes the R. arunco + R. spinulosa Groups (fig. 11). It is diagnosed by four phenotypic synapomorphies: (1) the supraorbital flange on frontoparietal does not extend laterally beyond the lateral margin of the sphenethmoid (char. 17.0, with instances of homoplasy in R. quechua and some outgroups); (2) the m. extensor digitorum on the metacarpophalangeal joint of digiti III (char. 44.1, with instances of homoplasy in Anaxyrus woodhousii [polymorphic], Rhinella hoogmoedi, R . jimi, and R. rumbolli); (3) parotoid gland round to ovoid, mostly symmetrical (char. 56.2, with instances of homoplasy in R. bergi and several species of the R. margaritifera Clade); and (4) vocal sac absent in adult males (char. 58.0, with instances of homoplasy within Rhinella and outgroups). The other subclade is well supported (JGC and JAF = 99%) and includes the R. crucifer, R . granulosa, and R. marina Groups, and the GIM (figs. 11, 12). Three phenotypic synapomorphies are recovered for this subclade: (1) occurrence of a well-developed supraorbital crest (char. 1.2, with instances of homoplasy in several bufonids); (2) occipital artery pathway completely covered with bone (char. 10.2, with instances of homoplasy in bufonids); and (3) general pattern of coloration of caudal musculature of larvae uniformly dark except an unpigmented longitudinal stripe along the inferior edge (char. 69.1, with instances of homoplasy in R. quechua, R . veraguensis, and some outgroups).
Previous phylogenetic studies including less complete sampling of Rhinella (Pramuk, 2006; Pramuk et al., 2008; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Portik and Papenfuss, 2015; Pereyra et al., 2016a; Jetz and Pyron, 2018) never found a sister relation between the clade composed of the R. arunco + R. spinulosa Groups and the clade composed of the R. granulosa + (R. crucifer + R. marina) Groups. Instead, these studies found the R. arunco and R. spinulosa Groups as: (1) the sister clade of the species of the R. margaritifera Clade as defined here (Pramuk, 2006), (2) as sister clade of the remaining species of Rhinella (Pramuk et al., 2008; Pereyra et al., 2016a), or (3) as successive sister clades of the remaining species of Rhinella (van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Portik and Papenfuss, 2015; Jetz and Pyron, 2018). In contrast to all these analyses, our ML analysis recovers the R. arunco Group as the sister clade of the remainder of the R. marina Clade, whereas the R. spinulosa Group is the sister taxon of the clade composed of R. granulosa + (R. crucifer + R. marina) Groups. This last clade has always been recovered as monophyletic in previous phylogenetic analyses (Pramuk, 2006; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Portik and Papenfuss, 2015; Pereyra et al., 2016a; Jetz and Pyron, 2018).
The Rhinella arunco Group
Diagnosis: Two phenotypic synapomorphies diagnose this strongly supported (JGC and JAF = 100%) species group: (1) jaw articulation opposite to the fenestra ovalis (char. 16.1, with instances of homoplasy in some species of the Rhinella granulosa Group, the R. margaritifera Clade, and in Nannophryne variegata); and (2) anterior edge of sacral diapophyses perpendicular to the midline axis of the vertebral column (char. 30.1, with instances of homoplasy in R. crucifer, R . quechua, R . rubescens, R . spinulosa, and R. vellardi). The presence of an insertion of the m. extensor digitorum longus on metatarsophalangeal joint of digit V (char. 40.1) and the presence of an insertion of the m. extensor digitorum on metacarpophalangeal joint of digiti V (char. 45.1) could represent two additional synapomorphies of this group or an internal clade. Moreover, species of the R. arunco Group can be distinguished from members of the other species groups of Rhinella by the following combination of character states: (1) preorbital crest weak (char. 0.1), (2) occipital artery pathway uncovered with bone (char. 10.0), (3) frontoparietal that does not extend laterally beyond the lateral margin of the sphenethmoid (char. 17.0), (4) medial ramus of the pterygoid fused with the anterolateral margin of the parasphenoid (char. 21.1), (5) m. extensor digitorum longus with an insertion on the metatarsophalangeal joint of the digit IV (char. 39.1), (6) m. extensor digitorum with an insertion on the metacarpophalangeal joint of digiti III (char. 44.1), (7) inguinal fat bodies present (char. 51.1), (8) row of dorsolateral tubercles absent (char. 57.0), (9) vocal sac absent in adult males (char. 58.0), and (10) eggs biserially disposed in strings (char. 87.1).
Sister clade: The Rhinella spinulosa Group.
Contents (3 species): Rhinella arunco (Molina, 1782), R. atacamensis (Cei, 1962), and R. rubropunctata (Guichenot, 1848).
Distribution: Species of the Rhinella arunco Group are distributed in Argentina and Chile: Rhinella arunco and R. atacamensis in the Atacama Desert region, R. rubropunctata in the Austral Temperate Forest region (Cei, 1962, 1980; Correa et al., 2013). See map 1 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: Pramuk (2006) found the Rhinella spinulosa Group (sensu Duellman and Schulte, 1992) as nonmonophyletic and excluded the species now placed in Nannophryne (i.e., N . apolobambica, N . cophotis, N . corynetes, and N . variegata; see Frost et al., 2006;; Frost, 2020). The remaining species constituted a well-supported clade in her combined (morphological and molecular) analysis, being the sister taxon to all the remaining species of Rhinella. A subsequent molecular phylogeny (Pramuk et al., 2008) considering a similar taxon sampling and mitochondrial genes, but several different nuclear genes with respect to Pramuk (2006; cxcr4 and rag1-a vs pomc and rag1-a), recovered this redelimited group as monophyletic with poor support. Previous and subsequent analyses with slightly increased taxon and gene sampling, however, found this group as paraphyletic with respect to all remaining species groups of Rhinella (Frost et al., 2006;; van Bocxlaer et al., 2010; Pyron and Wiens, 2011; Pyron, 2014; Portik and Papenfuss, 2015; Jetz and Pyron, 2018), or as the (poorly supported) sister taxon of all other species of Rhinella (Pereyra et al., 2016a). In our TE analysis, the former R. spinulosa Group (including R. gallardoi, see below) was recovered as monophyletic but poorly supported (JGC = 68%, JAF = 82%). Moreover, the individual monophyly of its sister subclades is strongly supported (both with JGC and JAF = 100%) and can be diagnosed by phenotypic synapomorphies (see Diagnosis of both groups). Our ML analysis found the former R. spinulosa Group paraphyletic with respect to the remaining species groups of the R. marina Clade (supplementary data 4.2). Based on these observations, we restrict the R. spinulosa Group to the strongly supported clade containing most species of the former R. spinulosa Group (and including R. gallardoi), and exclude the extra-Andean species R. arunco, R . atacamensis, and R. rubropunctata that constitute another well-supported clade, herein recognized as the R. arunco Group. The southernmost distributed species R. arunco and R. rubropunctata are recovered as sister taxa, although with poor support (JGC = 25, JAF < 50%). The three species of this group show a high genetic differentiation in comparison to other species groups of the R. marina Clade (see tables 3–6). Natural hybridization between R. arunco and R. atacamensis was reported by Correa et al. (2012, 2013), but they did not find mitochondrial and nuclear introgression outside a narrow hybrid zone.
The Rhinella spinulosa Group
Diagnosis: The following character states optimize as phenotypic synapomorphies of this strongly supported group (JGC and JAF = 100%) in our TE analysis: (1) pretympanic crest absent or indistinguible (char. 2.0, with instances of homoplasy in Rhinella arunco, R . castaneotica, R . festae, and some outgroups); (2) nasal and frontoparietal bones articulating only laterally (char. 8.1, homoplastic in R. quechua, R . rubropunctata, R . veraguensis, R . yanachaga, Rhinella sp. 14, and some outgroups); (3) lightly exostosed dermal roofing bones (char. 9.1, homoplastic in the R. festae Group, in several species of the R. marina Group, and outgroups); and (4) slightly enlarged otic ramus of squamosal, overlapping with the dorsal surface of the crista parotica (char. 11.1). In addition, species of the R. spinulosa Group can be distinguished from members of the other species groups of Rhinella by the following combination of character states: (1) occipital artery pathway not covered by bone (char. 10.0), (2) frontoparietal that does not extend laterally beyond the lateral margin of the sphenethmoid (char. 17.0), (3) medial ramus of the pterygoid fused with the anterolateral margin of the parasphenoid (char. 21.1), (4) m. extensor digitorum longus with an insertion on metatarsophalangeal joint of digiti IV (char. 39.1), (5) m. extensor digitorum with an insertion on the metacarpophalangeal joint of digiti III (char. 44.1), (6) inguinal fat bodies present (char. 51.1), (7) multiserial configuration of eggs in the jelly string (char. 87.2), (8) tarsal fold present (char. 65.1), and (9) adhesive gland divided after fusion of the operculum in embryo (char. 82.2).
Sister clade: The Rhinella arunco Group.
Contents (9 species): Rhinella achalensis (Cei, 1972b)), R. altiperuviana (Gallardo, 1961)) new status, R. amabilis (Pramuk and Kadivar, 2003), R . gallardoi (Carrizo, 1992), R. limensis (Werner, 1901), R. papillosa (Philippi, 1902), new status, R. spinulosa (Wiegmann, 1834) [including R. arequipensis (Vellard, 1959), new synonymy, see below], R. trifolium (Tschudi, 1845) new status, and R. vellardi (Leviton and Duellman, 1978).
Distribution: This species group is mostly distributed in arid regions along the Andes of Argentina, Bolivia, Ecuador, Chile, and Peru, except Rhinella gallardoi that inhabits the humid subandean forest of Argentina (Vellard, 1959; Córdova, 1999; Pramuk and Kadivar, 2003; Lavilla and Cei, 2001). Rhinella achalensis and R. limensis are the only species of this group with an extra-Andean distribution in the Sierras Pampeanas Centrales in the Pampas region of Argentina and Atacama Desert of Peru respectively (Vellard, 1959; Cei, 1972b)). See map 2 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: The Rhinella spinulosa Group as redelimited here is composed of some taxa with controversial taxonomies that are discussed in the context of our results. The widespread, polytypic, and poorly defined species R. spinulosa is recovered as nonmonophyletic, with R. achalensis, R . gallardoi, and R. arequipensis nested within it (fig. 11). Based on our results and considering that “Peru” is the type locality of R. spinulosa (and putatively confined to southern Peru, see Vellard, 1959), we restrict the species R. spinulosa s.s. to the well-supported lineage (JGC = 95%, JAF = 97%), composed of the populations from southern Peru and northern Bolivia. The lineage containing these populations of R. spinulosa also includes the sampled specimen of R. arequipensis from Departamento Arequipa, Peru. Rhinella arequipensis was originally described as a subspecies of R. spinulosa based only on differences in coloration and density of granular formations in the dorsal tegument (Vellard, 1959). Morrison (1992, 1994), Córdova (1999), and Aguilar and Gamarra (2004) did not find morphological, osteological, karyological, or larval differences that could discriminate between R. spinulosa and R. arequipensis. According to these observations and our results, we consider Bufo spinulosus arequipensis Vellard, 1959, a junior synonym of Rhinella spinulosa (Wiegmann, 1834). Thus, the species R. spinulosa is restricted to the populations distributed mainly along the Andean Puna of Peru and adjacent Bolivia.
