Head in the clouds: a new dwarf frog species of the Physalaemus signifer clade (Leptodactylidae, Leiuperinae) from the top of the Brazilian Atlantic Forest

In an elevational gradient, the mountain top generally presents a reduced species diversity. However, it is there where we often fi nd microendemic and quite often still undescribed species. That European Journal of Taxonomy 764: 119–151 (2021) 120 prediction is very common in underexplored Neotropical mountains, like those of the Caparaó National Park – a protected area that includes the highest peak of the Atlantic Forest, a megadiverse domain. Up in its top, we found a dwarf frog of the genus Physalaemus (Anura, Leptodactylidae, Leiuperinae), belonging to the P. signifer clade. After an integrative (morphological, bioacoustical, and genetic) analysis, we were able to describe it as a new species and found it to be sister to P. maculiventris. Due to its very restricted distribution at a site with extreme environmental conditions (which includes fi res and frosts) and current instability in national environmental policy, we suggest this to be classifi ed as an endangered species. A brief description of its natural history and the description of the species itself will now enable its proper conservation status categorization and the future planning for conservation actions.


Introduction
The Neotropical genus Physalaemus Fitzinger, 1876 (Leptodactylidae Werner, 1838: Leiuperinae Bonaparte, 1850) is one of the most speciose genera in Leptodactylidae, with 49 recognized species (Frost 2021). The genus comprises two large main clades, the Physalaemus signifer and Physalaemus cuvieri clades, both supported on the basis of molecular evidence (Lourenço et al. 2015). Although both clades still lack recognized morphological synapomorphies, species of the Physalaemus signifer clade may be identifi ed by the presence of an arrow-shaped blotch on the dorsum of the body combined with the absence of tarsal tubercle (Leal et al. 2020). Moreover, the telocentric chromosome 11 has been suggested as putative synapomorphy of the P. signifer clade (Lourenço et al. 2015).
The Mantiqueira Range, southeastern Brazil, is one of the largest and the highest Atlantic Forest mountain ranges, with altitudes reaching up to 2891 m a.s.l. (Alvarenga 1997), harboring an extremely diverse anuran fauna (Silva et al. 2018). In an elevational gradient, although the mountain tops generally present a reduced species richness (Rahbek 1995;Lomolino 2001), it is there where we often fi nd narrow endemic and quite often still undescribed species (e.g., Leite et al. 2012;Walker et al. 2018;Santos et al. 2020). During fi eld surveys in the highlands of the Serra do Caparaó (see Zornosa-Torres et al. 2020), in Adobe Audition CC). Calls were normalized, removing DC off set (mean amplitude displacement from zero), centering on 0.0 vertically, and to the maximum amplitude of -3.0 dB, using the software Adobe Audition CC. We analyzed the recordings of advertisement calls in Raven Pro ver. 1.4 (Center for Conservation Bioacoustics 2011) using the following confi gurations: 90% brightness, 70% contrast, and Fast Fourier Transform window size (FFT) of 512. Aggressive calls were analyzed in Raven Pro ver. 1.6 (Center for Conservation Bioacoustics 2019), using the following confi gurations: 53% brightness, 70% contrast, and FFT of 512.
We followed the terminology and defi nitions presented by Köhler et al. (2017), using the note-centered approach, in which one entire series would be a call, and each sound unit a note, that could be further subdivided into pulses. A note is defi ned as a subunit of the call with 100% amplitude modulation and short intervals between them. Note and call intervals were considered as equal or longer than the length of the entire note or call, respectively (Köhler et al. 2017). We measured the following acoustic parameters (Tables 2-3): call duration, note duration, intercall interval, internote interval, call rate (only for advertisement call), note rate (number of notes per call), dominant frequency (peak frequency function in Raven), minimum frequency (frequency 5% function in Raven), maximum frequency (frequency 95% function in Raven), and frequency bandwidth (only for advertisement call; defi ned as the diff erence between frequency 95% and frequency 5%; Charif et al. 2010). Spectral parameters were measured in the spectrogram and temporal parameters were measured in the oscillogram.

Natural history
We recorded 24 hours of vocal activity using an autonomous recorder (Song Meter SM3, Wildlife Acoustics) situated at the margin of the Lagoa da Sombra (Supp. fi le 1, arrow indicates the autonomous recorder), where males were observed calling. Recordings were made continuously from midday 11 October 2017 to midday of the next day. Lower frequencies fi lter and calls normalizations were done as reported above in the bioacoustics section. We counted the number of calls of each hour using the interactive detectors (Band Limited Energy Detector) on the software Raven Pro ver. 1.5 (Center for Conservation Bioacoustics 2011) using the following parameters: minimum frequency of 1 kHz; maximum frequency of 4 kHz; minimum duration of 0.202 s; maximum duration of 0.501 s; minimum separation interval of 0.053 s. After running the automatic detection, we visually corrected the selections and counted the number of calls per hour. Visual confi gurations adopted were: 50% brightness, 70% contrast, and FFT of 512. Analyzed calls may be for one or several individuals, as it was not possible to distinguish the number of calling males. With this information we plotted male vocal activity (number of calls / hour) along 24 hours and fi tted a normal distribution.

