A new species of Brachycephalus (Anura: Brachycephalidae) from the northern portion of the state of Rio de Janeiro, southeastern Brazil

Brachycephalus is a genus of small ground-dwelling anurans, endemic to the Brazilian Atlantic Forest. Recent molecular analyses have corroborated the monophyly of three species groups within this genus ( B. ephippium , B. pernix , and B. vertebralis ). In the meantime, the genus has been targeted as a group with recent taxonomic issues owing to its interspecific morphological similarity and genetic conservatism. Herein, we describe a new species of Brachycephalus from the northern portion of Serra do Mar mountain range, in the state of Rio de Janeiro, Brazil. It belongs to the B. vertebralis species group, exhibiting moderate hyperossification of the skull and vertebral column. The new species can be distinguished from all other congeners based on morphological, acoustic, and molecular data. Furthermore, we provide information on osteology and natural history of the new species.

The first published phylogenetic hypothesis focused on Brachycephalus (Clemente-Carvalho et al. 2011), based on all 14 species known at the time, indicated the existence of two main clades: one containing the brightly-colored species with hyperossification of skull and skeleton and the other containing the brightly-colored non-hyperossified ones; however, the relationships of the two cryptically colored species formerly included in the genus Psyllophryne Izecksohn, 1971 (see Kaplan 2002) were not consistent, with B. hermogenesi (Giaretta & Sawaya, 1998), in particular, being recovered in different positions in different gene trees (always with strong support, and usually within one of the brightly-colored species groups).Subsequently, it was proposed to allocate the species of Brachycephalus into three species groups based on morphological characters: the B. didactylus, B. ephippium, and B. pernix groups (Pie and Ribeiro 2015;Ribeiro et al. 2015;Bornschein et al. 2016).However, recent molecular phylogenetic studies, recovered only two of these groups as monophyletic (the B. ephippium and B. pernix groups), while species previously assigned to the B. didactylus group were not recovered as a clade (e.g., Condez et al. 2020;Folly et al. 2020Folly et al. , 2022;;Dos Reis et al. 2021;Almeida-Silva et al. 2021).Lyra et al. (2021) recovered a similar topology in a phylogenomic study including nine Brachycephalus species.

Morphological data
For morphometrics, a single person (M.Folly) took 18 measurements (in mm) from each of the 17 collected specimens with an ocular micrometer in a Leica MZ-6 stereomicroscope (0.01 mm): snout-vent length (SVL; ventral distance from the tip of the snout to cloacal opening); head length (HL; dorsal distance from the tip of the snout to angle of jaw); head width (HW; greatest width of head located between angles of jaw); nostril diameter (ND; maximum width of the nostril); inter-nostril distance (IND; taken between inner margins of nostrils); nostril-tip of snout distance (NSD; from anterior margin of nostril to the tip of the snout); interorbital distance (IOD; interval between the inner edges of the orbits); eye diameter (ED); eyelid width (EW); eye-nostril distance (END; from anterior corner of the eye to posterior margin of nostril); arm length (AL; distance between axilla and elbow); forearm length (FAL; distance between elbow and wrist); hand length (HAL; distance between the wrist and the tip of Finger III); Finger-III length (FIL; insertion between Fingers II-III to the tip of Finger III); thigh length (THL; distance from the cloaca to the knee); shank length (SL; distance from the ankle to the knee); foot length (FL; distance between the ankle and the tip of Toe III); and Toe-III length (TL; insertion between Toes II-III to the tip of Toe III).Except for FL, which is modified to include tarsus length, all these measurements follow Duellman (1970), Cei (1980), andHeyer et al. (1990).Some measurements (ND, NSD, ED, FIL, and TL) were also included because they are informative and currently adopted by taxonomists working with the genus Brachycephalus (Condez et al. 2016).The sex of specimens was determined by gonad examination.To evaluate sexual dimorphism, we tested differences in body size between males and females using the T-test (R Core Team 2022).For comparisons with other species in the genus we used data both from specimens examined in collections (see Table S1) and from the literature (especially original descriptions and redescriptions of taxa).We followed Sabaj (2016) for museum acronyms.Descriptive terminology of the snout follows Heyer et al. (1990).The fingers were numbered II-V following Fabrezi and Alberch (1996).

