Geographical Variation in Morphological and Bioacoustic Traits of Pseudopaludicola mystacalis (Cope, 1887) and a Reassessment of the Taxonomic Status of Pseudopaludicola serrana Toledo, 2010 (Anura: Leptodactylidae: Leiuperinae)

1 Pós-Graduação em Biologia Animal, Universidade Estadual Paulista (UNESP), Rua Cristóvão Colombo 2265, Jardim Nazareth, 15054-000 São José do Rio Preto, SP, Brazil 2 Pós-Graduação em Ecologia e Conservação da Biodiversidade, Instituto de Biociências, Universidade Federal de Mato Grosso (UFMT), Avenida Fernando Correa da Costa 2367, Boa Esperança, 78060-900 Cuiabá, MT, Brazil 3 Laboratório de Zoologia, Instituto Básico de Biociências, Universidade de Taubaté (UNITAU), Avenida Tiradentes 500, 12030-180 Taubaté, SP, Brazil 4Departamento de Ciências Básicas e Produção Animal, Faculdade de Agronomia, Medicina Veterinária e Zootecnia, Universidade Federal de Mato Grosso (UFMT), Avenida Fernando Correa da Costa 2367, Boa Esperança, 78060-900 Cuiabá, MT, Brazil


Introduction
Advertisement calls are important for mate recognition in anurans and hence contribute to premating isolation among sympatric species [1]. Comparison of advertisement calls can be used to determine differences and affinities among different anuran species and so to determine taxonomic identity and phylogenetic relationships [2,3]. Such comparisons may be especially useful for situations in which the taxonomic identity of a species is uncertain or contested involving synonymies, redescriptions, and species resurrections. Indeed, in the absence of consistent morphological characters for interspecific distinctiveness, advertisement calls have proven to be of great help in revealing and supporting hypotheses of putative new species of Pseudopaludicola [4][5][6][7][8][9].
Objectives of the present article are (1) to provide and to discuss the intraspecific variation in morphological and bioacoustic traits of P. mystacalis along its distribution range in Brazil; (2) to reassess the taxonomic status of P. serrana, based on the reinterpretation of the literature data and on the examination of new materials (specimens and calls), freshly obtained at the type-locality and/or its vicinities of P. serrana and related species.
Since the available diagnoses for P. serrana and P. murundu do not allow clearly distinguishing specimens from these two species, new material from additional localities was putatively associated with each of them based solely on the positioning of collecting and/or recording sites: populations from easternmost localities in Minas Gerais were considered to represent P. serrana and populations from westernmost localities in Minas Gerais (and those from the state of São Paulo) were considered to represent P. murundu.
Vocalizations of specimens of P. murundu, P. mystacalis, P. saltica, and P. serrana were recorded by André Pansonato with a professional digital recorder Marantz PMD 660 equipped with a Yoga EM-9600 external directional microphone. Digital recordings were sampled at 44.1 kHz sampling rate and 16 bit resolution and saved in uncompressed wave files. Spectrograms were edited using Raven Pro 1.3 software with the following configuration: for P. mystacalis, brightness 75%; contrast 80%; DFT size 512 samples; 3 dB filter bandwidth 112 Hz; time grid overlap 50%; for P. murundu, P. saltica, and P. serrana, brightness 70%; contrast 80%; DFT size 512 samples; 3 dB filter bandwidth 124 Hz; time grid overlap 50%.
We analyzed 1894 notes (Tables 3 and 4): 503 from P. murundu, 444 notes from P. mystacalis, 577 from P. saltica, and 370 from P. serrana. The following temporal variables were measured from the waveform: number of pulses per note; note and pulse duration (ms); internote and interpulse interval (ms). Note that repetition rate was calculated in notes per minute. Dominant frequency (note peak frequency; Hz) was obtained from the spectrogram. Terminology for bioacoustic variables follows Magalhães et al. [7], Pansonato et al. [8], and Heyer et al. [20]. Vocalizations are archived in the "Banco de Registros Bioacústicos" and housed at the "Laboratório de Herpetologia do Instituto de Biociências da Universidade Federal de Mato Grosso" (LH, Cuiabá, Mato Grosso).