Karyotype
We obtained metaphases from testicular cell suspensions of two males and intestinal epithelium suspension of one tadpole for the karyotype description. The individuals were anesthetized with 2% lidocaine (50 mg/g body weight -cutaneous administration) after pre-treatment in vivo with 2% colchicine (0.02 ml/g body weight) for 4 h. Chromosome preparations were obtained following King & Rofe (1976) with modifi cations from Gatto et al. (2018) and Schmid et al. (1979), stained with 10% Giemsa solution and C-banded according to King (1980). We took chromosome measurements from nine metaphases (Table 5) using DRAWID (DRAWing IDiogram; Kirov et al. 2017), and the chromosomes were classifi ed following Green & Sessions (1991).

Phylogenetic and genetic distance analyses
To assess the interspecifi c phylogenetic relationships of the new species, we conducted a phylogenetic analysis based on mitochondrial DNA sequences that comprised 12S rRNA, tRNA-val, and 16S rRNA genes (H1 fragment sequences). Our dataset included one specimen of the new species (ZUEC-AMP 24120) and representatives of other 44 species of Physalaemus, for which DNA sequences were available in the GenBank (Supp. fi le 2). Because the new species is closely related to P. maculiventris, we improved the sample of this species by adding one topotype (Paranapiacaba, SP, Brazil), one specimen from Iporanga (SP, Brazil), and another from Curucutu (Serra do Mar State Park, Núcleo Curucutu, São Paulo, SP, Brazil) (Supp. fi le 2). As outgroup, we included nine species of Engystomops Jiménez de la Espada, 1872 and one of Edalorhina Jiménez de la Espada, 1870 (for details, see Supp. fi le 2), which have been inferred as being closely related to Physalaemus (Fouquet et al. 2013;Lourenço et al. 2015).
To obtain the DNA sequences from the new species and P. maculiventris specimens, genomic DNA was extracted from liver tissue preserved in 100% ethanol using a TNES protocol (as performed by Lourenço et al. 2015). Polymerase chain reactions (PCR) with the primer pairs MVZ 59 (Graybeal 1997) -Titus I (Titus 1992) and 12L13 (Feller & Hedges 1998) -16Sbr (Palumbi et al. 2002) were performed to amplify the H1 mitochondrial fragment of interest. The PCR products were purifi ed using the Wizard SV Gel and PCR Clean-up System (Promega). The resulting fragments were bidirectionally sequenced using the BigDye Terminator kit (Applied Biosystems), with the above-mentioned primers and the primers MVZ50 (Graybeal 1997), 16SL2a (Hedges 1994), 16H10 (Hedges 1994), and 16Sar (Palumbi et al. 2002). The nucleotide sequences were read using an ABI 3730xL DNA Analyzer automatic sequencer (Applied Biosystems). Full-length sequences were obtained after assembling multiple reads of each specimen using the BioEdit Sequence Alignment Editor software (Hall 1999).
The nucleotide sequences were aligned using MAFFT ver. 7 (Katoh et al. 2019; https://maff t.cbrc.jp/alignment/server/), under the G-INS-i strategy. The resulting data matrix, composed of 71 terminals (Supp. fi le 2) and 2426 characters, was used to infer phylogenetic trees under maximum parsimony criterion using the software TNT ver. 1.1 (Goloboff et al. 2008) and by Bayesian analysis in MrBayes ver. 3.2.5 (Ronquist et al. 2011). To fi nd the most parsimonious trees, we used a heuristic search method performed through the option ʻNew Technology Searchʼ, including sectorial searches, ratchet, tree drifting, and tree fusing. The best length was hit 100 times from 5 random seeds. Gaps were considered as fi fth character state. A strict consensus cladogram was obtained from the 20 most parsimonious trees (with 6640 steps) and bootstrap values of the branches were calculated using traditional search method and 1000 pseudoreplicates. The Bayesian analysis was executed with the GTR+I+G model, which was inferred in MrModeltest ver. 2.3 (Nylander 2004) as the best-fi tting model for our dataset. Two simultaneous analyses were run, each with four chains (three heated and one cold) and 2 million generations. One tree was sampled every 100 generations. Consensus topology (50% majority-rule consensus tree) and posterior probabilities were produced after discarding the fi rst 25% of the trees generated. The average standard deviation of split frequencies (ASDSF) value was below 0.01 and the Potential Scale Reduction Factor values were approximately 1.000. The stabilization of posterior probabilities was checked using Tracer ver. 1.6 (Rambaut et al. 2014). The inferred cladograms were edited using the software FigTree ver. 1.4.3 (http://tree.bio.ed.ac.uk/software/fi gtree). Uncorrected genetic distances (p-distances) were estimated from the 12SrRNA, tRNA-val-16SrRNA fragment and also the partial segment of 16S rRNA gene fl anked by the primers 16Sar and 16Sbr (Table 6), which has been largely used for evaluating interspecifi c variation (see Fouquet et al. 2007 andLyra et al. 2017). For this analysis, we used the software MEGA ver. 7.0. (Kumar et al. 2016) and gaps were not considered in pairwise comparisons.

