Out of sight, but not out of mind: a name for the Stefania (Anura: Hemiphractidae) from the summit of Murisipán-tepui (Bolívar State, Venezuela)

Abstract. Previous molecular analyses of the frog genus Stefania have shown that species boundaries in that group are often difficult to delineate when solely based on morphology. As a consequence, “taxonomically cryptic” species are not uncommon in the genus. Several highland Stefania species remain to be described, some potentially critically endangered due to their highly restricted geographic ranges. One case is the microendemic Stefania population from the summit of Murisipán-tepui, a poorly explored table-top mountain in the Los Testigos Massif, a small tepui mountain range located north to the much larger Chimantá Massif in southern Venezuela. That population, mistaken as S. satelles for two decades, was later reported as Stefania sp. 2 and belongs to the S. ginesi clade. The new species is phylogenetically distinct but phenotypically similar to S. satelles, a taxon restricted to its type-locality, i.e. the summit of Aprada-tepui in Venezuela. The new species is described based on morphology and cranial osteology. Molecular divergences with S. satelles are high (> 8%) in the barcoding fragment of 16S rRNA. Amended definitions for the two other described species in the S. ginesi clade (S. ginesi and S. satelles) are also provided. The new species should be listed as critically endangered according to IUCN criteria.


Introduction
The hemiphractid frog genus Stefania currently contains 20 species (Kok 2023) of small (max. 54.8 mm in S. ayangannae) to large (max. 97.5 mm in S. evansi) terrestrial or semi-arboreal, direct-developing frogs. Eggs and juveniles are carried on the back of the mother until fully developed (Duellman 2015). Average clutch size in the genus is reported to be ten (Salerno & Pauly 2012), with up to 30 near-term juveniles observed carried by a large female of S. evansi (Kok & Benjamin 2007). The genus Stefania is one among many ancient (near-) endemic lineages of amphibians inhabiting the biodiverse Pantepui biogeographical region in the Guiana Shield Highlands of northern South America. Pantepui is so remote that it is often referred to as the "Lost World", in reference to the famous novel by Arthur Conan Doyle (1912) in which the author depicts the exploration of an isolated mountain plateau harbouring a relict fauna. Murisipán-tepui is a small, poorly surveyed sandstone table-top mountain (tepui) located in the Los Testigos Massif in the Bolívar State of Venezuela, ca. 43 km north of the Chimantá Massif SPECIAL ISSUE: NEOTROPICAL HERPETOFAUNA RESEARCH PAPER Fig. 1. Distribution map of the Stefania ginesi clade as currently understood. Locality data are based on specimens examined (see Appendix S1) and literature records (Señaris et al. 1997, Kok et al. 2016, 2017. Photos Philippe J.R. Kok. and ca. 38 km east of Auyán-tepui, airline (Fig. 1). Its fractured summit plateau reaches a maximum elevation of ca. 2,440 m (Figs. 2-3), covers a total area of ca. 0.5 km 2 above 2,000 m elevation (Kok 2015), and is only accessible by helicopter (although technical climbs are possible). I explored the disjunct southern part of the summit of Murisipán-tepui for four days in June 2012, which resulted in the collection of one amphibian and two reptile taxa (one of the latter new to science, see Kok 2015). This small collection included five specimens of a frog species reported in the literature as S. satelles (e.g. Señaris et al. 1997, McDiarmid & Donnelly 2005. Previous authors had raised concerns about the identity of the Stefania populations from the Los Testigos Massif, noting a difference in tympanum size compared to S. satelles from the type locality (Señaris et al. 1997, Gorzula & Señaris 1999. Kok et al. (2016Kok et al. ( , 2017 demonstrated that four distinct species are likely confused under that name, all of them seemingly distributed on different tepui summits, with S. satelles being restricted to its type locality (Aprada-tepui, Bolívar State, Venezuela). The population from Murisipántepui was recovered as sister to three undescribed species in a clade that is sister to the clade containing S. ginesi and S. satelles from the type localities (Kok et al. 2016, 2017). Kok et al. (2017)

µCT scanning, 3D reconstructions, and osteology
The male holotype (IRSNB 4208) and one female paratype (IRSNB 4210) were µCT-scanned at the Natural History Museum, London (NHM, UK) CT Lab facility using a Nikon HMX225; osteological images were exported from the virtual 3D models, which were reconstructed and segmented using VGStudio MAX version 2.1. Comparative specimens of S. ginesi and S. satelles have been µCT-scanned, reconstructed, and segmented (using Dragonfly) by J. Brecko at the IRSNB's CT Lab facility using a RX EasyTom150. Three-dimensional mesh files of crania have been deposited either on the MorphoSource (holotype and paratype) or the Sketchfab platform (comparative specimens) (Table S1). Osteological terminology followed Trueb (1973) and Duellman (2015). The degree of contact between bony structures followed Kok et al. (2020), i.e. contacting/in contact = contact between structures with a visible suture line and fused = contact between structures with a suture line being barely visible or absent.

