Out of Bound: A New Threatened Harlequin Toad (Bufonidae, Atelopus) from the Outer Borders of the Guiana Shield in Central Amazonia Described through Integrative Taxonomy

We used integrative taxonomy to describe a new species of Atelopus from the lowlands of Central Amazonia in the region of Manaus, Amazonas, Brazil. The new species is geographically isolated from the southernmost species of Atelopus of the Guiana Shield. Atelopus manauensis species nova (sp. nov.) is characterized by the combination of the following characteristics: male snout-vent length range (SVL = 19.1–26.4 mm; n = 11); dorsal and lateral skin smooth; ventral surface entirely white or white with cream-colored gular region; fingers and toes lacking subarticular tubercles and fringes. The advertisement call of the new species has a call duration of 689–840 ms, contains 15–26 pulses, is emitted at an average pulse rate of 25.5 pulses per second, and has a dominant frequency ranging 3088–3610 Hz. The genetic divergence between the new species and its morphologically most similar congeners (A. spumarius and A. pulcher) is greater than 4%. Atelopus manauensis sp. nov. is closely related to species of the A. hoogmoedi complex inhabiting the Guiana Shield. The new species has a small geographic distribution (approximately 4500 km2) in a landscape that is strongly threatened by the growth of Manaus, the largest city in Brazilian Amazonia. The new species is considered critically endangered and in need of urgent conservation measures.


Introduction
The genus Atelopus comprises 96 [1] diurnal (except for A. nocturnus) and small-sized (17-50 mm) species known as harlequin frogs, which are distributed from Central America to northern South America [2]. Although few species have been included in recent phylogenetic studies, recent phylogenies show Atelopus to contain two major clades: the Andean-Chocó-Central American clade, and the Amazonian-Guianan clade. The latter comprises the tricolor subclade (pre-Andean region of Peru and adjacent Bolivia) and the flavescens-spumarius subclade [3]. The flavescens-spumarius subclade includes species distributed in the upper Amazon Basin (Ecuador, Peru, and Colombia), the Guiana Shield (Guyana, Suriname, French Guiana, and Brazil) and the lowlands of Central Amazonia (Brazil). The alpha taxonomy of the species distributed in the Guiana Shield and surrounding lowlands is complex due to morphological, bioacoustic, and genetic similarities [3,4], and different recognized The specimens were killed with a 5% lidocaine topical solution, fixed in 10% formaldehyde and preserved in 70% alcohol. Muscle tissue was extracted prior to fixation and kept in absolute alcohol. In addition to the advertisement call (type 1 call), three other call types were recorded and analyzed: type 2, type 3, and type 4. Following Lötters et al. [25], the type 2 call represents the pure tone call; the type 3 call represents the short pure tone call; and the type 4 call represents the pulsed short call. A type 2 call was recorded from a male in the drainage system on the east bank of the middle Puraquequara River located in CIGS/EB (SVL = 20.3 mm; 26.5 • C; 2 • 51 50.4" S, 59 • 48 54.0" W; 108 m a.s.l.; INPA-H 041372). Type 3 calls were recorded from the holotype in RFAD and from a male at the headwaters of the upper Cuieiras River located in Fazenda Dimona (SVL = 20.9 mm; 26.5 • C; 2 • 19 43.3" S, 60 • 4 41.46" W; 95 m a.s.l.; INPA-H 041373). The type 4 call was emitted by the holotype and recorded in laboratory. Type 2 and type 3 calls emitted by the holotype were recorded with a Sennheiser K6/ME66 directional microphone coupled to a Marantz PMD660 digital recorder. Type 2 and type 3 calls of other males and type 4 calls were recorded with a SONY PX333 digital recorder (format mp3, 192 kbps) with an internal microphone at 30 cm from the individuals. Files in mp3 format were converted to wav format (sampling rate 44.1 kHz, sample size 16 bits) to obtain acoustic parameters.