Comparison with other species
Physalaemus araxa sp. nov. may be set apart from all the species of the Physalaemus cuvieri clade (sensu Lourenço et al. 2015) by simultaneously having (1) the presence of an arrow-shaped blotch on the dorsum of the body and (2) absence of tarsal tubercle (Leal et al. 2020). (3) The gular region and chest predominantly yellow-colored in live individuals, pale cream in preserved specimens, distinguish the new species from all the other species of the P. signifer clade (gular region and chest predominantly dark brown in live and preserved individuals of those species). Additionally, live individuals of Physalaemus araxa sp. nov. have (4) belly with dark vermiculation on a pale cream, slightly bluish background, distinguishing it from P. angrensis, P. atlanticus, P. nanus, and P. spiniger (belly with fl ashy orange blotches in those species), from P. maculiventris (which has a pale chest and belly, with posterior region of belly and ventral surface of thigh showing bold black blotches), and from P. obtectus (which has belly with fl ashy red blotches). (5) The lack of aposematic coloration on ventral surface of hand and foot in live individuals distinguishes the new species from P. angrensis, P. atlanticus, P. spiniger (orange aposematic coloration present on ventral surface of hand and foot in live individuals of those species) and from P. claptoni, P. deimaticus, P. erythros (red aposematic coloration present on ventral surface of hand and foot in live individuals of those species). (6) By having adult males of intermediate size within the P. signifer clade (SVL = 17.4-21.5 mm), P. araxa sp. nov. is set apart from P. bokermanni (which is smaller, SVL = 15.3-17.0 mm) and from P. caete, P. camacan, P. moreirae, P. nattereri, and P. obtectus (which are larger, combined SVL = 22.3-50.6 mm). (7) The presence of a brown, divided, nuptial pad in males distinguishes P. araxa sp. nov. from P. claptoni (which has a nuptial pad not divided) and from P. rupestris (which has a white cream nuptial pad). (8) The lack of supernumerary tubercles on foot distinguishes P. araxa sp. nov. from P. angrensis, P. caete, P. camacan, P. crombiei, P. irroratus, P. moreirae, P. signifer, and P. spiniger (supernumerary tubercles present on the foot in those species). (9) The lack of a tarsal fold distinguishes P. araxa sp. nov. from P. atlanticus, P. bokermanni, P. camacan, P. crombiei, P. irroratus, P. nanus, P. obtectus, P. signifi er, and P. spiniger (tarsal fold present in those species). (10) Texture of posterior region of belly and ventral surface of thigh smooth in P. araxa sp. nov. distinguishes it from P. camacan and P. irroratus (posterior region of belly and ventral surface of thigh granulated in those species). (11) The duration of the advertisement call of the new species ranges from 69-304 ms, setting it apart from P. angrensis, P. atlanticus, P. bokermanni, P. caete, P. camacan, P. claptoni P. crombiei, P. moreirae, P. rupestris, and P. signifer (which have longer advertisement calls, combined minimum advertisement call duration from 324-2,130 ms). (12) Tadpoles of P. araxa sp. nov. have a proportionally larger body (BL/TL = 0.39-0.43), diff ering from those of P. atlanticus, P. bokermanni, P. maculiventris, P. moreirae, and P. spiniger (0.34 in P. atlanticus, 0.35 in P. bokermanni, 0.33 in P. maculiventris, 0.27-0.37 in P. moreirae, 0.37 in P. signifer, 0.34 in P. spiniger). (13) Tadpoles with dextral vent tube distinguish the new species from P. atlanticus, P. caete, P. camacan, P. nanus, P. rupestris, and P. spiniger (vent tube medial in those species). (14) By presenting dorsal and ventral fi ns of the same height tadpoles of P. araxa sp. nov. diff er from those of P. angrensis, P. atlanticus, P. caete, P. camacan, P. crombiei, P. erythros, P. irroratus, P. moreirae, P. nanus, P. rupestris, P. signifer, and P. spiniger (dorsal fi n higher than ventral one in those species) and from P. caete (which has dorsal fi n lower than ventral one). (15) The broadly rounded tail tip also diff ers tadpoles of P. araxa sp. nov. from most species of the P. signifer clade, such as P. atlanticus, P. bokermanni, P. caete, P. camacan, P. crombiei, P. maculiventris, P. spiniger (tail tip pointed in those species) and P. angrensis, P. erythros, P. irroratus, P. moreirae, P. nanus, P. rupestris, P. signifi er (tail tip nearly rounded in those species). (16) Tadpoles presenting submarginal papillae arranged in small rows set P. araxa sp. nov. apart from the remaining species of the P. signifer clade (which have submarginal papillae scattered in the lateral portions of the oral disc), except from P. erythros which has a similar condition. (17) The A2 tooth row conspicuously longer than A1 distinguishes tadpoles of P. araxa sp. nov. from those of P. atlanticus, P. bokermanni, P. camacan, P. erythros, P. maculiventris, and P. moreirae (A1 = A2 in those species) and from P. angrensis, P. caete, and P. spiniger (A1 > A2 in those species). Additionally, tadpoles of P. araxa sp. nov. also diff er from those of P. maculiventris by (18) the absence of a dermal fold at the body-tail junction (dermal fold present in P. maculiventris), by (19) the external margins of the fi ns slightly convex (fi ns markedly convex in P. maculiventris) and by (20) the gular region convex (gular region straight in P. maculiventris).

