A Review of the Fossil record of Gymnophiona (Tetrapoda; Lissamphibia) with Comments on Its Use to Calibrate Molecular Timetrees

Gymnophiona, the most poorly known group of extant amphibians, includes elongated limbless tetrapods, with compact ossified skulls and reduced eyes, mainly adapted to fossorial life (only the Typhlonectidae exhibits adaptations for an aquatic or semiaquatic behavior). Caecilians are poorly represented in the fossil record, and despite the low number of fossil specimens described until now (only four taxa, in addition to indeterminate fragmentary material), their fossils play a key role in the knowledge of Lissamphibia origin and evolution, as well as contribute directly to a better understanding of phylogeny, taxonomy and biogeography of extant gymnophionan taxa. These records are scattered throughout geological time (from the Jurassic to the Neogene) and space (they are represented only on North and South America and Africa). Here, we revisit the caecilian fossil record, providing a brief description of all known extinct taxa described so far, along with general remarks about their impact on systematics, time range and geographic distribution of the clade, as well as prospects for future research. Possible calibration constraints based on the caecilian fossil record are provided.


Introduction
The crown-clade Lissamphibia (see Laurin et al., 2020 for a review; but see Dubois, 2004 for an opposing view on the use of this nomen) comprises the extant taxa Anura, Urodela, and Gymnophiona. Although lissamphibians are diverse in present day biotas (Frost, 2020), their fossil record is relatively scarce, and includes only a few, but important, specimens whose preservation status is sufficiently satisfactory to allow detailed diagnoses (Schoch & Millner, 2004;Marjanović & Laurin, 2019). This scarcity is particularly pronounced for gymnophionans. For many years, only one gymnophionan was known in the fossil record (Estes, 1981), and to date only four extinct taxa originally assigned to this group were erected and described in details (Estes & Wake, 1972;Jenkins & Walsh, 1993;Evans & Sigogneau-Russel, 2001;Pardo et al., 2017).
The clade Gymnophiona is moderately diverse, with approximately 214 known extant species (Frost, 2020). Popularly known as caecilians, these animals are well adapted to a fossorial existence, as shown by their elongated body, the absence of limbs and girdles, a compact and well-ossified skull, and reduced or vestigial eyes (Duellman & Trueb, 1994;Wilkinson & Nussbaum, 2006;Wilkinson et al., 2011). However, a subgroup of gymnophionans, the typhlonectids, exhibits an aquatic or semi-aquatic lifestyle (Taylor, 1968). Other distinct characteristics of caecilians include a dual mechanism for jaw closing and a pair of sensitive organs between the eyes and nostrils, known as tentacles (Wilkinson & Nussbaum, 2006).
The first caecilian fossil species described was Apodops pricei Estes & Wake, 1972, a crown-gymnophionan from the Early Eocene of Brazil consisting only of an isolated pre-cloacal vertebra (Duellman & Trueb, 1994). Later, Eocaecilia micropodia Jenkins & Walsh, 1993, found in Early Jurassic rocks of the United States, was described based on numerous specimens with cranial and postcranial elements, including limbs and girdles, both completely lost in all extant species, but predictable in stem-gymnophionans (Jenkins et al., 2007). Subsequently, a taxon from the Lower is more likely to be a stereospondyl, but its affinities with gymnophionans are more dubious. Similarly, Carroll & Currie (1975), and more recently Anderson et al. (2008), suggested a sister-group relationship between the Early Permian lepospondyl Rhynchonkos and caecilians. However, a detailed CT-scan analysis of its morphology suggested that similarities previously regarded as synapomorphies between the recumbirostran microsaurs and gymnophionans result from ambiguities in previous character definitions and convergent evolution due to a fossorial ecomorph (Szostakiwskyj et al., 2015).
Although the fossil record of caecilians is undoubtedly scarce, our present knowledge provides sufficient clues of their past history to allow a more comprehensive approach combining information from both extinct and extant taxa. Here, we provide a review of the gymnophionan fossil record and discuss aspects of the anatomy, taxonomy, phylogeny and biogeography of extinct groups, as well as their implications for our understanding of extant gymnophionans.

