Insights from Integrative Systematics Reveal Cryptic Diversity in Pristimantis Frogs (Anura: Craugastoridae) from the Upper Amazon Basin

Pluralistic approaches to taxonomy facilitate a more complete appraisal of biodiversity, especially the diversification of cryptic species. Although species delimitation has traditionally been based primarily on morphological differences, the integration of new methods allows diverse lines of evidence to solve the problem. Robber frogs (Pristimantis) are exemplary, as many of the species show high morphological variation within populations, but few traits that are diagnostic of species. We used a combination of DNA sequences from three mitochondrial genes, morphometric data, and comparisons of ecological niche models (ENMs) to infer a phylogenetic hypothesis for the Pristimantis acuminatus complex. Molecular phylogenetic analyses revealed a close relationship between three new species—Pristimantis enigmaticus sp. nov., P. limoncochensis sp. nov. and P. omeviridis sp. nov.—originally confused with Pristimantis acuminatus. In combination with morphometric data and geographic distributions, several morphological characters such as degree of tympanum exposure, skin texture, ulnar/tarsal tubercles and sexual secondary characters (vocal slits and nuptial pads in males) were found to be useful for diagnosing species in the complex. Multivariate discriminant analyses provided a successful classification rate for 83–100% of specimens. Discriminant analysis of localities in environmental niche space showed a successful classification rate of 75–98%. Identity tests of ENMs rejected hypotheses of niche equivalency, although not strongly because the high values on niche overlap. Pristimantis acuminatus and P. enigmaticus sp. nov. are distributed along the lowlands of central–southern Ecuador and northern Peru, in contrast with P. limoncochensis sp. nov. and P. omeviridis sp. nov., which are found in northern Ecuador and southern Colombia, up to 1200 m in the upper Amazon Basin. The methods used herein provide an integrated framework for inventorying the greatly underestimated biodiversity in Amazonia.

Information on species diversity is particularly important for regions such as the upper Amazon Basin in the western lowlands of Colombia, Ecuador and Peru, which holds the richest, most diverse, and complex amphibian assemblages from any area on Earth [17][18][19][20][21][22][23]. Amazonia is also an extremely threatened ecosystem, with about 1.8 million ha/year of primary forest lost since 1980, resulting in fragmentation of continuous forest into islands [24,25]. Knowledge of the ecology, biogeography and diversification patterns of herpetological assemblages in this threatened core of mega-diversity remains poor [26,27], despite characterization of the Amazon Basin as one of the best studied regions in South America [22].
Frogs of the genus Pristimantis (Terrarana: Craugastoridae) comprise one of the most striking, richest and understudied groups in the Neotropics [11,28,29]. Similar to other Terrarana, Pristimantis have direct development (no tadpole phase), which is associated with terrestrial habitats [30]. With nearly 469 species (~7% of amphibians worldwide) distributed mainly in South America [31,32], this group is considered highly threatened, with at least 35% of the species included in the Red List [33]. Most of the threatened species show a marked endemism in the tropical Andes, and are considered sensitive to environmental disturbances and habitat loss [34][35][36]. Species delimitation is particularly difficult within young evolutionary radiations such as Pristimantis [11,[37][38][39][40]. For example, Pinto-Sánchez et al. [41] demonstrated that most of phenetic taxonomic species of Central American Pristimantis are not monophyletic The Canelos Robber Frog, Pristimantis acuminatus Shreve [42], is a widely distributed species in Amazonia [30]. Curiously, as noted in the original description ( [42] page 217), the holotype of P. acuminatus lacks a distinct tympanic annulus, whereas the annulus is distinct in the two paratypes. Since then, the acuminate snout shape and the absence of the tympanum have been used commonly as diagnostic characters to assign similar greenish Pristimantis to P. acuminatus [43][44][45]. However, it is now suspected that P. acuminatus is a complex of species [30].
To resolve this taxonomic problem, we reviewed the type specimens of P. acuminatus as well as a broad sample of specimens from field and museum collections. We herein describe three new species based on an integrative approach that incorporates phylogenetic, morphometric and ecological data.

Ethics statement
Voucher specimens and tissue samples were obtained following ethical and technical protocols [46]. Vouchers were euthanized with lidocaine hydrochloride 2%, fixed in 10% buffered formalin and then later transferred to 70% ethanol. Liver and thigh muscle were preserved in 95% Kansas (KU). Each locality from museums databases was carefully reviewed (lat-long coordinates) to correct imprecise geo-references in decimal degrees, based on the WGS 84 datum.

DNA amplification
DNA was extracted from most tissue samples using a single-step method with acid guanidinium thiocyanate [50] or by using a UltraClean 1 Tissue & Cells DNA Isolation Kit (MO-BIO Laboratories, Inc., Carlsbad, CA, USA), following the manufacturer's manual. Three mitochondrial genes-16S rRNA (16S), 12S rRNA (12S), and the Folmer Region or ''Barcode of Life" fragment of the Cytochrome Oxidase sub-unit I (COI; [51]) gene-were amplified (S3 Table). Polymerase chain reaction was carried out under locus-specific optimal annealing temperatures following protocols detailed by Pinto-Sánchez et al. [41]. PCR products were cleaned using the UltraClean PCR Clean-Up Kit (MO-BIO Laboratories, Inc., Carlsbad, CA, USA) or by Exo I/SAP digest, and sequenced in both directions by Macrogen Co. Ltd. (South Korea). Sequences were edited and aligned in GENEIOUS v5.4.7 (Biomatters, Auckland, New Zealand). Multiple sequence alignments were generated using MAFFT v7.017 [52] with default gap opening cost and other settings configured in GENEIOUS. Leading and trailing ends were trimmed manually to remove any missing data. To identify related sequences, a Nucleotide Blast search was carried out using the NCBI database (http://blast.ncbi.nlm.nih.gov/Blast.cgi).

Phylogenetic analyses
Because our combined data set comprised two ribosomal genes with secondary structure (12S and 16S) and one protein-coding mitochondrial gene (COI), application of a single nucleotide substitution model was unlikely to provide a particularly good fit to the data [53]. Partitions were defined a priori and Bayes factors were used to choose among alternative partitioning strategies. Three distinct partitioning strategies were evaluated: 1) one partition (three genes concatenated), 2) three partitions, by gene (12S, 16S, and COI), and 3) five partitions (12S, 16S, and COI further partitioned by codon position). Bayes factors were calculated using twice the difference in the marginal model likelihoods [2ln(B 10 )] as estimated from the harmonic mean of the likelihoods of the posterior sample of trees for the simplest model (M 0 ) against the more complex model (M 1 ) [53,54]. A Bayes factor greater than 10 was considered as very strong support for the more complex model [55]. For each partitioning scheme, the matrix included the same number of terminals and characters.
Phylogenetic analyses were conducted using Maximum Likelihood (ML) and Bayesian Methods (BM) on individual genes and on concatenated datasets. Prior to ML and BM analyses, we used JModeltest 2.3.1 [56] through the Phylemon 2.0 Server [57] to select the optimal model for each gene and codon position for COI (Table 1). Due to the small sample size in our matrix among genes and partitions (12S = 13 sequences; 16S = 28 sequences; COI = 27 sequences), we used a corrected Akaike Information Criterion (AICc) to select the best-fitting model [58].
ML was carried out in GARLI v2.0 (Genetic Algorithm for Rapid Likelihood Inference; [59]) through the CIPRES portal (http://www.molecularevolution.org/index). We ran 5 independent searches, whereas support for the nodes were calculated by a search using 100 bootstrap replicates. Tree searches were performed with stepwise-addition starting trees (streefname = stepwise), 5000000 generations as maximum for each run (stopgen = 5000000), saving every 100 generations (saveevery = 100), a run termination threshold of 20000 generations without topology improvement (genthreshfortopoterm = 20000), and a termination threshold value of 0.01 in the increase in lnL required for any new topology (significanttopochange = 0.01); other parameters were used with the default setting [59]. Mesquite [60] was used to generate a majority-rule consensus tree from the bootstrap replicates. The Bayesian phylogenetic analysess were implemented in MrBayes v3.2.2 [60] in the CIPRES portal [61] (http://www.molecularevolution.org). The search consisted of two parallel runs for 10 million generations each. Three heated chains (heating parameter = 0.2) were used for each run Trees and their associated parameters were sampled every 1000 generations. Convergence of the two runs was judged sufficient using 0.008 as the cut-off for the average deviation of the split frequencies. The sampled log-likelihood values were visualized using Tracer v1.6 [62], and adequate mixing of the chain was assessed using an effective sample size (ESS) >200 and a Potential Scale Reduction Factor (PSRF) value near 1.0 as criteria [63]. The first 25% of generations were discarded as burn-in. Gene tree concordance was assessed by analyzing each locus individually using Bayesian analysis.
To estimate species-trees, we used the coalescent model in Ã BEAST 1.8.2 [64] using a yule prior for tree topology [65,66]. To allow for differences in the molecular characteristics of each sequence, the substitution and clock models parameters were set as unlinked for COI, 12S, and 16S. We estimated genetic distances between species for each gene using the uncorrected p-distance in MEGA 6 [67]. Pinto-Sánchez et al. [41] suggested an estimated root for this complex in about 4.16 Myr old. We used this date to infer specific nucleotide substitution rates for each locus (i.e. 0.004 Myr −1 for the 12S; 0.0025 Myr −1 for 16S; and 0.0097 Myr -1 for COI) to estimate lineage divergence (time to most recent common ancestor, TMRCA). These mutation rates were set as priors for the ucld.mean parameter in the Ã BEAST analyses. We performed two independent runs of 120 million generations with a 10% burnin to reach reliable ESS values (>200). We use TRACER v1.6. [66] to assess stationarity and convergence of runs.