Molecular data and phylogenetic analysis
In this study, we obtained novel molecular information from nine specimens within the new species.Three mitochondrial gene segments were selected: 12S ribosomal RNA (12S rRNA), 16S ribosomal RNA (16S rRNA), and cytochrome b (cyt b).GenBank accession numbers are listed in Table S2.We performed PCRs using a PCR Master Mix and a pair of primers for each gene segment, respectively listed here: 12SJL forward (AAAGRTTTG-GTCCTRRSCTT) and 12SKH reverse (TCCRG-TAYRCTTACCDTGTTACGA) published in Goebel et al. (1999); 12SM forward (GGCAAGTCGTAACAT-GGTAAG) and 16SA reverse (ATGTTTTTGGTAAA-CAGGCG) published in Darst and Cannatella (2004); MVZ15-L forward (GAACTAATGGCCCACACWW-TACGNAA) published in Moritz et al. (1992) and Cyt-bAR-H reverse (TAWAAGGGTCTTCTACTGGTTG) published in Goebel et al. (1999).Thermocycling conditions for 12S rRNA and 16S rRNA amplification began with a denaturation at 94 °C (5 min), followed by 35 cycles of denaturation at 94 °C (1 min), annealing at 50-52 °C (1 min), and a final extension step at 72 °C after the final cycle (5 min).Amplification conditions for cyt b were the same as described above, except that the annealing temperatures were adjusted to primer-specific temperatures ranging from 50 to 55 °C.PCR products were visualized in 1% agarose gels and sent to Macrogen Inc. (Seoul, Republic of Korea) for purification and sequencing reactions.Resulting electropherograms for both DNA strands were analyzed using Chromas Lite 2.01 and edited using MEGA X and adjusted manually to generate consensus sequences for each specimen.Sequences were checked with basic local alignment search tool (Altschul et al. 1997) against the GenBank nucleotide database to ensure that the amplified product was correct and not contaminated.We also confirmed the authenticity of the cyt b gene fragment by amino acid translations.
We used PartitionFinder 2.1.1 (Lanfear et al. 2017) to select the optimal partition scheme and nucleotide substitution models for each fragment, except for the protein coding genes, which had individual codon positions treated as separated partitions.We adopted the corrected Akaike Information Criterion (AICc) to select the best-fitting model for each gene, considering all possible models (JC, K80, SYM, F81, HKY, and GTR, with +G, +I, and +I +G as additional parameters), linked branch lengths, and the greedy search algorithm.After this, we performed a Bayesian Inference analysis in MrBayes 3.2.6 (Ronquist et al. 2012), running two independent analyses with four chains each, adopting the Markov Chain Monte Carlo approach (MCMC).We ran 100 million generations, sampling every 10,000 generations, and discarding the first 25% of generations as burn-in.Parameter convergences were checked in Tracer 1.6 (Rambaut et al. 2014).Trees were summarized in a majority-rule consensus tree where the node support values correspond to the Posterior Probabilities.Tree topology was edited using FigTree 1.4.3 (Rambaut 2016) and Illustrator (Adobe Inc. 2019).
MAFFT, PartitionFinder and MrBayes analyses were run at the CIPRES Science Gateway Portal (Miller et al. 2010).Additionally, we calculated the uncorrected pairwise genetic distances (p-distances) within the new species sequences and among its related taxa.We conducted this analysis using Mega X (Kumar et al. 2018;Stecher et al. 2020), considering the pairwise deletion of gaps and missing data.