Genetic distances
Uncorrected pairwise genetic distances between Stefania sp. 2 and other species in the S. ginesi clade (including undescribed taxa) were calculated in MEGA X (Kumar et al. 2018) using 16S sequences deposited by Kok et al. (2012Kok et al. ( , 2016 and available on GenBank. Sequences were aligned using MAFFT v7.490 (Katoh & Standley 2013) on the CIPRES Science Gateway (Miller et al. 2010) with appropriate strategies automatically selected. Ambiguously aligned regions were manually removed, resulting in a final alignment of 533 nucleotides.

Results
Stefania lathropae sp. nov. Stefania satelles (in part) Señaris et al. 1997: 33-37 Stefania satelles (in part) Gorzula & Señaris 1999 In contact or fused but low and barely visible (in life/preservative); 7) frontoparietal crests barely developed (on cranium); 8) constriction of the frontoparietal bones at the level of the anterior epiotic eminence; 9) low, sometimes extensive, exostosis on the cranium; 10) premaxillae slightly projecting posteriorly in lateral view; 11) posterodorsal projection of maxilla absent or highly reduced, not in contact with orbital/zygomatic ramus of squamosal; 12) maxillary process of the nasal not in contact with the maxilla; 13) horizontal length of tympanum more than 50% horizontal length of eye in both sexes; 14) vomerine teeth 3-8; 15) toes basally webbed, no significant difference in toe webbing between sexes; 16) dorsal skin ( Thenar tubercle large, distinct, transversely oval; palmar tubercle distinct, about half the size of thenar tubercle (Fig. 6). Subarticular tubercles distinct, large, projecting, and single in appearance, except on FIV of the left hand where two basal tubercles are present (Fig. 6). Supernumerary tubercles numerous, most of them large and projecting. Relative finger lengths II < I < IV < III; adpressed second finger barely reaches the first finger's disc. Fingers with lateral fringes; narrow webbing between FIII-FIV, which are fused proximally. Finger discs large, transversely oval, max. two times wider than adjacent phalange, smallest and equal in length on FI and FII, largest and equal in length on FIII-IV. Largest disc width 68% of horizontal length of tympanum. No nuptial excrescences detectable.
Inner metatarsal tubercle large, oval, distinct; outer metatarsal tubercle about 2.5 times smaller than inner, round, distinct, very similar to surrounding supernumerary tubercles but distinctly darker (Fig. 6). Subarticular tubercles single, round, distinct, except the distal one on TIV, which is slightly bifid. Supernumerary tubercles numerous, round, distinct, most of them similar in size. Relative lengths of toes I < II < III < V < IV; adpressed fifth toe distinctly longer than third. Toes basally webbed, webs tapering to lateral fringes. Webbing formula I 2 + -3 II 2 -3 + III 2 + -3 ¾ IV 3 ¼ -2 V. Toe discs oval, slightly wider than adjacent phalange, largest toe disc on toe IV, distinctly smaller than largest finger disc. Heels overlapping when hindlimbs are folded at right angles to sagittal body plane.