Species delimitation
We used a Poison tree processes (PTP) model for species delimitation [68] to infer the most likely species number in our data, as implemented in bPTP server (http://species.h-its.org/ptp/ ). This method has commonly used to explore putative species boundaries using only nucleotide substitution on a given phylogenetic tree, implementing a model assuming gene tree branch lengths generated by two independent Poisson process classes (within-and among-species substitution events). PTP is a single-locus species delimitation method that outperforms the commonly used Generalized Mixed Yule Coalescent (GMYC), without requiring an ultrametric tree [68,69]. As input, we used a makimum likelihood best solution tree of concatenated dataset, estimated by GARLI. We ran the PTP analysis using 100,000 MCMC generations, with a thinning value of 100, a burn-in of 0.1, and opted for removing the outgroup to improve species delimitation. Convergence of MCMC chain was confirmed visually as recommended [68].

Morphological analyses
We selected specimens for morphological analysis after a second step reduction of taxon sampling, based on qualitative morphological differences, congruent with the phylogenetic analyses (see details in the Integrative Working Protocol, S1 Table). We used the characters, terminology and format of Duellman and Lehr [30]. Measurements were taken from the right side of specimens, and, if this was not measurable, from the left side. Sex was determined by direct inspection of gonads. Measurements were taken on 14 morphometric characters with the aid of dial calipers (~0.1 mm precision): (1) snout-vent length (SVL) = distance from tip snout to posterior margin of vent; (2) head width (HW) = greatest width of head at level of jaw articulation; (3) head length (HL) = distance from the tip of snout to posterior angle of jaw articulation; (4) horizontal eye diameter (ED) = distance between anterior and posterior margin (corner) of eye; (5) Interorbital distance (IOD) = the width of the braincase between the orbits; (6) eye-nostril distance (EN) = distance from posterior margin of nostril to anterior margin of eye; (7) width of upper eyelid (EW) = horizontal length of the upper eyelid; (8) tympanum diameter (TD) = distance between external anterior and posterior margins of tympanic annulus (not used for multivariate analyses because tympanum is absent in P. acuminatus sensu lato); (9) femur length (FL) = length of femur from vent to knee; (10) tibia length (TL) = length of flexed leg from knee to heel; (11) foot length (FtL) = distance from heel to tip of toe IV, including in the measurement the length of tarsus and foot; (12) hand length (HdL) = distance from proximal border of thenar tubercle to tip of Finger III; (13) disc diameter of Finger III (F3) = horizontal width of the disc of Finger III; and (14) disc diameter of Toe IV(T4) = horizontal width of the disc in Toe IV. We conducted multivariate analyses with morphological measurements to reduce morphometric variables (principal component analysis, PCA) and to delimitate species (discriminant analysis). We performed normed principal component analyses as implemented in the ade4 R package [70]. We evaluated the effect of variables on percent explained variance for males and females separately. Additionally, we performed a discriminant analysis (DA) to identify morphometric traits that contribute most to species separation. Specimens used in DA were defined following the phylogenetic analyses. To avoid size-dependent correlation effects, regression residuals on log-transformed data were calculated using snout-vent length (SVL) as variable. We applied a forward stepwise procedure (tolerance = 1.0), with residuals and the natural logarithm of SVL as variables, to evaluate whether species were separated in morphological space and which morphometric characters contribute to the separation. A matrix of squared Mahalanobis distances was used to compare differences between species and to classify cases assigned by the DA.
Finally, we applied the Kruskal-Wallis non-parametric test (KW-test, corrected with exact Monte Carlo Test, 10000 samples and 95% confidence intervals) to compare overall morphological variation among species. We preferred this test over a parametric test (e.g. MANOVA), due to the nature of the data (e.g. unbalanced data, small sample size, lack of normality). For variables with significant p-values from the KW test, we applied non-parametric Mann-Whitney pairwise comparisons. All analyses and statistics were developed in PASW Statistics v18.0 (WinWrap Basic).

Environmental species delimitation by ecological niche models (ENMs)
Similar to morphological analysis, after a second step reduction of taxon sampling, we selected collection localities for ecological analysis congruent with the phylogeny (see details in Integrative Working Protocol). The ecological niche models (ENMs) were developed under the assumption that organisms have distinct ecological requirements that determine their occurrences in time and space [71]. It has been argued the ENMs are useful for delimiting cryptic species [8,72], but this is controversial. We collected georeferenced data from 130 specimens known from 93 unique localities available in herpetological collections (S2 Table).
We conducted multivariate analyses with 19 environmental variables from WorldClim project [73]. These parameters incorporate annual trends (e.g. mean annual temperature, annual precipitation), aspects of seasonality (e.g. annual range in temperature and precipitation) and extreme or potentially limiting environmental factors (e.g. temperature of the coldest and warmest months, and precipitation of the wettest and driest months). We performed a normed Principal Component Analyses (PCA-env) and Discriminant Analyses (DA-env) to define environmental traits that are most informative for distinguishing the ENMs of each species. The localities used in DA-env were defined based on clades recovered in the phylogenetic analyses. A matrix of squared Mahalanobis distances was used to compare differences among species and determine the number of cases correctly and incorrectly assigned by DA-env.
We also modeled habitat suitability for population of a species within the Pristimantis acuminatus complex using MaxEnt Software version 3.3.3a [74,75]. MaxEnt estimates the probability of distribution that has maximum entropy by applying the following principle: the expected value for each feature (e.g. climatic variables) must equal the empirical average value for points relating to known presence. The algorithm performs a certain number of iterations until reaching a convergence limit. The final map represents a favorability rating ranging from 0 (unsuitable) to 1 (perfectly adequate) [75]. The program uses two input resources: localities of the species record (presence-only data) and digital layers of the environmental conditions of a given area. The set of localities was randomly partitioned for each species, in 70% as training data and 30% for testing the model (see below concerning ROC curve criteria to validating data). The environmental variables that were relevant to each model's reconstruction are reported based on the multivariate analyses and results of the jackknife test calculated by Max-Ent [76]. This allowed us to reduce over-fitting of the distribution models generated for each species [7,77]. Resolution grid cell size, or pixel size, was 0.0083 degrees, which corresponds tõ 1 km 2 in each raster. The overall predictive distribution models for each species were generated with 5000 iterations of the complete training dataset. The analysis was done without "clamping" or "extrapolation" to avoid unsupported extrapolations on the extremes of the ecological variables. All other parameters of MaxEnt were maintained as default settings. To aid model validation and interpretation, it is usually desirable to distinguish suitable from unsuitable areas by setting a decision threshold above which model output is considered to be a prediction of the species presence. There is no rule for setting these thresholds because their values depend on the data used or the purpose of the map, which will vary from species to species [78]. Because we used all training datasets to generate validated ENMs, we decided to apply the Minimum Training Presence (MTP) value as the threshold to convert the probabilistic values (logistic ranges from 0 to 1) into a binary presence-absence map. The occurrence extension range was created from a convex hull polygon derived from the union of all points from verified localities. Using a polygon might underestimate the distribution of the species, especially when additional localities of occurrence are expected to be found. Nonetheless, we applied this method because it is commonly used to evaluate and compare the extension range for threatened species [79,80]. Spatial analyses and map algebra were done using ArcMap 10 Software; the convex hull polygon was calculated from Minimum Bounding Geometry routine in ArcTool Box [81].
The performance of MaxEnt models are usually evaluated using a ROC curve (Receiver Operating Characteristic curve; [75]), a statistical technique that has become a dominant tool in evaluating species distribution model [82]. However, several problems have been associated with this technique [83,84]; one is that the two error components (omission and commission) are inappropriately weighted equally. Accordingly, we use partial-area ROC, which evaluates only over the spectrum of the prediction and allows for differential weighting of the two error components [13,84]. The Area Under the ROC Curve (AUC) was limited to the proportional areas over which models actually made predictions and only omission errors of <5% were considered [84]. We calculated partial AUCs with the Tool for Partial-ROC V. 1.0. [85] using 30% of the original data for independent model evaluation. We present the partial ROC results as the ratio of the model AUC to the null expectation (AUC ratio; [84]). Bootstrapping analyses of AUCs were done by resampling (with replacement) 50% of the points 1000 times from the overall pool of data. One-tailed significance of differences in AUC (e.g. deviation from the line of null expectation) was assessed via fitting a standard normal variate (z-score) and calculating the probability that the mean AUC ratio is 1 [84].