Geographical Variation in Pseudopaludicola mystacalis
Discriminant function analysis (DFA) with our data on P. mystacalis did not group localities, neither regarding morphometric characters (males and females) nor regarding bioacoustic variables. The first function (DF1) of DFA with morphometric data explained 63% of the variation among males of different populations (Figure 3(a)) and higher loadings corresponded to thigh length (10.5), hand length (9.8), and foot length (8.4). The DF1 explained 85% of the variation among females (Figure 3(b)) and higher loadings corresponded to hand length (12.3), foot length (10.7), and snout-vent length (8.8). The DF1 of DFA with bioacoustic variables explained 79% of the total variation in our sample Advances in Zoology 5  ( Figure 3(c)) and higher loadings corresponded to note rate (38.5), internote interval (29.5), and note duration (22.1). No distinct groups were formed along the first function axis of the DFA. Toledo, 2010. Advertisement calls of Pseudopaludicola serrana recorded at the type-locality (Brumadinho, Minas Gerais) and nearly 120 km southwestwards at Serra do Lenheiro, Minas Gerais were similar to those of topotypes of P. murundu and distinct from topotypes of P. saltica, both in spectral and temporal parameters (Figures 4 and 5). The intraspecific variation in bioacoustic parameters of P. murundu and P. serrana, presented below, is based on calls from topotypical specimens and on calls recorded in additional localities listed in Appendix B.

Taxonomic Status of Pseudopaludicola serrana
Advertisement calls of P. serrana were composed of series of pulsed notes with nonconcatenated pulses. Each note consisted of 3.6 ± 0.6 (range 2-5) pulses. Mean note duration was 82 ± 58 ms (26-131) emitted, on average, at intervals of 167 ± 58 ms (63-288). Mean duration of each pulse was 9 ± 3 ms  A discriminant function analysis (DFA) with morphometric data revealed that P. serrana and P. murundu overlapped along the first function (DF1) of DFA, which explained 98% of the total variation (Figure 6(a)). While the DFA did not separate P. serrana from P. murundu, both were separated from a third species (Pseudopaludicola saltica) whose individuals also have long hindlimbs. Higher loadings corresponded to eye diameter (12.1), tarsus length (7.5), and tibia length (4.6). DFA from bioacoustic data also separated Pseudopaludicola saltica from P. murundu and P. serrana (Figure 6(b)). The first function (DF1) of DFA explained 99% of the total variation and higher loadings corresponded to dominant frequency (64.1) and number of pulses per note (6.1).
DFA of morphometric measurements using all specimens assigned 100% of the individuals of Pseudopaludicola saltica, 86% of the individuals of P. serrana, and only 16% of the individuals of P. murundu to the correct species (Table 5). DFA of bioacoustic variables using all recorded males correctly assigned 89% of the specimens of P. saltica, 83% of the specimens of P. murundu, and 40% of P. serrana (Table 5). In contrast, DFA with both morphometric and bioacoustic data using specimens of P. saltica and a mixed sample of specimens attributable to either P. murundu or P. serrana assigned 100% of these latter to a single taxon.
The overall similarity in morphology and general structure of the call of P. murundu and P. serrana, evidenced by the results of DFA with 12 morphometric and six bioacoustic variables, strongly support the rejection of the specific status of Pseudopaludicola serrana, here considered to be a junior synonym of Pseudopaludicola murundu. Geographical distribution of Pseudopaludicola murundu, as presently recognized, is not anymore restricted to its type-locality [16] but includes instead twelve distinct localities in southeastern Brazil in the states of Minas Gerais and São Paulo (Figure 7). (Cope, 1887). Among the 18 valid species of Pseudopaludicola, P. mystacalis is currently the one with the most widespread distribution range. Extent of occurrence in Brazil is nearly two million square kilometers encompassing areas in three different ecoregions along all the "diagonal of open formations" from South America [23]: Caatinga, Cerrado, and Pantanal. It is also found in areas of Chaco in Argentina [24] and Paraguay [25]. Although plausible, the occurrence in Bolivian Chaco was based on material that morphologically and bioacoustically would correspond to P. ameghini [12,26]. Habitats where individuals of this species were found include "restinga" [10, present study], "campo sujo" in elevated plateaus of the Cerrado Domain, and seasonally flooded fields in the Pantanal lowlands [8].

