Reassessment of the taxonomic status of Pseudopaludicola parnaiba (Anura, Leptodactylidae, Leiuperinae), with the description of a new cryptic species from the Brazilian Cerrado

The Neotropical frog genus Pseudopaludicola includes 25 species distributed throughout South America. Herein we review the taxonomic status of P. parnaiba relative to P. canga and the specific identity of the population treated in previous studies as Pseudopaludicola sp. 3 from Barreirinhas in the Brazilian state of Maranhão. The lack of differentiation in advertisement call, morphology, and mitochondrial markers from topotypes and different populations rejects the status of P. parnaiba and Pseudopaludicola sp. 3 from Barreirinhas as distinct species. For these reasons, we suggest to formally European Journal of Taxonomy 679: 1–36 (2020) 2 consider P. parnaiba as a junior synonym of P. canga. We also found that a population previously reported as P. facureae from central Brazil (Palmeiras de Goiás, Goiás) corresponds to a cryptic species that we describe here as a new species. Lastly, we provide for the first time the phylogenetic positions of P. giarettai, P. llanera and P. pusilla.


Introduction
Cerrado is the largest savanna formation in South America and is among the most threatened biodiversity hotspots on Earth (Myers et al. 2000), mainly due to habitat loss caused by agribusiness expansion, infrastructure development and limited conservation incentives (Strassburg et al. 2017). In the few last decades, the knowledge on anuran species richness of the Cerrado has expanded increasingly fast, indicating that species composition in the region needs to be urgently documented given the intensive anthropogenic activities developed in the region (Valdujo et al. 2013;Strassburg et al. 2017). The increase in species richness is partly due to the recognition of cryptic species in this formation (e.g., Vaz-Silva & Maciel 2011;Haga et al. 2017). Fišer et al. (2018) highlighted the importance of research efforts for fully integrating cryptic species into biodiversity science, thereby fostering a better understanding of the heterogeneous role of speciation in biodiversity pattern and process. Integrative taxonomy can notably improve the knowledge of frog diversity through species descriptions and delimitation, especially when dealing with morphologically cryptic species groups, recurrently reported for the frog genus Pseudopaludicola Miranda-Ribeiro, 1926 (e.g., Andrade et al. 2016Andrade et al. , 2018aAndrade et al. , 2018bAndrade et al. , 2019Pansonato et al. 2016;Cardozo et al. 2018).
Pseudopaludicola parnaiba Roberto, Cardozo & Ávila, 2013 is known only from its type locality, the municipality of Ribeiro Gonçalves, PI, Brazil (Roberto et al. 2013). This species is supposedly closely related to P. canga, P. facureae and P. atragula, but, to date, there is no molecular evidence available for a phylogenetic assessment. Based on acoustic and morphological traits, Carvalho et al. (2015a) stated that this species could not be distinguished from P. canga, and suggested that an integrative reassessment of the taxonomic status of P. parnaiba in relation to P. canga was necessary. Pseudopaludicola giarettai Carvalho, 2012 is a well-characterized species, based on acoustic and morphological data (Carvalho 2012;Carvalho et al. 2015b). However, the phylogenetic position of the species remains to be tested (Andrade et al. 2018a(Andrade et al. , 2018b(Andrade et al. , 2019. Carvalho et al. (2015a) also characterized the call of a population of Pseudopaludicola from the municipality of Palmeiras de Goiás, GO, Brazil, with the same trilled advertisement call pattern as P. facureae. Herein we combined acoustical, morphological and genetic evidence to (1) review the taxonomic status of P. parnaiba based on novel information from topotypes and additional populations of P. canga; (2) evaluate the specific identities of the population treated as Pseudopaludicola sp. 3 (sensu Veiga-Menoncello et al. 2014) and of the population from Palmeiras de Goiás; and (3) assess for the first time the phylogenetic positions of P. giarettai, P. pusilla (Ruthven, 1916) and P. llanera Lynch, 1989, providing the most complete mitochondrial phylogeny of the genus so far, with 21 species sampled. Our results revealed a cryptic species closely related to P. atragula and P. facureae, which we describe here as new.
Specimens were collected under authorization number #30059-12 issued by SISBio / ICMBio (Instituto Chico Mendes de Conservação da Biodiversidade). According to current legislation, the access to the National System for the Management of Genetic Heritage and Associated Traditional Knowledge was registered (SISGen #A2FCFCC). Individuals were euthanized by applying 5% lidocaine to the skin. After that, we collected muscle tissue for genetic analyses, fixed specimens in 10% formalin and transferred them to 70% ethanol for permanent storage. The new species hypothesis is in accordance to the General Lineage Concept, which treats species as separately evolving metapopulation lineages (de Queiroz 1998(de Queiroz , 2007.