Ecological equivalency of niche models
Schoener's D metrics [86,87] were calculated from ENMs for each pair of species using ENMtools software [6]. It has been argued that the ecological interpretation of this index suggest that the suitability scores generated by MaxEnt are relative proportional to species abundance [6]; thus we prefer this metric over the similarity statistic I-index [87]. The D-index measures the overlap or similarity of the suitability area predicted by MaxEnt for pairs of species (considering the logistic probabilities of pixels) and represents the proportion of niche coincidence between them. We applied a randomization test proposed by Warren et al. [87], the Identity Test, to explore whether pairs of species´models were more different than would be expected given the underlying environmental differences between the areas in which they occur. This test assumes that probabilities of ENMs produced by two populations are identical (= equivalent) if sampling is unbiased with respect to the species environmental tolerances [6]. The hypothesis of niche identity is rejected when the observed D-value is significantly lower than the values expected from the pseudoreplicate data [6,87]. The significance of differences in Schoener's D metric from the null expectation (one-tailed) was assessed by counting the number of bootstrap replicates with lower values than the observed D-index.

Phylogenetics and species delimitation
The genetic sampling corresponds to 33 specimens (12S: 13 sequences, 16S: 27 sequences, COI: 27 sequences; see S1 Table) identified as P. acuminatus sensu lato and P. tantanti from the Amazon basin of Ecuador and Peru. The parameter estimates for the best-fit models for each mitochondrial gene are summarized in Table 1. The best topology (log likelihood = -6762.95, ML analysis) was obtained from a 5-partitioned matrix of combined mtDNA (S4 Table), with 1997 characters, under a GTR+G (12S, 16S) and TrNef+I, F81, and HKY substitution models for the first, second, and third positions of COI. The phylogeny derived from concatenated sequences shows remarkably high divergences in mitochondrial genes (Fig 1, S1 Fig), among populations of "P. acuminatus", with four distinct and well-supported clades (Bayesian posterior probabilities, pp = 1.0; non-parametric bootstrap, bs = 89-100). Uncorrected p-distances between species are detailed in Table 2. Distances ranged from 1.8-4.7% (3.3±1.5 standard error) for 12S, 1.6-3.8% (2.8±0.8 SE) for 16S, and 4.7-8.7% (7.3±1.5 SE) for COI. Distances between the four clades and their sister taxon, Pristimantis tantanti, ranges from 5.3-11.5% The Ã BEAST analysis recovered the same topology estimated by the phylogenetic analyses (2a). We found congruence among individual mitochondrial trees, recovering a sister-group relationship between Pristimantis limoncochensis sp. nov. (clade A) and P. omeviridis sp. nov. (clade B), and a sister-group relationship for P. acuminatus (clade C) + P. enigmaticus sp. nov. (clade D), all of them with high support (pp = 1.0). All gene-tree reconstructions and consensus trees (Fig 2b and 2c) show a strong support for the monophyly of the complex, with a TMRCA for the entire ingroup estimated to be 10.86 Myr (8.01-14.1 95% HPD). The divergence time between northern and southern populations in the Amazonia of Ecuador is estimated to be 6.21 Myr (4.65-8.02 95% HPD). The TMRCA of Pristimantis limoncochensis sp. nov. and P. omeviridis sp. nov. is estimated to be 3.15 Myr (1.94-4.39 95% HPD), whereas the time of divergence between P. acuminatus and P. enigmaticus sp. nov. is estimated to be 3.91 Myr (2. 41-5.41 Myr 95% HPD).
The most relevant results are: (i) The phylogeny recovered four divergent lineages, supported by the PTP model, for specimens identified as P. acuminatus in the upper Amazon Basin, and (ii) the phylogenetic tree shows that populations located in the northern portion of  Table). Colors of clades in the phylogenetic tree correspond to colors of polygons and dotted circles on the map.
the Amazon basin of Ecuador and southern Colombia group in clades A and B which are sister to populations in the southern part of Ecuador and northern Peru (clades C and D).

Morphometric species delimitation
A total of 62 adult specimens from the Pristimantis acuminatus complex were examined for morphometric analyses. Data used for principal component and discriminant analyses are provided in the S5 Table. A summary of the descriptive statistics for morphometric measurements of males and females of the P. acuminatus complex is presented in Table 3. Loadings, eigenvalues and percentage of variance explained by Principal Components and Functions in DA are provided in S5 and S6 Tables. Projections of morphometric variables in morphological space are represented in Fig 3. The normed PCA shows limited evidence of grouping. A slight separation of male specimens between clade A and clade C can be in males (S2 Fig). Variables involved in such separation are body size (snout-vent length) and Foot+Tibia length (S6 Table). Species can be successfully separated based on quantitative traits by means of a discriminant function (Fig 3  and S7 Table). In males, 93% of specimens were correctly assigned to each species; body size (snout-vent length), head traits (head length, eye diameter, and inter-orbital distance) and disc diameter (Finger III and Toe IV) were the variables that explained 84% of the classification by the functions 1 and 2 (Fig 3 and Table 4). In females, the classification value falls to 81% success of correctly assigning the specimens to the species (Table 4). Variables involved in discrimination of the females were mainly related to foot length, hand length, head traits (eye-nostril distance, eye diameter, and eyelid width), with 84% of variance explained by the functions 1 and 2 (Fig 3). A Kruskal-Wallis analysis between morphometric variables revealed significant differences among species within the Pristimantis acuminatus complex (Table 3).

Environmental species delimitation
The phylogenetic analysis placed 28 localities in clade A, 24 in clade B, 18 in clade C and 23 in clade D, all in the Upper Amazon Basin of Ecuador and Peru. A summary of environmental The number of base substitutions per site from averaging over all sequence pairs between species is shown as an uncorrected p-distance value (%) by gene.
doi:10.1371/journal.pone.0143392.t002  variables, loadings, eigenvalues and percentage of variance explained by principal components (PCA-env) and functions in discriminant analysis (DA-env) is presented in S8 Table. Projections of climatic variables in environmental space are presented in Fig 3. The normed PCA-env shows limited evidence to separate localities by each species group, in spite of high correlation with axis 1 (96% of variance explained; S2 Fig). Annual Rainfall, Precipitation of Wettest Quarter, Precipitation of Warmest Quarter and Precipitation of Coldest Quarter are contribute to the two most important principal components. In contrast to PCA-env, the DA-env discriminates species shown a better classification in a context of discriminant function (Fig 3), 90% of localities were correctly assigned to each species (Table 4). The first two axes of the DA-env analysis explain nearly 90% of overall variance (

Ecological niche models
Descriptive statistics and spatial representation of the ecological niche models are shown in Fig  4 and Table 5, respectively. Models were constructed with 21 points for clade A (7 for testing), 18 points for clade B (6 for testing), 13 points for clade C (5 for testing), and 17 points for clade D (6 for testing). The AUC-ratios for the ENMs range from 1.9±0.053 (clade A) to 1.99±0.007 (clade B). Each AUC-ratio is statistically differeny from the null model AUC ratio of 1.0 (z-test, P<<0.01). Variables with the most important contribution to models were Precipitation of the  . The ercentage of non-overlapping or exclusive areas for each model ranges from 2%, in clade A to 27% in clade D (Table 5).