Phylogenetic Analyses
To infer the phylogenetic relationships of the new species and its congeners, 28 16S ribosomal RNA gene (16S) sequences and seven cytochrome c oxidase subunit I gene (COI) sequences were selected from GenBank of the National Center for Biotechnology Information (NCBI). All sequences used in the current study were obtained from GenBank. The sequences correspond to 15 species of Atelopus [22 16S sequences of 13 described Atelopus species + one sequence of an undescribed species from the Diversity 2020, 12, 310 4 of 25 south bank of the Amazon River (municipality of Anapu, state of Pará, Brazil) + four sequences from individuals of the new species, and four COI sequences from individuals of the new species + two from A. hoogmoedi from REBIO Uatumã + one from A. spurrelli], including two 16S sequences of A. hoogmoedi from Guyana, two from French Guiana, two from REBIO Uatumã, and three from the municipality of Monte Alegre, state of Pará, Brazil. A 16S sequence and another COI sequence of Rhinella marina were selected for rooting the phylogenetic trees. See Table 1 for detailed information on the specimens used in the phylogenetic analyses and accession numbers for all sequences in GenBank. After concatenating and aligning the total of 28 sequences in Geneious 4.8.2 using default parameters [26], a final matrix was obtained containing 1274 base-pairs (bp) (614 bp 16S and 660 bp COI). PartitionFinder2 [27] was used to infer the best partition scheme and its respective models of nucleotide evolution through PhyML 3.0 [28] and Bayesian Information Criterion (BIC). Suggested partitions and models were as follows: 16S = GTR + G; COI\1 = SYM; COI\2 = F81; COI\3 = HKY. We used for phylogenetic inference the same 20 16S and all COI sequences used in Jorge et al. [29], including those sequences of the new species, only adding eight new sequences, as follows: A. hoogmoedi (three sequences from the municipality of Monte Alegre, state of Pará, Brazil-IDs 8-10 in Table 1), A. sp. Anapu, A. barbotini "B", A. oxapampae, A. tricolor, and R. marina (IDs 7; 14; 26-28 in Table 1, respectively). Overall, we obtained the same phylogenetic relationships and statistical support as obtained by Jorge et al. [29], with a slight difference regarding the relationships of the French Guiana species, which is likely because of the addition of new samples (see Results section, Section 3.1). Phylogenetic relationships were estimated using Maximum Likelihood (ML) and Bayesian Inference (BI). The ML phylogenetic tree was estimated using IQ-TREE [36] with 10,000 ultrafast bootstraps; 10,000 iterations; a minimum correlation coefficient of 0.99; and 10,000 Shimodaira-Hasegawa approximate likelihood ratio replicates (SH-aLRT). The Bayesian phylogenetic tree was inferred using MrBayes 3.2.6 [37] through four runs of 10 million generations each using Monte Carlo Diversity 2020, 12, 310 5 of 25 via Markov Chains (MCMC) with four chains each. Probabilities were sampled every 1000 generations. The tree was summarized after burning of 25% of the sampled trees. Uncorrected genetic distances (p-distance) and Kimura-2-parameters (K2P; [38]) were calculated in MEGA 6.0 [39] using the 16S alignment.
Populations of Atelopus hoogmoedi of REBIO Uatumã and Pitinga River are the southernmost populations of the genus of the Guiana Shield and the geographically closest populations to the new species described here. For this reason, we used morphometric measurements of these populations to distinguish the new species from A. hoogmoedi. Three cluster and classification analyses were used for this purpose: (1) Principal Component Analysis (PCA); (2) Discriminant Analysis of Principal Components (DAPC; [45]); and (3) Random Forest (RF; [46]). The analyses were conducted in R [47] using SVL and 12 morphometric ratios (HW/SVL, HL/SVL, EYDM/SVL, EYNO/SVL, TL/SVL, IOD/SVL, ITNA/SVL, RDUL/SVL, FOOT/SVL, HAND/SVL, THBL/SVL, and SW/SVL) of 11 males of the new species (holotype, type series, paratypes, and paratopotypes) and eight males of A. hoogmoedi collected in REBIO Uatumã (4) and at Pitinga River (4). Paratype and paratopotype specimens were included in the analyses so that the extreme minimum and maximum body sizes observed among individuals of the new species could be included, and for the same reason, they were not included in the "type series". The PCA was performed using the "prcomp" function of the stats package using the parameters "scale = T" and "center = T" to scale and center the morphometric variables. A Multivariate Analysis of Variance (MANOVA) was performed to test whether the multidimensional morphometric space occupied by the new species differs significantly from that occupied by A. hoogmoedi in relation to the first two axes of the PCA. The DAPC was performed using the "dapc" function of the adegenete 2.1.2 package [45]. As recommended by Jombart et al. [45], all principal components (PCs) and discriminant analysis eigenvalues (DAE) were used in the DAPC. Random Forest is a learning algorithm that builds classification trees from random subsamples of a database to then aggregate results and classify samples (specimens) into groups (species) [45]. Random Forest was conducted with 4000 trees using the "randomForest" function of the randomForest 4.6-14 package [48]. Results are presented in the subsection "Morphological and Bioacoustic Analyses" of the section "Results". Raw morphologic measurements can be found in Supplementary Materials Table S1.