Morphometry
We measured 11 adult males and five adult females (type series) of the new species under a stereo microscope Zeiss Stemi 2000 coupled to an ocular micrometer; except snout-vent length which was taken with a Mitutoyo Absolute digital caliper (to the nearest 0.1 mm) under a stereo microscope. Twelve morphometric traits were measured following Watters et al. (2016): snout-vent length, head length, head width, eye diameter, interorbital distance, eye-nostril distance, snout length, internarial distance, hand length, thigh length, tibia length and foot length. Tarsus length was measured following Heyer et al. (1990). Terminal phalanges or expanded toe tips were verified by clearing and staining. These procedures were conducted following the protocols of Taylor & Van Dyke (1985). Shape of the snout in dorsal and lateral views follows Heyer et al. (1990). Further details on examined specimens are in Appendix 1.

Bioacoustics
We recorded vocalizations with two digital recorders at sampling rate of 44.1 kHz and a sample size of 16 bits: Marantz PMD 661MKII (Marantz, Japan) and M-audio Microtrack II (M-audio, USA), both with a Sennheiser ME66/K6 or ME 67/K6 directional microphones (Sennheiser electronic GmbH & Co. KG, Germany). Directional microphones were positioned about 1.5 m from the calling male. We analyzed calls with Raven Pro 1.5, 64-bit version (Bioacoustics Research Program 2014) with the following settings: window type = Hann, window size = 256 samples, 3 dB filter bandwidth = 248 Hz, Fig. 1. Partial map of South America showing the Brazilian domains and samples of the species included in our molecular, morphological and acoustic comparisons. The type localities of the species are indicated with stars: Pseudopaludicola coracoralinae sp. nov. in Palmeiras de Goiás, GO (red star), P. facureae Andrade & Carvalho, 2013 in Uberlândia, MG (blue star), P. canga Giaretta & Kokubum, 2003 in Marabá, PA (yellow star) and P. parnaiba Roberto, Cardozo & Ávila, 2013 in Ribeiro Gonçalves, PI (green star). Municipalities: 1 = Barreirinhas (MA); 2 = Santo Amaro do Maranhão (MA); 3 = Aragominas (TO); 4 = Palmas (TO); 5 = Mateiros (TO); 6 = São Desidério (BA). Veiga-Menoncello et al. (2014) first noticed a taxonomic unit which they called as Pseudopaludicola sp. 3 from Barreirinhas, MA. Since then it has been treated as a species not yet formally described. It is represented here as P. canga from the localities 1 and 2. brightness = 50%, contrast = 50%, overlap = 85% (locked), DFT size = 1024 samples (locked) and grid spacing (spectral resolution) = 43.1 Hz. Raven obtained the peaks of dominant frequency through its "Peak Frequency (Hz)" function. The frequency values with 5 and 95% of call energy were obtained by "Frequency 5%" and "Frequency 95%" functions, and were considered as the minimum and maximum frequencies (Hz), respectively. We assessed frequency modulation through the "1 st Quartile Frequency" and "3 rd Quartile Frequency" functions; these Raven functions provide the frequencies that divide the selection into two frequency intervals containing 25 and 75% of the energy in the selection, respectively (Charif et al. 2010). We generated call figures using the Seewave ver. 1.6 package (Sueur et al. 2008) in R ver. 3.5.3 64-bit (R Core Team 2019). Seewave settings were: Hanning window, 90% overlap and 256 points resolution (FFT). We also assessed the between-male call variation through the coefficients of variation (CV = (SD / mean) × 100). We considered only the stereotyped non-pulsed notes to calculate the CV values of the species, not the introductory notes. Gerhardt (1991) reported that between-male coefficients of variation of static acoustic properties were less than 11%, whereas coefficients of variation of dynamic properties exceeded 15%.
Temporal traits were measured on oscillograms and the spectral traits were measured on spectrograms. Details for acoustic terminology employed here for the species are available in Supplementary file 1. Pulse terminology follows Magalhães et al. (2014); note and call terminologies follow Köhler et al. (2017). We calculated means and standard deviations considering mean values of individual males, whereas the range (variation) encompasses the minimum and the maximum values for all call samples. For multivariate analyses, we considered only the stereotyped notes, not the introductory notes, because the introductory notes are very irregular and have no clear pattern. Sound files are deposited in Arquivo Sonoro da Coleção de Anuros da Universidade Federal de Uberlândia at UFU and in Fonoteca Neotropical Jacques Vielliard (FNJV) at UNICAMP, both in Brazil.
among those randomly chosen at each node) then generating classifiers and aggregating results by voting to classes (further details in Liaw & Wiener 2002). The function proximityPlot (rfPermute ver. 2.1.6 package; Archer 2018) creates a plot of RF proximity scores using multi-dimensional scaling. The direct or indirect packages for this discriminant analyses were run in R ver. 3.5.3 64-bit (R Core Team 2019).
For the morphometric multivariate analysis between new species and P. facureae, we used snout-vent length, head length, head width, eye diameter, eye-nostril distance, internarial distance, hand length, thigh length, tibia length and foot length. For the acoustic multivariate analysis and statistical tests, we used note duration, internote interval, number of notes per minute, number of notes per series, series duration, interseries interval, number of series per call, peak of dominant frequency, and minimum and maximum of dominant frequency. Acoustic traits were tested for statistical significance of differences between species through the "Exact Wilcoxon Mann Whitney Rank Sum Test", function wilcox_test of the package Coin (Resampling Statistics model; Hothorn et al. 2008) in R. We considered significance when P ≤ 0.05.