Ecological niche equivalence
Warren's Identity test revealed that most pairs of ENMs are not equivalent (i.e., identical), but are nonetheless highly similar (

Taxonomic considerations
The available name for the populations sampled in our phylogeny is Pristimantis acuminatus Shreve [42]. The holotype (MCZ A19951) is an adult male (SVL = 22.13 mm) in a relative good status of conservation (Fig 5), collected from the southern lowlands in the Amazon basin of Ecuador. Regarding comparisons with the holotype of Pristimantis acuminatus, integrative information from combined genetic, morphological, distributional and ecological evidence, we assign this name to clade C in our phylogeny (Figs 1 and 2, Table 2). The holotype share with specimens from clade C in our phylogeny, the following specific diagnostic characters: 1) the tympanic annulus is not visible on skin (= tympanum absent, sensu [30]), 2) tarsus lacking assigned to different species (Clades C and D, respectively) by the discriminant analysis (A). The 93% of the males and 81% of the females were morphological successfully assigned to species, whereas the 88% of the localities were successfully classified in the environmental space.
The numbers in the cells indicate the number of individuals, followed by the percentage in parentheses. Overall percentages of successful classification are given in parentheses for males, females and localities.  folds or tubercles, 3) skin on dorsum shagreened, 4) lacking supernumerary plantar tubercles, and 5) bearing short vocal slits, vocal sac small and nuptial excrescences cream.
The main diagnostic characters used herein to distinguish P. acuminatus from other species in the complex are: 1) the tympanic annulus is not visible on skin, 2) tarsus lacking tubercles or folds, 3) genetic p-distances ranging from 1.8-4.7% (3.3±2.1) for 12S, 2.1-3.8% (2.96±0.9) for 16S gene, and 4.7-7.9% (6.7±1.8) for COI. Accordingly, the northern Amazonian populations of the Pristimantis acuminatus complex (clades A and B), and southern populations (clade D) in Ecuador and Peru require formal description as different species. Based on the current systematic work, we provide a schematic working protocol and diagnostic characters to delimit these species (Fig 6 and Table 7).  Minimum training presence (MTP) was used as the threshold to define the suitability areas in the models, whereas the occurrence area was estimated with a convex hull polygon. * indicates that the z-score is statistically significant at P<<0.01 compared to null model (AUC ratio = 1) using a Partial-ROC criterion.

