Species limits within the widespread Amazonian treefrog Dendropsophus parviceps with descriptions of two new species (Anura, Hylidae)

Abstract The genus Dendropsophus is one of the most speciose among Neotropical anurans and its number of described species is increasing. Herein, molecular, morphological, and bioacoustic evidence are combined to assess species limits within D. parviceps, a widely distributed species in the Amazon Basin. Phylogenetic relationships were assessed using 3040 bp sequences of mitochondrial DNA, genes 12S, ND1, and CO1. The phylogeny shows three well-supported clades. Bioacoustic and morphological divergence is congruent with those clades demonstrating that Dendropsophus parviceps is a species complex. Dendropsophus parviceps sensu stricto occurs in the Amazon basin of Ecuador, northern Peru, southern Colombia and northwestern Brazil. It is sister to two previously undescribed species, D. kubricki sp. n. from central Peru and D. kamagarini sp. n. from southern Peru, northeastern Bolivia, and northwestern Brazil. Genetic distances (uncorrected p, gene 12S) between D. parviceps and the new species is 3 to 4%. Dendropsophus kamagarini sp. n. can be distinguished from D. parviceps by having a prominent conical tubercle on the distal edge of the upper eyelid (tubercle absent in D. parviceps). Dendropsophus kubricki sp. n. differs from D. parviceps by having scattered low tubercles on the upper eyelids (smooth in D. parviceps). Dendropsophus parviceps and both new species differ from all their congeners by their small size (adult maximum SVL = 28.39 mm in females, 22.73 mm in males) and by having a bright orange blotch on the hidden areas of the shanks and under arms. The advertisement call of the two new species has lower dominant frequency relative to D. parviceps. Probable speciation modes are discussed. Available evidence indicates that ecological speciation along an elevation gradient is unlikely in this species complex.


Introduction
The upper Amazon Basin harbors the highest diversity of amphibian species in the world (Bass et al. 2010;Duellman 1999). In the last decade, the use of genetic characters in amphibian taxonomy has helped to discover a large number of cryptic species through the upper and lower Amazon Basin (e.g., Almendáriz et al. 2014;Brown et al. 2008;Brown and Twomey 2009;Caminer and Ron 2014;Elmer and Cannatella 2008;Fouquet et al. 2015;Moravec et al. 2014;Páez-Vacas et al. 2010;Rivera-Correa and Orrico 2013;Rojas et al. 2015;Rojas et al. 2016;Twomey and Brown 2008). The use of genetic characters in combination with morphological and bioacoustic evidence allows unambiguous delimitation of species under the evolutionary species concept (de Queiroz 1998;de Queiroz 2007;Padial et al. 2009).
Dendropsophus Fitzinger 1843 is the most speciose genus of hylid frogs in the Neotropics. Currently it has 102 formally described species (Frost 2017). The few systematics studies of Dendropsophus that have included genetic evidence have resulted in the discovery of a large number of undescribed species (e.g., Fouquet et al. 2015;Gehara et al. 2014;Motta et al. 2012;Rivera-Correa and Orrico 2013). These studies underscore the need of genetics-based taxonomic reviews in the genus Dendropsophus.

Bioacoustic analyses
Recordings were made with two digital recorders Olympus LS-10 and Marantz professional PMD620MKII Handheld Solid State Recorder attached to a directional microphone Sennheiser K6-ME67. We also included published recordings from Peru, Tambopata (Cocroft et al. 2001), and Bolivia, Cobija (Márquez et al. 2002). Recordings are deposited at the on QCAZ collection.
Calls were analyzed using software Raven 1.3 (Charif et al. 2004) at a sampling rate of 44100 Hz and a resolution of 16 bits. Spectral parameters were obtained using a Fast Fourier Transformation (FFT) of 4096 points, a frequency resolution of 10.8 Hz, window type Hann and filter bandwidth of 52.2 Hz.
Terminology for call parameters follows Köhler et al. (2017) and Toledo et al. (2015). We measured the following variables: (1) call duration: time from the beginning to the end of the call; (2) note duration: time from beginning to end of the note; (3) rise time: time from the beginning of the note to the point of maximum amplitude; (4) number of pulses: number of pulses in the note; (5) pulse rate: number of pulses per note duration; (6) interval between notes: time from the end of one note to the beginning of the next; (7) dominant frequency: frequency with the most energy, measured along the entire call; (8) initial frequency: frequency at the beginning of the note; and (9) final frequency: frequency at the end of the note. If available, several calls or notes were analyzed per individual to calculate an individual average.
The calls of members of the Dendropsophus parviceps group (sensu Fouquet et al. 2015) consist of one high-pitched pulsed trill followed or not by a series of clicks: D. counani (Fouquet et al. 2015), D. bokermanni (Duellman and Crump 1974;Fouquet et al. 2015;Read and Ron 2012), D. brevifrons (Fouquet et al. 2015;Read and Ron 2011), and D. koechlini (Duellman and Trueb 1989;Fouquet et al. 2015). Therefore, we used a note-centered approach to define what is considered a call and a note (sensu Köhler et al. 2017).
A Principal Components Analysis (PCA) was conducted to evaluate call differentiation between species. We also performed Student's t-test to assess differences between species in the acoustic variables. For the PCA, only components with eigenvalues > 1 were retained. All statistical analyses were performed using JMP® 9.0.1 (SAS Institute 2010). Some recordings did not have temperature registered but temperature variation in equatorial rainforests at night is low (Duellman 1978) and therefore unlikely to severely influence the analyses.