Phylogeny and Classification of caecilians
The classification and definition of Gymnophionan clades varies according to authors. Trueb & Cloutier (1991) proposed to restrict the term Apoda Oppel, 1811 for the crown-group of caecilians and Gymnophiona Rafinesque-Schmaltz, 1814 for the stem-group including Apoda. However, the fact that the former name was preoccupied by several earlier nomina (Dubois, 2004), along with a possible misunderstanding in some statements about caecilian characteristics (such as the generalization of the limbless condition of gymnophionans), led some authors to reject these definitions (e.g. Dubois, 2004;Wilkinson & Nussbaum, 2006). Furthermore, as pointed out by Wilkinson et al. (2011), the use of the name Gymnophiona for the crown-group is already well established in the literature, and a change in it will probably bring more problems than create solutions. To avoid this problematic situation, for the stem-based clade that comprises extant caecilians plus extinct taxa, such as E. micropodia and R. monbaroni, Marjanović & Laurin (2008a) proposed the term Gymnophionomorpha.
According to the stem-based definition of Gymnophionomorpha, this clade comprises all lineages more closely related to the crown-clade Gymnophiona than to Batrachia. Therefore, this taxon encompasses E. micropodia, R. monbaroni and extant caecilians ( Figure 1). Under the phylogeny proposed by Pardo et al. (2017), it would also include C. jenkinsi, and all other stereospondyls, plus other (but not all) temnospondyls, such as archegosaurids and eryopoids. Members of the Gymnophionomorpha (under their currently accepted delimitation) are characterized by the presence of pseudodentary and pseudoangular forming the lower jaw, os basale and absence of tympanic ear.
Recent large-scale molecular analyses strongly corroborate the monophyly of extant Lissamphibia with respect to Amniota, and most also find caecilians placed as the sister-group of Batrachia, which includes Anura and Urodela (Frost et al., 2006;Pyron & Wiens, 2011). These results also stand in a total evidence analysis (Pyron, 2011) based on a molecular data set designed to be combined with a mainly fossiloriented data matrix (Vallin & Laurin, 2004). However, recent morphological approaches designed to test the phylogenetic affinities of lissamphibians within an expanded taxon sampling of Paleozoic tetrapods resulted in fundamentally distinct hypotheses on the origin of the group (see Ruta & Coates, 2007;Marjanović & Laurin, 2019). Currently three main hypotheses try to explain this question, all of which display minor variants (Figure 2). The first ( Figure 2A) considers that Lissamphibia is a monophyletic group inside Temnospondyli (e.g. Trueb & Cloutier, 1991;Ruta & Coates, 2007). The second ( Figure 2B) also recognize the monophyly of lissamphibians, but nested within Lepospondyli (e.g. Marjanović & Laurin, 2008a, 2019. The third (Figures 2C and 2D) suggest that extant amphibians do not actually form a monophyletic group, because frogs and salamanders are temnospondyls and caecilians lepospondyls (e.g. Anderson et al., 2008). In some variants, gymnophionans are more closely related to amniotes than to batrachians (e.g. Anderson et al., 2008), whereas this is contradicted by nearly all molecular (e.g. Irisarri et al., 2017), total evidence (Pyron, 2011) and some paleontological (Marjanović & Laurin, 2009, 2019 phylogenies. The hypothesis recently proposed by Pardo et al. (2017) is compatible with molecular phylogenies to the extent that extant amphibians form a clade that excludes amniotes ( Figure 2E). According to this hypothesis, caecilians and batrachians had separate origins, with caecilians being nested among stereospondyls, whereas batrachians are dissorophoids. However, as showed by Marjanović & Laurin (2019) after a reanalyzes of the data, this topology is not robust. In any case, there is no consensus about the phylogenetic relationships between the three extant groups of amphibians and their Paleozoic relatives, and more evidence from distinct data sources, such as developmental biology, CT-Scan and molecular data, can be used to discriminate between the various hypotheses (e.g. Szostakiwskyj et al., 2015).