Populations of Rhinella spinulosa that had been considered as R. s . trifolium were recovered as a distinct and strongly supported lineage (JGC and JAF = 98%) sister to a poorly supported clade (JGC <25%, JAF <50%) containing R. spinulosa s.s. and several other species of the group (see below). There are several morphological differences between R. s . trifolium and R. spinulosa s.s. Vellard (1959) pointed out the disposition of the dorsal glands (longitudinal rows in R. s . trifolium and a uniform distribution in R. s . spinulosa) and the occurrence of a middorsal vertebral line in R. s . trifolium, as the main distinguishing characters. Morrison (1992, 1994), Sinsch (1986), Haas (2002), and Pramuk and Kadivar (2003) considered R. spinulosa s.s . and R. s . trifolium (and also R. s . flavolineata) as variations of a single species (see below), although all but Haas failed to provide detailed justification. The morphological comparisons were some superficial and a detailed reevaluation of the specimens and comparisons with topotypes is needed. Córdova (1999) and Aguilar and Gamarra (2004) did not find karyological or larval differences between R. s . spinulosa and R. s . trifolium; however, these character systems are conserved in related species of Rhinella (see Tolledo and Toledo, 2010; Kolenc et al., 2013; Blotto et al., 2014). The UPDs between the specimens of R. s . trifolium and R. spinulosa s.s. are relatively high for this species group (1.11%–1.30%, see table 4). Consequently, the differences in adult morphology proposed by Vellard (1959) and their genetic divergence support the recognition of Rhinella trifolium (Tschudi, 1845) as a distinct species.
Some populations currently assigned to Rhinella spinulosa s.l. from Jujuy (Argentina) and La Paz (Bolivia) were recovered as another distinct and strongly supported lineage (JGC and JAF = 100%; see fig. 4) with a low UPD between them (0.18%). In the intermediate area of Puna between these localities (∼ 800 km) lays the type locality of R. s . altiperuviana (Challapata, Oruro, Bolivia). Gallardo (1961) described this subspecies from two adult females; the characters used to differentiate it from R. spinulosa s.l. (i.e., tubercles structure, head shape, tarsal fringe development) show considerable variation, at least, in the studied female specimens from northwestern Argentina. Thus, we tentatively assign these populations to R. s . altiperuviana. In addition to the phylogenetic position, these specimens differ in UPDs (see table 4) and adult and larval external morphology (B.L.B., D.B., M.O.P., personal obs.) from other species of the group. For these reasons, these populations should be considered as a distinctive species, R. altiperuviana (Gallardo, 1961)) from the Andean Puna of Argentina and Bolivia. A detailed taxonomic revision is beyond the scope of this work but will be discussed in a subsequent contribution (B.L.B. and M.O.P., in prep.).
Populations of Rhinella spinulosa that had been considered as R. s . papillosa are recovered as a strongly supported lineage (JGC and JAF = 100%), sister taxon of R. achalensis. Both taxa differ in UPDs (1.10 to 1.47%, see table 4), and are morphologically differentiable from R. spinulosa s.s. (B.L.B. and M.O.P., in prep.). Thus, we consider R. papillosa (Philippi, 1902), a valid species from the austral Andes of Argentina and Chile.
Rhinella gallardoi is deeply nested within the R. spinulosa Group. In the original description, Carrizo (1992) highlighted the “broad skull” of this species over the general morphological similarity with the species of the R. spinulosa Group and assigned it to the “Bufo veraguensis-typhonius” complex. Moreover, R. gallardoi is the only species of the R. spinulosa Group inhabiting exclusively the Yungas of the Andes in northwestern Argentina.
TABLE 3
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella arunco Group
Values reported are mean (range).
The specimens of Rhinella amabilis, R . limensis, and R. vellardi are recovered as a strongly supported clade (JCG and JAF and = 100%), which is the sister taxon of the remaining species of the R. spinulosa Group as redefined here. Within this clade, the specimen of R. amabilis collapses into a basal polytomy with the specimens of R. limensis (the UPDs within the clade composed of these specimens are low [0.19%–0.44%]). Rhinella amabilis was differentiated from R. limensis by a few characters (development of cranial crests, presence of vocal slits, and shape of the parotoid gland). Rhinella amabilis is the only species of the R. spinulosa Group distributed north of the Huancabamba depression (Loja, Ecuador), and we could not obtain tissues from this area. The only specimen sampled of this species comes from a locality in the Huancabamba depression region but does not fully correspond with the morphological description of the species. An extensive revision of both species, including topotypical material and comparison with the holotypes is necessary to test the validity of R. amabilis.
The currently recognized subspecies Rhinella spinulosa flavolineata was not included in our analyses. This subspecies differs from R. trifolium only in the conspicuity and time of emergence of the vertebral line. Haas (2002) studied the development of specimens he assigned to the subspecies R. s . spinulosa, R . s . trifolium, and R. s . flavolineata from the same locality (Man-taro valley, between Concepcion and Huancayo, Junin department, Peru). This author reported that juveniles with variable development (or even absence) of this vertebral line could be obtained from a single clutch, hence, this character seems not to be relevant in differentiating these taxa. The occurrence of R. spinulosa s.s. in that locality is debatable (see Vellard, 1959, for comments on the distributions of these taxa) and it is possible that Haas (2002) assigned specimens of R. trifolium with poorly defined vertebral line to R. spinulosa s.s. (see Haas, 2002: fig. 1). In any case, the results of that study demonstrate that the tempo and level of development of the vertebral line are highly variable. Considering that the different morphs found by this author correspond to intraspecific variation within R. trifolium, we consider Bufo spinulosus flavolineatus Vellard, 1959, a junior synonym of Rhinella trifolium (Tschudi, 1845). Rhinella trifolium is considered to inhabit the Central Andean Wet Puna (Vellard, 1959), but additional studies are necessary to determine the precise limits of its geographic distribution and variation with respect to R. spinulosa s.s.
The Rhinella granulosa Group
Diagnosis: This species group is recovered as monophyletic with strong support (JGC and JAF = 100%) as in previous analyses (Pramuk, 2006; Pereyra et al., 2016a). Four phenotypic synapomorphies are recovered for this group: (1) anterior end of the septomaxilla developed (previously considered to be the prenasal bones; see discussion of this character in List and Description of Characters) (char. 14.1); (2) sacral diapophyses with the maximum width greater than its maximum length (char. 29.1), with several instances of homoplasy in Rhinella and outgroups; (3) submarginal papillae in the larval oral disc absent (char. 71.0), with instances of homoplasy in several bufonids; and (4) two posterior labial tooth rows in the larval oral disc (char. 72.0), that revert in an internal clade of this group. Moreover, nine additional characters might represent synapomorphies of this group or an internal clade depending on their occurrence in R. bernardoi and R. dorbignyi, where they are still unknown: (1) anteriorly oriented alary process of the premaxilla (char. 13.2), which also optimizes as a synapomorphy of the R. margaritifera Clade and is homoplastic in Incilius coniferus, Schismaderma carens, and some species of the R. marina Clade; (2) articulation of the zygomatic ramus of the squamosal with the maxilla (char. 15.1), homoplastic in Peltophryne lemur and R. sternosignata; (3) articulation of the jaw anterior to the fenestra ovalis (char. 16.2), homoplastic in Melanophryniscus gr. stelzneri and Peltophryne lemur; (4) bony sphenethmoid reaching the level of the premaxillae anteriorly (char. 18.2); (5) posterior lobe in the anterolateral process of hyoid absent (char. 25.0), homoplastic in Rhaebo ceratophrys, Rhinella acrolopha, and in the R. margaritifera Group; (6) vocal sac projected anteriorly when fully inflated (char. 60.1), homoplastic in some species of Anaxyrus; (7) short third pair of gills in the embryos (char. 79.1), homoplastic in Melanophryniscus gr. stelzneri and Schismaderma carens; (8) short dorsal line of hatching glands in the embryos (char. 80.0), with an instance of homoplasy in R. marina; and (9) type-A adhesive glands in the embryos (char. 81.0).
TABLE 4
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella spinulosa Group
Values reported are mean (range).
The species of the Rhinella granulosa Group can be distinguished from members of the other groups of Rhinella by the following combination of character states: (1) preorbital crest well developed (char. 0.2), (2) supraorbital crest well developed (char. 1.2), (3) pretympanic crest well developed (char. 2.2), (4) nasal and frontoparietal articulate along the entire margin (char. 8.3), (5) occipital artery pathway completely covered with bone (char. 10.2), (6) medial ramus of the pterygoid fused and extending medially along approximately half the length of the parasphenoid alae (char. 21.2), (7) anterior edge of sacral diapophyses perpendicular to the midline axis of the vertebral column (char. 30.1), (8) inguinal fat bodies present (char. 51.1), (9) tarsal fold absent (char. 65.0), (10) caudal musculature of larvae uniformly dark except an unpigmented longitudinal stripe along the inferior edge (char. 70.1), (11) occurrence of irregular transverse whitish stripes in the caudal musculature of larvae (char. 70.1), (12) short third gill pair in the embryo (char. 79.1), and (13) adhesive gland divides immediately before the gills reach their maximum development (char. 82.1).
Sister clade: The clade composed of the GIM (see below) and the Rhinella crucifer and R. marina Groups.
Contents (13 species): Rhinella azarai (Gallardo, 1965); R. beebei (Gallardo, 1965); R. bergi (Céspedez, 2000); R. bernardoi Sanabria et al., 2010; R. centralis Narvaes and Rodrigues, 2009; R. dorbignyi (Duméril and Bibron, 1841) [including R. fernandezae (Gallardo, 1957) new synonymy, see below]; R. granulosa (Spix, 1824); R. humboldti (Gallardo, 1965); R. major (Müller and Hellmich, 1936); R. merianae (Gallardo, 1965); R. mirandaribeiroi (Gallardo, 1965); R. nattereri (Bokermann, 1967); and R. pygmaea (Myers and Carvalho, 1952).
Distribution: Species of this group are widely distributed in open areas of Amazonia, Atlantic Forest, Caatinga, Cerrado, Chaco/Pantanal, Chocó, and Pampas regions and in Panama (Narvaes and Rodrigues, 2009; Sanabria et al., 2010; Pereyra et al., 2016a; Murphy et al., 2017). See map 3 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: In a previous molecular phylogenetic analysis of the Rhinella granulosa Group, Pereyra et al. (2016a) recognized 12 putative phenotypic synapomorphies for the group (three of these were first proposed by Pramuk, 2006, and one by Blotto et al., 2014). Eight of these character states were included as part of homology hypotheses (characters) in our TE analysis and only three were corroborated as synapomorphies of the R. granulosa Group in all the MPTs (all the remaining were recovered as synapomorphies in some MPTs). The remaining four character states (i.e., the presence of an expanded flag-shaped dorsal crest of the ilium in lateral view; nasal bone articulates with the dorsal margin of the pars facialis of the maxilla from the preorbital process to the posterior margin of the narial opening; occipital condyles widely separated; and ability to build and inhabit holes in the ground) were not included due to the lack of detailed descriptions or preparations for many species of Rhinella. However, these character states are unique of the R. granulosa Group among the most closely related groups and are consequently considered putative synapomorphies of this group.
Taxonomic, genetic, and biological aspects of the Rhinella granulosa Group were addressed in detail by Pereyra et al. (2016a), but some differences need to be stressed. First, we found variations in the inferred relationships among the earlier diverging clades/species of this group. Our TE analysis recovered a basal polytomy that comprises: (1) R. bernardoi, (2) R. dorbignyi (including R. fernandezae, see below), and (3) a poorly supported clade (JGC = 66%, JAF = 74%) composed of the remaining species of the group. Pereyra et al. (2016a) found R. major to be the sister species of a poorly supported clade (JAF <50, no JGC value reported) comprising all the remaining species of the group. Although in both analyses the interspecific relationships are poorly supported in general, we presume that these differences are due to the inclusion of phenotypic characters, the inclusion of sequences of R. humboldti, the denser outgroup sampling in this study, and the inclusion of a contaminated fragment of cytochrome b (KP684992; contaminated with R. icterica) in the dataset of Pereyra et al. (2016a).