Species Accounts
+ indicates the overlapped area related with at least one of the ENMs generated by species within the Pristimantis acuminatus complex.
doi:10.1371/journal.pone.0143392.t005 Table 6. Niche overlap (Schoener's D index) and Identity Test for clades within Pristimantis acuminatus complex. Characteristics: Pristimantis acuminatus is characterized by: (1) skin of dorsum shagreened, with scattered small tubercles in males; dorsolateral folds absent; skin of belly areolate; discoidal fold prominent; (2) tympanic annulus and membrane not visible, covered by muscle beneath skin; (3) snout long, acuminate (females) to subacuminate (males) in dorsal view, truncated and posteriorly inclined in profile, bearing a rostral papilla in some; lips flared, canthus rostralis angular in dorsal and lateral view, loreal region concave; (4) upper eyelid about 60% of inter-orbital distance, lacking tubercles; (5) dentigerous processes of vomer small, triangular, bearing 3-4 teeth; (6) males with short vocal slits, extending from mid-lateral base of tongue to the angle of the jaw; vocal sac small, nuptial excrescences cream; (7) fingers large and slender, first shorter than second; discs on outer fingers expanded, bluntly truncated, about 1.5x the width of digit proximal to pad; supernumerary tubercles prominent, rounded; (8) fingers bearing lateral fringes; (9) forearm bearing 1-4 ulnar tubercles, small; (10) heel lacking tubercles; inner and outer border of tarsus smooth; tarsal folds absent; (11) two metatarsal tubercles; inner elliptical, about 5x the outer tubercle; supernumerary plantar tubercles absent; (12) toes with lateral fringes; webbing absent; discs equal in size or slightly smaller than those on fingers; Toe V longer than Toe III; (13) in life, dorsum bright greenish yellow with scattered black, orange or brown blotches; groin and anterior surfaces of thighs uniformly yellowish;    Diagnosis: Among greenish Amazonian Pristimantis, P. acuminatus is most similar to P. limoncochensis sp. nov., P. padiali, and P. tantanti by lacking tympanum and an externally visible tympanic annulus (it is concealed by muscle). Ulnar and tarsal tubercles in P. acuminatus are small or absent, whereas they are prominent in a row in P. padiali or coalescing forming a fold in P. tantanti. Furthermore, P. padiali have a dark brown plantar and palmar coloration (tan in P. acuminatus), whereas the relative size among inner/outer metatarsal tubercles is 2x smaller in P. tantanti compared with the 5x in P. acuminatus. P. acuminatus and P. limoncochensis sp. nov. having small ulnar tubercles (not forming a row or fold), but tarsal and supernumerary plantar tubercles are absent in the former species. Furthermore, dorsal skin is shagreened in P. acuminatus, but smooth in P. limoncochensis sp. nov. Males in Pristimantis acuminatus have vocal slits and nuptial pads, which are absent in males of P. limoncochensis sp. nov., P. padiali and P. tantanti ( Table 7). Comparisons of diagnostic characters between Pristimantis acuminatus and the new species are detailed in Table 2.
Variation: Measurements and proportions are provided in Table 3. Sexual dimorphism is evident in this species, males are smaller (22.8±1.12 mm; 20.91-24.01 mm) than females (30.5 ±2.96 mm; 27.1-33.45 mm). Furthermore, adult females present a well-defined acuminate shape in snout, compared with the subacuminate snout in males. In life, most specimens present a well-defined canthal stripe, which is less visible in preserved material, with a dark brown to black snout (e.g. QCAZ 53263, 53845). The holotype has a small tubercle (rostral papilla) on the tip of snout which is variable in prominence among specimens.
Coloration in life: Body coloration is darkest in specimens observed by day, and brightest in specimens observed at night. At night, dorsum and flanks are bright greenish yellow; black dots and orange blotches are present on dorsum; the legs of males are densely flecked. Interobital bar, subocular stripes, scapular and sacral marks absent; sides of head colored as dorsum, black canthal stripe continuing until the mid-flank; groin and anterior surfaces of thighs yellowish white; posterior surfaces of thighs uniform yellowish cream; ventral surfaces of belly yellowish cream; throat, foot and hand yellowish tan. Dorsal surfaces of pads on Fingers II and III distinctly yellow, dark brown on Fingers III and IV; iris coppery red. By day, all bright surfaces turn into dark greenish yellow.
Coloration in preservative: Dorsum cream, stippled with minute brown flecks. Sides of head dark brown to black, canthal stripe barely visible. Forearms and hind limbs with barely defined dark brown marks. Venter, throat, chest, ventral surfaces of limbs, and palms, cream, densely stippled with minute brown flecks (visible under magnification); posterior surfaces of tarsus and plantar surfaces uniform cream.
Natural history and distribution: Pristimantis acuminatus is known from 13 localities along Amazonian evergreen lowland forest of southern Ecuador in Pastaza and Morona-Santiago Provinces, and five localities from northern Loreto Department in Peru, up to 1123 m elevation. Specimens examined, but not included in morphometric or phylogenetic analyses, from Leticia [7-11 km road to Tarapacá (ICN 11187)] and from Parque Nacional Amacayacu (IAVH 4628), Amazonas Department are tentatively assigned to this species (Fig 1). The area of known occurrence is calculated to be about 50,951 km2 in lowland and piedmont evergreen forest in eastern Ecuador and Peru, but a wider distribution is likely, up to 201,174 km2 in the upper Amazon Basin (Fig 4). According to field notes and museum databases, specimens of P. acuminatus were found active at night on leaves of low vegetation, up to 3 m above ground. It is suspected that is an inhabitant on forest canopy. Calling and reproductive behavior are unknown.
Remarks: According with our phylogenetic analyses, P. acuminatus is sister to P. enigmaticus from populations along the southern Amazonia of Ecuador and northern Peru.
Pristimantis      Characteristics: Pristimantis enigmaticus is characterized by: (1) skin of dorsum shagreened; dorsolateral folds absent; skin of belly areolate; discoidal fold absent; (2) tympanic annulus and membrane visible; (3) snout long, acuminate in dorsal view, bearing a pointed papilla in adult females; truncated and posteriorly inclined in profile; lips not flared, canthus rostralis angular in dorsal and lateral view, loreal region concave; (4) upper eyelid about 60% of inter-orbital distance, lacking tubercles; (5) dentigerous processes of vomer large, transverse, bearing 6-8 teeth; (6) males lacking vocal slits, vocal sac, and nuptial excrescences; (7) fingers large and slender, first shorter than second; discs on outer fingers expanded, bluntly rounded, about 1.5x the width of digit proximal to pad; supernumerary tubercles barely visible, small, rounded; (8) fingers bearing lateral fringes; (9) forearm bearing 2-3 ulnar tubercles, small; (10) heel lacking tubercles; outer border of tarsus smooth; inner border of tarsus bearing a small tarsal fold; (11) two metatarsal tubercles; inner elliptical, about 4x the outer tubercle; lacking supernumerary plantar tubercles; (12) toes with lateral fringes; webbing absent; discs equal in size than those on fingers; Toe V longer than Toe III; (13) in life, dorsum bright greenish yellow or pale orange to pink, with or without scattered blackish blotches; groin and anterior surfaces of thighs uniformly pale yellow; belly and throat pale cream to white; ventral surfaces of hands and feet tan; black canthal stripe continuing posterior to the eyes, reaching mid-flank; iris coppery red, finely reticulated with black. In preservative, all yellow or green areas fade into cream or tan; anterior and posterior surfaces of thighs uniformly tan; venter cream, throat finely stippled with minute brown flecks; black canthal stripe; (14)  Diagnosis: Among greenish Amazonian Pristimantis, P. enigmaticus is most similar to P. omeviridis sp. nov., P. pseudoacuminatus and P. olivaceus in having a differentiated tympanic annulus and tympanum. Pristimantis enigmaticus is like P. omeviridis (characters in parentheses) in coloration and snout shape, but the former species differs by having a shagreened dorsum (smooth), lacking discoidal fold (present), bearing small ulnar tubercles (large) and tarsal fold (tubercles). Males of P. pseudoacuminatus and P. olivaceus have vocal slits and nuptial pads, which are absent in males of P. enigmaticus. Furthermore, the smaller P. pseudoacuminatus has a distinctive discoidal fold, brown vertical bars below eye, and tubercles on dorsum and heel; these are absent in P. enigmaticus. Pristimantis olivaceus is distributed in the southern Amazon Basin, and differs from P. enigmaticus in having an olive body coloration (greenish yellow in P. enigmaticus), having small tarsal tubercles (tarsal fold) and by a distinctive discoidal fold (absent).
Description of the holotype: Body slender; head wider than body; slightly longer than wide, about 30% of SVL; snout long, acuminate in dorsal view, truncated and posteriorly inclined in profile; distance from nostril to corner of eye equal in length to diameter of eye; canthus rostralis straight in dorsal view, angular in cross section, sloping gradually to lips; lips not flared; internarial area not depressed, nostrils not protuberant, directed anterolaterally, situated about three-quarters distance from eyes to tip of the snout; interorbital area flat, IOD 40% of head width; eye large, protuberant, its diameter about 3x depth of lip below eye, and about 30% of head length; upper eyelid about 65% of inter-orbital distance, lacking tubercles; no interocular fold; cranial crests absent. Tympanic membrane and annulus prominent, round in shape, with supratympanic fold partially obscuring upper and posterodorsal edges; horizontal diameter of tympanum 30% of eye diameter, separated from eye by a distance of one-third tympanum length; postrictal tubercles constricted and fused to form a barely visible short ridge extending ventrolaterally from tympanum; choana small, rounded, not concealed by palatal shelf of maxillary arc; dentigerous processes transverse, angled postero-medially and closely separated; bearing eight small teeth; tongue elliptical, posterior border notched, not adherent to floor of mouth for about 40% of its length. Skin on dorsum shagreened; no occipital ridges or dorsolateral folds; skin on flanks shagreened to areolate; ventral surfaces of belly coarsely areolate; skin on ventral surfaces of chest, throat, and thighs smooth; discoidal folds barely visible; no thoracic fold. Forearm slender; fingers large and slender, all with oval (broader than long) pads, Fingers III-IV with large pads, all fingers with large discs; pad on Finger III about 1.5x wider than narrowest portion of penultimate phalanx; disc on Finger I distinctly smaller than those on other fingers; relative length of fingers I<II<IV<III; subarticular tubercles large, subconical; supernumerary tubercles prominent, subconical; palmar tubercle bifid, 2x size of oval thenar tubercle; anterbrachial tubercle small; three conical ulnar tubercles are present along inner edge of forearm; outer edge of forearm shagreened, tubercles absent; Hind limbs slender; tibia length about 50% of SVL; knee and heel lacking tubercles; foot length about 67% of SVL; outer border of tarsus smooth; inner border of tarsus bearing four tubercles, small; inner metatarsal tubercle oval, 3x size of rounded outer tubercle; supernumerary tubercles rounded, small; subarticular tubercles subconical, rounded; toes with non-crenulate lateral fringes; webbing absent between toes; pads of Toes III−V large, in all other pads and discs of toes like those of fingers; relative lengths I<II<III<V<IV; Toe III extending to one half the distance between penultimate and ultimate subarticular tubercles on Toe IV; Toe V extending to distal edge of ultimate subarticular tubercle on Toe IV. Vent opening puckered, shagreened, not extended, lacking tubercles on its border, located at upper level of thighs.