Colour of the holotype in life: Dorsum golden brown with a few dark brown spots, and white dorsolateral stripes extending from the level of the tympanum to the sacrum (Fig. 2) Colour of the holotype in preservative: After 11 years in ethanol preservative (Fig. 5), the overall colouration changed considerably compared to the condition in life. All upper surfaces became dark brown, with some light brown spots on the back. The dorsolateral The skull is large, apparently slightly decalcified, slightly exostosed (covered with low bony growths) on the frontoparietals, widest at the level of the articulation of the quadratojugal and maxilla and wider than long (longest width ca. 130% of medial length). The braincase appears mostly ossified, except for the sphenethmoid complex, which is not fully ossified and not dorsally invested by the nasals. The exoccipital-prootic complex is not fully ossified and not laterally overlapped by the otic ramus of the squamosal. The paired septomaxillae are well developed and lie dorsal to the palatine process and posterolaterally to the articulation between the maxilla and premaxilla. The columellae (stapes) are ossified, formed by the synostotic fusion of the long, thin pars media plectri (stylus) and the pars interna plectri (baseplate), which is curved.
Cranial osteology: The nasals are broad, not exostosed, not contacting medially. The posteromedial margins of the nasals do not contact the sphenethmoid. The maxillary process of the nasal is long, acuminate, slightly curved in dorsal view (concave), not in contact with the maxilla. Lateral margins of the frontoparietal are slightly exostosed. The frontoparietal completely roofs the central braincase from the anterior level of the orbit to the level of the tectum synoticum posteriorly. Lamina perpendicularis poorly developed along the anterior orbital margin of the frontoparietal, but sharply expanding posteriorly. The frontoparietal slightly expands dorsolaterally to form a low frontoparietal crest, which bears low dermal sculpturing (exostosis).

Ventral investing and palatal bones:
The parasphenoid is nib-shaped, forming the floor of the braincase. The tip of the pointed cultriform process appears to be in contact with but not overlapping the sphenethmoid. The parasphenoid alary processes provides the floor for the otic capsules and are approximately perpendicular to the cultriform process. The posteromedial process of the parasphenoid is broadly acuminate and almost reaches the margin of the foramen magnum. The lateral arms of the parasphenoid are not in contact with the long medial ramus of the pterygoid. The massive neopalatine is fused posteromedially to the sphenethmoid. Postchoanal vomers are straight, clearly distinguishable, anterior to the neopalatine. Each vomer bears ca. 7-8 teeth. The neopalatine connects to the vomer via the postchoanal process and does not connect to the inner surface of the maxilla.
Maxillary arcade: A likely congenital maxillary prognathism is noted in the holotype. Both maxillae and premaxillae are dentate, apparently not exostosed. The premaxillae are separated medially, slightly inclined anteriorly. The alary processes of the premaxillae are broad and acuminate posteriorly, diverging from the midline. The alary processes are directed posterodorsally and have a moderate indentation along their anteromedial base. In dorsal view, the alary processes reach the level of the anteriormost margin of the maxillae. The palatine (medial) process of the premaxillae is long, acuminate, directed posterodorsally. The lateral process of the premaxillae is short, about half the size of the palatine process, acuminate and directed posterolaterally. The premaxillae do not contact the maxillae. The maxillae are greatly expanded and the pars facialis is well developed but not in contact with the maxillary process of the nasal. Anteriorly and in lateral view, the maxillae are squarish and more than twice as high as posteriorly. The maxillae possess a highly reduced posterodorsal projection directed towards the zygomatic ramus of the squamosal with which it is not in contact. Posteriorly, the maxillae are contacting but not fused to the robust quadratojugals.