Sequence analyses and phylogenetic inferences
For the taxonomic evaluations, we collected new tissue samples for specimens from Palmeiras de Goiás, GO; topotypes of P. parnaiba; specimens of P. canga from three different localities in TO; specimens We extracted total DNA from newly collected samples using a standard ammonium acetate precipitation method (Maniatis et al. 1982; adapted by Lyra et al. 2017). We amplified a fragment of mitochondrial DNA including the partial sequences of 12S rRNA, tRNA-val and 16S rRNA genes (H1 fragment, ~ 2450 bp; see Appendix 4 for primers used) for all species. PCR products were purified using enzymatic reaction and sent to Macrogen Inc. Republic of Korea, to be sequenced in an ABI 3730 automated DNA sequencer. New DNA sequences were edited for quality and assembled using Geneious ver. 11 (Biomatter) and submitted to GenBank (Appendix 3).
The new sequences were combined with the sequences available in GenBank for Pseudopaludicola spp. from previous works and 19 outgroups (Appendix 3), totalizing 94 samples. The H1 fragment was aligned using MAFFT ver. 7.25 using E-INS-I strategy (Katoh & Standley 2013). We performed Maximum Likelihood (ML) analysis with RAXML ver. 8.2.12 (Stamatakis 2014), searching for the most likely tree with 1000 replicates and using the GTRCAT substitution model. We then estimated node support with 1000 non-parametric bootstrap replicates under the same model. Analyses were run in the CIPRES Science Gateway (Miller et al. 2010). We edited the most likely tree in FigTree ver. 1.4.2 (http://tree.bio.ed.ac.uk/software/figtree).
The maximum genetic distances within species and / or populations and minimum genetic distances between species were calculated for the 16S fragment flanked by primers 16Sar-L and 16Sbr-H, since this fragment was available for all samples included in the analyses. Estimates were done using the package Spider in R ver. 3.6.1 (Brown et al. 2012; R Core Team 2019), uncorrected p-distances and the alignment obtained with MAFFT. Gaps and missing data were treated as pairwise deletions in uncorrected p-distances.