Measurements ( Table 3. Sexual dimorphism is evident in this species, males being smaller (20.7±2.32 mm; 18.51-24.79) than females (30.8±2.92 mm; 26. 38-36.37). Furthermore, adult females present a well-defined acuminate snout, bearing a small papilla. Males vary in snout shape, from sub-acuminate to acuminate. Ulnar tubercles are visible in adult males and females, but absent in juveniles. In life, most specimens show chromatic variation, being bright green by night and dark greenish brown by day.
Coloration in life (Figs 8 and 9): At night, dorsum bright green with or without black dots and orange blotches; flanks greenish white. Interobital bar, subocular stripes, scapular and sacral marks absent; sides of head colored as flanks, black canthal stripe continuing until reach the mid-flank; groin and anterior surfaces of thighs yellowish white; posterior surfaces of thighs uniform yellowish cream; ventral surfaces of belly yellowish cream; throat, foot and hand yellowish tan. Dorsal surfaces of pads on Fingers II and III distinctively cream white; iris coppery red. By day, all bright surfaces turn into dark greenish, orange blotches to pink; flanks, dorsal surfaces of limbs turning to dark brown with white flecks.
Coloration in preservative: Dorsum cream or tan with or without brown flecks. Sides of head similar to dorsum, with a well-defined canthal stripe. Forearms and hind limbs with barely defined dark brown marks. Venter, throat, chest, ventral surfaces of limbs, and palms, cream densely stippled with minute brown flecks (visible under magnification); posterior surfaces of tarsus and plantar surfaces uniform cream.
Etymology: The specific name is derived from the Latin adjective "aenigmaticus" ("puzzling, obscure, enigmatic") and refers to the fact that this species was hidden for decades in a complex of cryptic species. The epithet is an adjective.
Natural history and distribution: Pristimantis enigmaticus is known from 17 localities along Amazonian evergreen lowlands and piedmont of southern Ecuador in Morona-Santiago, Orellana and Pastaza Provinces, and five localities from northern Loreto Department in Peru, up to 960 m elevation. The area of occurrence is calculated to be about 53,202 km 2 in lowland and piedmont evergreen forest in eastern Ecuador and Peru, but a wider distribution is likely, up to 370,019 km 2 in the upper Amazon Basin (Fig 4). According to field notes and database records, specimens of Pristimantis enigmaticus were found active at night on leaves of low vegetation, up to 3 m above ground, but it is suspected that it is an inhabitant of forest canopy. Calls and reproductive behavior are unknown.
Remarks: Lynch and Duellman [89] proposed and discussed varying conditions of tympanum, which were widely followed in works of systematics and taxonomy of direct-developing frogs [29,30,90]. Following several important papers in the upper Amazon Basin of Ecuador and Peru [20,30,45], the diagnostic character of "Tympanum concealed beneath skin" was frequently applied to separate this species from congeners [43,44]. However, as noted in the original species description, the holotype of P. acuminatus (MCZ 19951) lacks a defined tympanic annulus, which is present on two paratypes (MCZ A19949-50); we re-identified them as Pristimantis enigmaticus sp. nov. According to the phylogenetic analyses, P. enigmaticus is sister to P. acuminatus. Specimens CORBIDI 2537 (Sierra del Divisor, S°6.21361, W°73.2391, 500 m, Loreto, Peru) and MHNC 11178 (Reserva Comunal Machiguenga, S°12.1789, W°73.0814, Cusco, Peru) were not included in phylogenetic analyses, but match most of the morphological traits described for P. enigmaticus, except for the rounded shape of snout. Both specimens are tentatively considered representatives of different species in southern and central Peru, maybe related to P. olivaceous [91].
Pristimantis limoncochensis sp. nov. Ortega-Andrade, Rojas-Soto, Valencia, Espinosa de los Monteros, Morrone, Ron, and Cannatella. urn:lsid:zoobank.org:act:A0F026EA-05AB-4E3A-928D-AD0B35BE7112 (Figs 7-10 Characteristics: Pristimantis limoncochensis is characterized by: (1) skin of dorsum smooth; dorsolateral folds absent; skin of belly areolate; discoidal fold barely evident; (2) tympanic annulus and membrane not visible, covered by muscle; (3) snout long, acuminate in dorsal view, truncated and posteriorly inclined in profile; lips not flared, canthus rostralis angular in dorsal and lateral view, loreal region concave; (4) upper eyelid about 60% of interorbital distance, lacking tubercles; (5) dentigerous processes of vomer small, transverse, bearing 4-5 teeth; (6) males lack vocal slits, vocal sac and nuptial excrescences; (7) fingers large and slender, first shorter than second; discs on outer fingers expanded, bluntly truncated, about 1.5x the width of digit proximal to pad; supernumerary tubercles large, rounded; (8) fingers bearing narrow lateral fringes; (9) forearm bearing 3-5 ulnar tubercles, small; (10) heel lacking tubercles; outer border of tarsus smooth; inner border of tarsus bearing a small tarsal fold; (11) two metatarsal tubercles; inner elliptical, about 3x diameter of outer tubercle; supernumerary plantar tubercles absent or barely visible; (12) toes with lateral fringes; webbing absent; discs equal in size or slightly smaller than those on fingers; Toe V longer than Toe III; (13) in life, dorsum bright greenish yellow with or without scattered black, orange, or brown blotches; groin and anterior surfaces of thighs uniformly greenish yellowish; belly bright yellow to cream; throat, foot and hand greenish tan; black canthal stripe extending from behind the eyes until the mid-flank; iris coppery red. Well-defined yellow line present along the border or upper eyelid. In preservative, all yellowish areas fading to cream, anterior and posterior surfaces of thighs uniformly tan; venter cream; canthal stripe black, snout brown; (14)  Diagnosis: Among greenish Amazonian Pristimantis, P. limoncochensis sp. nov. is similar to P. acuminatus, P. padiali, and P. tantanti in lacking a differentiated tympanic annulus and membrane, which are covered by muscle. Ulnar and tarsal tubercles in P. limoncochensis are small, whereas they are prominent in a row in P. padiali or coalescing forming a fold in P. tantanti. Furthermore, P. padiali has a dark brown plantar and palmar coloration (greenish tan in P. limoncochensis), whereas P. tantanti lacks a distinctive discoidal fold (present in P. limoncochensis). P. limoncochensis and P. acuminatus have small ulnar tubercles (not forming a row or fold), but tarsal and supernumerary plantar tubercles are present in the former species. Furthermore, the dorsum in P. limoncochensis is smooth compared with the shagreened dorsum in P. acuminatus, whereas males of the former species lack vocal slits and nuptial pads (both present in P. acuminatus).
Description of the holotype: Body slender; head wider than body; slightly longer than wide, about 40% of SVL; snout long, acuminate in dorsal view, truncated and posteriorly inclined in profile; distance from nostril to corner of eye slightly shorter than diameter of eye; canthus rostralis straight in dorsal view, angular in cross section, sloping gradually to lips; lips not flared; internarial area not depressed, nostrils slightly protuberant, directed anterolaterally, situated about three-quarters the distance from the eyes to the tip of the snout; interorbital area flat, IOD 43% of head width; eye large, protuberant, its diameter is about 3x depth of lip below eye and about 40% of head length; upper eyelid about 60% of inter-orbital distance, lacking tubercles; no inter-ocular fold; cranial crests absent. Tympanic annulus and membrane not visible, covered by muscle beneath skin; lacking postrictal tubercles; choana small, rounded, not concealed by the palatal shelf of maxillary arc; dentigerous processes transverse, angled postero-medially and closely separated; bearing four small teeth, visible; tongue elliptical, posterior border not notched, not adherent to the floor of the mouth for about 40% of its length; vocal slits, vocal sac and nuptial excrescences absent. Skin on dorsum smooth; no occipital ridges or dorsolateral folds; skin on flanks shagreened; ventral surfaces of belly areolate; skin on ventral surfaces of chest, throat and thighs smooth; discoidal folds barely visible; no thoracic fold. Forearm slender; fingers large and slender, all with oval (broader than long) pads, Fingers III-IV with large pads, all fingers with large discs; pad of Finger III about 1.5x wider than narrowest portion of penultimate phalanx; disc of Finger I distinctive smaller than those on other fingers; relative length of Fingers I < II< IV< III; subarticular tubercles large, subconical; supernumerary tubercles prominent, elliptical; palmar tubercle bifid, 1.5x size of oval thenar tubercle; antebrachial tubercle small; two ulnar tubercles are present on the anterior part of forearm; outer edge of forearm shagreened, tubercles absent.
Hind limbs slender; tibia length about 53% of SVL; knee and heel lacking tubercles; foot length in about 70% of SVL; outer border of tarsus smooth; inner border of tarsus bearing a small tarsal fold; inner metatarsal tubercle oval, 1.5x size of round outer; supernumerary tubercles barely visible, rounded, small; subarticular tubercles subconical, rounded; toes with noncrenulate lateral fringes; webbing absent between toes; pads of Toes III−V large, all other pads and discs of toes like those of fingers; relative lengths I<II<III<V<IV; Toe III extending to proximal edge of penultimate subarticular tubercle of Toe IV; Toe V extending to distal edge of ultimate subarticular tubercle of Toe IV. Vent opening puckered, shagreen, not protruding, lacking tubercles on its border, located at upper level of thighs; testis cream.
Measurements ( Coloration in life (Figs 8 and 9): Body coloration is darkest in specimens observed by day, and brightest in specimens observed by night. By night, dorsum and flanks are bright greenish yellow; with or without black dots or orange blotches. Interobital bar, subocular stripes, scapular and sacral marks absent; a well-defined yellow line is present along the border or upper eyelid; sides of head colored as dorsum, black canthal stripe continuing to the mid-flank; groin and anterior surfaces of thighs are greenish yellow; posterior surfaces of thighs uniform is yellowish cream; ventral surfaces of belly are yellowish cream; throat, foot and hand are greenish tan. Dorsal surfaces of pads on Fingers II and III are pale yellow; iris coppery red stippled with gold and narrow black reticulations. By day, all bright surfaces on head, throat, flanks and limbs turn into dark greenish brown.