Etymology
The specifi c epithet ʻaraxaʼ, is the combination of the Tupi-Guarani indigenous language words ʻaraʼ (meaning ʻworldʼ) and ʻeçaʼ (meaning ʻto seeʼ) meaning ʻthe fi rst place where the sun can be seenʼ, in

Description
Holotype (ZUEC-AMP 24095) Adult male (Figs 1-2), SVL 21.0 mm. Head slightly wider than long. Head width 38.0% SVL and length 36.0% SVL. Snout rounded in dorsal view and rounded to truncated in lateral view ( Fig. 2A-B). Canthus rostralis distinct, rounded; loreal region slightly concave. Snout protruding beyond lower jaw. Nostril dorsolaterally oriented, faintly protruding. Internarial region fl at; top of the head slightly concave. Eye slightly prominent, anterolaterally oriented, its diameter 5% larger than END. Tympanum indistinct externally. Supratympanic fold distinct, thick, extending from the posterior corner of the eye to the shoulder. Dentigerous process of vomer absent. Premaxillary and maxillary teeth absent. Choanae rounded, separated from each other by a distance as large as four times its diameter. Tongue elongated, constricted on its anterior third, wider on its posterior half, free around lateral and posterior margin. Vocal slit present, longitudinal, originating on the sides of the tongue in its anterior third, and extending towards the corner of the mouth. Vocal sac single and subgular, faintly diff erentiated externally. Dorsolateral fold present, weakly distinct, from the posterior corner of the eye to the inguinal region. Forearm hypertrophied in relation to upper arm; upper arm slender, short. Fingers thick, without webs, relative lengths I < II = III < IV; fi nger tips not expanded. Proximal subarticular tubercles large, simple, prominent, and rounded; distal subarticular tubercles present on fi ngers III and IV, approximately of the same size as the proximal ones; supernumerary tubercles large, rounded, low, more distinct in the left hand. Inner and outer metacarpal tubercles ovoid, large, prominent. Nuptial pad divided, densely covered by dark keratinized spicules, present on the dorsal and lateral surfaces of the thumb (except for the distal phalange) and on the internal surface of the internal metacarpal tubercle. Tibia length 39% SVL; foot length 60% SVL. Toes thick, without webs, relative lengths I < II = V < III < IV; toe tips not expanded. Subarticular tubercles distinct, simple, prominent, and rounded; supernumerary tubercles absent. Tarsal fold absent; tarsal tubercle absent. Inner metatarsal tubercle distinct, ovoid; outer metatarsal tubercle distinct, rounded. Inguinal gland well developed, oval. Cloacal opening directed posteriorly at upper level of thighs. Region below the cloaca with low and faintly distinct tubercles scattered, encroaching the thighs on its posteroventral edge, where tubercles become increasingly less distinct. In preservative, texture of ventral, lateral, hidden, and dorsal surfaces are smooth. Measurements of the holotype in Table 1.

Color of the holotype preserved in alcohol 70%
In preservative (Figs 1-2), dorsal background color of the body, head, and limbs dark brown. All dorsal blotches dark brown, darker than background color. Interorbital blotch triangular-shaped, one vertex over each eyelid, the third vertex connected with the tip of an arrow-shaped blotch medially located on the dorsum of the body. Middle of the arrow-head with a small light spot, posterior portion of the arrow extends transversely towards each inguinal gland. Urostyle region bears a longitudinal blotch with anterior margin poorly defi ned, centered by a longitudinal light blotch. Some poorly defi ned and irregularly shaped small blotc hes scattered through dorsal surfaces of head and body. Inguinal gland ca 90% covered by a black ocellus. Dorsal surfaces of thigh, tibia, and foot with transversal blotches. Many white dots scattered throughout dorsal surface of body and limbs, mainly outlining the more distinct blotches and the black ocelli over the inguinal gland. Heel, anterior part of knee, and forearm with a black blotch. Upper arm with a longitudinal black blotch covering its posterior surface and the elbow. Dorsolateral black blotches, dorsally outlined by small white dots, extending from the posterior margin of the eye through the supra tympanic fold and reaching the second third of the fl ank. Elongated black blotches extends from the tip off the snout to the eye, passing over the nostril and canthus rostralis. Irregularly shaped black blotches on the loreal region and upper lip. Region between eyes and the insertion of the arms, pale cream. Ventral surfaces of the limbs brown with scattered light spots; ventrolateral edge of forearm with a black blotch over glandular tissue. Black blotch above the cloaca, dorsally outlined by small white spots, curved down toward the back of the thighs. Gular region and chest predominantly pale cream, stained with very small dark dots. Ventral border of mandible brown colored, without distinct blotches. Belly with dark vermiculation pattern on a pale cream background. Iris dark brown with a faintly visible black vermiculation; pupil black and horizontal.