Bioacoustics
Calls were analyzed in Raven 1.5 [49] and configured as follows: Blackman window; 3 dB Filter Bandwidth of 80 Hz; overlap of 80%; hop size of 4.1 ms; and discrete Fourier transform (DFT) size of 2048 samples. Temporal parameters were measured using oscillograms, and spectral parameters were measured using "power spectrum" graphs. Different sets of acoustic parameters were measured for each of the four call types.
Ten advertisement calls (type 1 call) from three males of the new species and 35 calls from four males of Atelopus hoogmoedi from REBIO Uatumã and three males from Pitinga River were analyzed. Eleven parameters were measured in the advertisement calls, as follows: (1) call duration (Call_dur); (2) number of pulses (N_pulses); (3) pulse duration of first pulse (Puls_dur_1); (4) pulse Diversity 2020, 12, 310 6 of 25 duration of central pulse (Puls_dur_2), which usually precedes the last portion of the call when the pulses are emitted in a shorter time interval; (5) pulse duration of last pulse (Puls_dur_3); (6) pulse period of first three pulses, measured from the beginning of the first pulse to the end of the third pulse (Dur_3_first_puls); (7) pulse period of last three pulses, measured from the beginning of the antepenultimate pulse to the end of the last pulse (Dur_3_last_puls); (8) inter-pulse interval between first two pulses (Interpulse_1); (9) inter-pulse interval between central pulse and next pulse (Interpulse_2); (10) inter-pulse interval between last two pulses (Interpulse_3); (11) call dominant frequency measured using the "Peak Frequency" function (Domi_frequency); and (12) bandwidth. We also measured the advertisement calls of four males of A. hoogmoedi recorded by Costa-Campos and Carvalho [50] in Pedra Branca do Amapari (state of Amapá, Brazil), which is the closest population to the type locality of this species (Mont Atachi-Bacca, French Guiana) with data available to be re-analyzed in the same manner as the calls of the new species (recordings provided by the authors).
As with the morphological data, three cluster and classification analyses (PCA+MANOVA, DAPC, and RF) were used to test whether the new species can be distinguished from Atelopus hoogmoedi through the advertisement call. Analyses were performed using the average of each acoustic parameter for each male, three of the new species and 11 of A. hoogmoedi: four from REBIO Uatumã, three form Pitinga River, and four from Amapá recorded by Costa-Campos and Carvalho [50]. Results are presented in the subsection "Morphological and Bioacoustic Analyses" of the section "Results".
Six parameters were measured in type 2-4 calls, as follows: call duration, intercall interval, low frequency, high frequency, dominant frequency, and bandwidth. Low and high frequencies, as well as the bandwidth, were measured 20 dB below the peak frequency in order to avoid background noise. Dominant frequency and bandwidth were measured using the functions "Peak frequency" and "Bandwidth 90%" in Raven, respectively. In addition, we measured the intercall interval within a series of type 2 calls, the number of pulses per call, and the pulse duration of the first and last pulse of type 4 calls. The nomenclature of bioacoustic traits followed Köhler [51].

Phylogenetic Analyses
Both Bayesian Inference (BI) and Maximum Likelihood (ML) supported the flavescens-spumarius clade as monophyletic (Figure 1   Genetic distances between the new species and the other species of Atelopus included in the analyses ( Table 2) ranged 2.3-11.6% (p-distance) and 2.3-12.7% (Kimura-2-parameters, K2P). Genetic distances were relatively high between the new species and those with which it has been confused in the past: A. spumarius SS (p-distance = 4.7%; K2P = 4.9%) and A. pulcher (p-distance = 3.9%; K2P = 4.0%). Although A. hoogmoedi at REBIO Uatumã is the geographically closest population to the new species, the genetic distances between them varied between 2.5% (p-distance) and 2.6% (K2P). Similar values differentiate the new species and A. hoogmoedi of French Guiana (p-distance = 2.6%; K2P = 2.7%). The lowest values of genetic distance were between species of the A. hoogmoedi complex occurring in the Guiana Shield, which varied between 0.2-1.7% (p-distance) and 0.2-1.7% (K2P). Genetic distances between the new species and the other species of Atelopus included in the analyses ( Table 2) ranged 2.3-11.6% (p-distance) and 2.3-12.7% (Kimura-2-parameters, K2P). Genetic distances were relatively high between the new species and those with which it has been confused in the past: A. spumarius SS (p-distance = 4.7%; K2P = 4.9%) and A. pulcher (p-distance = 3.9%; K2P = 4.0%). Although A. hoogmoedi at REBIO Uatumã is the geographically closest population to the new species, the genetic distances between them varied between 2.5% (p-distance) and 2.6% (K2P). Similar values differentiate the new species and A. hoogmoedi of French Guiana (p-distance = 2.6%; K2P = 2.7%).