Coloration in preservative: Dorsum cream or tan, with or without brown flecks. Sides of head greyish brown; black canthal stripe well-defined. Forearms and hind limbs with or without dark brown marks. Venter, throat, chest, ventral surfaces of limbs, and palms cream; posterior surfaces of tarsus and plantar surfaces uniform cream. Darker coloration in preserved specimens might depend of the time in which each specimen was prepared as a voucher.
Etymology: Named for the Reserva Biológica Limoncocha, located in northern Amazonia of Ecuador. This area harbors the last remnant of natural forest in the Limoncocha Lagoon. Conservation programs are currently being developed together with local communities and Ecuadorian governmental entities such as the Ministerio del Ambiente del Ecuador.
Natural history and distribution: Pristimantis limoncochensis is known from 27 localities along the Amazonian evergreen lowland forest of southern Ecuador in Napo, Sucumbíos, and Orellana Provinces, and two localities from Putumayo and Caquetá Departments in Colombia, up to 593 m elevation. The area of occurrence is estimated to be about 28,139 km 2 in lowland evergreen forest in eastern Ecuador and southern Colombia, but a wider distribution is likely, up to 177,042 km 2 in the upper Amazon Basin (Fig 4). According to field notes and database records, specimens of P. limoncochensis were found active at night on leaves of low vegetation, up to 3 m above ground. It is suspected that is an inhabitant on forest canopy. Duellman [20], under the name of Eleutherodactylus acuminatus, describe the mating call like a "short, high whistle repeated infrequently".
Characteristics: A member of the Pristimantis acuminatus complex characterized by: (1) skin of dorsum smooth; dorsolateral folds absent; skin of belly coarsely areolate; discoidal fold barely evident; (2) tympanic annulus and membrane visible; (3) snout long, acuminate in dorsal view in females, sub-acuminate in males and juveniles; truncated and posteriorly inclined in profile; lips not flared, canthus rostralis angular in dorsal and lateral view, loreal region concave; (4) upper eyelid about 65% of inter-orbital distance, lacking tubercles; (5) dentigerous processes of vomer small, transverse, bearing 5-8 teeth; (6) males lacking vocal slits, vocal sac and nuptial excrescences; (7) fingers large and slender, first shorter than second; discs on outer fingers expanded, bluntly rounded, about 1.5x the width of digit proximal to pad; supernumerary tubercles large, rounded; (8) fingers bearing lateral fringes; (9) forearm bearing 2-4 ulnar tubercles, large; (10) heel lacking tubercles; outer border of tarsus smooth; inner border of tarsus bearing 3-4 tubercles or large tarsal fold; (11) two metatarsal tubercles; inner elliptical, about 2x the outer tubercle; supernumerary plantar tubercles small; (12) toes with lateral fringes; webbing absent; discs equal in size to those on fingers; Toe V longer than Toe III; (13) in life, anterior dorsum tends to be brighter greenish yellow than posterior portion and limbs, with or without scattered black, orange or brown blotches; groin and anterior surfaces of thighs uniformly pale yellow; belly and throat pale cream to white; ventral surfaces of hands and feet tan; black canthal stripe continuing posterior to the eyes until reaching the mid-flank; iris bronze, finely reticulated with black. In preservative, all yellow or green areas turn into cream or tan, anterior and posterior surfaces of thighs uniformly tan; venter cream; black canthal stripe, brown snout; (14) SVL in adult males 20.8±1.77mm (17.77-23.31); females with 28.6 ±1.92mm (26.1-30.91).
Diagnosis: Among greenish Amazonian Pristimantis, P. omeviridis sp. nov. is most similar to P. enigmaticus sp. nov., P. pseudoacuminatus, and P. olivaceus by having a differentiated tympanic annulus and membrane on skin. Pristimantis omeviridis is similar to P. enigmaticus in coloration and snout shape, but the P. omeviridis differs by smooth dorsum (shagreened in P. enigmaticus), having discoidal fold (absent), bearing large ulnar (small) and tarsal tubercles (small tarsal fold). Pristimantis omeviridis is like P. pseudoacuminatus in having a coarsely areolate belly, but the former species can be differentiated by bearing ulnar tubercles (absent), having smooth dorsum (with warts or conical tubercles), males lacking vocal slits (present) and by uniform color below eyes (with vertical brown bars), and larger body size (female SVL = 20.05 −21.2 mm, male SVL = 13.7−16.2 mm). Pristimantis omeviridis can be distinguished from P. olivaceus by having smooth dorsum (shagreened), males lacking nuptial pads and vocal slits (present), enlarged ulnar and tarsal tubercles (small) and bright greenish yellow body (olive green with dark spots).
Description of the holotype: Body slender; head wider than body; slightly longer than wide, about 30% of SVL; snout long, acuminate in dorsal view, truncated and posteriorly inclined in profile; distance from nostril to corner of eye equal in length than diameter of eye; canthus rostralis straight in dorsal view, angular in cross section, sloping gradually to lips; lips not flared; internarial area not depressed, nostrils not protuberant, directed anterolaterally, situated about three-quarters the distance from the eyes to the tip of the snout; interorbital area flat, IOD 40% of head width; eye large, protuberant, its diameter about 3x depth of lip below eye and about 30% of head length; upper eyelid about 65% of inter-orbital distance, lacking tubercles; no interocular fold; cranial crests absent. Tympanic membrane and annulus prominent, round in shape, with supratympanic fold partially obscuring upper and posterodorsal edges; horizontal diameter of tympanum 30% of eye diameter, separated from eye by a distance of one third-tympanum length; postrictal tubercles fused to form a barely visible short ridge extending ventrolaterally from the tympanum; choana small, rounded, not concealed by the palatal shelf of maxillary arc; dentigerous processes transverse, angled postero-medially and closely separated; bearing 8 small teeth, visible; tongue elliptical, posterior border notched, not adherent to the floor of the mouth for about 40% of its length. Skin on dorsum shagreened; no occipital ridges or dorsolateral folds; skin on flanks gradually shagreened to areolate; ventral surfaces of belly coarsely areolate; skin on ventral surfaces of chest, throat and thighs, smooth; discoidal folds barely visible; no thoracic fold. Forearm slender; fingers large and slender, all with oval (broader than long) pads, Fingers III-IV with large pads, all fingers with large discs; pad of Finger III about 1.5x wider than narrowest portion of penultimate phalanx; disc of Finger I distinctive smaller than those on other fingers; relative length of Fingers I<II<IV<III; subarticular tubercles large, subconical; supernumerary tubercles prominent, subconical; palmar tubercle bifid, 2x size of oval thenar tubercle; antebrachial tubercle small; three ulnar tubercles are present along inner edge of forearm, conical; outer edge of forearm shagreened, tubercles absent; Hind limbs slender; tibia length about 50% of SVL; knee and heel lacking tubercles; foot length about 67% of SVL; outer border of tarsus smooth; inner border of tarsus bearing four tubercles, small; inner metatarsal tubercle oval, 3x size of round outer tubercle; supernumerary tubercles round, small; subarticular tubercles subconical, rounded; toes with non-crenulate lateral fringes; webbing absent between toes; pads of Toes III−V large, in all other pads and discs of toes like those of fingers; relative lengths I<II<III<V<IV; Toe III extending to one-half the distance between penultimate and ultimate subarticular tubercles on Toe IV; Toe V extending to distal edge of ultimate subarticular tubercle on Toe IV. Vent opening puckered, shagreen, not protruding, lacking tubercles on its border, located at upper level of thighs.
Measurements ( Variation: Measurements and proportions of the specimens examined are given in Table 3. Sexual dimorphism is evident in this species, males being smaller (20.8±1.77 mm, 17.77-23.31 mm) than females (28.6±1.92 mm, 26.1-30.91 mm). Juveniles and males commonly show faded interorbital bar and dark brown flecks on dorsum. Most specimens have a pointed snout tip, whereas some have subacuminate to truncated snouts (e.g. QCAZ 10564, 17866). Ulnar tubercles are prominent in living individuals, whereas they tend to be less evident in preserved specimens (e.g. QCAZ 10564).
Coloration in life (Figs 7 and 8): At night, dorsum and flanks bright greenish yellow, with or without black dots or orange blotches. Interobital bar barely visible in males and juveniles; subocular stripes, scapular and sacral marks absent; a well-defined yellow line present along the border or upper eyelid; sides of head darker colored than dorsum, black canthal stripe continuing until reaching the mid-flank; groin and anterior surfaces of thighs pale yellow; posterior surfaces of thighs uniform yellowish cream; belly and throat pale yellow to white; ventral surfaces of hands and feet tan. Dorsal surfaces of pads on Fingers II and III greenish yellow; iris bronze with narrow black reticulations. At day, all bright surfaces on head, throat, flanks and limbs turn into dark greenish brown.
Coloration in preservative: Dorsum cream to tan, with or without brown flecks. Sides of head greyish brown, with a well-defined black canthal stripe. Forearms and hind limbs with or without dark brown marks. Venter, throat, chest, ventral surfaces of limbs, and palms cream; posterior surfaces of tarsus and plantar surfaces uniform cream. Darker coloration in preserved specimens might depend of the time in which each specimen was prepared as a voucher.
Etymology: The new species is named in honor of the Huaorani indigenous nation (including non-contacted Tagaeri and Taromenane people), and recognizes the jungle as essential for their physical and cultural survival in northern Amazonia of Ecuador. The specific name is derived from ömë, meaning "forest" or "jungle" in Huao terero, the Huaorani language, and the Greek viridis, meaning "green" in allusion to the green forest which is home to this group of mostly green species.
Natural history and distribution: Pristimantis omeviridis is known from 23 localities in the Amazonian evergreen lowland forest of northern Ecuador in Napo and Orellana Provinces, and one locality from Loreto Department in Peru, up to 382 m elevation. The area of occurrence is calculated to be about 22,950 km 2 in lowland evergreen forest in eastern Ecuador and southern Colombia, but a wider distribution is likely, up to 147,787 km 2 in the upper Amazon Basin (Fig 4). Specimens of P. omeviridis were found active at night on leaves of low vegetation, up to 3 m above ground, in primary and secondary forest. It is suspected that is an inhabitant of forest canopy.