Geographical Variation in Pseudopaludicola mystacalis
Advertisement calls of Pseudopaludicola mystacalis from different Brazilian localities are composed of series of notes with concatenated pulses. Maximum variation in dominant frequency of the calls of all specimens analyzed is 861.3 Hz (from 4478.9 Hz, in Cuiabá, Mato Grosso, to 5340.2 Hz, in Balsas, Maranhão; Table 3 Table 3 and [8]).   variation in the number of notes per minute and in the number of pulses in four localities from the state of Mato Grosso is 81 notes/min in Cuiabá and 5 pulses/note in Cuiabá and Nossa Senhora do Livramento (see Table 3 and [8]). Mean note duration does not differ significantly in all the examined calls (Table 3; = 4.9; df = 2; = 0.03). Acoustic analysis of geographical variation in species with an extent of occurrence of the magnitude seen in P. mystacalis sometimes reveals cryptic taxa being treated under a single name. However, this was not the case in P. mystacalis, a taxon that remained erroneously characterized and poorly recognized in the literature until recently [8]. Our results reveal a relatively small variation in morphological and acoustical traits of P. mystacalis and allow confirming the presence of the species in a wide geographical area along the Brazilian open ecosystems.  Toledo, 2010. When describing P. serrana, Toledo [15] used as morphological characteristics separating the three species belonging to the P. saltica group the aspect (relative size and coloration) of nuptial pads in the external part of finger I and aspect of the vocal sac. Such characteristics allow differing P. serrana and P. murundu from P. saltica but do not allow differing P. serrana from P. murundu [15]. Previously, Haddad and Cardoso [17] had called attention to the fact that nuptial pads were more developed in specimens of P. saltica from the type-locality (Chapada do Guimarães, Mato Grosso) than in specimens collected and recorded in localities in the states of Minas Gerais and São Paulo. In the same article, Haddad and Cardoso [17] also presented a spectrogram of the call of a specimen from Campinas, São Paulo. Based on such reported morphological evidence and acoustic parameters, we here argue that specimens from Minas Gerais and São Paulo attributed to P. saltica in Haddad and Cardoso [17] in fact correspond to P. murundu. The only external morphological characteristic used by Toledo [15] to differ P. serrana from P. murundu was a shorter head length/head width ratio (HL/HW = 0.97±0.06 in the five individuals of P. serrana evaluated; 1.11±0.07 in 11 individuals of P. murundu). After having analyzed a higher number of specimens attributed to P. serrana ( = 21), we found that HL/HW (1.10 ± 0.06; see Table 6) in this sample does not differ from the values presented in the original description, and confirmed herein, for P. murundu.

Taxonomic Status of Pseudopaludicola serrana
The description of P. serrana was also based on physical characteristics of its advertisement calls [15]. However, the original dataset is relatively low: 12 notes of a single male of P. saltica, 14 notes of 2 males of P. murundu, and 15 notes of 3 males of P. serrana were analyzed and compared. Only one out of four acoustic variables (pulse duration) was considered diagnostic for the new species. Toledo [15] also showed that dominant frequency range of P. serrana was completely nested within the dominant frequency range of P. murundu. However, when later proposing the synonymization of Pseudopaludicola riopiedadensis Mercadal de Barrio and Barrio, 1994 to Pseudopaludicola ternetzi Miranda-Ribeiro, 1937, Cardozo and Toledo [5] based their decision on "several important overlaps", including dominant frequency ranges, between the two species. Indeed, our data showed that dominant frequency contributed to higher loadings to separate species in the P. saltica group. Nonoverlapping dominant frequencies would therefore be expected between related but distinct species of Pseudopaludicola.
Bioacoustic data have proven to be useful for uncovering morphologically cryptic species in the genus Pseudopaludicola [4-9, 15, 16]. Advertisement calls of species in this genus might be divided into three distinct groups according to type of notes: nonpulsed notes, notes with concatenated pulses, and notes with nonconcatenated pulses [7]. Each group might be diagnosed by temporal (pulses per note, note and pulse duration, internotes and interpulses intervals, note rate, and pulse rate) and spectral variables (dominant frequency and harmonics), which in combination allow characterizing all the species for which calls were already analyzed, except for P. serrana and P. murundu. As evidenced by the recent synonymization of Pseudopaludicola riopiedadensis to P. ternetzi [15], and reinforced herein by the proposed synonymization of P. serrana to P. murundu, morphologically similar species of Pseudopaludicola presenting the same type of note and an overlap in their spectral acoustic variables are expected to represent the same taxon. Therefore, no arguments remain to consider P. serrana as a distinct species from P. murundu and the former must be considered as a junior synonym of P. murundu.

Conclusions
We here report on the variation in morphometric characters and variation in bioacoustic variables of the advertisement calls in fifteen Brazilian populations of Pseudopaludicola mystacalis. Although limited, our sample evidenced no statistical differences in any of the evaluated attributes and allowed us to confirm that this species is widely distributed in Brazil. Morphometric and bioacoustic data did not differ, also, in populations currently attributed to Pseudopaludicola serrana and Pseudopaludicola murundu, and the former taxon is here considered as a junior synonym of P. murundu. The data in the present work aims at contributing to a better understanding of the diversity and distribution patterns in the genus Pseudopaludicola.