Taxonomic status of Pseudopaludicola parnaiba and Pseudopaludicola sp. 3 of Veiga-Menoncello et al. (2014) from Barreirinhas, state of Maranhão
Based on acoustics traits, we were unable to discriminate topotypical males of P. parnaiba from those of Pseudopaludicola sp. 3 from Barreirinhas and Santo Amaro do Maranhão (Table 1). The RF multivariate approach applied to acoustic data indicated a broad overlap between these two partitions ( Fig. 2), with a considerable classification error (Table 2). Moreover, the results were very similar when comparing both P. parnaiba and Pseudopaludicola sp. 3 with P. canga (topotypical and non-topotypical males from TO), with a broad overlap of these four partitions (Fig. 2). All three topotypes of P. canga were correctly classified, while the other three groupings had classification errors (Table 2). In contrast, the raw data of their variables overlapped with those of P. parnaiba, Pseudopaludicola sp. 3 and P. canga from TO (Table 1).
Furthermore, the specimens of Pseudopaludicola sp. 3 and P. parnaiba were nested together with specimens of P. canga from TO in the topology of the phylogenetic tree ( Fig. 3). Based on the 16S fragment, the minimum uncorrected p-distance was 0.21% between Pseudopaludicola sp. 3 and P. parnaiba, and 2.27% between Pseudopaludicola sp. 3 and P. canga from the type locality (Supplementary file 2). The genetic distance between P. parnaiba and P. canga from the type locality was 1.86% (Supplementary file 2).
In short, the acoustic and genetic evidence did not support a novel specific identity for Pseudopaludicola sp. 3 of Veiga-Menoncello et al. (2014). Also, the values of all traits of the analyzed calls of the two males from São Desidério, BA, overlapped with those described for P. canga in the present study (Table 1). Therefore, P. canga is the most suitable taxonomic identity for the populations of Pseudopaludicola sp.

Phylogenetic inference for Pseudopaludicola
The final alignment used for phylogenetic inference contained 2499 bp, and the tree obtained ( Fig. 3) recovered basically the same topologies and interspecific relationships inferred in previous analyses of Pseudopaludicola (Veiga-Menoncello et al. 2014;Andrade et al. 2016Andrade et al. , 2018aAndrade et al. , 2018bAndrade et al. , 2019 Number of series per call 12.9 ± 7.3   Fig. 3). In this subclade, P. canga was recovered as a sister taxon of P. facureae + P. atragula + Pseudopaludicola sp. (Palmeiras de Goiás, GO), with a maximum bootstrap support for these relationships (Fig. 3); and P. giarettai was recovered as a sister taxon of P. mystacalis + P. jazmynmcdonaldae, with a moderate support (BS = 88; Fig. 3). The four sequences of the specimens of P. giarettai from the type locality and Grande Sertão Veredas National Park were nested together in the topology (Fig. 3). This park is 400 km north of the type locality.

Synonymy
The acoustic and genetic analyses did not allow the discrimination between P. parnaiba and P. canga. Given that the phylogenetic positions and call data could not distinguish these two species, no current evidence remains to consider P. parnaiba as a different species from P. canga, and we consider that P. parnaiba Roberto, Cardozo & Ávila, 2013 should be treated as a junior synonym of P. canga Giaretta & Kokubum, 2003

Diagnosis
Pseudopaludicola coracoralinae sp. nov. is assigned to Pseudopaludicola by having a hypertrophied antebrachial tubercle (see Lynch 1989;Lobo 1995) and by its phylogenetic position within the genus.
The new species is characterized by the following combination of characters: (1) upper eyelids smooth, without enlarged palpebral tubercles; (2) heel smooth, without conical tubercle; (3) single, subgular vocal sac, cream-colored with white or off-white warts; (4) terminal phalanges knobbed, without T-shaped terminal phalanges or expanded toe tips; (5) relative short hind limbs (tibio-tarsal articulation just reaching the corner of the mouth); (6) trilled advertisement call pattern, composed of 2-6 welldefined series of tonal notes, having each series of 7-116 notes, emitted at rates of 1485-2077 notes per minute.

Etymology
The specific name honors Anna Lins dos Guimarães Peixoto Bretas, better known by her pseudonym Cora Coralina. She was a simple woman, a Brazilian candy maker, writer and poetess. She was born and raised on the banks of the Vermelho River, in the municipality of Goiás, GO, and lived apart from urban centers. Cora Coralina studied until the third year of elementary school and did a typing course at the age of 70, due to a requirement of the publisher that would publish her first book. She is considered one of the most influential Brazilian writers. Although Cora Coralina wrote her first verses during her adolescence, she had her first book (Poemas dos Becos de Goiás e Estórias Mais) published in June 1965, when she was 75 years old. In 1984, the Brazilian Union of Writers awarded her the "literary personality of the year". Following that honor, Carlos Drummond de Andrade, another distinguished Brazilian poet, said: "I admire Cora Coralina and her mastery of living in a state of grace with her poetry. Her verse is like running waters, her lyricism has the power and delicacy of the natural world."