The lowest values of genetic distance were between species of the A. hoogmoedi complex occurring in the Guiana Shield, which varied between 0.2-1.7% (p-distance) and 0.2-1.7% (K2P).

Morphological and Bioacoustic Analyses
The multidimensional morphometric ( Figure 2A) and bioacoustic ( Figure 2B) spaces occupied by Atelopus manauensis sp. nov. and the two geographically closest populations of A. hoogmoedi (REBIO Uatumã and Pitinga River) were distinct for the first two axes of the PCAs (MANOVA morphology: Pillai trace = 0.69, degrees of freedom (df) = 16, p < 0.000001; MANOVA bioacoustic: Pillai trace = 1.00, df = 22, p < 0.002). The DAPCs were congruent with the results of the PCAs and showed that the two species can be distinguished from each other with maximum values of posterior probability of membership through morphology ( Figure 2C) and advertisement calls ( Figure 2D).

Morphological and Bioacoustic Analyses
The multidimensional morphometric ( Figure 2A) and bioacoustic ( Figure 2B) spaces occupied by Atelopus manauensis sp. nov. and the two geographically closest populations of A. hoogmoedi (REBIO Uatumã and Pitinga River) were distinct for the first two axes of the PCAs (MANOVA morphology: Pillai trace = 0.69, degrees of freedom (df) = 16, p < 0.000001; MANOVA bioacoustic: Pillai trace = 1.00, df = 22, p < 0.002). The DAPCs were congruent with the results of the PCAs and showed that the two species can be distinguished from each other with maximum values of posterior probability of membership through morphology ( Figure 2C) and advertisement calls ( Figure 2D). The variables that contributed most to the morphology DAPC were, in decreasing order, ITNA, THBL, and HW, while for the bioacoustic DAPC, they were the pulse duration of the central pulse, number of pulses, and pulse duration of the last pulse (Table 3). The multidimensional acoustic space occupied by Atelopus manauensis sp. nov. and that occupied by A. hoogmoedi of REBIO Uatumã, Pitinga River and Amapá in the first two axes of the PCA ( Figure 2B) were distinct (MANOVA bioacoustic: Pillai trace = 1.00; df = 22; p = 0.002). The DAPC with advertisement call variables obtained 100% accuracy in attributing the calls of the analyzed individuals to their respective species ( Figure  2D). The contribution of each acoustic variable and the explanation value of the first two PCs are provided in Table 3. The variables that contributed most to the morphology DAPC were, in decreasing order, ITNA, THBL, and HW, while for the bioacoustic DAPC, they were the pulse duration of the central pulse, number of pulses, and pulse duration of the last pulse (Table 3). The multidimensional acoustic space occupied by Atelopus manauensis sp. nov. and that occupied by A. hoogmoedi of REBIO Uatumã, Pitinga River and Amapá in the first two axes of the PCA ( Figure 2B) were distinct (MANOVA bioacoustic: Pillai trace = 1.00; df = 22; p = 0.002). The DAPC with advertisement call variables obtained 100% accuracy in attributing the calls of the analyzed individuals to their respective species ( Figure 2D). The contribution of each acoustic variable and the explanation value of the first two PCs are provided in Table 3. Table 3. Contribution of each variable to the first two axes of the PCAs in multidimensional morphological and acoustic spaces to explain the variation between Atelopus manauensis sp. nov. and A. hoogmoedi of REBIO Uatumã, Pitinga River and Amapá, Brazil. The total percentage of explanation of the variation between species and populations for each axis are provided at the bottom. Abbreviations: (SVL) snout-vent length, (SW) sacrum width, (HW) head width, (HL) head length, (EYDM) eye diameter, (EYNO) eye-to-nostril distance, (IOD) interorbital distance, (ITNA) internarial distance, (RDUL) length of flexed forearm, (HAND) hand length, (THBL) thumb length, (FOOT) foot length, (TL) tibia length. Bold numbers highlight five variables with highest contribution values for each Principal Component (PC).