Suspensory apparatus:
The triradiate pterygoid is moderately slender. The anterior ramus extends toward the braincase from the maxilla at the mid-orbit level and further extends against the anteroventral margin of the otic capsule via the long medial ramus. There is no contact between the otic capsule and the medial ramus. The posterior ramus is broad and flat, not in contact with the ventral ramus of the squamosal. The posterior ramus is slightly shorter than the medial one. The quadratojugal is robust, in contact with but not fused to the maxilla. The otic and ventral rami of the squamosal are well developed; the otic ramus does not extend over the lateral margin of the prootic. The ventral ramus of the squamosal is curved, narrow in lateral view, and extends from the quadratojugal to the posterodorsal margin of the orbit. The otic ramus is acuminate, smooth, and approximately half the length of the narrower, curved zygomatic ramus. The otic ramus bears a conspicuous low crest laterally. The zygomatic ramus is long, slim, and acuminate in lateral profile, not in contact with the maxilla. The zygomatic rami do not appear exostosed.
Mandible: The dentary is long and moderately stout, posteriorly acuminate, fused to the small, arcuate mentomeckelian bone anteriorly. The mentomeckelians are separated medially. The dentary overlaps almost half of the angulosplenial length. The main component of the mandible is the angulosplenial, which is long and weakly sigmoid, acuminate anteriorly. Anteriorly and in ventral view, the angulosplenial fails to extend near the maxillapremaxilla articulation. The coronoid process is dorsomedial and well-developed, about one-fourth of the posterior ramus.
Variation: Stefania lathropae sp. nov. is a polychromatic species (Figs. 2-3). Polychromatism was not found to be related to sex. Adult individuals range from plain medium brown, plain copper brown to golden brown with some more or less extensive dark brown mottling on the dorsal surfaces and flanks. Dorsal markings vary from none to a few dark brown ill-defined chevrons and conspicuous spots. In life, a canthal dark brown stripe is present in all specimens (faint in preservative), usually tapering towards nostrils. A faint whitish interorbital bar and white dorsolateral stripes are always present, although the latter vary greatly in length, sometimes not extending farther than the level of arm insertion, sometimes extending post-sacrum. The upper surface of limbs may be plain or with faint to conspicuous dark brown spots, but never bars. The dorsal skin is heavily tuberculate, but in some specimens these tubercles are very low. Males are distinctly smaller than females: 36.8-40.0 mm SVL in adult males (n = 2) vs. 50.6-55.5 mm SVL in adult females (n = 3). Unlike the holotype, the male paratype (IRSNB 4212) has white (in preservative) nuptial excrescences on the dorsal surface of the first phalange of the thumb. Head is slightly longer than wide, or slightly wider than long. The number of vomerine teeth varies between 3-8. There is no significant difference in toe webbing between the sexes; toe webbing formula for adult males (n = 2) is I 2 + -(3 -2 ½) II (2 + -2) -3 + III (2 + -2) -3 ¾ IV (3 ½ -3 ¼) -2 V; toe webbing formula for adult females (n = 3) is I (2 + -2) -(2 ¾ -2 ½) II 2 --(3 + -3) III 2 -(3 ½ -3 ¼) IV (3 + -3 ¼) -2 V. Although the cranium of the µCT-scanned paratype female specimen (IRSNB 4210) is slightly more calcified than the cranium of the holotype, no major difference has been detected in diagnostic cranial features. Exostosis is more extensive in the paratype specimen, as well as the sphenethmoid ossification. In the paratype specimen the sphenethmoid is dorsally invested by the nasals (vs. not invested in the holotype) and the exoccipital-prootic complex is more ossified and is laterally barely overlapped by the otic ramus of the squamosal (vs. no overlap in the holotype). No maxillary prognathism is noted in the paratype specimen and the premaxillary is slightly inclined posteriorly (vs. slightly inclined anteriorly in the holotype). Despite stronger ossification, the maxilla of the paratype specimen remains free of contact with the maxillary process of the nasal. Likewise, the orbital/ zygomatic ramus of the squamosal remains free of contact with the maxilla.