Color in life
Gular region, chest, axillary region, lower border of anterior half of the fl anks, and region between eyes and the insertion of the arms yellow (Fig. 3B-D). Inguinal region pink posteriorly to the black ocelli and yellow around its anterior edge (Fig. 3A, C-D). Ventral surface of limbs pinkish brown (Fig. 3C). Background color of belly pale cream, slightly bluish (Fig. 3C). Overall dorsal background and blotches coloration varying from brownish to greenish, background color with some sparkled areas of cream and orange brown.

Variation
Measurements and proportions of 13 adult males are presented in Table 1. In life, color of dorsal background of head, body, and limbs may vary from pale yellow to dark brown; some individuals had these areas, as well as the region between eyes and the insertion of the arms, stained by pinkish coloration in diff erent levels of area size and color intensity. Yellow coloration present on gular region, chest, axillary region, lower border of anterior half of the fl anks, and the region between eyes and the insertion of the arms may vary in intensity. Ventral surface of thigh may be pinkish. White dots outlining dorsal blotches may vary in size and number, forming distinct lines in some individuals; these white dots / lines may be greenish in live individuals. Few individuals have dorsal pattern weakly distinct in life and in preservative. Dorsal arrow-shaped blotch may present interruptions and be slightly irregularly shaped. Irregularly shaped small blotches scattered across dorsal surfaces of head and body may vary in number, size, shape, position, and distinctness from the background. Background color of belly may be bluish in diff erent intensities. Brown coloration of the ventral border of mandible is more visible in fi xed specimens, after the yellow coloration has faded. In fi xed specimens it varies from occupying just the edge of the mandible to the anterior half of gular region. Tympanum may be slightly discernible externally in some individuals. Snout may be rounded in lateral view. Dorsum of the head, body, thighs, and tibiae may be slightly rugose. Dark superfi cial keratinized layer of the nuptial pad may be peeled; however, nuptial pad remains visible.

Advertisement and aggressive call
We identifi ed two call types, the advertisement and an aggressive call (of undetermined specifi c function, see Toledo et al. 2015). Both the advertisement (n = 386 calls from seven individuals) and aggressive calls (n = 43 calls from four individuals) presented a pulsed structure. They diff erentiate from each other mainly by the call duration, note rate, and the context in which the recordings were obtained (advertisement calls recorded in the fi eld and aggressive calls recorded from the aqua-terrarium). The advertisement calls had variable number of notes (1-3), frequency, and amplitude of notes varying in the call ( Fig. 4; Table 2). These calls seemed to be composed of a harmonic structure, but the harmonics were not clear in most of the calls. The most common advertisement call structure was that with 3 notes (exhibited in 63.63% of calls, 34.1% had two notes and 2.27% had only one note) with the dominant frequency in the second note (48.64% of the calls, 35.13% of calls had the dominant frequency in the third note and 16.21% in the fi rst note). This call was emitted at rate of 0.87-2.04 calls / second, lasting between 0.07 to 0.3 seconds, with a dominant frequency of 0.94-2.63 kHz (average of 1.7 kHz ± 0.31 SD). On the other hand, we analyzed 1306 aggressive call notes from four males ( Fig. 5; Table 3). These calls were emitted by males close to other calling males, but it was not possible to determine a specifi c function (e.g., territorial, encounter or fi ghting calls). Aggressive calls were organized in groups of 9-118 notes, lasting 4.73-53.46 seconds, longer than advertisement calls, and with a dominant frequency of 1.59-2.30 kHz.