DNA extraction, amplification, and sequencing
Total DNA was extracted from muscle and liver preserved in 95% ethanol or tissue storage buffer using guanidine-thiocyanate extraction protocol of M. Fujita (unpublished). Polymerase chain reaction (PCR) was used to amplify the mitochondrial genes 12S rRNA (12S), Cytochrome Oxidase 1 (CO1), and a continuous fragment of 16S (partial sequence), tRNA Leu , NADH dehydrogenase subunit 1 (ND1), tRNA Ile , and tRNA Gln . PCRs were performed in 25 μl reactions using 2.5 μl of PCR buffer, 1.5 μl MgCl 2 , 0.5 μl of each primer, 0.5 μl of each dNTP, 0.25 μl of Taq polymerase, 1 U of DNA, and 18.25 μl dH 2 O. Primers are listed in Table 1. PCR amplification was carried under standard protocols. PCR products were visualized in 1% agarose gel, and primers residues and dNTPs were removed from PCR products using ExoSAP-It purification. Amplified products were sequenced by the Macrogen Sequencing Team (Macrogen Inc., Seoul, Korea).
New sequences were obtained from 61 specimens from the upper Amazon Basin of Ecuador and Peru. A sequence of Dendropsophus parviceps available in GenBank published by Faivovich et al. (2005) from Brazil (Acre) was also downloaded. Sequences of three closely related species, D. brevifrons, D. frosti, and D. koechlini were also included. Dendropsophus marmoratus and Xenohyla truncata were used as outgroups. Sequences of D. brevifrons, D. frosti, and X. truncata were published by Fouquet et al. (2015), Motta et al. (2012), andFaivovich et al. (2005), respectively.
Sequences were assembled and aligned in Geneiuos Pro v5.4.6 (Kearse et al. 2012) using the MAFFT plugin under the L-INS-i algorithm (Katoh et al. 2002). Manual adjustments to the alignment were made using Mesquite v3.04 (Maddison and Maddison 2015). ND1 and CO1 gene sequences were translated into amino acids in Mesquite to confirm the alignment and verify the absence of stop codons.

Phylogeny
Phylogenetic relationships were inferred using Maximum likelihood (ML) with software GARLI v2.0 (Zwickl 2006) and Bayesian inference with MrBayes v3.1.2 (Ronquist et al. 2012). The best partition strategy and the best-fit substitution model of DNA evolution for each partition were selected using PartitionFinder v1.1.0 (Lanfear et al. 2012) according to the Bayesian Information Criterion (BIC). We defined nine a priori partitions: 12S, 16S, tRNAs, and one partition for each codon position of ND1 and CO1.
Maximum likelihood analyses were performed with ten replicates starting from stepwise addition trees (streefname = stepwise). Other GARLI settings were set to default values (Zwickl 2006). Bootstrap support was evaluated through 500 replicates. The 50% majority rule consensus for the bootstrap trees was obtained with Mesquite v3.04 (Maddison and Maddison 2015). Bayesian analyses were performed with two  (1994) searches of 35 × 10 7 generations each with four Markov chains and trees sampled every 5000 generations; stationarity and convergence were assessed in Tracer v1.6 (Rambaut et al. 2014) examining the standard deviation of split frequencies and plotting the -lnL per generation. Trees generated before stationarity were discarded as burn-in. Additionally, pairwise genetic distances (uncorrected p) were calculated for 12S using MEGA 6.0 (Tamura et al. 2013).

Phylogenetic relationships
The total alignment of concatenated DNA sequences had 3040 base pairs from mitochondrial markers 12S rRNA (~895 bp), small fragment of 16S rRNA (~282 bp), portions of tRNA (~215 bp), ND1 (~961 bp) and CO1 (~687 bp) from 70 individuals. Genes sequenced and GenBank accession numbers are listed in Appendix 2. The best partition strategy and the best-fit model for each partition are shown in Table 2. The phylogenetic relationships strongly support Dendropsophus parviceps as monophyletic (posterior probability, pp = 1 and bootstrap = 99) (Fig. 1). There are three clades within D. parviceps, each strongly supported. One clade is distributed in southern Peru (e.g., Madre de Dios and Cusco regions) and northwest Brazil (Acre); we refer to this clade as the "Southern Clade" hereafter. The second clade is distributed in northern and central Peru (e.g., Sierra del Divisor, Río Tapiche, and Chambira) ("Central Clade" hereafter). The third clade is distributed in eastern Ecuador (called "Northern Clade" hereafter). Maximum pairwise uncorrected genetic distance for 12S between the Central Clade and the Southern Clade is 2.8%, between the Northern Clade and the Central Clade is 3.2% and between the Northern Clade and the Southern Clade is 3.7%.
Mean p genetic distance within the Central Clade is 1.3% (range 0-1.3%) while within the Southern Clade is 0.07% (range 0-0.15%). The Northern Clade is divided Table 2. Partition strategy and the best-fit model of substitution for each partition block used in phylogenetic analyses.

Morphological comparisons
Morphometric variables from adults are summarized in Two components with eigenvalues > 1.0 were extracted from the PCA. Both PCs account for 52.1% of the total variation for males (Table 4). Principal Component I has high positive loadings for femur length and tibia length and PC II for head width and internarial distance (Table 4). The morphometric space shows high overlap between clades (Fig. 3).
Two PCs with eigenvalues > 1.0 explain the 58% of total variation among females (Table 4). The highest loadings for PC I were head width, femur length, and tibia length; PC II has high loadings for eye to nostril distance and is negatively correlated with head length and foot length (Table 4). As in the PCA for males, there is high overlap between clades in morphometric space (Fig. 3).