Pereyra et al. (2016a) retained Rhinella dorbignyi and R. fernandezae as different species, although they noted the absence of reciprocal monophyly between both taxa and the very low genetic distances among the sampled specimens. Although we did not add additional specimens or sequences to our analyses (but a set of phenotypic characters was added in the TE analysis) and we recovered the same topology as Pereyra et al. (2016a) for the clade containing both taxa, we consider Bufo granulosus fernandezae Gallardo, 1957, a junior synonym of Rhinella dorbignyi (Duméril and Bibron, 1841). This decision is consistent with the criteria followed to synonymize other taxa of Rhinella (i.e., absence of reciprocal monophyly, absence of genetic differentiation, and absence of conspicuous differential morphological characters). Different populations of R. dorbignyi s.s. vary only in the level of development of the cranial crest, but not in other phenotypic or molecular characters. We hypothesize that local environmental factors through the area of distribution (i.e., Espinal, Humid Chaco, Humid Pampa, and Uruguayan Savanna) could affect the levels of ossification in the skull, resulting in differential development of cranial crests. The genetic and environmental causes of hyperossification are still not well understood in anurans (Paluh et al., 2020; Blotto et al., 2021). The differential patterns of bone deposition on the skull of R. dorbignyi are drastic and generate large morphological differences, making this species an excellent candidate to explore the role and impact of environmental factors on hyperossification.
We recovered Rhinella humboldti as distinct from R. beebei, as obtained by Murphy et al. (2017). However, both specimens of R. humboldti collapse in a polytomy together with the well supported R. centralis (JGC and JAF = 99%). Both taxa seem to differ in several morphological characters (Narvaes and Rodrigues, 2009; although these authors considered R. beebei and R. humboldti as a single taxon) and the UPDs between the specimens of both species are 1.04%–1.37% (see table 5). The poor internal resolution of this clade could be due to the reduced gene sampling for both specimens of R. humboldti (see appendix 2). However, a thorough analysis including additional molecular markers and morphological comparisons with R. humboldti s.s. is necessary to test the validity of R. centralis.
The Mitochondrial Lineage of Rhinella horribilis
The included mitochondrial sequences of Rhinella horribilis together with the R. crucifer + R. marina Groups constitute a strongly supported clade (JGC and JAF = 100%) in the TE analysis. Within this clade, they are recovered as sister taxon of a poorly suported clade (JGC = 56%, JAF = 72%) formed by the two aforementioned groups. Alternatively, this lineage is recovered in the ML analysis as the sister of the R. crucifer Group, with low support (44% ultrafast bootstrap support value). This clade is, in turn, sister to the R. marina Group (supplementary data 4.3). As we discuss below (see “Hybridization and genetic introgression in Rhinella”), the strong phylogenetic incongruence between mitochondrial and nuclear sequences of all the sampled specimens of R. horribilis is the result of a past hybridization with introgression event in which R. horribilis incorporated this mitochondrial lineage and completely replaced the original mtDNA of this species. We hypothesize that this mitochondrial lineage corresponds to a still unknown, or perhaps even extinct species of Rhinella, as we could not associate it to any of the 92 included species. In addition, two well-supported lineages are genetically differentiated within this mitochondrial clade according to the tree topology and proportionately large genetic distances (mean UPD = 4.19%, table 6): one includes most populations of R. horribilis from Colombia and Central America, which we associate to R. horribilis s.s., whereas the second lineage includes populations of R. horribilis from Ecuador that represent an undescribed species (Rhinella sp. 1). This structure is not recovered by the nuclear sequences of R. horribilis because they collapse in polytomy.
The Rhinella crucifer Group
Diagnosis: This species group was recovered as monophyletic and well supported (JGC and JAF = 100%), as in previous studies (Maciel et al., 2006; Thomé et al., 2010, 2012). Three phenotypic characters states optimize as synapomorphies of the Rhinella crucifer Group: (1) insertion of the m. extensor digitorum longus on the metatarsophalangeal joint of digiti IV absent (char. 39.0), which is homoplastic in a subclade of the R. granulosa Group, in the R. margaritifera Clade, and in some of the earlier-diverging bufonids; (2) lateral m. dorsometatarsalis proximalis digiti IV with a discrete tendon inserting on the proximal interphalangeal joint of digiti IV (char. 42.1), with an instance of homoplasy in Nannophryne variegata; and (3) the occurrence of a vertebral line (char. 54.1), with several instances of homoplasy within Rhinella. Other additional character states that could optimize as a synapomorphy of this group or an internal clade, depending on their occurence in R. casconi and R. henseli, that are still unknown: (1) dorsal protuberance of the illium small, low, and laterally projected (char. 32.1; condition within the group known only in R. crucifer); and (2) inguinal fat bodies absent (char. 51.0), with instances of homoplasy in R. achavali, R . rumbolli, in a subclade of the R. festae Group, in the R. margaritifera Group, and in several sampled outgroups.
Species of the Rhinella crucifer Group can be distinguished from members of the other groups of Rhinella by the following combination of character states: (1) supraorbital crest well developed (char. 1.2), (2) pretympanic crest weak (char. 2.1), (3) nasal and frontoparietal articulate along most of its margin but not completely (char. 8.2), (4) occipital artery pathway completely covered with bone (char. 10.2), (5) medial ramus of the pterygoid fused medially along approximately half the length of the parasphenoid ala (char. 21.2), (6) head of the m. extensor carpi ulnaris from the radioulna with an origin via a flat tendon (char. 47.1), (7) parotoid gland approximately ellipsoid (char. 56.0), (8) tarsal fold present (char. 65.1), (9) caudal musculature of larvae uniformly dark except an unpigmented longitudinal stripe along the inferior edge (char. 69.1), and (10) adhesive gland of the embryo divides after opercular fusion (char. 82.2).
TABLE 5
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella granulosa Group
Values reported are mean (range).
TABLE 6
Percentage of uncorrected p-distances between 16S sequences among terminals of the ghost introgressed mitochondrion
Values reported are mean (range).
Sister clade: The Rhinella marina Group.
Contents (5 species): Rhinella casconi Roberto et al., 2014;; R. crucifer (Wied, 1821); R. henseli (Lutz, 1934)); R. inopina Vaz-Silva et al., 2012; and R. ornata (Spix, 1824) [including R. abei (Baldissera et al., 2004), new synonymy, see below].
Distribution: These species are distributed mainly along the Atlantic Forest region, except R. inopina, which inhabits the Cerrado region (Baldissera et al., 2004; Thomé et al., 2010; Arruda et al., 2014; Roberto et al., 2014)). See map 4 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: The general internal relationships among the species are similar to those reported by Thomé et al. (2010, 2012). Previously, Maciel et al. (2006) recovered this group as monophyletic, but the internal relationships among the species were poorly supported. Based on our results and those of previous analyses, we discuss below several relevant taxonomic issues of this group.
Based on external morphology and morphometric analyses, Baldissera et al. (2004) reviewed the taxonomy of Rhinella crucifer. These authors resurrected two species (R. henseli and R. ornata) and recognized two new species (R. abei and R. pombali) for several populations previously considered within R. crucifer . More recently, two additional species were described, R. casconi and R. inopina (Vaz-Silva et al., 2012; Roberto et al., 2014)). Three of these species (i.e., R. casconi, R . crucifer, and R. henseli) were recovered as strongly supported lineages (JAF and JGC = 100%), and they have moderate UPDs with respect to other species (>0.98% see table 7).
Thomé et al. (2010, 2012) found Rhinella abei nested in R. ornata and stressed the need for including additional molecular markers before taking a taxonomic decision on this species. Our analyses, considering additional genes, recovered R. abei as nonmonophyletic and nested within R. ornata. Moreover, vouchers from multiple localities show no consistent differences in the morphological characters employed by Baldissera et al. (2004) to distinguish these species (e.g., color in preserved specimens, subocular band distinctiveness, head width, and forearm development; M.O.P. and D.B., personal obs.). Thus, we found no evidence to support the distinctiveness of R. abei, and consider Bufo abei Baldissera et al., 2004, a junior synonym of Rhinella ornata (Spix, 1824).
Rhinella ornata (including R. abei) is monophyletic, but poorly supported (JGC = 56%, JAF = 58%). Its sister taxon is R. inopina, a putatively independent lineage (see FCA analysis in Thomé et al., 2012) recovered with strong support (JGC and JAF = 99%) in the TE analysis. The genetic distances between R. ornata and R. inopina are very low for the R. crucifer Group (0.2%–0.7%) and cannot be attributable to evident mitochondrial introgression (see Thomé et al., 2012; fig. 9); some morphological characters (e.g., adult size, the coloration of marks on flanks, cloacae, and the posterior surface of thighs, and the disposition of parotoid macroglands) were proposed to differentiate both species. Considering the exceptionally low UPDs between R. ornata and R. inopina and the considerably wide range of R. ornata, further comparative studies accounting for geographical variation in these characters are necessary to definitely support or reject the status of R. inopina as a distinct species.
The Rhinella marina Group
Diagnosis: Our TE analysis recovered a poorly supported Rhinella marina Group (JGC = 63%, JAF = 79%) as in previous studies with less dense taxon sampling (e.g., Maciel et al., 2010; van Bocxlaer et al., 2010; Pyron, 2014). Two phenotypic synapomorphies support this species group: (1) the jagged or scalloped articulation between the medial ramus of pterygoid and parasphenoid alae (char. 22.1), with instances of homoplasy in R. atacamensis, R . achalensis, R . sternosignata, in a subclade of the R. festae Group, and in some species of the R. margaritifera Group, and (2) the sacral diapophyses with the anterior edge angled posteriorly to the midline axis of the vertebral column (char. 30.0), with instances of homoplasy in Rentapia hosii and Schismaderma carens.
Species of the Rhinella marina Group can be distinguished from members of the other species groups of Rhinella by the following combination of character states: (1) preorbital crest well developed (char. 0.2), (2) supraorbital crest well developed (char. 1.2), (3) pretympanic crest well developed (char. 2.2), (4) nasal and frontoparietal articulate along the entire margin (char. 8.3), (5) occipital artery pathway completely covered with bone (char. 10.2), (6) medial ramus of the pterygoid fused and extending medially along approximately half the length of the parasphenoid ala (char. 21.2), (7) m. extensor digitorum longus with an insertion on the metatarsophalangeal joint of digiti IV (char. 39.1), (8) inguinal fat bodies present (char. 51.1), (9) parotoid gland approximately ellipsoid, longer than wide or triangular and bulky (char. 56.0 or 56.3), (10) tarsal fold present (char. 65.1), (11) adhesive gland division after operculum fusion in embryo (char. 82.2), and (12) eggs biserially disposed in strings (char. 87.1).
Sister clade: The Rhinella crucifer Group.
Contents (10 species): Rhinella achavali (Maneyro et al., 2004); R. arenarum (Hensel, 1867); R. cerradensis Maciel et al., 2007; R. diptycha (Cope, 1862) [including R. jimi (Stevaux, 2002), new synonymy, see below]; R. horribilis (Wiegmann, 1833); R. icterica (Spix, 1824); R. marina (Linnaeus, 1758); R. poeppigii (Tschudi, 1845); R. rubescens (Lutz, 1925); and R. veredas (Brandão et al., 2007).
Distribution: These species are naturally distributed throughout all the main regions of the Neotropics, except in arid Andean areas and the Austral Temperate Forest region (Cei, 1980; De la Riva, 2002; Stevaux, 2002; Kwet et al., 2006; Brandão et al., 2007; Maneyro and Kwet, 2008; Santana et al., 2010; Acevedo et al., 2016; Saito et al., 2016; Venâncio et al., 2017). See map 5 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: Both recovered phenotypic synapomorphies were suggested as distinctive character states of this group by Pramuk (2006). Moreover, Maciel et al. (2010) proposed four osteological synapomorphies for the Rhinella marina Group (ventral ramus of the squamosal ventrolateral in posterior view; anterior extension of the cultriform process extends beyond the orbitonasal foramina; sphenethmoid lightly ossified; medial ramus of the pterygoid relatively narrow) and one skin-secretion compound (occurrence of a specific indolealkylamine). These characters were not considered in our TE analysis and should be reevaluated considering a denser sample of outgroups than the one employed by Maciel et al. (2010).