Tadpole
Maximum total length 28.2 mm, at stage 37. Body depressed (BH/BW = 0.73-0.85; Fig. 6A-B, H), 0.39-0.43 times TL; in dorsal view, ovoid with well-marked lateral constrictions at the spiracle level; in lateral view, ventral contour convex in gular and abdominal regions, with a well-defi ned constriction slightly anterior to the spiracle level. Snout truncated in lateral view and obtuse in dorsal view (BWN/ BWE = 0.71-0.75). Nostrils elliptical, small (ND/BL = 0.02-0.02), dorsally located, anterodorsally   Table 2. Temporal and spectral parameters of the advertisement call of Physalaemus araxa sp. nov. and P. maculiventris (Lutz, 1925). Values are presented as a range (average ± SD); n = number of males (number of calls analyzed).  (Fig. 6F). Vent tube dextral, posteriorly directed, short (VTL/BL = 0.07-0.10), with a large opening, fused to the ventral fi n, and positioned at its ventral margin (Fig. 6G). Tail moderately high, with about the same height as the body (MTH/BH = 0.95-1.00); tail musculature slender (TMH/BH = 0.29-0.39), straight, not reaching the broadly rounded tip of tail. Dorsal and ventral fi ns about the same height (DFH/TAL = 0.11-0.14; VFH/TAL = 0.11-0.14), with the external margins slightly convex. Dorsal fi n emerging on the posterior third of the body at a moderate sloping (DFIA = 15-17°); maximum height at the middle third of the tail. Oral disc medium-sized (ODW/ BW = 0.31-0.35, measured with oral disc closed), anteroventrally positioned (ODP = 33-43°), laterally emarginated (Fig. 6D); single row of conical and alternate marginal papillae interrupted anteriorly by a wide anterior gap (AGL/ODW = 0.58-0.60); few (3-4) small submarginal papillae arranged in a small row at the supra-angular region and other (1-3) aligned at the fold of the oral disc emargination. Labial tooth row formula (LTRF) 2(2)/3(1); A2, frequently irregular at the lateral portions, longer than A1, which is irregular along its length; P1 and P2 equal in length, slightly longer than P3; jaw sheaths wide, fi nely serrated on the margins (about 36 serrations on the upper sheath), upper jaw sheath M-shaped and lower jaw sheath V-shaped. Stitches of lateral line system not distinct; nerves of the ventral body-line and longitudinal oral line evident laterally, and nerves of dorsal and middle lines in the posterior portion of body. Intestinal tube circularly coiled (Fig. 6C), switchback point slightly dislocated from the center of the abdominal region. Measurements are shown in Table 4.

Tadpole coloration
In preservative, body densely covered by dark brown melanophores, except the gular region, which is pale (Fig. 6A-C); intestine tube barely visible, almost covered by melanophores; rectus abdominis visible from the region of posterior limbs to the peribranchial region; distal portion of spiracle not pigmented. Tail musculature cream, homogeneously pigmented by melanophores; fi ns translucent, fi nely reticulated with fi liform melanophores mainly the dorsal fi n.
In life, body dark brown, fi nely speckled with iridophores (Fig. 6H); spiracle translucent; venter cream in the gular and abdominal regions; iris black with golden dots scattered and a narrow golden rim surrounding the pupil. Tail musculature cream, homogeneously covered with melanophores except by few small, depigmented areas; fi ns translucent, fi nely reticulated with fi liform melanophores and golden dots; dorsal fi n more pigmented than ventral fi n.  Table 3. Temporal and spectral parameters of the aggressive call from four males of Physalaemus araxa sp. nov. Values are presented as a range (average ± SD).

Phylogenetic inferences and genetic distances
In both MrBayes and TNT analyses, the new species, together with Physalaemus cf. araxa from Santa Teresa, ES, composed a highly supported clade that was the sister group of P. maculiventris in the Physalaemus signifer clade (Fig. 8). Two genetic lineages were recognized in the P. maculiventris clade, one represented by the specimens from Bananal, SP, and another by specimens from three other localities (Fig. 8). High genetic distances were found between these P. maculiventris lineages and also between each of them and the new species (Table 6). In contrast, the p-distances estimated from the 16Sar-16Sbr and H1 fragments, between the specimen from Santa Teresa and the new species, were low, 1.48 and 2.57, respectively (Table 6).  Table 4. Measurements (in mm) and angles of tadpoles of Physalaemus araxa sp. nov. (n = 9; ZUEC-AMP 24214) for the stages 37-38 (Gosner 1960). Data are presented as range (mean ± SD). See Material and methods for abbreviations.   Table 5. Morphometric parameters of the chromosomes of Physalaemus araxa sp. nov.; chromosomal classifi cation according to their centromeric position followed Green & Sessions (1991). Abbreviations: AR = arm ratio; RL = relative length; CC = centromeric classifi cation: M = metacentric; SM = submetacentric; T = telocentric. Asterisks represent the pairs at the threshold between metacentric and submetacentric classifi cation.

Remarks
The phylogenies presented in Lourenço et al. (2015) and in Leal et al. (2020) have a terminal named "P. signifer (Congonhas do Campo, MG)". The city name "Congonhas do Campo" was incorrectly used as it has changed to "Congonhas" since 1948. Herein we fi xed it, changing the terminal name to "P. signifer (Congonhas, MG)" to avoid any misunderstandings regarding the city name.   Diagonal line presents uncorrected p-distances id entifi ed within each species or lineage, based on the H1 (left) and 16Sar-16Sbr (right) fragments. En-dash denotes that only one sequence is available.