Bioacoustic comparisons
The call of the Dendropsophus parviceps species complex consists of one pulsed trill (Fig. 4A, C, E). The pulsed trill is facultatively followed by one or more click notes (Fig. 4B, D, F).
The pulsed trill appears to function as advertisement call because males produce these calls repeatedly and antiphonally. Acoustic parameters for the advertisement calls and click notes are shown in Table 5    The dominant frequency of the advertisement call of the Northern Clade is higher (range 5081.8−6869.1 Hz) than that of the Southern Clade (range 3164.1−4306.6 Hz) and Central Clade (range 3542.2−4394.5 Hz). There are significant differences in dominant frequency for advertisement calls between the Northern Clade and the  Southern Clade (Student's t test, t = 13.68, df = 17, p < 0.001), and between the Northern Clade and the Central Clade (Student's t test, t = 9.94, df = 13, p < 0.001). The number of pulses of the advertisement calls of Southern Clade is larger (12-32) than that of the Northern Clade (8-25; differences are significant: Student's t test, t = -2.48, df = 17, p = 0.02).
The PCA for advertisement calls shows that the Northern Clade is acoustically distinct from the Southern and Central clades (Fig. 5). Two components with eigenvalues > 1.0 account for 87.7% of the acoustic variation (Table 8). The highest loadings for PC I were dominant frequency, initial frequency, final frequency, and number of pulses; the highest loadings for PC II were note duration, rise time, and pulse rate. The northern clade differs from the southern and central clades mostly along PC I, which mainly represents variation in call frequency (Fig. 5).

Species limits
The integrative analyses presented in this work show congruent differences in genetic, morphological, and bioacoustic characters that demonstrate the existence of three confirmed candidate species within "Dendropsophus parviceps": Northern, Central, and Southern clades. Because the type locality of Dendropsophus parviceps is in Amazonian Ecuador (Sarayaku), we consider that the Northern Clade is Dendropsophus parviceps sensu stricto. This assignment is confirmed by the lack of tubercles in the eyelid of the holotype, a character state unique to the Northern Clade. Therefore, the two Peruvian species are new and we describe them in the following section.
Color in life. Based on digital photographs (Fig. 6): dorsum varies from brown, tan, grayish tan to reddish brown, some individuals have few scattered dorsolateral dark brown flecks; dorsal markings are dark brown; flanks are white or creamy yellow with black or dark brown diagonal bars; dorsal surfaces of forelimbs and shanks have dark brown transversal bars; anterodorsally, thighs are black or dark brown with two or three white spots. The single suborbital bar is white. The venter is translucent gray mottled with black or dark brown; in some females venter is black; chest is white; in adult males, throat is olive tan mottled with dark brown flecks anteriorly and translucent gray posteriorly; in adult females, throat is grayish tan or olive brown, dark brown, or black anteriorly with a white blotch with stripes posteriorly; the ventral surfaces of the limbs are translucent gray or translucent white, thighs are mottled with dark brown posteriorly; there is one bright orange or amber blotch in ventral surface of shank next to the knee, and in the posterior arm, from the axillae to near the elbow. The iris is creamy white to reddish brown with brown or dark brown reticulations. Calls (Fig. 4A−B). Descriptive statistics of acoustic variables are provided in Table 5. Calls from ten individuals were analyzed. Three individuals (two of them unvouchered specimens and QCAZ 52753) were recorded at the type locality, Sarayaku, Pastaza Province, at night, on 6 April 2012 (QCAZ 52753 was recorded at 01:00h, temperature 22.4°C). Three individuals (QCAZ 52820, 52837 and one individual not collected) were recorded at Canelos, Pastaza Province, on 11 April 2012 (QCAZ 52820 recorded at 01:00h, 23.4°C). Two individuals (QCAZ 52017, 52918)  The advertisement call is a pulsed note (Fig. 4A−B). The amplitude increases gradually at the beginning and falls sharply towards the end. The advertisement call may be emitted alone or followed by one or more click notes. However, the click notes occasionally are emitted alone. The click notes may be non-pulsed or pulsed.
Call comparisons between populations. The advertisement calls from Río Verde separate along PC II from the calls of other populations ( Fig. 5; Table 8). Mean dominant frequency is 5364.7 Hz (SD = 167) at Río Verde and 6498.1 Hz (SD = 239) at the other populations. Mean pulse rate is 86.1 pulses/s (SD = 11.79) at Río Verde and 127.6 pulses/s (SD = 15.5) at others. Mean rise time is 0.1 (SD = 0.02) and mean advertisement call duration is 0.19 s (SD = 0.04) at Río Verde, while mean rise time is 0.06 (SD = 0.01) and mean advertisement call duration is 0.12 s (SD = 0.03) at the other populations.
Dendropsophus parviceps inhabits Amazonian lower montane forest, Amazonian foothill forest, and Amazonian evergreen lowland rainforest (habitat types based on . Dendropsophus parviceps is an opportunistic breeder and can be found in primary and secondary forest, temporary ponds, flooded areas, swamps, and artificial open areas. Calling activity starts at dusk (17-18h), but it is mainly nocturnal. According to Lynch (2005), D. parviceps is a canopy species that visits the lower forest strata for breeding.