The finding of a moderately supported Rhinella marina Group contrasts with previous studies that recovered it well supported (e.g., Maciel et al., 2010; van Bocxlaer et al., 2010; Pyron, 2014; Jetz and Pyron, 2018). Two distinctive moderately supported clades are evident in this genetically and taxonomically complex species group. The first roughly corresponds to the North-Central Clade of Maciel et al. (2010) and is composed of R. diptycha (including R. jimi), R. horribilis, R . marina, R . poeppigii, and R. veredas, but does not include R. cerradensis (although see MP tree in Maciel et al., 2010: fig. 3).
TABLE 7
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella crucifer Group
Values reported are mean (range).
Rhinella poeppigii, R . veredas, and R. marina s.s. (see Acevedo et al., 2016) are successive sister taxa of the remaining species of the North-Central Clade but their positions are poorly supported (JGC <38%, JAF <54%). Except for R. veredas, the monophyly of these species are poorly supported (JGC = 74%, JAF = 77 for R. poeppigii; JGC <25%, JAF <50% for R. marina). The relationships among R. diptycha, R . jimi, and the nuclear sequences of R. horribilis and Rhinella sp. 1 are poorly resolved (see fig. 12). The lack of resolution and support for the internal relations of the North-Central Clade could be due, at least in part, to (1) the nuclear markers employed that do not provide sufficiently informative variation to resolve the relationships in the absence of mitochondrial information for some terminals, and (2) the occurrence of putative past and present hybridization that could not be detected with the available molecular evidence.
The occurrence of a deep mitochondrial divergence in Rhinella horribilis and Rhinella sp. 1 precluded the combination of the mitochondrial and nuclear sequences into single terminals. This results in an unstable and poorly supported phylogenetic position for nuclear sequences of both species in the TE analysis (in a polytomy with R. jimi specimens). Because the nuclear sequences employed provided relatively few informative characters, their relationships should be better explored considering additional evidence. Moreover, R. horribilis diverges in several morphological characters (adult morphology, osteology, and larval morphology; see Savage, 2002; Stevaux, 2002; Kwet et al., 2006; Tolledo and Toledo, 2010; Acevedo et al., 2016) from the largely allopatric R. jimi. Otherwise, the position of R. jimi in a polytomy together with the specimens of R. diptycha and the very low UPDs (0%–0.19%) among the specimens of these species indicate that the taxa are conspecific. Rhinella diptycha and R. jimi are two morphologically similar species with large parotoid and tibial macroglands. Remarkably, R. jimi has distinctive glands on its forearms and on both sides of the cloaca that were considered as the only distinctive characters from R. diptycha (Stevaux, 2002; Kwet et al., 2006). Mailho-Fontana et al. (2018) found that both species have similar types and distribution of skin glands, although in different proportions. These authors proposed that this differential development could be related to the occupancy of xeric environments by R. jimi . We also found a greater glandular development in the forearms and both sides of the cloaca in some specimens of R. diptycha from different localities of the dry Chaco in Argentina (M.O.P. and D.B., personal obs.). Based on these observations, we consider Bufo jimi Stevaux, 2002, a junior synonym of Rhinella diptycha (Cope, 1862). More physiological and histological studies, investigating different populations from different environments, could help to understand the patterns of variation in the development of macroglands in this species.
The other clade of the Rhinella marina Group is composed of R. achavali, R . arenarum, R . cerradensis, R . icterica, and R. rubescens (fig. 12), and roughly corresponds to the South-Central Clade of Maciel et al. (2010). Within this clade, R. arenarum is supported as sister taxon of the remaining species of the clade with strong support (JGC and JAF = 100%). The sampled specimen from the populations that had been considered as R. arenarum mendocina is nested within the remaining specimens of R. arenarum.
The sister clade of Rhinella arenarum is well supported but it is internally poorly resolved. This includes R. achavali, R . cerradensis, R . icterica, and R. rubescens. Rhinella cerradensis and R. rubescens are reciprocally monophyletic, their UPDs are low (0.19%–0.74%, see table 8), and constitute a strongly supported clade (JGC and JAF = 99%) that collapses in a basal polytomy within the clade. Rhinella achavali was recovered nested in a poorly supported clade (JGC <25%, JAF <50%) composed of some populations of R. icterica and the UPDs within this clade are low (0.37%–0.76%, see table 8). Although R. icterica is quite variable morphologically (M.O.P. and D.B., personal obs.) and this species includes several synonymized forms (e.g., Bufo missionum; Faivovich and Carrizo, 1997), this taxon is divergent morphologically from R. achavali (see Maneyro et al., 2004; Kwet et al., 2006; M.O.P., F.K., and C.B., personal. obs.). Finally, some specimens tentatively assigned to R. cerradensis (R. aff. cerradensis) collapse into a basal polytomy within the sister clade of R. arenarum. We refrain from taking any decision regarding the taxonomy of these species pending more studies, particularly with respect to understanding the effect of genetic (e.g., nuclear and/or mitochondrial introgressions) and environmental (e.g., phenotypic plasticity) factors on their morphological variation.
The Rhinella margaritifera Clade
Diagnosis: This well-supported clade (JGC and JAF = 99%) is diagnosed by two phenotypic synapomorphies: (1) alary process of the premaxillae angled anteriorly to the anterior margin of the pars dentalis of premaxillae (char. 13.2), with instances of homoplasy in Incilius coniferus, Rhinella achalensis, R . ornata, R . poeppigii, and Schismaderma carens; and (2) skin of dorsum of females with small tubercles lacking cornified tips (char. 52.3).
Sister clade: The Rhinella marina Clade.
Contents: Rhinella sternosignata and the R. festae, R . margaritifera, and R. veraguensis Groups.
Distribution: The species of this clade are mainly distributed throughout Amazonia and montane humid forest of the Andes. Some species of this clade are also found in the Atlantic Forest, Caatinga, Cerrado, Chaco/Pantanal, and Chocó regions, and in Central America (Duellman, 1999).
Comments: Within this clade, Rhinella sternosignata is recovered as the sister taxon of a large, poorly supported clade (JGC = 49%, JAF = 71%). This last clade is supported by a single phenotypic synapomorphy (ventral ridges on the palatine absent; char. 20.0), which is homoplastic in several species of the R. marina Group and outgroups. The clade is composed of three strongly supported species groups (JGC and JAF = 100%): (1) the redefined R. veraguensis Group, (2) the redefined R. festae Group, and (3) the redefined R. margaritifera Group. The R. festae and R. margaritifera Groups were recovered as sister clades with moderate support (JGC = 81%, JAF = 89%) and five character states optimize as phenotypic synapomorphies of this clade: (1) discrete superficial cutaneous tendons absent (char. 33.0); (2) lateral slip of the m. interphalangeus proximalis digiti V (foot) absent, with instances of homoplasy in R. major and R. papillosa (char. 35.0); (3) m. abductor brevis plantaris hallucis absent (char. 36.0), with instances of homoplasy in Anaxyrus woodhousii, Peltophryne empusa (polymorphic), and R. mirandaribeiroi; (4) slip of the medial m. lumbricalis brevis digiti V originating from the distal carpal 3-4-5 absent (char. 43.0) with an instance of homoplasy in Nannophryne variegata (polymorphic); and (5) head of the m. extensor carpi ulnaris from the radioulna with a fleshy origin (char. 47.0), with an instance of homoplasy in P . empusa. A similar topology for the main internal clades of the R. margaritifera Clade was recovered in the ML analysis (supplementary data 4.4–4.5).
Rhinella sternosignata
Diagnosis: Rhinella sternosignata (Günther, 1858b) was recovered as the sister taxon of all other species of the R. margaritifera Clade, with poor support (JGC = 49%, JAC = 71%). Phenotypic autapomorphies are: (1) acuminate anterior margins of nasals (char. 6.1), with instances of homoplasy in Incilius coniferus and the R. margaritifera Group; (2) articulation of the zygomatic ramus of the squamosal with the maxilla (char. 15.1), with instances of homoplasy in the R. granulosa Group and Peltophryne lemur; (3) articulation between the medial ramus of the pterygoid and parasphenoid alae with a jagged suture (char. 22.1) with instances of homoplasy in R. achalensis, R . atacamensis, some species of the R. festae and R. margaritifera Groups, and in the R. marina Group; (4) parotoid gland round to ovoid mostly symmetrical (char. 56.2); (5) large size of adult males with respect to adult females (char. 66.1), with instances of homoplasy in R. yanachaga, and in several species of the R. marina Clade; and (6) unpigmented eggs (char. 88.0), with instances of homoplasy in Ansonia longidigita, Rhinella justinianoi, R . stanlaii, and in the R. festae Group.
Distribution: This species inhabits montane forests of the Cordillera de la Costa and the Andean Cordillera de Mérida of Venezuela (La Marca and Mijares-Urrutia, 1996; Barrio-Amorós et al., 2019). See map 5 (available at https://doi.org/10.5531/sd.sp.46) for type and sampled localities.
Sister clade: The clade composed of the Rhinella festae, R . margaritifera, and R. veraguensis Groups.
Comments: This species was tentatively associated with the Rhinella margaritifera (Cei, 1972a; Hoogmoed, 1990; Duellman and Schulte, 1992) or R. granulosa Groups (Gallardo, 1962). Pereyra et al. (2016a) rejected the inclusion of this species in any of these groups, but they could not determine its relationships rigorously due to the poor sampling of Rhinella. This species was wrongly reported for many localities outside the Cordillera de la Costa montane forests region in Venezuela as discussed by La Marca and Mijares-Urrutia (1996). Vélez-Rodríguez (1999) recorded this species in error for Colombia (see Vélez-Rodríguez, 2004b,, 2005). Additionally, there are a large number of recent reports of R. sternosignata for Colombia (Acosta-Galvis et al., 2006; Romero et al., 2008; Acosta-Galvis, 2012a, 2012b). Analyzed specimens tentatively assigned to this species from the eastern slope of the Cordillera Oriental in Colombia (MAR 1314, Boyacá and MAR 1955, Caquetá) were unrelated to the specimen of R. sternosignata from Venezuela in the phylogenetic analyses, and instead, they represent an undescribed species along with other specimens of the R. margaritifera Group from Loreto, Peru, and Miranda, Venezuela (Rhinella sp. 13, see below). These results, and the absence of comprehensive comparative studies considering topotypical material of R. sternosignata, indicate that there is no evidence to consider its occurrence in Colombia.
The Rhinella veraguensis Group
Diagnosis: No phenotypic synapomorphies were recovered for this strongly supported group (JGC and JAF = 100%). This is mainly due to the lack of detailed information for one of its two constituent clades (composed of Rhinella sp. 2 [see below], R. inca and R. leptoscelis). Nevertheless, some character states might represent synapomorphies for this group or a subclade: (1) the articulation of jaw opposite to the fenestra ovalis (char. 16.1), with instances of homoplasy in Nannophryne variegata, Rhinella beebei, R . merianae, R . yanachaga, and the R. arunco Group; (2) light-colored nuptial pads (char. 62.0); (3) larval peribranchial region with bulging regions lateral to the oral disc (char. 67.1); (4) larval oral disc with complete A2 labial tooth row (char. 73.0), with instances of homoplasy in Amazophrynella aff. minuta, Ansonia longidigita, Melanophryniscus gr. stelzneri, Phrynoidis juxtaspera, and Schismaderma carens; (5) the dextral opening of the vent tube (char. 75.1); and (6) eggs laid in open clumps (char. 86.1; structure of the spawn only known in R. rumbolli within the R. veraguensis Group).