Distribution and natural history: Stefania lathropae sp. nov. is currently only confirmed from the type locality, i.e. the summit of Murisipán-tepui between 2,401-2,419 m elevation (Figs. 1-2). The species might be present on adjacent tepuis of the small Los Testigos Massif (e.g. Kamarkawarai-tepui where S. "satelles" has been reported; Señaris et al. 1997, Gorzula & Señaris 1999, McDiarmid & Donnelly 2005, but this should be confirmed. The summit of Murisipán-tepui is a mix of large areas of exposed rocks with some richly vegetated areas and deep fractures (Figs. 2-3). A seasonal lagoon (Fig.  2) is present ca. in the middle of the highest part of the summit (Huber 1995, Kok 2015. The northernmost part of the tepui summit is inaccessible by foot from the southernmost part due to a large fracture (Kok 2015).
Specimens of S. lathropae sp. nov. were exclusively found hiding under rocks. No active specimen, nor female carrying eggs or juveniles, were observed during my short stay on the summit of Murisipántepui (15-18 June 2012, i.e. during the rainy season). IRSNB 4211 (a female) and IRSNB 4212 (a male) were found under the same rock. Amended definitions for the two other described species in the S. ginesi clade (exclusively based on specimens from the type localities): Stefania ginesi (Fig. 8) is characterized by the combination of the following morphological characters: 1) a small species of Stefania, max. SVL in preserved females 55.0 mm, 49.0 mm in preserved males; 2) head not distinctly longer than wide, about as wide as long; 3) canthus rostralis poorly distinct, round, slightly concave, heavily granular/ tuberculate, distinct canthal stripe absent in life; 4) loreal region with a large prominent tubercles; 5) upper eyelid with prominent tubercles, none of them distinctly enlarged; 6) frontoparietal ridges inconspicuous (in life/preservative); 7) frontoparietal crests projecting laterally (on cranium); 8) constriction of the frontoparietal bones at the level of the anterior epiotic eminence; 9) absence of extensive exostosis on the cranium; 10) premaxillae slightly projecting posteriorly in lateral view; 11) posterodorsal projection of maxilla absent or highly reduced, not in contact with orbital/zygomatic ramus of squamosal; 12) maxillary process of the nasal not in contact with maxilla; 13) horizontal length of tympanum more than 50% horizontal length of eye in both sexes; 14) vomerine teeth 4-6; 15) toes basally webbed; 16) dorsal skin (in life) heavily tuberculate; 17) ventral skin (in life) granular; 18) absence of conspicuous outer tarsal tubercles (in life); 19) absence of multiple conspicuous dark brown bars on flanks and lips, presence of illdefined, irregularly-shaped dorsolateral stripes (in life); 20) in living adults, iris unicolor, greyish blue, with extensive dark brown reticulations.
Stefania satelles (Fig. 8) is characterized by the combination of the following morphological characters: 1) a small species of Stefania, max SVL in preserved females 56.6 mm, 52.1 mm in preserved males; 2) head not distinctly longer than wide, about as wide as long; 3) canthus rostralis angular, slightly concave, with a few low tubercles, canthal stripe present in life, not reaching nostril; 4) loreal region with many small tubercles; 5) upper eyelid with a few low tubercles, none of them distinctly enlarged; 6) frontoparietal ridges inconspicuous (in life/preservative); 7) frontoparietal crests moderately developed and distinctly projecting dorsolaterally (on cranium); 8) constriction of the frontoparietal bones at the level of the anterior epiotic eminence; 9) absence of extensive exostosis on the cranium; 10) premaxillae slightly projecting posteriorly in lateral view; 11) posterodorsal projection of maxilla fused with orbital/zygomatic ramus of squamosal; 12) maxillary process of the nasal in contact with maxilla; 13) horizontal length of tympanum more than 50% horizontal length of eye in both sexes; 14) vomerine teeth 4-7; 15) toes basally webbed, no significant difference in toe webbing between sexes; 16) dorsal skin (in life) tuberculate; 17) ventral skin (in life) granular; 18) absence of conspicuous outer tarsal tubercles (in life); 19) absence of multiple conspicuous dark brown bars on flanks and lips, presence or absence of irregularly-shaped dorsolateral stripes (in life); 20) in living adults, iris unicolour, copper, with extensive dark brown reticulations.