Conservation status. Its extent of occurrence is 256,944 km 2 . There is habitat degradation and fragmentation within its distribution as result of human activities, especially cattle rising, agriculture, and oil exploitation. Its presence in artificial open areas suggests that is tolerant of at least some level of habitat modification (Azevedo-Ramos et al. 2004). Its distribution range is large and includes extensive undisturbed areas (Ministerio de Ambiente Ecuador 2013). Therefore, we propose that D. parviceps should be assigned to the Least Concern category, following IUCN (2001) criteria.
Remarks. The advertisement call from Río Verde differs from other population calls ( Fig. 5; Table 5). However, low genetic and morphological differences between Río Verde and the other populations indicate that they are conspecific. The Brazilian record from Tarauacá, Río Uaupés (Amazonas State) is based on Melin (1941) who reported a juvenile specimen with SVL = 21 mm. This specimen could be an adult male because the throat is mottled with brown, characteristic of all adult males of Dendropsophus parviceps. Nevertheless, the SVL of the male from Taracuá falls above the range of variation of males of D. parviceps (14.3−18.7 mm) and it has a thoracic fold (fold absent in D. parviceps; see above in Diagnosis section). Therefore, the record from Tarauacá requires verification.
The holotype has SVL = 26.5 mm (adult female; Fig. 7A). This value is above the range of variation of females of D. parviceps reported in Table 3 (20.3−24.4 mm).
To confirm that the holotype falls within the range of variation of D. parviceps from Ecuador, we measured the SVL of the largest adult females from the QCAZ collection. We found three specimens with size close to the holotype: QCAZ 4340 (SVL = 26.13 mm) from La Selva (Sucumbíos Province), QCAZ 27028 (SVL = 26.03 mm) from Ahuano (Napo Province), and QCAZ 59772 (SVL 26.26 mm) from Comunidad Zarentza (Pastaza Province; Appendix 1). Although the holotype is the largest specimen known for D. parviceps, other specimens are smaller by just ~1% of SVL. Other characteristics of the external morphology of the holotype fall within the known variation of the Ecuadorian populations confirming that they are conspecific (Figs 7A, 8). Etymology. The specific name kamagarini is a noun derived from the Matsigenka language, which means demon or devil (Snell et al. 2011). The Matsigenka language is spoken by the Matsigenka people who inhabit the highlands and lowlands of southeastern Peru, in the departments of Cusco and Madre de Dios. Judeo-Christian religions depict the demon as a human figure with horns. The species name is in allusion to the prominent horn-like tubercles on the upper eyelid of D. kamagarini.

Dendropsophus kamagarini
Diagnosis. Throughout the species description, coloration refers to preserved specimens unless otherwise noted. The new species is assigned to the genus Dendropsophus based on our phylogenetic results (Fig. 1) and the overall similarity with D. parviceps and other species of the genus (Figs 10-11). Dendropsophus kamagarini is a mediumsized species, relative to other species in the D. parviceps group and is characterized by the following combination of traits: (1) size sexually dimorphic; mean SVL 19.9 mm in males (range 17.6-22.7; n = 35), 26.1 mm in females (range 24.0-28.1; n = 7); (2) throat brown mottled with white flecks posteriorly in males vs. white blotch with flecks or with stripes posteriorly in females (Fig. 11); (3) snout is short and truncate in dorsal and lateral views; (4) nostrils slightly protuberant; (5) tympanum visible, tympanic membrane non-differentiated, annulus distinct; (6) one prominent conical tubercle on the distal edge of the upper eyelid; (7) thoracic fold indistinct to barely evident; (8) ulnar tubercles and outer tarsal tubercles distinct; (9) axillary membrane present; (10) skin on dorsal surfaces smooth with scattered tubercles; skin on chest, belly, posterior surfaces of thighs, and subcloacal area coarsely areolate; skin on throat grooved with scattered tubercles; (11) dark brown markings on dorsum (Fig. 11); (12) thenar tubercle distinct; (13) hand webbing formula II1 --2 + III1 --1 -IV, feet webbing formula I1 1/2 -2 + II1 --1III1 --2 -IV2-1V; (14) in life, dorsum tan, brown or reddish brown; (15) orange to amber blotch on the proximal ventral surface of shanks and under arms, from the axillae to near the elbow, in life (white to creamy white in preserved); (16) one suborbital white bar present both in life and preserved; (17) thighs black to dark brown with two or three spots on the anterodorsal surfaces both in life and preserved; (18) iris in life creamy white with brown to reddish brown reticulations and a cream ring around pupil.
Comparisons with other species. Dendropsophus kamagarini is most similar to D. parviceps and D. kubricki sp. n. It can be distinguished from D. parviceps by having a prominent conical tubercle on the distal edge of the upper eyelid (tubercle absent in D. parviceps; Fig. 12) and a blunt and short snout in lateral view (slightly inclined posteroventrally in profile in D. parviceps; Fig. 12). Dendropsophus kamagarini is larger than D. parviceps ( Fig. 2; [Goin 1960;Duellman and Crump 1974], in D. brevifrons [Duellman and Crump 1974], in D. counani [Fouquet et al. 2015], in D. frosti [Motta et al. 2012] and in D. koechlini [Duellman and Trueb 1989]). Dendropsophus kamagarini also resembles D. pauiniensis, but it can be distinguished by the presence of an orange or amber blotch on the proximal ventral surface of shanks and a prominent conical tubercle on the distal edge of the upper eyelid (blotch and tubercle are absent in D. pauiniensis ;Heyer 1977).