Diagnosis: A small species of Atelopus; adult males SVL 19.1-26.4 mm (n = 11), adult females 27.9-28.8 mm (n = 2); interdigital webbing covering all of Finger I and a phalange of Finger II and rudimentary between other fingers; absence of serrated fringe on the sides of Finger III ( Figure 4A,B); first phalange of Toe I atrophied, completely hidden in skin of the foot (similar to a callus), with no visible phalanges ( Figure 5A,B); subarticular tubercles absent from hand and foot; palmar tubercles round and visible and plantar tubercles oval and poorly defined; the interdigital webbing reaches half of the third phalange of Toe IV and half of the first phalange of Toe V; in life, dorsum light brown to reddish brown with light yellow or light green reticulation network ( Figure 6A-D); ventral surface of feet, hands, half of the posteroventral portion of thighs and half or all of the posteroventral portion of the cloacal region red ( Figure 6E-H); throat, chest, and central portion of belly white or whitish cream without spots in males, spotted in females; advertisement call consisting of a single multipulsed note with a call duration of 689-840 ms consisting of 15-26 pulses, with a dominant frequency of 3,088-3,610 Hz and bandwidth of 633-915 Hz ( Figure 7A,B). Morphometric measurements of the holotype and type series are shown in Table 4.

Taxonomic Comparisons
Interspecific morphological and bioacoustic comparisons were made between Atelopus manauensis sp. nov. and the geographically and phylogenetically closest species. Characteristics of the compared species are presented in parentheses, unless specified.
Atelopus manauensis sp. nov. differs from the population of A. hoogmoedi of REBIO Uatumã and Pitinga River by lacking basal and subarticular tubercles on the hands [A. hoogmoedi basal tubercles present on hands ( Figure 8A) and feet and two subarticular tubercles on Toe IV ( Figure 8B  It differs from Atelopus hoogmoedi of Amapá (Brazil) by having a light brown to reddish brown dorsum with a light yellow or light green reticulation network (dark brown to black dorsum with yellow, orange, or pink reticulation network); venter all white or white with cream-colored gular region and head (bright yellow); red spot restricted to half of the posteroventral region of thighs and half or all of the posteroventral portion of the cloacal region red (red spot covering the entire posteroventral portion of the cloacal region extending to near the knee and to the initial portion of the belly; photos C.E. Costa-Campos at AmphibiaWeb/Atelopus hoogmoedi); advertisement call duration of 689-840 ms (873-1710 ms; [50]) consisting of up to 26 pulses (35 pulses; [50]), dominant frequency of 3088-3610 Hz (2812-2838 Hz; [50]), and a bandwidth of 633-915 Hz (280-568 Hz; [50]) ( Figure 7D,E).
It differs from Atelopus hoogmoedi of Guyana by having a light brown to reddish brown dorsum with a light yellow or light green reticulation network (dark brown to black dorsum with yellow dorsolateral bands and marks with black spots; [54]); venter all white or white with a cream-colored gular region and head (venter, throat, and head yellow, pink, or orange with irregular black marks; [54]); interdigital webbing covering all of Finger I and a phalange of Finger II and rudimentary between other fingers (fingers unwebbed; [54]); longer interdigital webbing between toes IV and V (toes moderately webbed; [54]); and Toe I reduced (Toe I and Toe II much reduced; [54]).
Atelopus manauensis sp. nov. differs from A. flavescens by the smaller size of females, maximum SVL of 28.8 mm (minimum SVL of 31.5 mm; [9]), and of males, maximum SVL of 26.4 mm (minimum SVL of 27 mm; [9]); shorter tibia in females of 12.1-12.9 mm (minimum tibia length 13.0-16.0 mm; [9]); dorsum light brown to reddish brown with light yellow or light green reticulation network (dorsum varies in light yellow tones with small brown or light red vermiculation; [9]); venter all white or white with cream-colored gular region and head in males and white with rounded spots in females (venter pink-salmon in females and pink in males; [8]); and advertisement call duration of 689-840 ms (1340-1820 ms; [7]) consisting of 15-26 pulses (45-58 pulses; [7]).
The new species differs from A. pulcher by its smaller size, with a maximum SVL for females of 28.8 mm (minimum SVL of 32.0 mm; [4]); longer interdigital webbing between toes IV and V (shorter interdigital webbing; [4]); no subarticular tubercles on the hand or foot (ill-defined subarticular tubercles on fingers II, III, and IV and toes II, III, IV, and V; [4]); venter all white or white with a cream-colored gular region and head in males and white with dark brown rounded spots in females (light red venter; [4]); red spot restricted to half of the posteroventral region of thighs and all or half of the posteroventral portion of the cloacal region red in males and females (red spot covering the entire ventral region of the thighs and all of the posteroventral portion of the cloacal region extending to the belly in males, and all of the venter, thighs, and head red with black spots in females; [4]); advertisement call duration 689-840 ms (1100-1300 ms; [4]) consisting of up to 26 pulses (47 pulses; [4]) and a dominant frequency of 3088-3610 Hz (2034-2824 Hz; [4]).