Description of the holotype
Body elliptic and broad (Fig. 4A-B; Table 3). Head elliptical, slightly wider than long. Snout subovoid in dorsal view and rounded in profile (Fig. 4C-D). Eye not protuberant. Eye diameter almost equal to interorbital distance. Interorbital area flat. Pupil rounded. Upper eyelid without tubercles. Nostril not protuberant and closer to snout tip than to eye. Canthus rostralis rounded, smooth. Loreal region slightly concave. Single subgular vocal sac, externally expanded with warty texture. Choanae rounded, well separated from each other. Vocal slits present. Tympanum membrane and annulus absent. Discrete tympanic ridge from behind eye to proximal portion of arm insertion. Mouth opening ventral. Vomerine teeth absent. Tongue ovoid, longer than wide, free posteriorly, without pigmentation at its base. Lateral of head and flanks with discrete granules. One ovoid antebrachial tubercle presents in first quarter of forearm. Finger and toe tips not expanded. Outer and inner metacarpal tubercles welldefined; outer metacarpal tubercle rounded and inner metacarpal tubercle ovoid. Fingers with single and rounded subarticular tubercles. Supernumerary tubercles absent on palm of hand. Thumb with discrete, keratinized, light brown nuptial pad, extending from base of hand to proximal limit of terminal phalanx, covering almost entire external portion of finger. Webbing absent between fingers. Relative finger lengths, when adpressed one to another: I < II < IV < III (Fig. 4E). Outer metatarsal tubercle well defined, conical, smaller than ovoid inner metatarsal tubercle. Toes with well-defined, single, enlarged, rounded subarticular tubercles. Supernumerary tubercles absent on sole of foot. Toes webbed basally and fringed along their sides to almost their tips. Fringes developed on all toes (mainly on II, III, IV and V). External fringe on Toe V continues almost to outer metatarsal tubercle. Well-developed fold from internal metatarsal tubercle to mid-ventral tarsus, ending in protuberant tarsal tubercle. Relative toe lengths, when adpressed one to another: I < II < V < III < IV (Fig. 4F). Hind limb robust with tibiotarsal articulation just reaching posterior margins of eye. Thigh shorter than tibia. Foot longer than thigh. Foot longer than tibia. Heel without tubercles. Belly skin smooth. Abdominal fold present and complete. Well-defined vertebral stripe from snout tip to vent. Dorsal surfaces of head, body and limbs smooth. Paravertebral chevron-shaped dermal ridge from behind eye to scapular region. Cloacal region smooth (Fig. 4B). Measurements of the holotype showed in Table 3.

Color pattern of the holotype in preservative
Dorsum brown with dark brown and black gray blotches. Belly whitish (unpigmented). Vocal sac creamcolored with white or off-white warts. Dorsum darker than dorsal surfaces of limbs. Region between upper lip and eye with several rounded white blotches with alternating vertical grey and light beige stripes. Ventral faces of arm and leg unpigmented. Palm of hand brown, pigmented. Sole of foot brown, pigmented and darker than hand, arm and hind limb. Color of sole of foot similar to that of dorsal region of hind limb. Dorsal face of arm light beige with several dark brown blotches. Dorsal face of limb light brown with dark brown transversal discontinuous stripes and with scattered brown blotches. Dark brown transverse stripes on arm (2-3), thigh (2-3), shank (2-3), foot (3-4). Light brown nuptial pads (Fig. 4).