Description of holotype. Adult male (Fig. 7B), SVL 19.6 mm. Head as wide as body, wider than long, widest below eyes; snout truncate and short in dorsal view, slightly inclined posteroventrally in lateral view; loreal region flat; lips thin; internarial region slightly concave; nostrils slightly protuberant dorsally and laterally; interorbital area flat; tympanum rounded distinct, tympanic annulus evident, tympanic membrane non-differentiated, supratympanic fold thin, restricted to upper edge of tympanum. Arms slender, not hypertrophied; axillary membrane extending to one third of upper arm; ulnar fold distinct, low ulnar tubercles present; fingers short, bearing small, round discs; relative length of fingers I < II < IV < III; subarticular tubercles small, round on fingers I and II, bifid on finger III, and divided on finger IV; supernumerary tubercles small, slightly evident; thenar tubercle distinct; palmar tubercle flat, round; webbing basal between fingers I and II; webbing formula of fingers II1 --2III1 1/2 -1 -IV. Hindlimbs long, slender; tarsal fold absent, outer tarsal tubercles present, low; calcar and heel tubercles absent; toes bearing round discs, smaller than those of fingers; relative length of toes I < II < III < V < IV; subarticular tubercles small, round; supernumerary tubercles indistinct; inner metatarsal tubercle small, flat, elliptical; outer metatarsal tubercle absent; webbing formula of toes I1 1/2 -1 -II1 --1III1 --2IV2-1 -V. Skin on dorsum, dorsal surfaces of limbs, flanks, and groin smooth; skin on head smooth with scattered tubercles and one prominent conical tubercle on the distal edge of the upper eyelid; skin on venter, chest, posterior surfaces of thighs, and subcloacal area coarsely areolate; skin on throat and ventral surfaces of limbs smooth. Cloacal opening directed posteriorly at upper level of thighs; cloacal sheath short; cloacal folds and tubercles absent. Tongue cordiform, barely free posteriorly; dentigerous process of vomers small, prominent, narrowly separated, each bearing three and two vomerine teeth (left/right), positioned obliquely to choanae; choanae small, rounded; vocal slits long, extending from midlateral base of tongue to angle of jaws; vocal sac single, median, subgular.
Color of holotype in preservative. Figure 7B. Dorsal surfaces of head, body, and limbs brown, grayish tan dorsolaterally with dark brown markings on dorsum consisting of median blotch anteriorly, transverse bars posteriorly; dark brown broad transverse bars on the forelimbs and shanks; anterodorsal surfaces of thighs black with three white spots; white suborbital bar. Ventral surface of belly white anteriorly, creamy mottled posteriorly with dark brown scattered flecks; chest white, throat brown anteriorly and white with brown flecks posteriorly; ventral surfaces of limbs creamy.    Measurements of holotype (in mm). SVL 19.6, HW 6.3, HL 5.9, END 2.1, IN 2.0, FL 10.4, TL 10.7, FL 8.6.
Color in life. Based on digital photographs (Fig. 10): dorsum varies from tan to brown or reddish brown; creamy tan or mustard brown dorsolaterally; dorsal markings are dark brown, some individuals have brown, creamy, or grayish tan stripes around markings; some individuals also have scattered dark brown flecks dorsolaterally; the flanks are white with black vertical bars; dorsal surfaces of forelimbs and shanks have dark brown transversal bars; the thighs are black with two or three spots on the anterodorsal surfaces. The single suborbital bar is white. The venter is white anteriorly and dark brown or black mottled with translucent gray posteriorly, with white scattered flecks; chest is white and mottled with brown anteriorly; throat is brown or dark brown anteriorly and spotted with white flecks posteriorly in males (posterior part of throat with white blotch with or without stripes in females); the ventral surfaces of the limbs are translucent gray, thighs are mottled with black or dark brown anteriorly and posteriorly the thighs are black with white flecks; the ventral surface of shanks, from the knee to one third or on half the length of the shank, and arms, from the axillae to near the elbow, have a bright amber or orange blotch. Vocal sac in males is olive tan. The iris is creamy white with brown to reddish brown reticulations or reddish brown with creamy white reticulations and a cream ring around pupil. Calls (Fig. 4C−D). Descriptive statistics of acoustic variables are provided in Table 6. We analyzed recordings from: (1)  The advertisement call is a pulsed note (Fig. 4C−D). The amplitude of the advertisement call gradually increases until three-quarters of the note duration to decrease abruptly until the end. The advertisement call may be followed or not by secondary click notes. Nonetheless, the click notes are occasionally vocalized alone. The click notes are pulsed except for the last note.
One recording from Cobija, Bolivia (Pando Department, Nicolás Suárez Province) by Márquez et al. (2002) falls within the range of variation of advertisement calls of Dendropsophus kamagarini from Peru (Table 6). In addition, the number of pulses (15) and the dominant frequency (4150 Hz) of the call described by Duellman (2005) fall within the range for calls of D. kamagarini (Table 6).
Distribution and ecology. Dendropsophus kamagarini occurs in the Amazon basin of southeastern Peru (Cusco and Madre de Dios regions; Fig. 9), northwestern Brazil (Acre and Rondônia states; Fig. 9), and northeastern Bolivia, from the Andean slopes to lowland tropical rainforest (Fig. 9). Localities with known elevation range from 150 m (Acre) to 1696 m (Ochigoteni) above sea level.