Finally, Atelopus manauensis sp. nov. differs from A. seminiferus by the smaller size of males and females: SVL 19.1-26.4 mm in males, 27.9-28.8 mm in females (SVL 33.8-35.2 mm in males, 40 mm in females; [56]); interdigital webbing covers two phalanges of Toe IV (interdigital webbing covers all of Toe IV; [57]); dorsum light brown to reddish brown with light yellow or light green reticulation network (dorsum uniformly dark brown or black with small yellow spots; [56,57]); venter all white or white with a cream-colored gular region and head (venter dark brown streaked with white and orange; [56]); and smooth sides (sides with tubercles; [57]).
Description of the Holotype: Body slender; neural spines not evident externally; head slightly longer than wide (HW/HL = 0.97); snout acuminate, with oval tip, dorsally concave; head length 30% of SVL; maxilla projected slightly over mandible; nostrils lateral and not visible from above; canthus rostralis concave between nostril and tip of snout and moderately straight between eye and nostril; loreal region concave; nostril closer to tip of snout than to eye; distance between nostrils and eye equal to eye diameter; distance between nostrils greater than eye diameter; tibia 51% of SVL; foot shorter than tibia (FOOT/TL = 0.74); relative sizes of fingers I < II < IV < III and toes I < II < III < V < IV, with first phalange of Toe I being atrophied and the toe hidden in the skin, similar to a callus; hand webbing rudimentary, present only between Finger I and Finger II; foot webbing, disregarding the Toe I in the form of a callus: II 0−2 − , III 1 1/2 -2 −1/2 , IV 2 −1/3 −1 + V; palmar tubercle well defined and rounded ( Figure 4A) and plantar tubercle poorly defined and oval ( Figure 5A); subarticular tubercles absent; thumb 44% of hand length (THBL/HAND = 0.44), covered by tiny brown keratinized spikes (nuptial pads); dorsal skin smooth; belly with tiny black keratinized dots; gular region smooth. In life, dorsal and lateral surfaces of the body with a lime-green reticulated network spread irregularly on a chocolate brown background. Ventral region white with a cream-colored gular region and head; lower part of limbs with brown bands on the sides. Red spot restricted to half of the posteroventral region of thighs and half of the posteroventral portion of the cloacal region ( Figure 6E). In preservative, the dorsal surface was brown and the reticulated network was pale yellow. Ventral region colored similar as in life, only the red spots on the hands, feet, posteroventral region of thighs, and posteroventral portion of the cloacal region turn reddish brown ( Figure 3A-D).
Vocalization: The advertisement call of Atelopus manauensis sp. nov. consists of a single multi-pulsed note issued at regular time intervals ( Figure 7A). The mean call duration is 744 ± 84 ms (689-840 ms) and consists of 19 ± 6 pulses (15-26 pulses) ( Figure 7B). Based on the inter-pulse interval, the song can be temporally divided into two portions, with the pulses being emitted more widely during the first two-thirds of the call than during the last third. The first pulse and central pulse have similar average pulse durations of 7 ± 1 ms (6-8 ms) and 6 ± 1 ms (5-7 ms), respectively. On the other hand, the last pulse is on average longer than all the other pulses in the song with a pulse duration of 14 ± 6 ms (8-19 ms), with downward frequency modulation ( Figure 7C). The inter-pulse interval at the beginning and center of the song lasts for 54 ± 16 ms (36-67 ms) and 50 ± 13 ms (40-65 ms), respectively, while the inter-pulse interval between the last two pulses lasts for 7 ± 3 ms (4-9 ms). The total pulse period duration of the first three pulses [132 ± 40 ms (92-172 ms)] is approximately three times longer than the pulse period duration of the last three pulses of the call [43 ± 6 ms (38-49 ms)]. The call has a dominant frequency of 3334 ± 263 Hz (3088-3610 Hz) and a bandwidth of 743 ± 151 Hz (633-915 Hz). The temporal and spectral parameters of the advertisement call of A. manauensis sp. nov., A. hoogmoedi of REBIO Uatumã, of Pitinga River, and of Amapá are provided in Table 5.