Advertisement call analysis
The advertisement call description was based on unvouchered recordings from Parque Estadual do Desengano, trail to Pico do Desengano, municipality of Santa Maria Madalena, state of Rio de Janeiro, Brazil (21°52'49.1"S,41°55'0.9"W;1167 m a.s.l.; WGS84).Recordings were made at 44.1 kHz sampling rate and sample size of 16 bit in the mono pattern.Even though no other species in the area has the same call characteristics, the taxonomic assignment was confirmed by the collection of specimens in the targeted area immediately after the recordings were taken.Sound files are deposited in Fonoteca Neotropical Jacques Vielliard, at Universidade Estadual de Campinas (respectively as FNJV 58772, 58773, and 58774).Prior to the analysis we tested 500-2000 Hz high-pass filters to sound files to try reducing the background noise using Audacity v2.1.1 (Audacity Team 2017).As we were still not able to analyze the oscillograms after background filtering, we kept the original recordings for subsequent analysis.We therefore recognize that further recordings are necessary to refine some of the temporal parameters presented for the advertisement call of this species.Bioacoustic analyses were performed using 20 calls per individual in the program Raven Pro 1.5 (Bioacoustics Research Program 2014).Spectrograms were produced with a Hann window type, window size 512 samples, overlap 90% (locked), hop size 51 samples, DFT size 1024 samples (locked), grid spacing 43.1 Hz, brightness and contrast in default, and color map Cool.We used the note-centered approach and quantified ten advertisement call parameters, according to Köhler et al. (2017;but see Hepp and Pombal 2019): (1) call duration (s), (2) inter-call interval (s), (3) number of notes per call (notes/call), (4) note repetition rate (notes/s), ( 5) number of pulses per note (pulses/note), ( 6) number of pulses/note (pulses/note), ( 7) pulse repetition rate (pulses/s), (8) dominant frequency (peak frequency in Raven Pro 1.5), (9) lower and (10) upper frequencies (kHz; frequency 5% and frequency 95% in Raven Pro 1.5, respectively).Values are presented as the parameter range plus the average ± standard deviation (in parenthesis).The sound graphics were prepared in R 4.0.1 (R Core Team 2020) using the package Seewave v2.1.0(Sueur et al. 2008), with the spectrogram settings Hann window, 90% overlap, 256 samples resolution, and a color scale defined by 50 relative dB.T. Moreira, from 13-17 May 2008. CFBH 28049, adult female, CFBH 28124, 28129-30, 28135, adult males, and CFBH 28128, 28132-34, 28136-38, individuals  Diagnosis.Brachycephalus herculeus sp.nov. is distinguished from all its congeners by the following combination of characters: (1) skin on head and dorsum with dermal ossification; (2) skull with hyperossification of postorbital crests, which can be seen externally; (3) fourth presacral vertebra with transverse processes hyperossified, not ornamented and not visible externally; (4) urostyle crest extending up to 2/3 of urostyle length; (5) long oesophageal process of hyolaryngeal apparatus; (6) general color in life orange with dorsal green irregular patch; (7) presence of osteoderms; (8) presence of black connective tissue scattered over dorsal musculature; (9) medium body size for the genus (SVL of adults: 11.8-14.7 mm for males and 13.9-15.2mm for females; (10) advertisement call characterized by one note repeated in sequence, commonly comprised of 8-12 pulses.

Brachycephalus herculeus sp. nov.
Comparisons with other species.The presence of hyperossification in the skeleton and skin with dermal ossification on the head and dorsum distinguish the new species from all members of the B. pernix species group (B. actaeus, B. albolineatus, B. auroguttatus, B. boticario, B. brunneus, B. coloratus, B. curupira, B. ferruginus, B. fuscolineatus, B. izecksohni, B. leopardus, B. mariaeterezae, B. mirissimus, B. olivaceus, B. pernix, B. pombali, B. quiririensis, B. tabuleiro, B. tridactylus and B. verrucosus;Pombal et al. 1998Pombal et al. , 2018;;Ribeiro et al. 2005Ribeiro et al. , 2015Ribeiro et al. , 2017;;Alves et al. 2006;Garey et al. 2012;Pie et al. 2015;Bornschein et al. 2016;Monteiro et al. 2018a;Mângia et al. 2023), and from B. clarissae, B. didactylus, B. hermogenesi, B. pulex, B. puri, and B. sulfuratus, all of which lack hyperossification (Izecksohn 1971;Giaretta and Sawaya 1998;Napoli et al. 2011;Condez et al. 2016;Almeida-Silva et al. 2021;Folly et al. 2022) B. crispus, by the presence of wart-like osteoderms on the skin of the dorsum, flanks, venter, arms, and legs (see Folly et al. 2021a).The new species' dorsal aspect is similar to that of B. nodoterga, which also presents greenish coloration on dorsal surfaces, but differs from the latter in that its dorsal color is leaf green (olive green in B. nodoterga) and does not extend to the flanks and to the sides of the head as it does in B. nodoterga.Brachycephalus herculeus sp.nov. is the sister species of B. bufonoides according to our molecular analysis (Fig. 3).These species exhibit some of the smallest interspecific genetic distances among all pairs of species within the Brachycephalus vertebralis group (Table 2).They also share a prominent pelvic region, which is externally marked as a v-shaped bulge (presumably representing the protrusion of the iliac bones underneath) whose vertex lies at the posterior end of the body and whose tips extend to the middle of the trunk on either side; this characteristic distinguishes Brachycephalus herculeus sp.nov.and B. bufonoides from the other species of the B. vertebralis group.Notwithstanding, Brachycephalus herculeus sp.nov. is easily distinguished from B. bufonoides by the presence of osteoderms on the skin (absent in B. bufonoides), by its greenish dorsal color in life (orange in B. bufonoides), and by the advertisement call comprised by notes with 10-12 pulses (notes with 13-17 pulses in B. bufonoides).Furthermore, in terms of osteology and internal anatomy, the new species can be also distinguished from B. bufonoides by its lateroposteriorly oriented parotic plate (posteriorly oriented in B. bufonoides), oesophageal process of hyolaryngeal cartilages longer than arytaenoids (shorter than arytenoids in B. bufonoides), absence of an elbow sesamoid (present in B. bufonoides), paravertebral plates not ornamented (ornamented in B. bufonoides), and dorsal crest of the urostyle extending for 2/3 of its length (extending for all its length in B. bufonoides; Folly et al. 2020Folly et al. , 2021a)).
Coloration of holotype in preservative (Fig. 1).Forearms and legs gray; hands, feet, arms, lateral and ventral sides of the body beige; two contiguous dorsal paired dark gray marks bordering the vertebral column; scattered whitish warts (osteoderms) on the dorsum, flanks, arms, and legs; a cream line below each eye.
Color in life (Fig. 4).General background color orange; central area of head and dorsum and dorsal part of thighs and shanks green.Osteoderms, appearing as light yellow "warts" distributed throughout most of the body and limbs.Hands, feet, elbows, knees, and ankles orange.Ventral surface of body bright yellow.Eyes entirely black, with no visible delimitation between pupil and iris.6).