Bolivian records are partly based on De la Riva et al. (2000) report of "Dendropsophus parviceps" in central northeastern Bolivia, Departments of Beni, Cochabamba, La Paz, Pando, and Santa Cruz. One photograph from Puerto Almacén (Santa Cruz Department; pp. 102 in De la Riva et al. 2000) and two photographs from Tahuamanu and El Negro (both from Pando Department; Fig. 10) show the conical tubercle on the upper eyelid characteristic of D. kamagarini. Dendropsophus parviceps distribution range is at a distance of over 1500 km, which make very unlikely that Bolivian records are conspecific. Thus, we propose that all historic records of "Dendropsophus parviceps" from Bolivia are assigned to D. kamagarini.
The call from Cobija (Pando Department) falls within the range of advertisement call of D. kamagarini (Table 6). The localities of El Negro, Tahuamanu, and Cobija are at a distance of 89 km, 158 km, and 203 km, respectively, to the type locality of D. kamagarini (Inotawa). In addition, specimens from Museo de Historia Natural Alcide d'Orbigny, Cochabamba, Bolivia, also have a prominent conical tubercle on the distal edge of the upper eyelid. These specimens are from Valle del Sacta and the confluence of the Altamachi and Ipiri rivers (both from Cochabamba Department; see Appendix 1). There is also one record from Santa Elena (16.6791°S, 66.6791°W, 600 m, Cochabamba Department, Ayopaya Province; Fig. 9) based on a locality record from Museo de Historia Natural Alcide d'Orbigny, Cochabamba, Bolivia. Additionally, the records from Acre, Brazil, of D. kamagarini are also supported by Cochran and Goin (1970) who examined one specimen (WCAB 2511) and report the presence of the conical tubercle on the upper eyelid.
Dendropsophus kamagarini congregates for breeding at temporary and permanent ponds in flooded forest and Terra Firme forest; it is an opportunistic breeder (Duellman 2005). Adults of both sexes were found at night perching on leaves of bushes and trees, on branches and on palm fronds. Males were calling perched from 2−3 m above the water.
Conservation status. Extent of occurrence (B1) is 637,800 km 2 . Dendropsophus kamagarini occurs in the following protected areas from Peru: Otishi National Park, Megantoni National Sanctuary, Amarakaeri Communal Reserve, Manu National Park, Tambopata National Reserve and Bahuaja-Sonene National Park, and protected areas from Bolivia: Manuripi-Heath Amazonian Wildlife National Reserve and Isiboro Sécure National Park and Indigenous Territory. Because its distribution range is large and occurs in several protected areas we suggest that D. kamagarini is assigned to the Least Concern category, following IUCN (2001) criteria.
Remarks. The specimens from Cochabamba Department (Appendix 1) are assigned as referred specimens because we lack genetic data. Márquez et al. (1993) report a maximum SVL = 24.6 for males from Puerto Almacén. This value is slightly above the maximum SVL of males of D. kamagarini (see Table 3). Márquez et al. (1993) also report the dominant frequency range of the advertisement calls (2476-3144 Hz), which is lower than the dominant frequency range of D. kamagarini (3164.1−4306.6 Hz). Therefore, further data is needed to determine the status of that population. We tentatively assign those specimens to D. kamagarini as referred material. Schlüter (1979)  Etymology. The specific name kubricki is a noun in the genitive case and is a patronym for Stanley Kubrick, an American filmmaker who is one of the most brilliant and influential film directors of all time. We dedicate this species to him for his legacy to film culture and science fiction.
Diagnosis. Throughout the species description, coloration refers to preserved specimens unless otherwise noted. The new species is assigned to the genus Dendropsophus based on our phylogenetic results (Fig. 1) and the overall similarity with D. parviceps and other species of the genus (Figs 13-14). Dendropsophus kubricki is a medium-sized species, relative to other species in the D. parviceps group and is characterized by the following combination of traits: (1) size sexually dimorphic; mean SVL 19.4 mm in males (range 18.3-20.1; n = 14), 26.0 mm in females (range 22.0-28.4; n = 8); (2) throat with white flecks posteriorly in males and white blotch with stripes posteriorly in females (Fig. 14); (3) snout truncate in dorsal view, rounded and inclined posteroventrally in lateral view; (4) nostrils slightly prominent; (5) tympanum distinct, rounded, concealed posterodorsally, tympanic membrane non-differentiated and annulus evident; (6) low tubercles on upper eyelid can be distinct or ill-defined; (7) thoracic fold slightly evident or indistinct; (8) ulnar tubercles and outer tarsal tubercles low; (9) axillary membrane present; (10) skin on dorsal surfaces smooth with scattered tubercles mainly on head; skin on throat areolate, skin on chest, belly, posterior surfaces of thighs, and subcloacal area coarsely areolate; (11) dark brown markings on dorsum consisting of chevrons and transverse blotches in variable arrangements (Fig. 14); (12) thenar tubercle distinct; (13) hand webbing formula II1 -−2 + III1 -−1 -IV, foot webbing formula I1 -−2 -II1 -−2 -III1 --2IV2−1 -V; (14) in life, dorsal surfaces reddish brown, brown, or grayish tan; (15) orange to amber blotch on the proximal ventral surface of shanks and under arms, from the axillae to near elbow, in life (white to creamy white in preserved); (16) one suborbital white bar present both in life and preserved; (17) anterodorsal surfaces of thighs are black to dark brown with two or three white spots, both in life and preserved; (18) iris in life is reddish brown, brown or silver gray.