TABLE 8
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella marina Group
Values reported are mean (range). Species with an asterisk include recent synonyms.
The species of the Rhinella veraguensis Group can be distinguished from members of the other species groups of Rhinella by the following combination of character states: (1) preorbital crest weak (char. 0.1), (2) supraorbital crest weak (char. 1.1), (3) pretympanic crest weak (char. 2.1), (4) medial ramus of the pterygoid fused and extending medially along approximately half the length of the parasphenoid ala (char. 21.2), (5) m. extensor digitorum longus without an insertion on the metatarsophalangeal joint of digiti IV (char. 39.0), and (6) tarsal fold absent (char. 65.0).
Sister clade: The clade composed of the Rhinella festae and R. margaritifera Groups.
Content (9 species): Rhinella chrysophora (McCranie et al., 1989); R. fissipes (Boulenger, 1903); R. gnustae (Gallardo, 1967)); R. inca (Stejneger, 1913); R. justinianoi (Harvey and Smith, 1994); R. leptoscelis (Boulenger, 1912); R. quechua (Gallardo, 1961)) [including R. amboroensis (Harvey and Smith, 1993), new synonymy, see below]; R. rumbolli (Carrizo, 1992); and R. veraguensis (Schmidt, 1857).
Distribution: All species of the Rhinella veraguensis Group are distributed in Andean humid forests of Argentina, Bolivia, and Peru, except R. chrysophora, which inhabits the Central American Atlantic moist forests in Honduras (Rodríguez et al., 1993; De la Riva et al., 2000; Köhler, 2000; Lavilla and Cei, 2001; Padial et al., 2009; McCranie, 2017). See map 6 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: The former Rhinella veraguensis Group (see appendix 5) is recovered as polyphyletic. To remedy this, we restrict its content to the clade of species most closely related to R. veraguensis . This group also comprises two species not included in the phylogenetic analyses that share several putative synapomorphies with the species sampled here (see below). Two clades were recovered within this redefined R. veraguensis Group. One clade is poorly supported (JGC = 67%, JAF = 76%) and comprises the southernmost distributed species of the group. In the second clade, we were unable to examine the voucher of R. amboroensis (MNK 5302), but this specimen was collected near the type locality of the species. The specimen was recovered as the sister taxon of R. quechua and the genetic distance between the specimens is 0% (see table 9). Both species are very similar morphologically and only a few morphological characters were proposed to differentiate the taxa (i.e., the extension of the foot webbing, ventral skin texture, and finger length). However, these difference are not consistently observed in specimens collected in the type locality of R. amboroensis (I.D.L.R., personal obs.) and they could simply represent variations within R. quechua. For these reasons, we consider Bufo amboroensis Harvey and Smith, 1993, a junior synonym of Rhinella quechua (Gallardo, 1961)).
The other clade in this group is strongly supported (JGC and JAF = 100%) and includes the northernmost distributed species, R. inca, R . leptoscelis, and an undescribed species from Oxapampa, Peru (Rhinella sp. 2). The UPDs among these three species are relatively low (1.16%–1.90%; see table 9).
Rhinella chrysophora and R. gnustae, two species not included in the phylogenetic analysis, are considered to belong to this species group. Rhinella chrysophora is known only from two localities in northern Honduras and is supposedly extinct, not collected since 1996 (McCranie and Castañeda, 2005; McCranie, 2017). This species was originally described as belonging to a distinct genus (Atelophryniscus; McCranie et al., 1989) of no evident relationships within Bufonidae. Pramuk and Lehr (2005), based on a morphological phylogenetic analysis, demonstrated that it is related to the species of the R. veraguensis Group s.l. Unfortunately, the character scores for R. chrysophora are not available and the condition of the double-stained specimen used in that study is very poor (J.J.O.-S., personal obs.). However, morphological evidence indicates that R. chrysophora belongs to the R. veraguensis Group, as it posses all its known putative synapomorphies (except for oviposition mode, which is unknown; McCranie et al., 1989; Lavilla and de Sá, 2001; Pramuk and Lehr, 2005).
Rhinella gnustae (Gallardo, 1967)) was described based on a single subadult specimen from an imprecise locality of Jujuy Province (Argentina) (Gallardo, 1967;; Cei, 1980; Lavilla and Cei, 2001; Lavilla et al., 2002). We tentatively assign this species to the R. veraguensis Group based on a combination of characters (although none of them synapomorphic) that occur in multiple species of this group: row of dorsolateral tubercles in the skin absent, tarsal fold absent, and small tubercles without a cornified tip.
The Rhinella festae Group
Diagnosis: This well-supported group (JGC and JAF = 100%) is diagnosed by the following five phenotypic synapomorphies: (1) skull lightly exostosed (char. 9.1), with instances of homoplasy in Rhinella achavali, R . rubescens, the R. spinulosa Group, and in several outgroups; (2) fusion of medial ramus of pterygoid with anterolateral margin of the parasphenoid ala (char. 21.1), with instances of homoplasy in Rhinella sp. 12, in the R. arunco Group, in some species of the R. spinulosa Group, and in several outgroups; (3) anterior margin of cultriform process of parasphenoid truncated (char. 23.2); (4) arboreal habits (char. 84.1) that revert in an internal clade of this group, and with instances of homoplasy in Incilius coniferus and Rentapia hosii; and (5) unpigmented eggs (char. 88.0), with instances of homoplasy in Ansonia longidigita, Rhinella justinianoi, R . stanlaii, and R. sternosignata. Other putative synapomorphies of this group or an internal clade are: (1) additional origin of the m. interosseus cruris from the tibiale absent (char. 38.0; known within the group only for R. paraguas); (2) m. extensor brevis medius hallucis absent (char. 41.0; known within the group only for R. paraguas); and (3) the terrestrial oviposition (char. 85.1; known within the group only in R. tacana). Moreover, species of the R. festae Group can be distinguished from members of the other species groups of Rhinella by the following combination of character states: (1) preorbital crest absent or indistinguible (char. 0.0), (2) supraorbital crest weak (char. 1.1), (3) discrete superficial cutaneous tendons absent (char. 33.0), (4) lateral slip of the m. interphalangeus proximalis digiti V absent (char. 35.0), (5) m. abductor brevis plantaris hallucis absent (char. 36.0), (6) m. extensor digitorum longus without an insertion on the metatarsophalangeal joint of digit IV (char. 39.0), (7) slip of the medial m. lumbricalis brevis digiti V originating from the distal carpal 3-4-5 absent (char. 43.0), (8) head from the radioulna of the m. extensor carpi ulnaris with a fleshy origin (char. 47.0), (9) nuptial pads dark colored (char. 62.1), and (10) tarsal fold absent (char. 65.0).
Sister clade: The Rhinella margaritifera Group.
Contents (18 Species): Rhinella acrolopha (Trueb, 1971)); R. arborescandens (Duellman and Schulte, 1992); R. chavin (Lehr et al., 2001); R. festae (Peracca, 1904); R. lilyrodriguezae Cusi et al., 2017; R. lindae (Rivero and Castaño, 1990); R. macrorhina (Trueb, 1971)); R. manu Chaparro et al., 2007; R. multiverrucosa (Lehr et al., 2005); R. nesiotes (Duellman and Toft, 1979); R. nicefori (Cochran and Goin, 1970); R. paraguas Grant and Bolívar-G., 2014; R. rostrata (Noble, 1920); R. ruizi (Grant, 2000); R. tacana (Padial et al., 2006); R. tenrec (Lynch and Renjifo, 1990); R. truebae (Lynch and Renjifo, 1990); and R. yanachaga Lehr et al., 2007.
TABLE 9
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella veraguensis Group
Values reported are mean (range).
Distribution: Mainly distributed in Andean humid forests of Bolivia, Colombia, Ecuador, and Perú (Trueb, 1971;; Duellman and Lynch, 1988; Lynch and Renjifo, 1990; Duellman and Schulte, 1992; Ruiz-Carranza et al., 1996; Lehr et al., 2001, 2005, 2007; Rueda-Almonacid et al., 2004; Chávez et al., 2013; Grant and Bolivar-G., 2014; Cusi et al., 2017). The only species distributed outside this region is Rhinella acrolopha, which inhabits the Chocó region (Darién, Panama; Trueb, 1971)). See maps 7 and 8 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: Analyses of our expanded sample of the Rhinella acrolopha Group corroborated the monophyly of that group; however, they also corroborated its placement among a subset of the species referred to the R. veraguensis Group s.l. Insofar as those species are paraphyletic with respect to the R. acrolopha Group, the only means of maintaining the current composition of the R. acrolopha Group would be to recognize two or more groups within the larger clade, which we believe to be unwarranted at this time. Consequently, we redefined the R. festae Group as was originally proposed by Moravec et al. (2014) to also include all the species previously referred to the R. acrolopha Group and three species (R. arborescandens, R . multiverrucosa, and R. tacana) of the former R. veraguensis Group.
Pramuk (2006) proposed a close phylogenetic relationship between Rhamphophryne and some species of the Rhinella veraguensis Group as formerly defined, but she did not provide a diagnosis for the inclusive clade. As defined in this study, the R. festae Group is a morphologically and ecologically diverse clade of toads; the lack of information for many aspects of these species is remarkable (e.g., adult osteology, adult musculature, larval morphology, natural history; see below).
Species of this group have notably high interspecific UPDs compared to species of other groups of Rhinella (see tables 3–11) except between the pairs R. chavin/R . multiverrucosa and R nicefori/R . ruizi. Sequences of the included specimen assigned to R. multiverrucosa (MUBI 11455) are identical (UPD = 0%) to the topotype of R. chavin (sequence DQ158441 from Pramuk, 2006). Although morphologically most similar to R. multiverrucosa, the specimen MUBI 11455 was not collected near the type locality of this species and was actually collected closer to the type locality of R. chavin (see map 8; available at https://doi.org/10.5531/sd.sp.46). Most of the characters that distinguish these two species involve differences in glandular development. Our results should be tested considering the existing morphological variation within R. chavin and including topotypes of R. multiverrucosa in a future revision of these species.
Similarly, the UPDs between the included specimen of Rhinella nicefori and topotypic specimens of R. ruizi is low (UPDs = 0.19%; see table 10). The two species were not explicitly differentiated in the original description of R. ruizi (Grant, 2000), but they differ in some characters (degree of cranial ornamentation, the occurrence of vocal slits in adult males, adult size). As we did not sample topotypical material of R. nicefori and cannot discard the occurrence of some additional variation in the diagnostic characters that differentiate the two species, the identity of the included specimen MHUA 4793 should be reevaluated. For this reason, we refrain from taking a taxonomic action, pending a detailed taxonomic evaluation of both species, considering topotypical material of R. nicefori and comparison with type specimens.
Two undescribed species within this species group are recovered in our TE analysis. Firstly, some specimens tentatively assigned to Rhinella manu from Madre de Dios and Cusco display high UPDs (3.37%) with respect to specimens of R. manu s.s., suggesting they might represent an undescribed species (Rhinella sp. 3). Second, the specimen of R. sp. “gr. acrolopha” (referred to Rhinella sp. C. by Machado et al., 2016) from Caldas (Colombia) is recovered as sister species of R. paraguas, and the genetic distance between them (UPDs = 5.73%–6.11%) is consistent with the hypothesis that it is an undescribed species (Rhinella sp. 4).
We could not include Rhinella rostrata and R. truebae in our analyses. Nevertheless, these species can be placed in the R. festae Group on the basis of several character states that are synapomorphies of this group or its internal clades: (1) skull lightly exostosed (char. 9.1); (2–4) columella, annulus tympanicus, and tympanic membrane absent (chars. 48.0, 49.0, and 50.0) in R. rostrata (present in R. truebae); and (5) finger webbing present (char. 63.1).