Variation. Morphometric variation is given in
Spinal plates weakly to well-delimited externally (Fig. 5).
In preservative, dorsal and lateral surfaces of body can be cream to gray, varying especially on the head and limbs.
A dissected preserved specimen (CFBH 28130), show scattered areas with pigmentation (dark tissue) in the dorsolateral region adjacent to the dorsal musculature, also around the line of the spinal vertebrae (Fig. 2).The visibility of black connective tissue scattered over the dorsal musculature is externally represented, in most specimens (70% of the type-series), by paired contiguous dorsal dark gray marks bordering the vertebral column (Fig. 5).
Distribution (Fig. 7).Brachycephalus herculeus sp.nov., is known only from the type-locality, the Parque Estadual do Desengano, in the municipality of Santa Maria Madalena, state of Rio de Janeiro, southeastern Brazil.This site is located within the northern portion of the Serra do Mar Mountain Range.
Natural history.The new species lives amidst the leaf-litter and can be found active during the day.Most individuals were found hidden amidst the leaf-litter, under fallen trunks, and among roots.The holotype and paratype (MNRJ 42407) were collected on the ground, during nocturnal surveys.The holotype was collected using quadrat sampling (see Siqueira et al. 2011) and thus may have been disturbed by the revolving of the leaf litter during the survey, so it is uncertain if it was active at night.The latter specimen may also have been found inactive under a decomposing leaf at the time.Paratype (MNRJ 42409) was found during the afternoon (at 1500 h) inside a bromeliad (Vriesea sp.) that was attached to a tree stump, ca.80 cm from the ground.The specimens collected in 2011,   S2 for the GenBank accession numbers.
during the rainy season, were on sloping terrain bordering trails in the forest.Males were not observed exposed when calling.An amplectant pair was found hidden under the leaf-litter in February 2011, during the morning after a rainy night.When exposed, the male changed its axillary amplexus to inguinal, as described for B. rotenbergae (Pombal et al., 1994, where it was called B. ephippium) and the pair moved away looking for shelter.Minutes after having hidden among the leaf litter, the female left the place and the male remained immobile.No eggs were found.
The main structure of the advertisement call of the new species, characterized by one note repeated in sequence, commonly comprised of 10-12 pulses, distinguishes it from the advertisement call of B. actaeus, B. albolineatus, B. hermogenesi, B. mirissimus, B. olivaceus, B. pernix, B. quiririensis, B. tridactylus (in which the notes are comprised of 1-4 pulses; Monteiro et al. 2018a,b;Bornschein et al. 2018;Pie et al. 2018;Bornschein et al. 2019)  the advertisement call of Brachycephalus herculeus sp.nov.overlap with those described for B. crispus, B. pitanga, B. bufonoides, and B. ibitinga (Condez et al. 2014;Oliveira and Haddad 2017;Folly et al. 2020;Condez et al. 2021).Among these species, the number of pulses per call ( 10 Etymology.Herculeus is a Latin adjective meaning very great, difficult, or dangerous; requiring the strength or courage of Hercules to challenge or accomplish a mission, as herculean labor or task (derived from Greek Mythology).The specific epithet was chosen to represent the "herculean" task that is, for such a small species, to survive in one of the most threatened forest environments in the world, the Brazilian Atlantic Forest.