Time Range of Gymnophionomorpha
Due to the scarcity of the amphibian fossil record, time divergence estimates are relatively inaccurate and vary considerably according to the methodology and data source used (Marjanović & Laurin, 2007). Some works suggest that the appearance of the amphibian crown occurred most likely in the Early Carboniferous, approximately 318-359 Ma (e.g. Pyron, 2011;Pardo et al., 2017). However, subsequent studies found a much younger origin for amphibians, in the Permian, approximately 300-250 Ma ago (e.g. Marjanović & Laurin, 2007;2008b). There is no consensus on this, but any further tests should use fossil data, including stem caecilians, to achieve robust results.
If C. jenkinsi (along with many other temnospondyls) is indeed a gymnophionomorph, the origin of Lissamphibia and Gymnophionomorpha occurred during the Late Carboniferous (Pardo et al., 2017). Although these results are congruent with some previous time divergence estimates based on molecular data (e.g. Roelants et al., 2007;Zhang & Wake, 2009;San Mauro, 2010), it is incompatible with others (e.g. Marjanović & Laurin, 2007). It is compatible with the divergence times obtained from total evidence tip dating of Pyron (2011), but it is incompatible with its topology.
Clearly, more evidence is required to corroborate this hypothesis.
The two other caecilian stem lineages, represented by E. micropodia and R. monbaroni, date from Early Jurassic and Early Cretaceous, respectively. The age of the crown-group Gymnophiona is poorly constrained, with estimates ranging from Early Jurassic, approximately 188 Ma (Kamei et al., 2012) to about 100 Ma, near the Jurassic/Cretaceous boundary (Marjanović & Laurin, 2007;Pyron, 2011). Fossils attributed to caecilian crown are limited to isolated remains, mainly vertebral elements too fragmentary to allow a more specific taxonomic assignment. They are known from the Cretaceous of Sudan and Bolivia (Evans et al., 1996;Gayet et al., 2001) Accounting for the gymnophionomorph fossil record at the beginning of the Mesozoic, a distribution concentrated at least in northern Pangea is well established (Pyron, 2014). However, either a northern origin followed by dispersal into austral lands or a southern origin and subsequent radiation to the septentrional areas were proposed, as observed by Evans & Sigogneau-Russel (2001). The gymnophionan crown-clade was probably already widespread in southern landmasses prior to its breakup during the Cretaceous, as shown by the predominantly gondwanan distribution of extant taxa and the Cretaceous record of R. monbaroni (Duellman & Trueb, 1994;Evans & Sigogneau-Russell, 2001). However, the presence of Eocaecilia in North America is compatible with a Laurasian origin of Gymnophionomorpha.
Therefore, because of its scarcity, the fossil record provides limited biogeographical data, and information from extant taxa, instead of fossils, are preferably used in biogeographic hypotheses (Gower et al., 2002). Mainly with the discovery of new and more complete caecilian crown fossils and ancient Gondwanan stem-group remains, paleontological data can help to elucidate the biogeographic patterns of Gymnophionomorpha evolution.