Natural history
Individuals of Physalaemus araxa sp. nov. reproduce aggregated in shallow temporary swampy ponds ( Fig. 9) in altitudinal grasslands (campo de altitude) with interspersed granitic outcrops, a typical phytophysiognomy of the highlands of the Mantiqueira mountain range (Saff ord 2007; Vasconcelos 2010). Individuals were registered between 2551 m a.s.l (Lagoa da Sombra) and 2656 m a.s.l. (Três Lagoas), P. araxa sp. nov. was the only anuran species found breeding at those high-altitude ponds (observation also corroborated by the autonomous recordings). After amplexus, pairs produced fl oating foam nests where eggs were laid. Two foam nests were collected on 5 October 2017. Ten days after, tadpoles started leaving the nests and at 12 November tadpoles reached the stage 25. We counted 187 tadpoles from those nests. Benthic exotrophic tadpoles (ecomorphological guild II: A:1 sensu Altig & McDiarmid 1999) were observed in the wild in October and December 2017. Calling activity, foam nests, and tadpoles were registered between October 2017 and December 2017. Immediately metamorphosed froglets (with no evidence of tail), raised in captivity, had between 7.76 and 8.85 mm (ZUEC-AMP 24403-5).
Males were found calling from the ground or on top of peat moss (Sphagnum sp.) hidden on the pond's margins. During the complete day (24 h) that we recorded P. araxa sp. nov. vocal activity (Fig. 10), males called during all 24 h, however, clearly presenting predominant nocturnal calling activity. Most of the calls were recorded from 18:00 to 3:00, reaching the peak of calling activity between 19:00 and 21:00. Fig. 9. Type locality of Physalaemus araxa sp. nov., Lagoa da Sombra, at Parque Nacional do Caparaó.

External morphology, phylogenetic inferences, and genetic distances
The ability of Physalaemus araxa sp. nov. to inhabit areas restricted to such high altitudes (2551-2656 m a.s.l.) in a harsh environment prone to constant frosts, is not its only highly distinctive characteristic.
The new species' gular region and chest predominantly yellow colored in live individuals is also a novelty in Physalaemus, distinguishing it from all congeners, which have the gular region predominantly dark.
During fi eld work we observed some males of P. araxa sp. nov. adopting an upright posture with vocal sac slightly infl ated, displaying its conspicuous yellow color ( Fig. 3B-C). This type of behavior has already been identifi ed as related to intraspecifi c communication in anurans, with the yellow coloration of the vocal sac contrasting with the background, enhancing the location of the individual (Hödl & Amézquita 2001;Hartmann et al. 2005). Rosenthal et al. (2004)  Physalaemus araxa sp. nov. was recovered in our phylogenetic inference fully supported as sister to a specimen collected at Santa Teresa, ES (Fig. 8). This sample from Santa Teresa was already included in the phylogenies presented by Lourenço et al. (2015) and Leal et al. (2020), where it was identifi ed as P. maculiventris and recovered as sister to specimens of P. maculiventris from the municipality of Bananal, SP. Besides the phylogenetic proximity, the values of genetic distance acquired from the 16S rRNA gene between the new species and the sample from Santa Teresa were surprisingly low, set at 1.48%, much lower than the 3% suggested by Fouquet et al. (2007) and Lyra et al. (2017) as a threshold for intraspecifi c variation. This value is also signifi cantly lower than the values found between other species of Physalaemus, such as the ones from the P. deimaticus species group (Leal et al. 2020). These data indicate that the sample from Santa Teresa could be from an individual belonging to the new species, instead of being a P. maculiventris as previously thought. We have tried to fi nd the voucher specimen for making a morphological evaluation, but we were unsuccessful in fi nding it. The tissue sample was obtained from the Célio F.B. Haddad Amphibian Collection but there was no information regarding the specimen collector or the location of the voucher individual. We even contacted the herpetologist J.L. Gasparini who worked at Santa Teresa but this did not help to either fi nd the voucher specimen or any information about this specifi c specimen. Although we could not fi nd the voucher specimen from the tissue sample, we analyzed seven specimens from Santa Teresa that had previously been identifi ed as P. maculiventris. Despite being from a location 740 km in a straight line far from the species type locality (Paranapiacaba, SP), those seven specimens proved to be morphologically indistinguishable from topotypes of P. maculiventris, sharing its singular characteristics of a pointed snout, dark throat, and black bold blotches over the belly, thus, conspicuously distinct from P. araxa sp. nov. Therefore, we decided to designate the specimen from Santa Teresa, sister to Physalaemus araxa sp. nov., as Physalaemus cf. araxa. Additionally, it is noteworthy that the environment occupied by P. araxa sp. nov. at its type locality is singular and does not match any environment found in Santa Teresa, including its altitudinal range of occurrence. Considering all these uncertainties regarding this occurrence record of Physalaemus cf. araxa, it is even possible that the collecting locality attributed to it is wrong. Thus, until a more reliable record of the new species is made for areas outside the type locality, we consider the species as restricted to the sites where the type series was collected.
We added three new samples of Physalaemus maculiventris to the molecular data matrix previously used by Lourenço et al. (2015) and Leal et al. (2020), one from a topotype specimen (Paranapiacaba, SP) and the others from Curucutu, SP (~ 40 km southwest of the type locality), and Iporanga, SP (~ 240 km southwest of the type locality) (Supp. fi le 2). These three specimens formed a fully supported clade, sister to a clade composed of two individuals of P. maculiventris from Bananal, SP. The gen etic distance of the 16S gene fragment was low among the specimens from the type locality (Paranapiacaba), Curucutu, and Iporanga (p-distance of 2%). On the other hand, the p-distance between these three specimens and the two specimens from Bananal was relatively high (p-distance of 3.72%). This relatively high value may indicate the existence of undescribed species within P. maculiventris and should be further investigated, especially considering the fact that we could not access the voucher specimens from Bananal nor recordings from this locality. We have examined adult specimens of P. maculiventris from many localities (see Appendix I for a complete list), from São Bento do Sul (Santa Catarina) to Santa Teresa (ES) and, even though it covered approximately 1150 km 2 of distribution range, we could not fi nd morphological characters that could be used as diagnosis to any of the analyzed populations. However, with such a wide distribution range and very few registers made, always all along the mountains of the Brazilian coastal area, this species seems also to be a good model for phylogeographic and biogeographic studies.