Molecular analysis.
We obtained a final aligned matrix of 5,456 base pairs considering the selected mitochondrial and nuclear gene fragments.The optimal partition scheme and nucleotide substitution models selected for each data block were GTR+I+G (12S rRNA, 16S rRNA), K80+G (fist codon position of COI), GTR+I co-   Folly et al. 2022).Brachycephalus sulfuratus was recovered as the sister species to a clade containing B. puri and two undescribed taxa (Brachycephalus sp.7 and Brachycephalus sp.8), but its position relative to the pumpkin-toadlets remains unstable (e.g., Condez et al. 2020;Folly et al. 2020Folly et al. , 2022;;Lyra et al. 2021).The position of B. puri in our tree was distinct from that presented at the original description of the species (Almeida-Silva et al. 2021), probably due to the inclusion of samples of the two flea-toad candidate species in our dataset.Brachycephalus herculeus, new species, was recovered as monophyletic in our phylogenetic reconstruction, and was nested within the B. vertebralis species group, as the sister species of B. bufonoides (Fig. 3).Based on the genetic distances for the 16S rRNA gene (1548 bp), the sequences of different individuals of the new species did not differ from each other (p-distance = 0; n = 11).Within the B. vertebralis species group, the smallest genetic divergences between the new species and other congeners were 0.5-0.8%(from its sister species B. bufonoides) and the greatest was 4.9% (from B. guarani; Table 2).(Figs 9,10).Shape and proportions: Isolated bony structures (osteoderms) are predominantly distributed on the lateral surfaces of the body.The skull is wider than long (length/width range 85-95%).The diameter of the orbit is 39-47% of the total length of the skull.The skull is widest at the prootics, and the jaw articulation lies anterior to the posterior end of the skull at the occipital condyles.The upper jaws are relatively short, with the posterior apex lying at the level of the optic fenestra.Large additional elements associated with the skull (parotic plate) and vertebrae (vertebral and paravertebral plates) are dermal bones.-Skull plate.Exostosis and co-ossification are present in the skull forming an ornamented ridge-like pattern.Thus, nasals, dermal sphenethmoid, frontoparietals, squamosals, and the parotic plate are synostosed and exostosed, forming a dorsal skull plate.A marginal extension of the frontoparietals is fused to the parotic plate, dorsally covering the exoccipitals and prootics.There is a space between the parotic plate and a post-orbital crest forming a small concavity that reaches half of the width of the orbit.The parotic plate is connected and fused medially to the frontoparietal and has a lateroposterior orientation related to the main axis of the skull.The anterior extension of the parotic plate is small, not covering the articulation between squamosal and prootic and the posterior extension does not reach the posterior border of the occipital condyle.The post-orbital crest is latero-posteriorly projected and well developed, reaching the height of occipital con- dyle.-Dermal investing bones.Nasals: Exostosed.These bones are completely synostosed with the sphenethmoid but are not in anteromedial contact.An attenuate maxillary process contacts the preorbital process of the pars facialis of the maxilla.Frontoparietals: Expanded and exostosed.-Neurocranium.Sphenethmoid: Expanded.This bone is dorsally covered by frontoparietals and nasals.The cartilaginous optic fenestrae are absent.Fused Exoccipitals and Prootics: Not expanded and nor exostosed.The epiotic eminence (anterior and posterior) is undetectable.These bones are dorsally covered by the skull plate (frontoparietal fused with parotic plate).-Ventral investing bones.Parasphenoid: The cultriform process is synostosed with the sphenethmoid (Fig. 9).The parasphenoid alae are broad and laterally oriented beneath the otic capsule; the distal margins of the alae are wider than the proximal margins, terminating between the first third and the midpoint of the otic capsule.The posterior margin (posteromedial process) of the parasphenoid terminates in a triangular apex that lye the foramen magnum.Neopalatine: Absent.Vomers: Reduced.The anterior process is triangular with a rounded tip and is around the same length of the prechoanal process.The reduced prechoanal process forms the anterior and anteromedial margins of the choana and terminates in a rounded tip.The postchoanal process forms the posteromedial margin of the choana.The posteromedial margins of the vomers are moderately separated from each other, diverging abruptly anteriorly.The pre and postchoanal processes are around the same length.The dentigerous process is absent.