Discussion
Incongruence between phylogenetic relationships and morphological traits among tepui summit species of Stefania in the Eastern Pantepui District (east of the Río Caroní sensu McDiarmid & Donnelly 2005) was first highlighted by Kok et al. (2016Kok et al. ( , 2017. According to Kok et al. (2017), the S. ginesi clade (endemic to the Aprada, Chimantá and Los Testigos Massifs, as well as Angasima-tepui and Upuigmatepui) diversified about 14.5 Mya (Miocene) into two main lineages. One of these lineages contains S. ginesi, S. satelles and an undescribed species sister to S. ginesi (S. sp. 1); the other lineage contains four undescribed species (Fig. 4), including S. sp. 2, which is described in this work. Although S. satelles and S. lathropae sp. nov. are phylogenetically distant (Fig.  4), they share similar phenotypic traits that caused taxonomic confusion for two decades. The fact that both species are polychromatic (Figs. 3, 8) probably added to the confusion, and the use of molecular data have been crucial to resolving phylogenetic relationships, simultaneously unravelling surprising cryptic diversity in the S. ginesi clade (Kok et al. 2016). The summits of Aprada-tepui and Murisipán-tepui are probably subject to similar ecological pressure and constraints, and the morphological similarity between S. satelles and S. lathropae sp. nov. could be the consequence of shared ancestry (symplesiomorphy) and convergent evolution (homoplasy).
The combination of high genetic divergence, molecular phylogenetic position, and morphological differences leaves no doubt that S. lathropae sp. nov. is distinct from S. satelles. Genetic divergence in the barcoding fragment of 16S rRNA between S. satelles and S. lathropae sp. nov. is remarkably high for tepui summit species (> 8%). Indeed, a low genetic divergence has been reported as extremely common among congeneric tepui top populations (Kok et al. 2012). Interspecific divergences (up to 9%) strongly exceed intraspecific genetic divergence (0-0.2%) in most species of the S. ginesi clade (see above and Table 2). Surprisingly, as noted by Kok (2013), divergence in phenotypic traits tends to be higher between tepui summit species that are genetically close than between genetically distant species. In the S. ginesi clade, this is exemplified by the case of S. lathropae sp. nov., which is genetically highly divergent from S. satelles (> 8% in 16S) despite being phenotypically similar, and by the case of S. sp. 4 (Kok et al. 2017), a species previously confused with S. ginesi, which is phylogenetically distantly related to S. ginesi (Fig. 4) but genetically very close to S. sp. 5 (< 1% in 16S), a species previously confused with S. satelles. This conundrum can only be resolved with molecular data, and these insights are of utmost importance from a conservation perspective of the unique biodiversity of the Pantepui.
It should be noted that two subtle osteological characters reported in this study -i.e. the distinctions between the absence of contact vs. the contact or fusion of the posterodorsal projection of the maxilla with the orbital ramus of the squamosal, and of the maxillary process of the nasal with the maxilla -remain speculative and could merely reflect different resolutions of the µCT scans (see Table S1 for details) or bias in the interpretation of µCT-scan reconstructions (see Kok 2023). The diagnostic value of these characters should be confirmed by analysing a larger number of specimens scanned at similar resolution.
In accordance with criteria B1 and B2a, b (IUCN 2022), i.e. an extent of occurrence < 100 km 2 , an area of occupancy < 10 km 2 , a single known location, and projected decline due to the effects of climate change and the possible increase in anthropogenic fires in the region, S. lathropae sp. nov. should be listed as critically endangered. Owing to its strong isolation, the species is currently not threatened by the introduction of pathogens, such as chytridiomycosis, which is present in remote areas of Pantepui that are subjected to tourism (Kok et al. 2022).

NOMENCLATURAL ACTS REGISTRATION
The electronic version of this article in portable document format will represent a published work according to the International Commission on Zoological Nomenclature (ICZN), and hence the new names contained in the electronic version are effectively published under that Code from the electronic edition alone (see Articles 8.5-8.6 of the Code). This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information can be viewed through any standard web browser by appending the LSID to the prefix http://zoobank.org/.