Variation in the type series
Dorsal surface of body varies from brown to dark brown, with black or dark brown irregular blotches (Fig. 5). The specimen ZUEC 24705 does not have well-defined light vertebral line. The specimens ZUEC 24705-06, 24708-10 and AAG-UFU 3396 have no paravertebral chevron-shaped dermal ridges from behind the eyes to the scapular region. The specimens ZUEC 24702-03, 24706, 24710-12 and

Advertisement call
The advertisement call of the new species (total duration: 1.3-25.8 s) consists of 2-6 series of stereotyped tonal notes (non-pulsed) that last 0.2-4.1 s, separated by intervals of 0.4-6.7 s. Before the emission of the series of stereotyped tonal notes, 12-40 (mean = 22.1, SD = 9.8) isolated notes with irregular structure, duration and interval are emitted, herein referred to as introductory notes (Fig. 6A). Introductory notes last 4-24 ms (mean 12, SD = 3), separated by intervals of 49-477 ms (mean = 146, SD = 27), and their dominant frequency peaks between 3.62-5.16 kHz (mean = 4.39, SD = 0.24). In contrast, within the series of stereotyped tonal notes, the notes have regular structure, duration and interval. These notes last 11-21 ms, separated by intervals of 12-61 ms, and are released at a rate of 1484.7-2076.6 notes per minute; notes have a slight increase in amplitude until the end of the first quartile of their durations, in the last quartile of their durations the notes suffer a decrease in amplitude (Fig. 6B). Dominant frequency peaks are between 4.18 and 5.06 kHz; the minimum frequency ranges between 3.84 and 4.59 kHz and the maximum frequency ranges between 4.41 and 5.44 kHz. The notes have a slight increase in frequency along their durations; on average, the notes have an increase of 275 Hz from the first to the third quartiles of frequencies (Table 4). The dominant frequency coincides with the fundamental harmonic, and the second harmonic ranges between 8.34 and 10.50 kHz (Fig. 6B). Air temperature of recorded calls varied from 22.2 to 26.0°C. Traits that were classified as static (between-male CV < 11%) to P. coracoralinae sp. nov. were note duration, notes per minute and all spectral traits. The other traits were classified as dynamic. Call quantitative traits and CV values are summarized in Table 4.

Differential diagnosis
Pseudopaludicola coracoralinae sp. nov. is promptly diagnosed from the P. pusilla species group (sensu Lynch 1989), which includes P. boliviana, P. ceratophyes Rivero & Serna, 1985, P. llanera, P. pusilla and P. motorzinho, by the absence of either T-shaped terminal phalanges or expanded toe tips (discs or pads). The new species has terminal phalanges knobbed, similar in shape to those of P. falcipes (Cardozo & Suárez 2012: fig. 2B). The new species is also distinguished from P. ceratophyes by having upper eyelids smooth; P. ceratophyes has upper eyelids with an enlarged palpebral tubercle (Lynch  1989). The new species also differs from P. boliviana, P. ceratophyes, P. llanera and P. motorzinho by having a smooth heel, without enlarged, conical tubercle (Lynch 1989;Pansonato et al. 2016).
Pseudopaludicola coracoralinae sp. nov. is promptly distinguished from the P. saltica species group that includes P. saltica, P. murundu and P. jaredi, by having short hind limbs (tibiotarsal articulation reaching near the corner of the mouth), whereas all three above-mentioned species have long hind limbs (tibiotarsal articulation extending beyond the tip of snout; Andrade et al. 2016).

Acoustic comparison with its sister species
Pseudopaludicola coracoralinae sp. nov. and P. facureae are indistinguishable in external morphology, but the new species was recovered as a sister species of P. facureae + P. atragula (Fig. 3). Furthermore, the RF multivariate approach applied to morphometric data indicated a broad overlap between the two partitions ( Fig. 7A-B), with a considerable classification error (Table 5). In relation to three species of Pseudopaludicola that share the trilled advertisement call pattern (P. hyleaustralis, P. facureae and P. canga), P. facureae is the one with the most similar call to that of P. coracoralinae sp. nov. The trait of notes per minute distinguishes the new species (1485-2077 notes per minute) from P. canga and P. hyleaustralis (368-1286 notes per minute; combined values, Table 1; see Carvalho et al. 2015a). The RF multivariate analysis on acoustic data indicated a complete segregation between P. coracoralinae sp. nov. and P. facureae, without any classification error (Table 5; Fig. 7C). Notes per minute (P. coracoralinae sp. nov. 1796 ± 123 (1485-2077) vs P. facureae 1383 ± 192 (512-1843)), number of series per call (P. coracoralinae sp. nov. 3 ± 1 (2-6) vs P. facureae 10 ± 6 (4-20)) and number of notes per series (P. coracoralinae sp. nov. 29 ± 16 (7-116) vs P. facureae 17 ± 18 (2-93)) were the main sources of variation in both variable importance measurements (Fig. 7D). In addition to these abovementioned traits, we found differences (P ≤ 0.01) between these two species in note duration, internote interval, series of notes duration, interseries interval and dominant frequency.