Comparisons with other species. Dendropsophus kubricki is most similar to D. kamagarini and D. parviceps. It is distinguished from D. parviceps by its larger size shanks and arms [orange blotches are absent in D. bokermanni (Goin 1960;Duellman and Crump 1974), in D. brevifrons Crump 1974), in D. counani (Fouquet et al. 2015), in D. frosti (Motta et al. 2012) and in D. koechlini (Duellman and Trueb 1989)]. Dendropsophus kubricki also resembles D. pauiniensis, but it differs by the presence of an orange blotch on the proximal ventral surface of shanks (absent in D. pauiniensis ;Heyer 1977).
Description of holotype. Adult male (Fig. 7C), SVL 19.0 mm. Head as wide as body, slightly wider than long, widest below eyes; snout truncate and short in dorsal view, moderately rounded and slightly inclined posteroventrally in lateral view; loreal region concave; lips thin; internarial region slightly concave; nostrils slightly protuberant dorsally and laterally; interorbital area flat; tympanum rounded and distinct, tympanic annulus evident, tympanic membrane non-differentiated, supratympanic fold thin, covering tympanum posterodorsally. Arms slender, not hypertrophied; axillary membrane extending along proximal one third of arm; ulnar fold distinct, low ulnar tubercles present; fingers short, bearing small, round discs; relative length of fingers I < II < IV < III; subarticular tubercles small, round on fingers I and II, bifid on finger III, and divided on finger IV; supernumerary tubercles small, slightly evident; thenar tubercle distinct; palmar tubercle flat, round; webbing basal between fingers I and II; hand webbing formula II1 -−2 -III1 --1IV. Hindlimbs long, slender; tarsal fold absent, low outer tarsal tubercles present; calcar and heel tubercles absent; toes moderately long, bearing round discs, smaller than those of fingers; relative length of toes I < II < III < V < IV; subarticular tubercles small, round; supernumerary tubercles indistinct; inner metatarsal tubercle small, flat, elliptical; outer metatarsal tubercle absent; foot webbing formula I1 --2II1 --2III1 --2IV2-1 -V. Skin on dorsum and head smooth with scattered tubercles, skin on dorsal surfaces of limbs, flanks and groins smooth; skin on venter, posterior surfaces of thighs and subcloacal area coarsely areolate; skin on chest and throat areolate; skin on other surfaces smooth. Cloacal opening directed posteriorly at upper level of thighs; cloacal sheath short; cloacal folds and tubercles absent. Tongue cordiform, barely free posteriorly; dentigerous process of vomers small, prominent, adjacent medially, each bearing three and five vomerine teeth (left/right), positioned obliquely to choanae; choanae small, rounded; vocal slits long, extending from midlateral base of tongue to angle of jaws; vocal sac single, median, subgular.
Color of holotype in preservative (Fig. 7C). Dorsal surfaces of head, body, and limbs brownish gray with scattered reddish brown flecks with melanophores and leucophores on dorsum, dark brown markings on dorsum consisting of a median blotch anteriorly and transverse bar posteriorly; dark brown broad transverse bars on the forelimbs and shanks; snout brown dorsally; scapular region gray; thighs black with three white spots on the anterodorsal surfaces; one small white suborbital bar. Ventral surface of belly dark brown mottled with white anteriorly and with translucent white posteriorly and laterally; chest white; throat dark brown anteriorly and white with dark brown flecks posteriorly; ventral surfaces of the forelimbs translucent white; thighs translucent white anteriorly and dark brown posteriorly.
Color in life. Based on digital photographs (Fig. 13): dorsum varies from reddish brown, brownish gray, brown to creamy tan suffused with gray and brown with scattered white flecks, with or without reddish brown or brown low tubercles; some individuals have few scattered dark brown flecks on the dorsum; flanks are white with black or dark brown vertical bars; dorsal markings are dark brown, some individuals are mustard brown dorsolaterally; dorsal surfaces of forelimbs and shanks have dark brown transversal bars; anterodorsal surfaces of the thighs are black to dark brown with two or three spots. The single suborbital bar is white. The venter is black to dark brown mottled with translucent gray, and with white flecks anteriorly; some individuals have scattered white flecks on venter; chest is white; throat is brown, reddish brown, or dark brown with white flecks posteriorly in males (posterior part of throat with white blotch with or without stripes in females), some individuals have a white medial spot adjacent to the chest; the ventral surfaces of the limbs are translucent gray, the thighs are mottled with black or dark brown anteriorly; posteriorly the thighs are black with white flecks; the proximal ventral surface of shanks and arms, from the axillae to near the elbow, have an orange or amber blotch. The iris is reddish brown, brown, or silver gray. Calls (Fig. 4E−F). Descriptive statistics of acoustic variables are provided in Table 7. We recorded three males (two not collected and CORBIDI 15778) at Campamento Wishuincho-Río Tapiche (Loreto Department, Requena Province, Peru) on 11 October 2014 at 12:53h and 12 October 2014 at 21:53h. We also recorded one male (not collected) at Cordillera Azul (7.8103°S, 75.9928°W, 725 m, San Martín Department, Picota Province, Peru).
The advertisement call is a pulsed note (Fig. 4E−F). The amplitude of the advertisement call increases gradually until three-quarters of the note duration to decrease abruptly until the end. This call may be followed by one or more click notes. These clicks sometimes are emitted alone. Moreover, the click notes are pulsed, except for the last one which is unpulsed.

Distribution and ecology.
Dendropsophus kubricki is distributed in the Amazon basin in northeastern and central Peru (Fig. 9), at elevations between 106 (Jenaro Herrera) and 725 m (Cordillera Azul). Dendropsophus kubricki was found in flooded forest. Specimens from Chambira were collected in a small pond in a Terra Firme forest. Males call at night while perching on leaves of bushes and trees. They were observed between 0.3 and 0.4 m above the water.