The Rhinella margaritifera Group
Diagnosis: No phenotypic synapomorphies were recovered in our TE analysis for this well-supported species group (JGC and JAF = 100%). However, given the lack of information (see Comments on the phenotypic evidence considered for Rhinella section) for its earlier diverging species (e.g., R. ocellata, R . yunga, and Rhinella sp. 5) or closely related clades (i.e., R. sternosignata, the R. festae and R. veraguensis Groups), the inclusion of additional observations in the phenotypic dataset could provide diagnostic synapomorphies for this clade. A putative synapomorphy for this species group (unknown condition in Rhinella sp. 5) is the acuminate anterior margins of nasals (char. 6.1), with instances of homoplasy in Incilius coniferus and R. sternosignata. Moreover, species of the R. margaritifera Group can be distinguished from members of the other species groups of Rhinella by the following combination of character states: (1) preorbital crest weak (char. 0.1), (2) medial ramus of the pterygoid fused and extending medially along approximately half the length of the parasphenoid ala (char. 21.2), (3) posterior lobe of the anterolateral process of hyoid absent (char. 25.0), (4) discrete superficial cutaneous tendons absent (char. 33.0), (5) lateral slip of the m. interphalangeus proximalis digiti V absent (char. 35.0), (6) m. abductor brevis plantaris hallucis absent (char. 36.0), (7) m. extensor digitorum longus without an insertion on the metatarsophalangeal joint of digiti IV (char. 39.0), (8) slip of the medial m. lumbricalis brevis digiti V originating from the distal carpal 3-4-5 absent (char. 43.0), (9) head of the m. extensor carpi ulnaris from the radioulna with a fleshy origin (char. 47.0), (10) inguinal fat bodies absent (char. 51.0), (11) tarsal fold absent (char. 65.0), and (12) submarginal papillae in the oral disc of larvae absent (char. 71.0).
Sister clade: The Rhinella festae Group.
Contents (17 species): Rhinella acutirostris (Spix, 1824); R. alata (Thominot, 1884); R. castaneotica (Caldwell, 1991); R. cristinae (Vélez-Rodríguez and Ruiz-Carranza, 2002); R. dapsilis (Myers and Carvalho, 1945) [including R. gildae Vaz-Silva et al., 2015, new synonymy, see below]; R. hoogmoedi Caramaschi and Pombal, 2006; R. iserni (Jiménez de la Espada, 1875) [including R. yunga Moravec et al., 2014 new synonymy, see below]; R. lescurei Fouquet et al., 2007a; R. magnussoni Lima et al., 2007; R. margaritifera (Laurenti, 1768) [including R. martyi Fouquet et al., 2007a, new synonymy, see below]; R. ocellata (Günther, 1858b); R. proboscidea (Spix, 1824); R. roqueana (Melin, 1941); R. scitula (Caramaschi and Niemeyer, 2003) [including R. paraguayensis Ávila et al., 2010, new synonymy, see below]; R. sclerocephala (Mijares-Urrutia and Arends, 2001); R. sebbeni Vaz-Silva et al., 2015; and R. stanlaii (Lötters and Köhler, 2000).
Distribution: Mainly distributed in Amazonia, but a few species also occur in the Andes, Atlantic Forest, Caatinga, Cerrado, Chocó, Chaco/Pantanal, and in Central America (Hoogmoed, 1986, 1990; Ruiz-Carranza et al., 1996; Caramaschi and Pombal, 2006; Köhler et al., 2006; Fouquet et al., 2007a; Moravec et al., 2014; Sugai et al., 2014; Santos et al., 2015; Ávila et al., 2018; Freitas et al., 2018; Silva et al., 2018). See maps 9 and 10 (available at https://doi.org/10.5531/sd.sp.46) for type localities and sampled localities.
Comments: This species group is particularly controversial regarding its diagnosis, content, and taxonomy of its species. The main revisions dealing with this group (e.g., Hoogmoed, 1986, 1990; Duellman and Schulte, 1992; Vélez-Rodríguez, 2004b;; Pramuk, 2006; Fouquet et al., 2007a) disagreed with respect to the inclusion of multiple species (e.g., Rhinella cristinae, R . iserni, R . ocellata; see appendix 5). Vélez-Rodríguez (2004b) performed a phylogenetic analysis of the group based on morphological characters and proposed the restriction of its content to a clade diagnosed by two synapomorphies: (1) m. depressor mandibulae composed of two slips with independent origins, on the posterior portion of the otic ramus of the squamosal and the anterior portion of the otic ramus of the squamosal and tympanic annulus; and (2) thickening of the ventral margin of the quadratojugal (our char. 24.2). This redefinition of the R. margaritifera Group was not supported by the combined (i.e., molecular + morphological characters) phylogenetic analysis of Pramuk (2006). Pramuk (2006) recovered two synapomorphies for the few exemplar species of this group that she included: (1) the expansion of the posterior ramus of the pterygoid and (2) the occurrence of a lateral articulation between the nasals and the preorbital processes of the maxillae (homoplastic). She also found R. ocellata to be the sister species of the R. margaritifera Group.
TABLE 10
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella festae Group
Values reported are mean (range).
continued
Subsequent studies dealing with the taxonomy of this group (e.g., Fouquet et al., 2007a; Ávila-Pires et al., 2010; Lavilla et al., 2013; Vaz-Silva et al., 2015; Avila et al., 2018)) did not explicitly follow a definition based on synapomorphies and thus its composition varied. Based on our results, we redefine the Rhinella margaritifera Group to include the species listed above that are now grouped exclusively on molecular evidence (although some character states may result in putative synapomorphies, see Diagnosis above). Also, the characters proposed by Vélez-Rodríguez (2004) and Pramuk (2006) should be reevaluated considering relevant species not included in these studies (e.g., R. sternosignata, R . yunga, and Rhinella sp. 5) to understand their polarity in the context of our results.
The lack of a precise type locality and reference specimens, the large intraspecific (including sexual dimorphism) and interspecific variation in adult size and cranial crest shape and development, and the occurrence of sympatry among some species of the Rhinella margaritifera Group turned its taxonomy chaotic and confusing (Hoogmoed, 1989; 1990; Lavilla et al., 2013). A detailed revision of this complex species group is beyond the scope of the present study. As a result, we have been cautious to take taxonomic actions only when evidence is decisive.
The nominal species of the group was described by Laurenti (1768) based on illustrations of Seba (1734) of a specimen from “Brasilia” as the type locality. The identity of this taxon remains unclear after more than two and a half centuries (see discussions in Hoogmoed, 1989; Vélez-Rodríguez, 2004; Fouquet et al., 2007a; Ávila-Pires et al., 2010; Lavilla et al., 2013, 2017). Ávila-Pires et al. (2010) designated the specimen depicted in Seba (1734: pl. 71, figs. 6, 7) as the lectotype of Rana margaritifera Laurenti, 1768, and considered the species to be conspecific with Rhinella martyi Fouquet et al., 2007a. Subsequently, Lavilla et al. (2013) invalidated the lectotype designation by Ávila-Pires et al. (2010) and, assuming that the type specimen of R. margaritifera was lost, designated and described a neotype for this species. More recently, Lavilla et al. (2017) noted that a previous publication (Milto and Barabanov, 2011) had reported the existence of the type of R. margaritifera, invalidating the neotype.
Milto and Barabanov (2011) mentioned two specimens (ZISP 257.1 and 257.2) within the type series of R. margaritifera without additional comments. Photographs of both specimens are inadequate to determine which one was used in the illustration of Seba (1734; or if both were used) because both are adult females that fully agree with the description and illustrations. Consequently, it is reasonable to consider the specimens found by Milto and Baravanov (2011) to indeed be those used by Seba (1734) and to arbitrarily designate the specimen ZISP 257.1 as lectotype of Rana margaritifera Laurenti, 1768.
Additionally, we follow Ávila-Pires et al. (2010) regarding the conspecificity of Rhinella margaritifera and R. martyi because the lectotype and paralectotype of R. margaritifera match almost all the characters used by Fouquet et al. (2007a) to differentiate R. martyi from other species of the group (heel extension with hind limbs adpressed and iris coloration unknown in the types of R. margaritifera). Thus, we consider Rhinella martyi Fouquet et al., 2007a, to be a junior synonym of R. margaritifera (Laurenti, 1768).
Having established the identity of Rhinella margaritifera, we now introduce our results regarding this species group. An undescribed species from Pastaza (Ecuador), Rhinella sp. 5, is recovered with low support (JGC = 28%, JAF = 58%) as the sister taxon of all other species of the clade. The sister group of next most inclusive clade is poorly supported (JGC = 32%, JAF = 60%) and composed of two morphologically and geographically divergent species, R. yunga and R. ocellata . Rhinella yunga was recently described from the montane forest of the Selva Central, Peru. Distinctive characters used to diagnose this species in the original description are also present in the poorly known R. iserni (skin of dorsum mostly smooth, degree of development of cranial crest, and especially the absence of all the structures of the tympanic middle ear; Jiménez de la Espada, 1875; Moravec et al., 2014; Hoogmoed, personal commun.; J.M. and M.O.P., personal obs.). In addition to their morphological resemblance, both species were described from nearby type localities from the Peruvian Yungas region. Thus, we consider Rhinella yunga Moravec et al., 2014, to be a junior synonym of Rhinella iserni (Jiménez de la Espada, 1875).
Rhinella magnussoni, R . cf . margaritifera from Amazonas (Colombia), specimens of “R. proboscidea” from Ecuador and Peru, and an undetermined specimen of the R. margaritifera Group from São Pedro (Amazonas, Brazil) compose a well-supported clade (JGC = 93%, JAF = 94%). Rhinella magnussoni and R. cf . margaritifera from Amazonas (Colombia) have a relatively high UPD (2.10%), which seem to support the specific distinctiveness of the latter (Rhinella sp. 6). The results of the phylogenetic analysis (see fig. 14) and UPDs among clades (2.60%–10.27%; see table 11) strongly suggest that the specimen from São Pedro and both populations of “R. proboscidea” from Ecuador (Sucumbios) and Peru (Loreto) correspond to three undescribed species (Rhinella sp. 7–9). The similarity of these undescribed taxa with the phylogenetically distantly R. proboscidea s.s. and R. castaneotica (see above) indicates the need for a thorough revision of the “R. proboscidea” complex.
Another well-supported clade (JGC and JAF = 99%) includes Rhinella acutirostris, R . alata, R . sclerocephala, R . sternosignata from Colombia, and several apparently undescribed species (see below). A nonmonophyletic Rhinella alata (sensu Santos et al., 2015) and R. sclerocephala constitute a well-supported subclade within this clade (JGC = 96%, JAF = 97%). The latter species was recovered as sister taxon of the Ecuadorian populations of R. alata with poor support (JGC = 61% JAF = 62%), and their UPDs are 1.31%–1.52%. Otherwise, the populations of Panama and Colombia were recovered as a poorly supported lineage (JGC = 55%, JAF = 58%) with UPDs of 0.56%–1.52% with respect to its sister clade. Santos et al. (2015) did not find any morphological character or evident genetic differentiation between the populations from Panama and those from Ecuador as to consider them as different taxa. Otherwise, R. sclerocephala was differentiated from R. alata by several characters such as its larger size, presence of well-developed cranial crests, vertebral apophysis, and bony knob (Mijares-Urrutia and Arends, 2001; Santos et al., 2015), although these characters vary extensively within species of the R. margaritifera Group and might be associated with particular environmental conditions over their areas of distribution (see Kutrup et al., 2006; Bandeira et al., 2016). The absence of evident differences between specimens of both clades of R. alata, the nested position of R. sclerocephala within this clade, and the low UPDs seem to support their conspecificity, but we refrain from synonymyzing both taxa due to the low support for the nested position of the specimen of R. sclerocephala and the several morphological differences. Nevertheless, it is evident that a detailed taxonomic revision of both taxa considering additional specimens and molecular evidence is required to better resolve the taxonomy within this clade.