Phylogenetic inference and genetic distances of the new species
Pseudopaludicola coracoralinae sp. nov. was recovered as a sister species of the P. atragula + P. facureae clade (Fig. 3). Uncorrected genetic distance between the P. coracoralinae sp. nov. and P. atragula was 4.5% (mean value), and from P. facureae, it was 4.9% (mean value). The maximum intraspecific distance was 0.4% (Supplementary file 2). No molecular data are available for P. ceratophyes, P. hyleaustralis and P. ibisoroca; however, the new species is easily diagnosed from these species by morphology and acoustics (see further details in Differential diagnosis section).

Natural history notes
Males of the new species were found calling in a partially flooded open area surrounded by a newly planted cornfield (corn stalk < 40 cm tall). We collected the holotype and ZUEC's paratypes at this site. The AAG-UFU's paratypes were collected in another partially flooded open area near to a permanent lagoon at the margins of the GO-156 highway. We observed three couples in axillary amplexus in the field. In our field recordings of vocalizations, the males were vocalizing well-spaced from each other, without any close-range encounters. The new species was observed syntopically with Leptodactylus fuscus (Schneider, 1799) and Physalaemus marmoratus (Reinhardt & Lütken, 1862) at its type locality. Curiously, the congener Pseudopaludicola mystacalis was observed about 50 meters in a similar partially flooded open area surrounded by the same cornfield. We heard and observed only P. mystacalis at this site, not P. coracoralinae sp. nov.

Distribution
Pseudopaludicola coracoralinae sp. nov. is known only from the type locality (Fig. 1). However, we are aware of other populations that have trilled advertisement calls similar to those of P. coracoralinae sp. nov. and P. facureae. These populations occur in Limeira do Oeste, MG (Andrade & Carvalho 2013); Goianésia, Piracanjuba and in the Altamiro de Moura Pacheco State Park, all in GO, Brazil (Guimarães et al. 2001;Carvalho et al. 2015a;Ramalho et al. 2018). Goianésia is about 180 km northeast from the type locality of P. coracoralinae sp. nov., Piracanjuba is about 100 km southeast and the Altamiro de Moura Pacheco State Park is about 80 km northeast. Limeira do Oeste is closer to the type locality of P. facureae, about 250 km east. However, the specific identities of these populations will only be confirmed when their genetic information is available because they are morphologically and acoustically cryptic.