Reasons for the scarcity of Gymnophionomorpha in the fossil record
Although the ecology and behavior of caecilians remain poorly documented (e.g., Jared et al., 1999Jared et al., , 2018Wilkinson et al., 2008), most are known to be fossorial, except for typhlonectids that are aquatic or semi-aquatic (Taylor, 1968;Ducey et al., 1993). The fossorial lifestyle could, under some circumstances facilitate fossilization by reducing significantly the negative effects associated with transport that generally occurs prior to burial. However, this would enhance fossilization only if caecilians lived in environments where sedimentation occurs. Given that fossorial caecilians live in the uppermost layers of emerged soil, they are unlikely to be fossilized there, and transport of the carcass to an environment more conducive to fossilization is unlikely, unless their body is exposed by a scavenger or by quick erosion prior to decay. Aquatic caecilians are more likely to be fossilized because some of their environments, like braided rivers and lakes, are often preserved in fossiliferous deposits (Behrensmeyer et al., 2000). Therefore, the gymnophionan fossil record is probably biased in favor of aquatic taxa and of the earliest (stem) caecilians that must have been surface dwellers.
Other factors may contribute to the scarcity of the caecilian fossil record. One being the fact that extant caecilians are mainly distributed in tropical regions around the world, a type of environment characterized by high levels of biological activity in decomposition and carbon cycling of remains in acid soils, hampering the fossilization process (Tappen, 1994, but see Peterhans, 1993 for a different perspective). The second is related with size, because usually, larger vertebrate fossils are more noticeable than smaller ones and have a greater fossilization potential (Behrensmeyer et al., 2000), although this effect should be offset to an extent by the much greater number of small animals, which reflects obvious resource limitations (Kozlowski & Gawelczyk, 2002).
Although a few caecilian species reach more than 1 m, almost the entire group is formed by smaller animals, with few decimeters of length (Renous & Gasc, 1989).
Finally, Gymnophiona represents one of the least studied tetrapod groups, with a limited number of scientists dedicated to their study (Wilkinson & Nussbaum, 2006). Thus, the combined effects of all these factors can help explain the rarity of caecilians in the fossil record.

Chinlestegophis: a true gymnophionomorphan?
The skull of the Triassic Chinlestegophis jenkinsi has been interpreted as displaying a combination of stereospondyli plesiomorphies, along with gymnophionan and lissamphibian synapomorphies, but also exhibiting uniquely derived features.
Autapomorphies include a dorsomedial orbital margin formed mainly by a long anterior process of the postfrontal, a short contact between parietal and tabular, and a finger-like process of the prefrontal connected with a notch on the postfrontal.
However, plesiomorphic features typical of triassic stegocephalians also concedes quite conservative traits to the skull of C. jenkinsi. For instance, unlike extant caecilians (Wake & Hanken, 1982;Nussbaum, 1983), the lower jaw of C. jenkinsi is composed by almost all typical tetrapod bones, including distinct dentary, coronoid, Permian Gerobatrachus and Batrachia. However, lissamphibians also display a large gap in their fossil record under the lepospondyl hypothesis (Marjanović & Laurin, 2008b, 2009, 2013, 2019. If C. jenkinsi is closely related to caecilians, features interpreted as adaptations for a fossorial lifestyle, including bone fusions and loss or reduction of limbs, girdles and orbits occurred more gradually than previously thought. But in any case, these characters show homoplasy and developed early because they are present in some Permo-Carboniferous lepospondyls (Carroll & Gaskill, 1978).
It seems that the affinities of C. jenkinsi still represents an open question that will need to be evaluated in subsequent phylogenetic analyses with further revised data matrices.

Stem-Gymnophionomorpha
Several features in caecilian morphology, such as the stegokrotraphic skull, fusion or loss of bones, and serpentiform body, were identified as adaptations for a fossorial lifestyle (Wilkinson & Nussbaum, 2006;Sherratt et al., 2014). According to phylogenies that include only extant taxa, a closed skull roof evolved later in caecilian lineages, while the primitive rhinatrematids retain the plesiomorphic zygokrotraphic pattern. This scenario of gradual evolution towards the closure of cranial fenestrae was not corroborated with the description of E. micropodia, which bears a well-ossified stegokrotraphic skull (Jenkins & Walsh, 1993). According to the accepted topology illustrated in Figure 1 Walsh, 1993). The presence of limbs is less certain in R. monbaroni because a femur was only tentatively attributed to it, based mainly in the presence of trochanteric crest, a trait also observed in E. micropodia (Evans & Sigogneau-Russell, 2001).