The other subclade includes Rhinella acutirostris and four undescribed species (see also ML topology in supplementary data 4.5): Rhinella sp. 10 from Napo (Ecuador), Rhinella sp. 11 from Amazonas (Brazil), Rhinella sp. 12 from Pando (Bolivia) and Madre de Dios (Peru), and Rhinella sp. 13 from some localities of Colombia (Boyacá and Caquetá), Peru (Loreto), and Venezuela (Miranda). This latter new species was previously recorded from Colombia as R. sternosignata (M.R., personal obs.; see comments for R. sternosignata).
The nonmonophyletic Rhinella castaneotica and R. proboscidea compose a well-supported clade (JGC and JAF = 96%). The internal relationships among the included specimens are poorly resolved and the UPDs are 0.19%–2.75%. Within this clade, we could not obtain samples from the type locality of R. castaneotica (“near Cachoeira Jaruá, Rio Xingu, Pará, Brazil”), but we included sequences from a relatively close locality (300 km airline distance) that could represent R. castaneotica s.s. (see Fouquet et al., 2012a: fig. S6). We also included sequences of specimens from Manaus (Amazonas, Brazil) and Floresta (Roraima, Brazil) that could be morphologically assigned to R. proboscidea . However, the imprecise type locality of this species is “flumen Solimoens (= Rio Solimões),” which comprises the Brazilian section of the Amazon River between the triple border of Brazil-Colombia-Peru and the city of Manaus and has an extension of approximately 1700 km (Vanzolini, 1981). Although we consider that phylogenetic evidence, UPDs (see table 11), and absence of distinctive adult and larval differential characters (see comparisons provided by Caldwell, 1991, and Menin et al., 2006) support their conspecificity, we continue recognizing both taxa pending a thorough revision, including additional samples and detailed comparison with the type material.
The sister taxon of the clade including the problematic Rhinella castaneotica and R. proboscidea, is well supported (JGC = 97%, JAF = 98%) and includes two successively diverging species (R. lescurei and R. hoogmoedi), and two subclades. One of these is well supported (JGC and JAF = 99%) and composed of R. paraguayensis, R . scitula, R . stanlaii, and two undescribed species: Rhinella sp. 14 from SE Peru (“Bufo sp. 6” sensu Vélez-Rodríguez, 2004b,, and “Bufo cf. margaritifer 5” sensu Pramuk, 2006), and Rhinella sp. 15 from La Paz, Bolivia (which corresponds to Bufo sp. 1 of Lötters and Köhler, 2000). The included specimen of R. paraguayensis was recovered in a polytomy with the specimens of R. scitula; the UPDs among these specimens are low (0.13%–0.3% in the complete the 16S rRNA gene). These parapatric species were differentiated mainly by adult size, crest development, and skin texture. All these characters have been demonstrated to be subject to variation due to specific environmental conditions throughout the distribution of some bufonids (see Kutrup et al., 2006; Bandeira et al., 2016). Thus, we consider Rhinella paraguayensis Ávila et al., 2010, to be a junior synonym of R. scitula (Caramaschi and Niemeyer, 2003).
The other subclade is also well supported (JGC and JAF = 98%) and includes Rhinella margaritifera s.s. and a poorly supported clade (JGC <25%, JAF <50%) composed of R. dapsilis, R . cf . dapsilis, R . gildae, and several divergent lineages of R. margaritifera s.l. (e.g., the lineages called Rhinella sp. A and Rhinella sp. B by Fouquet et al., 2007c). The clade includes specimens that vary in the degree of development of bony protrusions and cranial crests, dorsal coloration, occurrence of a developed proboscis, and dorsal skin texture (Myers and Carvalho, 1945; Dixon, 1976; Rodríguez and Duellman, 1994; Vaz-Silva et al., 2015; M.O.P., personal obs.). The UPDs within this clade are 0%–2.79% (mean UPD = 1.29%, see table 11). Thus, the absence of unequivocal morphological differences and very low genetic distances indicate that Rhinella gildae Vaz-Silva et al., 2015, is a junior synonym of R. dapsilis (Myers and Carvalho, 1945).
We could not obtain tissue samples of Rhinella cristinae to test its relationships. However, this species can be assigned to this group on the basis of the occurrence of the only putative phenotypic synapomorphy of the group (acuminate anterior margins of nasals; char. 6.1) and a combination of characters typical of this group: (1) preorbital crest weak (char. 0.1), (2) medial ramus of the pterygoid fused and extending medially along approximately half the length of the parasphenoid ala (char. 21.2), (3) posterior lobe of the anterolateral process of hyoid absent (char. 25.0), (4) inguinal fat bodies absent (char. 51.0), and (5) tarsal fold absent (char. 65.0).
Hybridization, Deep Mitochondrial Divergence, and “Ghost Introgression” in Rhinella
Reports on natural and artificial hybridization are well known in many bufonids including multiple species of Rhinella (e.g., Blair, 1972;; Green, 1996; Gergus et al., 1999; Malmos et al., 2001; Masta et al., 2002; Baldo and Basso, 2004; Yamazaki et al., 2008; Goebel et al., 2009; Fontenot et al., 2011; Correa et al., 2012, 2013; Pereyra et al., 2016a; Betto-Colliard et al., 2018). Explosive breeding events with intense male competition for mates and passive female choice (i.e., scramble competition; see Wells, 2007; Pereyra et al., 2016b) is common in many species of several genera of Bufonidae, and premating isolating mechanisms seem to be insufficient to avoid interspecific amplexus in these species (see Blair, 1958; Guerra et al., 2011). Malone and Fontenot (2008) also demonstrated that bufonids require a substantial genetic divergence to achieve postzygotic reproductive isolation. Under this scenario, the common occurrence of hybridization in this family is not surprising.
A particular situation of natural hybridization could happen in “Rhinella pombali” (Thomé et al., 2010, 2012), where all individuals of this taxon are considered hybrids between R. crucifer and R. ornata (Thomé et al., 2010, 2012) and our results are in agreement with this idea. The two included specimens of “R. pombali” in preliminary analyses were not recovered as monophyletic in the nuclear analysis and each of them has a unique mitochondrion (one from R. crucifer and the other from R. ornata). Moreover, available evidence is insufficient to test whether “R. pombali” could represent a species of hybrid origin (see Avise, 2008; Darras and Aron, 2015, Lavanchy and Schwander, 2019) and more detailed studies are necessary to explore this possibility.
Although we deliberately excluded the hybrid specimens from our analyses, the impact of natural hybridization in Bufonidae could be currently underestimated due to the difficulties in recognizing hybrids and/or past hybridization events. Introgressive hybridization (both nuclear and mitochondrial) could have an impact on bufonid evolution allowing a faster accumulation of genetic novelties than through mutation alone. The incorporation of additional genetic diversity could impact the acquisition of adaptive phenotypic traits and have a significant role in speciation as is common in diverse taxonomic groups (for reviews see Baack and Rieseberg, 2007; Schwenk et al., 2008; Toews and Brelsford, 2012; Abbott et al., 2016; Gopalakrishnan et al., 2018; Hill, 2019; Servedio and Hermisson, 2019).
Mitochondrial introgressions are more commonly reported than are nuclear introgressions and can be evidenced by genetic populational studies or by the discordance between phylogenetic trees inferred from separate analyses of both genomes (Toews and Brelsford, 2012; Bonnet et al., 2017). Within Rhinella, putative events of mitochondrial introgression were documented for R. marina (Sequeira et al., 2011; but also see Vallinoto et al., 2017 and Bessa-Silva et al., 2020), R. bernardoi (Pereyra et al., 2016a), and R. horribilis s.l. (present study), and they occur without noticeable evidence of nuclear introgression, as was also reported in other vertebrates (Alves et al., 2006; Chen et al., 2009; Schwarzer et al., 2012).
TABLE 11
Percentage of uncorrected p-distances between 16S sequences among species of the Rhinella margaritifera Group
Values reported are mean (range).
continued
Several populations of Rhinella marina from south of the Amazon River seem to have similar mitochondrial lineages as R. diptycha, in contrast to populations northward. As nuclear loci of specimens of both populations of R. marina were similar, and divergent from R. diptycha, the occurrence of an extensive mtDNA unidirectional introgression from R. diptycha into R. marina was hypothesized (Sequeira et al., 2011). However, this hypothesis was not conclusively corroborated in a subsequent study because an additional mitochondrial clade, found for some populations of R. marina, obscured the direction of the introgression between these species (Vallinoto et al., 2017). A similar situation of possible unidirectional mitochondrial introgression from R. dorbignyi to R. bernardoi was reported by Pereyra et al. (2016a). Evidence that supports this hypothesis comes from the well-supported incongruence between the independent analyses of the mitochondrial and nuclear genes: R. bernardoi is deeply nested within R. dorbignyi in the mitochondrial analysis, but not in the nuclear analysis.
Our results from independent mitochondrial and nuclear analyses (rMD and rND, respectively) also show incongruence in the position of the specimens of Rhinella horribilis s.l. We recover this species deeply nested within (morphologically similar) species of the R. marina Group in the rND analysis, whereas in the rMD analysis it is recovered as sister of all the species of the R. crucifer + R. marina Groups. Another striking characteristic of this case of hybridization is the origin of these mitochondria, which is not traceable to any known extant species. These particular forms of deep mitochondrial divergence were denominated “ghost introgressions” (see Zhang et al., 2019). This kind of event involving deep mitochondrial divergence that implies past mitochondrial introgression from an unknown and not closely related species is uncommon in anurans. Historical interspecific introgressions events were reported in several groups of Anura: Ameerega (Dendrobatidae; Brown and Twomey, 2009); Anaxyrus, Bufo, and Bufotes (Bufonidae; Malmos et al., 2001; Yamazaki et al., 2008; Colliard et al., 2010; Dufresnes et al., 2019); Bombina (Bombinatoridae; Hofman and Szymura, 2007; De Cahsan et al., 2019); Dyscophus (Microhylidae; Orozco-terWengel et al., 2013); Hyla (Hylidae; Lamb and Avise, 1986; Bryson et al., 2010, 2014; Klymus et al., 2010); Mantella (Mantellidae; Crottini et al., 2019); Pelophylax and Rana (Ranidae; Liu et al., 2010; Zhou et al., 2012; Eto et al., 2013); Quasipaa (Dicroglossidae; Zhang et al., 2018); and Scutiger (Megophryidae; Chen et al., 2009). However, most of these events (except in Bombina, Bufotes, Quasipaa, and Scutiger) occurred among closely related species. Another striking characteristic of this phenomenon in Rhinella horribilis s.l. is that after the ancient introgression, the GIM (i.e., the mitochondrial DNA) diversified into two divergent clades (UPDs >3.33%). We consider most plausible the hypothesis that these mitochondrial clades represent two different species (R. horribilis s.s. and Rhinella sp. 1) that are not fully detectable (e.g., recovered as monophyletic) with our limited nuclear dataset. More intense genomic and phylogeographic sampling will be necessary to eventually solve the taxonomic status and puzzling history of R. horribilis and its lineages.
The reports of hybridization and mitochondrial introgression in Rhinella suggest the need for an extensive and careful exploration of these phenomena in other lineages of Bufonidae. The particular reproductive biology (i.e., scramble competition), the occurrence of broad sympatric areas between related species, and genetic features (i.e., complete reproductive isolation with relatively high levels of genetic divergence; Malone and Fontenot, 2008) of many bufonids may facilitate the occurrence of these phenomena. The identification of foreign mitochondrial genomes is particularly relevant to avoid errors both in phylogeographic and taxonomic studies (especially DNA barcoding studies) and phylogenetic inferences (Ballard and Whitlock, 2004; Alves et al., 2006; Obertegger et al., 2018; Barley et al., 2019). Moreover, the identification of mitochondrial introgressions will serve, among other things, as a base for future studies on adaptive coevolution between these organelles and the nuclear components of the oxidative metabolism of the cell (Hill, 2019).