Bioacoustics and natural history
The advertisement call of Physalaemus araxa sp. nov. can be distinguished from that of populations of P. maculiventris by the higher note rate (number of notes per call), which is commonly three notes in calls of P. araxa sp. nov. , while P. maculiventris presented only one note, as also reported by Hepp & Pombal (2020). Even with more notes, the advertisement call of P. araxa sp. nov. was still shorter than that of P. maculiventris. Additionally, aggressive calls were longer than advertisement calls and present a higher note rate, as observed in other species that have these two types of calls (P. maculiventris, P. erythros, P. signifer, and P. spiniger; Hepp & Pombal 2020).
Males of Physalaemus araxa sp. nov. could be found hidden in the border vegetation of the ponds; therefore, the calls we analyzed based on the recordings of the autonomous recorder are likely to be emitted from diff erent individuals. In spite of that, data presented showed to be an adequate temporal proxy for the species activity.

Tadpole
Fourteen of the 18 species of the P. signifer clade had their tadpoles characterized in the literature (Rugeri & Weber 2012;Ceron & Santana 2017;Leal et al. 2020). The tadpoles of Physalaemus araxa sp. nov. diff er from the other species of the clade by a combination of characters states of body (i.e., relative body size), vent tube (i.e., vent opening position), tail (i.e., shape of the tail tip), fi ns (i.e., dorsal and ventral height), and oral disc (i.e., relative size of anterior rows and submarginal papillae confi guration). Worth to note that the tadpoles of P. araxa sp. nov. are quite distinct from the one of P. maculiventris described and illustrated by Bokermann (1963). Physalaemus araxa sp. nov. has a proportionally larger body than P. maculiventris (i.e., BL/TL = 0.39-0.43 in P. araxa sp. nov.; 0.33 in P. maculiventris), lacking the well-defi ned cutaneous fold in the body-tail junction, which is distinct in P. maculiventris (see Bokermann 1963: fi g. 2). The gular region contour is also distinct between these species, being convex in P. araxa sp. nov., while it is straight in P. maculiventris. Moreover, in P. araxa sp. nov. the tail has the same height as the body and a broad rounded tip (tail higher that the body and with pointed tip in P. maculiventris; Bokermann 1963). The submarginal papillae arranged in small rows and the A2 tooth row conspicuously longer than A1 also diff er P. araxa sp. nov. from P. maculiventris.

Conservation
This is another example of micro-endemic anuran species, restricted to a mountaintop. Both the extent of occurrence and area of occupancy, as defi ned by IUCN guidelines (IUCN 2019), are less than 10 km 2 . That fact, allied to constant natural frosts and occasional anthropogenic fi res at the Caparaó National Park (Zornosa-Torres et al. 2020), may constitute a threat to P. araxa sp. nov. Therefore, this species could be classifi ed as endangered, under the category Vulnerable (VU), with the criterion D2, i.e., which is related to a species with a very restricted distribution and immediacy of threat. Besides the formal conservation assessment to this species, made by the Brazilian government or the IUCN, we indicate the long-term monitoring of the species and a practical fi re control specifi cally at the Caparaó National Park top regions.  ZUEC 2264ZUEC , 2618CFBH 1503CFBH , 2107.  7650-7652, 18842, 18843, 18845-18847, 18849, 19512. Bombinhas: UFMG 9251. Brusque: UFMG 10074, 10076, 10078-10081. Florianópolis: UFMG 2717, 2719-2721, 4608-4616, 4640, 4641, 7090, 7091, 19686, 19687. Rancho Queimado: UFMG 2716, 2718.