-Maxillary arcade.Premaxillae: Each premaxilla is broad and separated from each other by a short gap in anterior view.The pars dentalis of each premaxilla lack teeth.In anterior view, the height of the alary process corresponds to around the length of premaxillae, and the distal tip of each alary process is pointed.The basal parts of the alary processes converge medially.The distal tips diverge from one another.Laterally, the alary processes are curved.The distal end of each pars palatina is sharp and converges to each other almost touching one another.Maxillae: The anterior end of the maxilla overlaps the posterolateral end of the premaxilla.Maxilla toothless; its posterior end is rounded and does not reach the ventral ramus of the squamosal.The pars facialis is high and thin, extending for less than half the length of the maxilla.Quadratojugals: Absent.-Suspensory apparatus.Pterygoids: The anterior ramus of each pterygoid is long, cylindrical, terminating in a truncated apex and extending to the final third of the maxilla.The medial ramus is short, slightly laminar, has a truncated end, and invests the prootic.The posterior ramus is long, triangular, and has a truncated point.Squamosals: Expanded and exostosed.They are composed of three rami that render each squamosal as a "T" shaped element composed by the ventral ramus, the otic ramus (posterior ramus), and the zygomatic ramus (anterior ramus).The ventral ramus is the longest of the three and has a rounded distal end, which is wider than its proximal portion.In lateral view, the zygomatic ramus is reduced, shorter than the ventral ramus and the distal end is rounded towards the maxilla, but without contacting it.In lateral view, the otic ramus is longer than the zygomatic ramus, and its distal end is rounded.In dorsal view, the distal end of the otic ramus is truncated, towards the parotic plate, overlapping it.Mandible: The mentomeckelian bones are in the anterior part of the mandible and are separated from each other by a short space.In anterior view, the mentomeckelians are sharp anteriorly and are fused to the dentaries laterally.(Fig. 10).The hyoid plate is longer than wide, with its length around four times its smallest width.The anterior processes are long and form a deep hyoglossal sinus, which deepens to nearly the height of the alary processes.The alary and posterolateral processes are much reduced, and the former is longer than the posterolateral processes.The posteromedial processes diverge widely to embrace the broad larynx.The arytenoids are short, semicircular and narrowly separated from each other, while the oesophageal process is longer than the arytenoids.-Postcranium (Fig. 9).Pectoral girdle: The clavicle, procoracoid, epicoracoid, coracoid, and scapula are completely ossified and fused, with a large fenestra separating the clavicle from coracoid; suprascapula not expanded, with anterior half ossified as cleithrum; omosternum and sternum absent.The two sides of the coracoid are well-devel- Sesamoids are present on both sides of sacral diapophyses.There are two types of bony elements associated with the vertebral column: (1) spinal plates, which lie dorsal to the vertebra and are generally ornamented (except for the posteriormost ones, which might not be ornamented); and, (2) the paravertebral plates, which are associated with the transverse processes of vertebrae IV; they are poorly developed, not covering the distal part of the transverse processes, and are not ornamented.The dorsal crest of the urostyle extends for 2/3 of its length.The ilium crest extends for the whole length of this bone.Manus: Phalangeal formula of hand 1-2-3-1.The carpus is composed of a radiale, ulnare, element y fused with Carpal 2, and a large postaxial assumed to represent a fusion of centrale with Carpals 3-5.Prepollex with two elements ossified and very reduced.Tips of the terminal phalangeal elements of fingers arrow shaped.A palmar sesamoid is present.Pes: Phalangeal formula of foot 1-2-3-4-1.Tarsus composed of tibiale, fibulare, three individual elements, including distal tarsal 2-3, distal tarsal 1, element y.Distal tarsal 1 is the smallest and articulates with element y, distal tarsal 2-3, and metatarsal I and II.Distal tarsal 2-3 articulates mainly with metatarsal III, but also with metatarsal II and IV, and with distal tarsal I. Prehallux has a single element.Tips of the terminal phalangeal elements of toes II-V arrow-shaped, toes I and V reduced, digit IV elongated.Plantar and cartilage sesamoids are present in all specimens.Three sesamoids are present at the knee region, the larger one might be the graciella sesamoid.