Comments on the Phenotypic Evidence Considered for Rhinella
Our phenotypic sampling results in some synapomorphic and/or diagnostic characters for several internal clades of Rhinella, including most of the species groups. However, an evaluation of the available information for the character systems used clearly shows large gaps in the knowledge within each species group/clades (see fig. 15).
In general, there is relatively more information available for species of the Rhinella marina Clade. Within the R. margaritifera Clade, characters on adult osteology and musculature, natural history, and larval morphology are poorly known, and characters of larval chondrocranium, cytogenetics, and embryonic morphology are virtually unknown. This is a source of ambiguity in the reconstruction of ancestral character states for many characters that optimize in more inclusive nodes (e.g., oviposition mode within the R. festae and R. veraguensis Groups).
With the exceptions of foot and hand musculature, external larval, and embryonic morphologies described for several species of Rhinella (e.g., Mercês et al., 2009; Tolledo and Toledo, 2010; Blotto et al., 2014; Vera Candioti et al., 2016, 2020; Grosso et al., 2020; B.L.B., personal obs.), detailed descriptions considering ontogenetic variation, sexual dimorphism, and inter-population variations are still largely necessary. It must be noticed that these and many other character systems are promising as additional sources of evidence to be included in future phylogenetic analyses. Some examples of variation within species of Rhinella were reported on bio-acoustics (W.F. Martin, 1972; De la Riva et al., 1996; Guerra et al., 2011; Andrade et al., 2015; Valencia-Zuleta et al., 2020); integument and parotoid macroglands structure (O'Donohoe et al., 2019); anatomy of urogenital and digestive systems (Stohler, 1932; Lynch and Renjifo, 1990; but see Grant, 2000); clutch and egg size (Liedtke et al., 2014; Pereyra et al., 2015); mandibular, pelvic, and thigh musculature (Noble, 1922; Limeses, 1964, 1965; Trueb, 1971;; Winokur and Hillyard, 1992; Grant and Bolívar-G., 2014); and secretions (Cei et al., 1968; Maciel et al., 2006; Rodríguez et al., 2017). An inclusive sampling considering all these characters will contribute to the study of patterns of evolution of different character systems and their functional and adaptive significance.
The Fossil Record of Rhinella and Calibration Points
As is common for most neobatrachian anuran families, the pre-Pleistocene fossil record of Bufonidae is deficient, and most of these specimens lack an apomorphy-based diagnosis to unambiguously assign them to particular nodes or species (see Parham et al., 2012). The currently known pre-Pleistocene fossils of Rhinella are phylogenetically concentrated within the R. marina Group: (1) R. arenarum (as R. pisanoi) from Pliocene outcrops (3.9–3.2 Ma) of coastal Buenos Aires Province, Argentina (Casamiquela, 1967; Pérez-Ben et al., 2014); (2) R. loba, an extinct species putatively related to R. arenarum, from the late Pliocene (4.5–3.2 Ma) from the Chapadmalal Formation of Argentina (Pérez-Ben et al., 2019); (3) R. marina from the mid Miocene (13.8–11.8 Ma) from La Venta fauna of Colombia (Estes and Wassersug, 1963); and (4) R. aff. arenarum and Rhinella sp . marina Group from the upper Oligocene (29–26 Ma) of Salla, Bolivia (Báez and Nicoli, 2004). Another fossil from the upper Paleocene (59.2–56 Ma) from Itaboraí, Brazil (Estes, 1970) was also considered as related to some of the South American species groups of the former Bufo, but all proposed associations are vague and tentative (see Estes, 1970; Estes and Reig, 1973; Báez and Gasparini, 1977); even an association with Rhinella is controversial. Only the Miocene specimen of R. marina has been used as a calibration point (along with few other non-Rhinella bufonid fossils) in divergence dating analyses of Bufonidae or its internal clades (e.g., Pramuk et al., 2008; Maciel et al., 2010; van Bocxlaer et al., 2010; Liedtke et al., 2016, 2017). These studies differ in the sampled taxa and genes, and their results are not fully congruent, but the divergence-time estimates, considering relaxed molecular clocks and similar calibration points, indicate a split time between Rhinella and Anaxyrus + Incilius around 41 Ma (34–47 Ma; Pramuk et al., 2008) and 38.7 Ma (28.5–51.8; van Bocxlaer et al., 2010). However, the absence of an understanding of the interspecific osteological variation in species of Rhinella and the absence of an apomorphy-based determination of the fossils could result in the association of fossils to a lower-level taxon than the data can demonstrate (see Bever, 2005; Parham et al., 2012). Consequently, a critical reexamination of the available pre-Pleistocene fossils of Rhinella, along with an extensive study of living species of all the species groups is necessary before their defensible use as calibration points in divergence dating analyses. If material from Itaboraí can be unambiguously associated with Rhinella, its inclusion will provide a crucial point of calibration that could modify extensively our current understanding of the patterns of diversification of Rhinella and also of Bufonidae.
CONCLUSIONS
Our results provide a general framework for better understanding the evolution and taxonomy of Rhinella and its internal clades. The main results of our work include: (1) the generation of a well-supported phylogenetic hypothesis of the genus resulting from a total evidence analysis of most of its specific diversity, (2) the redefinition and morphological diagnosis of its species groups, (3) the demonstration of hybridization and mitochondrial introgression between some species, and (4) evaluation of the taxonomic status of several species. Nevertheless, many challenges are still pending. For example: (1) the taxonomic revision of many clades, including the designation of neotypes for several taxa; (2) the evaluation of the ontogenetic and intersexual variation in several problematic taxa; (3) the use of denser gene sampling (with high throughput sequencing) to better understand the evolutionary relationships in poorly supported clades and evaluate the role of the introgressive hybridization in the evolution of some lineages of Rhinella; and (4) the incorporation of more phenotypic characters to better understand their evolution in this group and define many morphologically and ecologically diverse clades of the genus. Future studies addressing these problems would result in crucial contributions in the knowledge of the diversity of Rhinella.
ACKNOWLEDGMENTS
We sincerely thank José M. Padial (Bronx Community College, New York); Thelia M. Céspedes Alejabo (CINBIOTYC, Peru); Guarino Colli and Helga Wiederhecker (Coleção Herpetológica da Universidade de Brasília, Brazil); Jimena Grosso (Fundación Miguel Lillo, Argentina); Juan M. Daza (Grupo Herpetológico de Antioquia, Colombia); Diego A. Barrasso and Leonardo Cotichelli (Instituto de Diversidad y Evolución Austral, Centro Nacional Patagónico, Argentina); Albertina P. Lima (Instituto Nacional de Pesquisas da Amazônia, Brazil); Antoine Fouquet (Laboratoire Evolution & Diversité Biologique, Université Toulouse, France); Juan M. Boeris, Andrés E. Brunetti, Dario E. Cardozo, and Juan M. Ferro (Laboratorio de Genética Evolutiva, Instituto de Biología Subtropical, Argentina); Sebastián Barrionuevo, Nadia G. Cervino, Laura Nicoli, and Santiago J. Nenda (Museo Argentino de Ciencias Naturales, Argentina); César Aguilar (Museo de Historia Natural, Universidad Nacional Mayor de San Marcos, Lima, Perú); Beatriz Álvarez Dorda and Isabel Rey Fraile (Museo Nacional de Ciencias Naturales, Spain); personnel of the herpetological collection of Museo de Biodiversidad (MUBI, Perú); Andrés Pansonato and Tiago Pezzuti (Universidade Federal de Minas Gerais, Brazil); Francisco Brusquetti, Ariadne Fares Sabbag, and Mariana Lyra (Universidad Estadual Paulista, Brazil); and Pedro H. Dias, Juliana Jordão, Julia S. Parreiras, Paulo D.P. Pinheiro, and Miguel Rodrigues (Universidade de São Paulo, Brazil) for sharing with us specimens, tissue samples and/or DNA sequences. For access to collections and specimen loans we thank Ariane Silva and Fernanda Werneck (Instituto Nacional de Pesquisas da Amazonia, Brazil); Rafe Brown, Richard Glor, Linda Trueb, and Luke Welton (Kansas University, Lawrence); Marta Calvo (Museo Nacional de Ciencias Naturales, Spain); Aline Staskowian Benetti (Museu de Zoologia da Universidade de São Paulo, Brazil); Carol Spencer and Natasha Stepanova (Museum of Vertebrate Zoology, Berkeley, California); and Eric Smith (University of Texas at Arlington). Pedro H. Dias; Ana Duport Bru, M. Laura Ponssa, and M. Florencia Vera Candioti (Fundación Miguel Lillo, Argentina); and Marinus S. Hoogmoed (Museu Paraense Emilio Goeldi, Brazil), contributed with valuable information on taxonomy or morphology of several specimens. Michael Jowers (Porto University) helped us to clarify the identity of some sequences of R. beebei and R. humboldti. M. Jimena Gómez Fernández, M. Daniela Pereyra, and Araceli Seiffe (Museo Argentino de Ciencias Naturales, Argentina) provided technical laboratory assistance. Several colleagues kindly shared with us photographs and/or data of species of Rhinella: César L. Barrio-Amorós (R. sternosignata), Michelle Castellanos (R. sclerocephala), Andrés Cecconi (R. diptycha), Pedro H. Dias (R. acutirostris), José Gerardo Espinoza (Rhinella sp . 14), Peter Janzen (R. crucifer), Konstantin D. Milto (type material of R. margaritifera), Arturo Muñoz (R. quechua), Roberto L.M. Novaes (R. icterica), Mirco Solé (R. hoogmoedi), Mauro Texeira Jr. (R. ornata), Rodrigo Tinoco (R. ocellata), Vicente Valdés Guzman (R. atacamensis), and Mario Yañez-Muñoz (R. cf. roqueana). David Blackburn, contributors, and the staff of Morphosource (Duke Library Digital Repository) very kindly made available µCT images of relevant species. TNT was provided free by the Willi Hennig Society. We thank Consejo Nacional de Investigaciones Científicas y Técnicas (CONICET), ANPCyT, Fundação de Amparo à Pesquisa do Estado de São Paulo for financial support: PIP 11220110100889; PICT 2013-0404, 2015-0813, 2015-0820, 2015-2381, 2017-2437, and 2018-3349; and grants #2012/10000-5, #2013/20423-3, #2013/50741-7, #2014/03585-2, #2015/11237-7, #2016/25070-0, and #2018/15425-0 from São Paulo Research Foundation (FAPESP). M.R. was supported by a PNPD post-doctoral fellowship from the Brazilian Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES grant 2016.1.263.41.6). J.M. thanks Ministry of Culture of the Czech Republic (grant DKRVO 2021/6.VI.c National Museum Prague, 00023272). S.C.F. was supported by CNPq (grant 312744/2017-0) and the PrInt program of Coordenação de Aperfeiçoamento de Pessoal de Nivel Superior, CAPES, Brazil (grant 88887.508359/2020-00). T.G. was supported by research fellowships from the Brazilian Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq grant 306823/2017-9). C.F.B.H. thanks CNPq for a research fellowship (grant 306623/2018-8). The Centro de Estudos de Insetos Sociais, I.B., UNESP, Rio Claro allowed access to its molecular laboratory facilities for the production of some sequences used in this study. Field and laboratory work in Ecuador were funded by Secretaría Nacional de Educación Superior, Ciencia, Tecnología e Innovación del Ecuador SENESCYT (Arca de Noé initiative; S.R.R. and Omar Torres principal investigators) and grants from Pontificia Universidad Católica del Ecuador, Dirección General Académica. B.L.B. acknowledges Esteban O. Lavilla for his support during initial studies on the R. spinulosa Group. We greatly appreciate the critical reviews of the manuscript by Aaron Bauer and Joseph Mendelson.