Discussion
The acoustic characterization of P. parnaiba showed here is in accordance with that of Carvalho et al. (2015a), and we were also unable to found reliable diagnostic characters between P. parnaiba and P. canga. In addition, the phylogenetic positions of the topotypes of P. parnaiba did not support the hypothesis that it evolved independently of P. canga. Therefore, P. parnaiba is regarded herein as a junior synonym of P. canga. Hence, the distribution of P. canga has increased considerably, occurring in four Brazilian states: PA (Marabá, type locality), MA (Barreirinhas and Santo Amaro do Maranhão), TO (Aragominas, Mateiros and Palmas) and PI (Ribeiro Gonçalves).
We analyzed calls from São Desidério, BA, and concluded that these individuals represent P. canga, since the values of all their traits overlapped those described for this species in the present study. Therefore, this is the first record of P. canga for BA about 900 km southeast from the type locality. Oliveira et al. (2013)   Pseudopaludicola coracoralinae sp. nov. and P. facureae are morphometrically indistinguishable from each other. However, the phylogenetic positions of these two cryptic sister species provides sufficient evidence to support our hypothesis that these lineages evolved independently. Moreover, based on the RF results of the acoustic comparison between P. coracoralinae sp. nov. and P. facureae, notes per minute was the main source of variation in both variable importance measurements. This trait was classified as static in P. coracoralinae sp. nov. (sensu Gerhardt 1991), due to its low between-male variability. Gerhardt (1991) pointed out that the spectral and fine-scale temporal traits of the frog calls are usually important for species recognition, whereas variable temporal traits may be important for mate choice. Therefore, P. coracoralinae sp. nov. differs statistically from P. facureae in a temporal trait that is expected to be associated with their species recognition.
In addition, we found differences in the environment occupied by P. coracoralinae sp. nov. and P. facureae. The marshy areas of both sites where we recorded and collected P. coracoralinae sp. nov. in Palmeiras de Goiás were temporary (see details in the Natural history notes section). On the other hand, P. facureae was always found in permanent marshy areas, both in disturbed areas and along palm marshes  Fišer et al. (2018) stated that evolutionary mechanisms leading to morphological similarity are heterogeneous, comprising recent divergence, niche conservatism and morphological convergence. Yet, those authors also argued that the biodiversity science is only just beginning to understand the 'invisible' world of cryptic species. They indicated that integrative approaches can reveal generalities in the speciation process, improving our understanding of the heterogeneity (species properties) in speciation, allowing a better integration in biodiversity science. Closely related species of Pseudopaludicola have similar external morphology and high intraspecific variation on dorsal color patterns (e.g., Andrade et al. 2017a), therefore, the association of multiple datasets is crucial for unequivocal identifications of Pseudopaludicola and to elucidate the hidden diversity within this Neotropical frog clade (Andrade et al. 2018a(Andrade et al. , 2018b(Andrade et al. , 2019. Thus, the identification of specimens of Pseudopaludicola based solely on their external morphology should be avoided in most cases. As an example, two recent studies failed in reporting the occurrence of P. falcipes for the Brazilian states GO and MG (Neves et al. 2019;Oliveira et al. 2019), because it is well-known that it does not occur in the Brazilian Cerrado (Langone et al. 2015(Langone et al. , 2016. The topology and interspecific relationships of the phylogeny proposed here for Pseudopaludicola species corroborated those topologies and relationships recovered previously (Veiga-Menoncello et al. 2014;Andrade et al. 2016Andrade et al. , 2018aAndrade et al. , 2018bAndrade et al. , 2019. As in previously phylogenetic inferences, we were not able to find good support for nodes of some subclades of Pseudopaludicola (e.g., subclades of P. mineira + P. matuta and P. florencei + P. restinga + P. pocoto, and P. saltica species group). These low bootstrap values can be explained by the existence of unknown species that are not represented in the topologies available so far, i.e., the species richness of Pseudopaludicola still remains underestimated.
Here we include for the first time P. pusilla and P. llanera in a molecular phylogenetic inference of the genus. Although they are nested together with the other sampled species from P. pusilla group, it is worth highlighting the low support for this clade. However, we recall here that we obtained only the 16S mitochondrial sequences of P. pusilla and P. llanera from the GenBank database, which were made available by Guarnizo et al. (2015). These authors preliminary identified and allocated to nominal species of their collected specimens using external morphology. They collected two specimens of P. pusilla in San Vicente, Santander, in the Magdalena River valley, western slope of the Eastern Cordillera (Colombia). Lynch (1989) restricted the distribution of P. pusilla to the lower and middle Magdalena River valley and to the Caribbean lowlands of northern Colombia and adjacent Venezuela. Therefore, we emphasize the need to sequence more markers for these Colombian specimens of P. pusilla, and further taxonomic studies with the species of this group, especially P. llanera, P. pusilla and P. ceratophyes. For example, there is no genetic information for P. ceratophyes yet. Also, acoustic data would be very important to better elucidate these taxonomic issues.
The increase of sampling efforts and the use of multiple datasets in Pseudopaludicola taxonomical studies have uncovered the striking species richness of these fascinating dwarf swamp frogs, especially in the last decade (e.g., Toledo et al. 2010;Pansonato et al. 2014aPansonato et al. , 2016Andrade et al. 2016Andrade et al. , 2018aAndrade et al. , 2018bAndrade et al. , 2019Cardozo et al. 2018). On the other hand, unsustainable agricultural activities, particularly soy production and cattle ranching, as well as burning of vegetation for charcoal, make the Cerrado one of the most threatened biodiversity hotspots (Strassburg et al. 2017). All these actions continue to pose a major threat to the Cerrado's biodiversity and despite its environmental importance and uniqueness, it is one of the least protected formations in Brazil. The recognition of the P. coracoralinae sp. nov. is important to the knowledge of the frog richness and diversification patterns that operated in this region. Future phylogeographic studies would be valuable to shed light on the evolutionary process of P. coracoralinae sp. nov., P. facureae and P. atragula.