Discussion
Osteological, bioacoustic and genetic data seem to be important to elucidate species limits, especially when considering extremely diverse lineages (i.e., Monteiro et al. 2018a).Brachycephalus herculeus sp.nov. is distinguished from all its congeners based on morphological, molecular, and acoustic evidence, relying on divergence of both internal and external phenotypic traits that characterize the pumpkin toadlets.Within the B. vertebralis species group, a strong set of morphological characters distinguishes the new species from its sister taxon B. bufonoides, confirming its specific distinctness.
The genus Brachycephalus represents an interesting case among anurans, as it contains populations/species that are barely or not distinguishable in external morphology, yet separated by relatively deep genetic distances, as well as species that are distinct from each other morphologically but have very shallow genetic divergence (Condez et al. 2020;Folly et al. 2020Folly et al. , 2022;;Bornschein et al. 2021;Almeida-Silva et al. 2021).The former case is frequent among the cryptically colored species with "flea-toad" phenotype, whereas the latter case occurs mainly among the brightly colored species with "pumpkin toadlet" phenotype.For example, the very similar-looking flea-toads B. hermogenesi and B. sulfuratus diverge by 10% in the 16S gene (and are not even sister species), whereas the morphologically distinct pumpkin toadlets B. nodoterga and B. toby have practically no divergence (Almeida-Silva et al. 2021).Brachycephalus herculeus sp.nov.and B. bufonoides represent yet another case of a sister-species pair whose members differ considerably in external appearance but present little genetic divergence (<1%).
This is even more remarkable when we consider other direct-developing taxonomic groups, such as the genera Ischnocnema (Gehara et al. 2013;Oswald et al. 2023), Pristimantis (Padial et al. 2009), Euparkerella (Fusinatto et al. 2013) and Adelophryne (Lourenço-de-Moraes et al. 2018), as well as groups with non-free living tadpoles such as Adenomera (Fouquet et al. 2014), which tend to exhibit high genetic distances among species.Considering the natural history and biology of species in the genus Brachycephalus, which are slow-moving, small-bodied and lack free-swimming larvae (traits suggestive of low dispersal ability and limited gene flow among spatially separated populations), one would expect significant genetic distances among its species.Yet, the genetic distance between Brachycephalus herculeus sp.nov.and B. bufonoides is similar to the ones observed among some species of Alsodes (Blotto et al. 2013), a group with free-living tadpoles inhabiting streams.As in previous studies (Almeida-Silva et al. 2021), we also recorded  2), despite their morphological distinctiveness.The extremely low genetic divergences among some species of pumpkin toadlets with a high number of diagnostic characters, including osteological ones, is remarkable and deserves to be further investigated.It hints at a very fast rate of morphological differentiation in the genus Brachycephalus, or at least in some clades within it, though for reasons not yet understood.
Many species in the genus Brachycephalus have increased cranial ossification while retaining a variety of juvenile features, with the neopalatine, quadratojugal and columella liable to be lost in the adults within different lineages (Trueb and Alberch 1985;Alves et al. 2006;Bornschein et al. 2016).The ancestral form of the genus is presumed to have had a miniaturized body and to have lacked paravertebral plates, and the evolution of those plates seems to be correlated with an increase in body size (Condez et al. 2020;Dos Reis et al. 2021).The absence of ornamentation on the paravertebral plates of the new species could be seen as a juvenile feature following the ontogenetic sequence of Campos et al. (2010).However, the five individuals whose skeletons were analyzed, all with mature gonads and skull presenting adult features, confirm this character as diagnostic.Such a mosaic of paedomorphic and non-paedomorphic features in the morphology of adults, as well as the intraspecific variability in the miniaturized species of Brachycephalus reiterates the importance of including more than one analyzed skeleton in studies of that genus (Hanken 1984;Yeh 2002).
Histological analysis of dissected individuals of Brachycephalus darkside shows the pigmentation in tissue following adjacent muscle fibers, i.e., a connective tissue covers the dorsal musculature (epimysium) and is present between muscle fibers (perimysium and endomysium).This tissue exhibited spots of extracellular matrix containing dark pigments, which were reflected in the macroscopy of fixed specimens as well (Guimarães et al. 2017).Concerning the presence of dark-pigmented tissue within the B. ephippium species group, we add that there is no pigmentation in B. ephippium, while the black connective tissue covers all sides of dorsal muscles in B. darkside, and a few scattered areas with pigmentation are observed in the dorsolateral region adjacent to the dorsal musculature in B. garbeanus and B. margaritatus (Guimarães et al. 2017;Folly et al. 2021b).Within the B. vertebralis species group (sensu Folly et al. 2022), there is black connective tissue covering all sides of dorsal muscles in Brachycephalus herculeus sp.nov.(present study), little dark pigmentation internally around the line of the spinal vertebrae in B. vertebralis, and no pigmentation in B. pitanga (Guimarães et al. 2017).An absence of pigmentation was also observed for species of the B. pernix species group (B. izecksohni and B. pernix;Guimarães et al. 2017).Currently, there are still open issues regarding the intraspecific variation of dark pigmentation within each of those species, as well as information about other species in the genus Brachycephalus.