Variability
Male INPA-H 041376 is 4.6 mm larger than average in relation to total body size (SVL = 26.4 mm); its nuptial callus is a little evident; and the spot on the lower ventral portion was light red and changed to cream-color in preservative, similar to the two females. In general, the other males are similar to the holotype: nine individuals have an all-white venter and two have a white belly and cream-colored gular region and head; all males have the hands, feet, and half of the posteroventral portion of the thighs colored red, while seven individuals had this spot covering the entire posteroventral portion of the cloacal region and four had it covering only half of this region. There is a variation in the light yellow or light green reticulation network on a background of different shades of brown on the dorsum; each individual has its own unique pattern, as if it were an individual fingerprint ( Figure 6A-D). In preservative, red spots on the hands, feet, posteroventral region of the thighs, and posteroventral portion of the cloacal region turn reddish brown in males and cream in females. Some males (n = 3) have an eye diameter greater than the distance from the eye to the nostril, but in general, it is the same or slightly smaller (ED/EYNO = 0.87-0.96). Females are considerably larger than males. Nuptial pads and vocal slits are present in males. Females have a wrinkled gular region, whereas this region is smooth and white or cream-colored in males; and females have a venter without spicules and with brown spots, unlike males, who possess a venter with spicules and is rarely spotted.

Distribution and Natural History
The distribution of Atelopus manauensis sp. nov. is limited to the interfluve between the Negro and the Uatumã Rivers in the Alter do Chão formation (absent from the Guiana Shield and the Trombetas Group), within a narrow altitudinal range (61-125 m a.s.l.), and restricted to the micro-interfluve between the east bank of the Cuieiras River and the west bank of the Urubu River (Figure 10), which are tributaries of the Negro and the Amazon Rivers, respectively [29]. In this region, Atelopus manauensis sp. nov. inhabits the banks of small streams in dense ombrophilous forests of lowlands not flooded by large rivers [22]. Most of the individuals can be found within a four-meter strip along the banks of these streams, but some can be seen up to 10 m from the bank. Reproductive activity likely occurs from the beginning of the rainy season in the region (November) until March. On cloudy days, males are less exposed, but when the sun's rays start to warm the forest floor, they look for sun spots and then start vocalizing at any time of the day. They vocalize and forage on roots, accumulated leaves, fallen tree trunks on the bank, or over streams, at the base of herbaceous plants or stemless palms and even climbing inclined trunks of small shrubs, Figure 10. Distribution map for species of Atelopus used in the calculations of genetic distances and phylogenetic inferences, with colored circles representing species distributed to the west of Amazonia, on eastern and western versants of the Andes and Andean highlands; the squares representing species distributed in the Guiana Shield, and the green diamond a candidate species distributed on the south bank of the Amazon River, municipality of Anapu, state of Pará, Brazil [30]. The shaded polygon represents the portion of the Amazonian plain where there has been no record yet of a species of Atelopus, indicating a distribution gap for the genus (adapted from Lötters et al. [32]). Insert in the lower right shows the limits of the geographical distribution of Atelopus manauensis sp. nov. as represented by the yellow polygon between the Urubu and the Cuieiras Rivers; black stars refer to the type of locality where the holotype, allotype, and paratypes (RFAD) and individuals used for the genetics analyses (FEUFAM and LBA/ZF2) of Atelopus manauensis sp. nov. were collected; numbered dark green squares represent populations of Atelopus hoogmoedi of REBIO Uatumã (1) and of Pitinga River (2), on the east bank of the Uatumã River, the geographically closest Atelopus populations to the new species. Two specimens of A. hoogmoedi from REBIO Uatumã were used for genetic analyses.
Reproductive activity likely occurs from the beginning of the rainy season in the region (November) until March. On cloudy days, males are less exposed, but when the sun's rays start to warm the forest floor, they look for sun spots and then start vocalizing at any time of the day. They vocalize and forage on roots, accumulated leaves, fallen tree trunks on the bank, or over streams, at the base of herbaceous plants or stemless palms and even climbing inclined trunks of small shrubs, usually during the day (07:00-10:00 and 15:00-17:00) with milder temperatures. During the night, it is common to see individuals resting more than 1 m high on leaves of herbaceous plants on the banks of streams. The variety of calls emitted by Atelopus species is related to intra-and interspecific communication, but the functions of these calls have been poorly studied [25]. Type 1 calls (advertisement call) are used, as with other anurans, in the context of attracting reproductive mates, whereas type 2 calls (pure tone call) are likely used in an aggressive context such as in intraspecific interaction. Type 3 calls (pure tone short calls) are mostly of unknown function but may have a function between aggressive and release calls in the context of close-range interactions between conspecifics males with physical contact. Type 4 calls (pulsed short calls) clearly represent release calls [25]. Information on eggs and tadpoles of Atelopus manauensis sp. nov. can be found in Gascon [23]. There is no information on the eggs or tadpoles of closely related species. Lötters et al. [4] recorded a female and a male of A. pulcher in auxiliary amplexus, and after 2-3 weeks, c.a. 600 unpigmented eggs were deposited in the water arranged in a single chain or string-like fashion [4].