Conservation status. Extent of occurrence (B1) is 53,548 km 2 . Dendropsophus kubricki occurs in the following protected areas: Sierra del Divisor National Park, Cordillera Azul National Park, and Cordillera Escalera Regional Conservation Area. Because its distribution range is large and occurs in protected areas, we recommend that D. kubricki is assigned to the Least Concern category, following IUCN (2001) criteria.
Remarks. Specimens from Chambira (Picota Province) are closely related to Río Tapiche and Jenaro Herrera specimens (both localities from Requena Province) (Fig. 1); genetic distances between these populations are low (mean p-distances 1.3% in mitochondrial gene 12S). Therefore, we include them tentatively in Dendropsophus kubricki. However, individuals from Chambira and Tarapoto (Picota and San Martin provinces, respectively) show differences in coloration because their dorsal tubercles have the same color as the background, white flecks and reddish brown low tubercles on the dorsum and dorsal surfaces are absent both in life and preservative (present in Río Tapiche and Jenaro Herrera individuals; Figs 13−14). In addition, there is segregation in acoustic space between advertisement calls from Cordillera Azul (Picota Province) and the type locality ( Fig. 5; Table 7). Therefore, it is conceivable that they represent a separate species.

Discussion
Our genetic, morphologic, and bioacoustic data demonstrated that Dendropsophus parviceps, as previously defined, was a complex of three cryptic species. Duellman (2005) noted the morphological distinctiveness of populations from Peru and suggested that they may represent separate species. However, without genetic information the definition of species limits within "D. parviceps" was difficult because the morphology of the three species is highly conserved. The dark brown markings on dorsum, the suborbital bar, spots on the anterior surfaces of thighs, and orange or amber blotches on shanks and arms are shared between the three species. Differences between D. parviceps, D. kamagarini, and D. kubricki are limited to body size, skin texture, and advertisement calls. This pattern of highly conserved morphology has also been reported in other species complexes in Amazonian amphibians (e.g., Caminer and Ron 2014;Elmer et al. 2007;Fouquet et al. 2012;Funk et al. 2012;Kieswetter and Schneider 2013;Moravec et al. 2014;Rojas et al. 2016).
The pattern of variation in bioacoustic, and quantitative and qualitative morphological characters found in the D. parviceps species complex is not unusual among closely related species of Amazonian amphibians. Genetic divergence usually covaries with size, bioacoustic, and qualitative morphological characters (e.g., skin ornamentation and coloration) while it has low covariation with size-corrected morphometric variables (Caminer et al. 2017;dos Santos et al. 2015;Fouquet et al. 2012;Funk et al. 2012;Ortega-Andrade et al. 2015;Ron et al. 2016, but see Acevedo et al. 2016). This suggests that advertisement calls and qualitative morphological characters are among the first components of the phenotype to diverge during speciation. In contrast, body shape, as quantified in linear morphometric analyses, is highly conserved and of limited value to assess limits among closely related species.

Speciation modes
Several authors have discussed the role of niche evolution in the speciation of vertebrates in tropical mountains (e.g., Cadena et al. 2012;Kozak and Wiens 2007). Some studies have shown that sister species tend to segregate along environmental gradients suggesting ecological speciation and niche lability (Arteaga et al. 2016;Graham et al. 2004;. Other studies have shown that sister species tend to be allopatric but with similar environmental niches (Cadena et al. 2012;Ortega-Andrade et al. 2015). Those results imply vicariant speciation and niche conservativism. Clearly, both speciation mechanisms have contributed to the high diversity of the Andes and adjacent Amazon lowlands. Our results with the Dendropsophus parviceps complex suggest niche conservativism and vicariance speciation. We base this conjecture in the elevational distribution of the three species and in the intraspecfic genetic variation among populations of D. parviceps.
Dendropsophus parviceps and D. kamagarini show wide and overlapping elevation ranges: 151 m to 1600 m in D. parviceps and 150 m to 1696 m in D. kamagarini. Fewer localities are known for D. kubricki but its known range (106-725 m) overlaps with the ranges of the other two species. Because elevation is the most influential variable defining the environmental niche in tropical regions, overlapping elevation ranges suggest conserved environmental niches. The allopatric distribution of the three species also indicates that vicariant speciation with latitudinal replacement is more likely than ecological speciation with elevational replacement.
The lack of importance of elevation in promoting genetic differentiation is also suggested by interpopulation genetic differentiation in D. parviceps. We sampled 55 populations encompassing an elevation range of 186-1600 m. If disruptive selection across the elevation gradient were generating genetic isolation, we would expect to find parapatric clades segregating by elevation. Instead, two parapatric clades were found that segregate latitudinally, each occurring across a wide range of elevations. Both clades have a contact zone in central Amazonia, Ecuador (Fig. 1). A northern subclade occurs in Napo, Orellana, Sucumbíos provinces reaching marginally Pastaza Province; a southern subclade occurs in Pastaza, Tungurahua, and Amazonian Peru. Both clades have a narrow zone of contact near the limit between Napo and Pastaza provinces at an elevation of 900 m. Genetic distances between both clades are moderate (range for gene 12S is 0.4-1.2%), but there is strong support for each clade indicating structured genetic differentiation. Currently, there are not conspicuous geographic barriers between both clades leaving as an open question the processes that promoted genetic divergence within D. parviceps. Overall, the available evidence suggests that species of the D. parviceps complex speciated allopatrically instead of ecologically along an elevation gradient.