Figure 1 .
Figure 1.Holotype of herculeus, new species (MNRJ 42408).A Dorsal, B ventral, and C lateral views of the body; D ventral view of foot, and E ventral view of hand.Scale bar = 1 mm.

Figure 2 .
Figure 2. Ventral and dorsal view of dissected preserved specimens of Brachycephalus herculeus, new species (CFBH 28130), showing scattered areas with pigmentation (dark tissue) in the dorsolateral region adjacent to the dorsal musculature, also around the line of the spinal vertebrae.

Figure 3 .
Figure 3.The majority rule consensus tree resulting from the Bayesian Inference analysis based on the concatenated alignment of 5456 bp (four mitochondrial and three nuclear genes) showing the relationships within Brachycephalus.Sequences of the new species are highlighted in blue.Numbers associated to nodes represent Bayesian Posterior Probabilities < 0.90; all other nodes were fully supported (BPP = 1.0).Voucher numbers of specimens are provided at each terminal, when available.See TableS2for the GenBank accession numbers.
positions of COI and cyt b), GTR+G (three codon positions of Tyr; first codon positions of RAG1 and RAG2; third positions of COI and cyt b), SYM+I+G (first codon position of cyt b), HKY (second codon positions of RAG1 and RAG2), HKY+G (third codon position of RAG1), K80+I (third codon position of RAG2).Our main results corroborate many of the previously published phylogenetic hypotheses for the genus Brachycephalus but differs from those with respect to the position of the flea-toads B. clarissae, B. hermogenesi, B. puri, B. sulfuratus, and the candidate species B. sp.7 and B. sp. 8 (Fig. 3).Nevertheless, our hypothesis supports the position of B. pulex as sister to all other Brachycephalus species, as suggested in previous studies (Almeida-Silva et al. 2021; Dos Reis et al. 2021; Lyra et al. 2021).It also corroborates the position of B. didactylus as the sister species of the clade comprised of the B. ephippium and B. vertebralis species groups (Condez et al. 2020; Dos Reis et al. 2021; Almei-

Figure 7 .
Figure 7. Geographic distribution of the hyperossified species within the Brachycephalus vertebralis species group.Note the occurrence of Brachycephalus herculeus, new species, in the state of Rio de Janeiro, southeastern Brazil.

Figure 9 .
Figure 9. High-resolution computed tomography (Micro-CT) scans of paratypes of Brachycephalus herculeus, new species, showing osteological features.A Dorsal, B ventral views of the skeleton (CFBH 28131); C left hand in palmar view (CFBH 28133); D right foot in plantar view (CFBH 28048 and 28137, respectively); E dorsal, F ventral, and G lateral views of the skull (without the lower jaw; CFBH 28133); H dorsal view of vertebral column (CFBH 28137 and 28133, respectively).The dots surround the skeleton in (A) and (B) are osteoderms.Scale bars = 1 mm.

Table 2 .
Interspecific uncorrected p-distances based on the 16S rRNA gene for the species within the B. vertebralis group.Minimum and maximum values are given as percentages (%).The number of analyzed individuals for each species (in parentheses), followed by their intraspecific distances are presented in bold.