Discussion
We used integrative taxonomy to describe a new species of Atelopus with a distribution mostly restricted to the municipality of Manaus, Brazil, including two of the most intensively studied forests in Amazonia (RFAD and PDBFF). Atelopus manauensis is the first species of the genus to be formally described for Brazil. Another Atelopus species known from Brazil is A. hoogmoedi, which is distributed in the states of Amapá and Pará, but its type locality is in French Guiana. As is the case for several of its congeners, A. manauensis has a small area of occurrence (approximately 4500 km 2 ) and is critically threatened by urban expansion of the largest city in Brazilian Amazonia-Manaus [29].
Atelopus manauensis has ecological, morphological, and genetic similarities with species of Atelopus distributed c.a. 1500 km straight-line distance from Manaus, so the threats for extinctions identified for those species are likely to be relevant to A. manauensis as well. For example, the pathogenic fungus Batrachochytrium dendrobatidis (BD) has caused the extinction of numerous species of Atelopus [58], and the deforestation that is reaching Central Amazonia via highway BR-174 (Manaus-Boa Vista) could very likely facilitate its invasion into the area of occurrence of the new species. Although BD has not been found in populations of A. manauensis [29], an ongoing study indicates that BD infection would cause high mortality rates for the new species [59]. In addition to the threat of BD, the urban expansion of Manaus and other nearby cities is fragmenting areas with aggregations of individuals of A. manauensis. The conservation status of A. manauensis was recently assessed as "Endangered" according to the criteria established by the International Union for Conservation of Nature, which are as follows: reduction of area of occurrence, area of occupation, and habitat quality; projected decline in the area of occurrence, area of occupation, and habitat quality; and area of occurrence less than 5000 km 2 (approximately 4500 km 2 ; [29]).
The interspecific phylogenetic relationships reconstructed in this study are in agreement with those of previous studies with regard to the monophyly of the clade distributed in the Guiana Shield [3, 11,29]. Our phylogenetic analysis and that of Jorge et al. [29] were the first to show that Atelopus manauensis is sister to the clade containing the species of Atelopus distributed in the Guiana Shield. The genetic distances inferred between these species and A. manauensis vary between 2.3-2.8% (K2P) and 2.3-2.7% (p-distance) and are close to the average observed for anurans using 16S markers [60,61]. On the other hand, the genetic distances among species distributed in the Guiana Shield (A. franciscus, A. flavescens, A. barbotini, and A. hoogmoedi) were very low, ranging between 0.2-1.5% (p-distance) and 0.2-1.3% (K2P). This finding, together with the lack of bioacoustic divergences among the species of the Guiana Shield as reported by Costa-Campos and Carvalho [50], corroborate the suggestion of Kok [5] and Lötters et al. [3] that these species actually represent a single polymorphic species and that they should be synonymized.
In the past, Atelopus manauensis has also been mistaken for A. pulcher and A. spumarius. However, A. spumarius SS from Peru is the sister of A. seminiferus, which together form a clade positioned as the sister of the clade containing the species of the Guiana Shield and A. manauensis. Atelopus pulcher was finally allocated as a basal species to the aforementioned clade. The genetic distances between the new species and A. pulcher and A. spumarius SS were moderately high at 3.9% and 4.0% (p-distance) and 4.0% and 3.9% (K2P), respectively. In addition, species distributed in western Amazonia (i.e., A. pulcher and A. spumarius SS) and in the Guiana Shield (i.e., A. hoogmoedi) are geographically and ecologically isolated from each other and from A. manauensis due to historical biogeographical barriers, such as large rivers and mountains, and ecological barriers, such extensive areas of open, floodable forests that contain acidic water drainage systems to which the tadpole of A. manauensis is probably sensitive [22], preventing any possible contact of the new species with the genetically closest species of the A. hoogmoedi complex or with the morphologically most similar species of the A. spumarius complex.