Crown-Gymnophiona
With exception of the cranial material from Uganda, all crown-gymnophionan fossils are limited to isolated vertebrae. Due to their typical morphology, caecilian vertebrae are easily distinguishable, bearing an amphicoelous, medially constricted centrum, large parapophysis, low and flat neural arch, neural spine short and a welldeveloped ventral keel (Wake, 1980). However, the caecilian postcranial elements, with exception of the atlas and other anteriormost vertebrae (Taylor, 1977), are quite conservative among gymnophionan subgroups, according some authors (e.g. Wilkinson et al., 2011). Therefore, the attribution of such fossils to Gymnophiona seems to be unequivocal, but a more accurate and specific identifications are uncertain. Evans et al. (1996) noted that the Sudanese fossil trunk vertebrae lack a characteristic common in Scolecomorphidae, the presence of a posteriorly projected process in the basapophyses, and thus cannot be assigned to, at least, the scolecomorphid crown. However, exclusively shared features with other African taxa are absent. The vertebra of Tiupampa described by Rage (1991) exhibits an amphicoelous centrum and a well-developed parapophyses, however is too much damaged to allow a more detailed identification.
In the description of the vertebra assigned to A. pricei, Estes & Wake (1972) noticed that its morphology and size proportions closely resemble extant genera of West Africa (Geotrypetes) and Central America (Dermophis and Gymnopis). Curiously, similarities with Siphonops and other taxa commonly found in Brazil were considered less compelling. The Colombian fossil vertebrae described by Hecht & LaDuke (1997) are morphologically similar to extant South American species, except for its size, as they are three to four times larger than the compared taxa. Remarks: According the several morphologic characteristics mentioned above, R.
monbaroni was interpreted as a stem caecilian, but closer related to crown groups than  Besides the well-developed ventral spine, the amphicoelous centrum also exhibits a pronounced medial constriction. Even though most of the parapophyses have been lost during fossilization process, the broad base can be used to infer its large size.
The neural arch is flat and low, with a short neural spine limited to its anterior half and two lateral deep groves that extends to the rib-bearing surface. The vertebra also bears two large flanges connecting pre and postzygapophyses. Remarks: According to Taylor (1977), anterior trunk vertebrae and more importantly the atlas of gymnophionans contain phylogenetic information. Unfortunately, none of these materials were preserved in a crown caecilian fossil. Therefore, due to the lack of most specific diagnostic characters, all these specimens were assigned to Gymnophiona indet.

Possible indeterminate record
In a faunal list of Maboko Island, originally published by Andrews et al. (1981, table 1), a record assigned to a Miocene Nectridia was reported, but without images or detailed descriptions. Due to the significant temporal gap that this record would imply in the nectridian fossil record (nectridians were presumably extinct at the end of the Permian), subsequent works considered that this material is actually a lissamphibian, probably a salamander (Van Dijk, 1995) or a caecilian (Gardner & Rage, 2016).
However, until the material is reevaluated, such assignments remain uncertain, though both hypotheses are much more likely than the initial nectridian assignment.

Possible calibration points for Gymnophionomorpha
Calibration constraints based on the caecilian fossil record are uncommon.
Probably due to its scarcity, usually other taxa are used, such as batrachians (e.g. San Mauro et al., 2014). Following the recommendations of Parham et al. (2012), here we provide calibrations for four nodes of the gymnophionomorphans. Calibrations are highly dependent of phylogeny and stratigraphy, and thus the latest, best-supported dating and phylogenetic hypotheses were considered.

Oldest fossil: Eocaecilia micropodia Jenkins & Walsh, 1993
Phylogenetic Justification: Phylogenetic analyzes have repeatedly confirmed the relationship between E. micropodia and extant caecilians (e.g. Maddin et al., 2012), and its position as a sister group of all other gymnophionomorphs is supported by many characters (discussed in more detail above) and is widely accepted. Gymnophionans remain the most poorly known group of tetrapods, particularly in aspects of their evolutionary history. Despite the paucity of its fossil record, fossil caecilians directly affect our understanding of taxonomy, phylogeny and biogeography of extant caecilians, and they help to discriminate between hypotheses about the origin of Lissamphibia.
The gymnophionan fossil record shows wide temporal gaps, even in the Cenozoic, in which (not considering A. pricei) no diagnostic material has been assigned to an extant species, genus or even family so far. Therefore, the discovery of new, more complete and diagnostic fossils assigned to the caecilian crown-group may contribute to resolve phylogenetic and biogeographic questions about caecilian clades and better constrain molecular clocks.