Discussion
Pristimantis acuminatus was considered a widely distributed nominal species in the upper Amazon Basin, but was suspected to be a complex of cryptic species [30]. This suspicion is corroborated by our study based on the review of its type series and the application of an integrative analyses (phylogenetic, morphological and ecological data) to a broad sample of specimens. As a result, we have described three new species and present additional information on the identity of P. acuminatus. We encourage examination of museum collections, accompanied by molecular analyses, and framing geographic distributions in an ecological context using ENMs are invaluable for the identification and validation of nominal species. In this context, following a first step reduction of taxon sampling is important to formulate a starting taxonomic framework. The protocol used herein has proven to be a very informative approach for evaluating the limits of cryptic species and could be applied to other species complexes (Fig 6).

Integrative species delimitation
Delimitation of cryptic species within the P. acuminatus complex is supported by combined qualitative (morphological traits), quantitative (PCA-ordination and discriminant analyses on morphological and ecological space) and strongly supported molecular phylogenies. Although there are high levels of genetic diversity within Amazonian Pristimantis [37], the small genetic distances (1.6-3.8% for 16S) fall within typically used values species delimitation in other Neotropical frogs [93,94]. Nevertheless, genetic thresholds for setting species boundaries remain highly subjective; thus it is necessary to consider other sources of data to justify taxonomic action, including the description of new species [12,48,95].
Due the high diversity and variation within Pristimantis [19,29], several phylogenetic analyses have demonstrated lack of agreement with taxonomies based on morphology [11,37,96]. For example, Duellman and Lehr [30] firstly recognized P. acuminatus as a complex of species, because of the variation in tympanic condition within the nominal species. Here, we conclude that the condition of the tympanum is a useful character for diagnosing species within the P. acuminatus complex (Fig 10), but a combination of other traits such as skin texture, ulnar/tarsal tubercles and sexually dimorphic characters (vocal slits and nuptial pads in males), is preferable for confident delimitation of these species (Table 7).
Atympanic amphibians typically detect low-frequency sounds, in contrast to tympanic species, which process high-frequency sounds [97]. The absence of a tympanum in anurans from the tropical Andes seems to be relatively common, and has been suggested to be an adaptation to high-altitude habitats [98]. Nevertheless, nearly 25% of Pristimantis species in the Amazonian lowlands apparently lack a distinct tympanum (condition D sensu [30]) in Amazonian lowlands. The lack of a tympanum has been claimed to inhibit anuran vocalization [99], and nontympanic pathways of sound reception and ultrasonic signal systems may compensate for degeneration of the middle ear complex [100]. The evolutionary and functional implications of tympanum absence in frogs are intriguing due to the essential role in the middle ear in mating behavior, social communication, sexual selection and territoriality [101,102].
Multivariate analysis of quantitative data is useful when cryptic species cannot be identified based on qualitative morphological differences. PCA of morphometric traits indicated that snout-vent length and other traits related to limbs (foot, tibia, and femur length) and head (head length and inter-orbital distance) explain most of the variance. Based on phylogenetic analyses, we obtained a better discrimination of species limits by a consideration of sexual dimorphism (Fig 3). Consideration of morphometric variation among populations yielded successful discriminant classifications ranging from 81 to 93% (Table 4). Thus, statistical analyses enhance our diagnosis of species, even though occasional individuals are not correctly classified (e.g. juveniles, subadults).
The distribution of mitochondrial genetic variation within the P. acuminatus complex revealed a parallel phylogeographic history for monophyletic sister species (Fig 1). The partial overlap of distributions among northern and southern Amazonian populations suggests a history of parapatric divergence (clades A-B with respect to clades C-D) (Figs 1-5). The common ancestor of the complex split in northern and southern lineages approximately 6.2 Mya, at the end of late Miocene and early Pliocene [103]. During this timeframe, critical changes in climate, tectonics, river drainages, and biodiversity in concert with the major uplift of northern Andes, including extensive marine transgressions into the Amazon basin [26]. These events may have promoted biotic isolation in Amazonia, along with the effects of river barriers and vegetation changes during periods of aridity [104]. On the other hand, sister species demonstrate sympatry (between clades A-B and C-D), with the split from common ancestors during the late Pliocene (3.2 and 3.9 Mya, respectively). Accordingly, the diversification of sister species took place by isolation into the Napo Refugia [104], where environmental stability allowed extensive development of the Amazon terra firme forests [105].
In interpreting the ENMs, it is important to distinguish between the species' known distribution, as indicated by the species records, the realized niche, and their potential distribution, the fundamental niche [71,106]. The former is that part of the fundamental niche in which a species has positive population growth rates, given the constraining effects of biological interactions and dispersal limitations (e.g. competition, habitat preferences). The fundamental niche refers to all the requirements for maintaining a positive population growth rate, disregarding biotic interactions [106]. We evaluated ENMs in the context of the fundamental niche, which identifies probabilistic suitability areas in geographic space (Fig 4). The extent of suitability areas varied among species, with a narrow ENM for Pristimantis omeviridis and a broad ENM for P. enigmaticus (Table 5). A reduction of about 75-86% in the extent of ENMs is observed when compared with occurrence areas as delimited by the convex hull polygon criterion (Table 5). Furthermore, distinctiveness among environmental niches is supported by PCA-env and discriminant analysis ( Table 4, S8 Table).
The ENMs provides complementary information for species separation (Fig 6) in spite of their relatively high similarity ( Table 6). The only exception is between P. limoncochensis from P. omeviridis and P. enigmaticus, in which the D value is not significantly different from random (Fig 4, Table 6). A debate on the concept of niche conservatism/divergence and its relation to ENM similarity indices has promoted hypotheses in an ecological-evolutionary framework [7,87,106]. From an evolutionary perspective, the close phylogenetic relationship between species of the P. acuminatus complex could explain the high similarity of the values calculated for ENMs (Table 6), suggesting niche conservatism (Table 6), suggesting niche conservatism [107]. However, a broader analysis including as larger sample size of related species of Pristimantis is needed to test this hypothesis.
The new species are suspected to be cryptic inhabitants of forest canopy, associated with bromeliads [44]. This bias in sampling and the resulting underestimation of the canopy fauna might explain the rarity of records in museum collections [108,109]. Despite the dearth of ecological data for Pristimantis, these frogs are thought to have small home ranges and low vagility [26,110], which is supported by genetic data in a very few species [111]. Local competition, source partition and microhabitat preferences can be leading ecological niche shift promoting speciation [112], in spite of the high values of environmental niche similitude found in this study [9,113].
A wider distribution is suspected for these species based on discovery of unknown populations in unexplored areas with a high probability of suitability. Examination of museum specimens reveals that several leaflitter dwellers, and canopy species were uncommon in herpetological collections [109,114,115]. New Amazonian species discovered and described in the past decade typically were based on few specimens [21,116,117], and the dearth of specimens was attributed mainly to sampling bias. This provides a challenge to biologists to critically evaluate the taxonomic status of widely distributed species, considering the growing evidence that an inventory of the amphibian fauna of Amazonia is far from complete, especially in complex and poorly surveyed microhabitats.
Despite the interest in developing methods of delimiting species, differences in methodologies, datasets and species recognition criteria often make species delimitation challenging in species-rich evolutionary radiations [1,12]. This is true for Amazonian Pristimantis frogs [11,37,118], where many cryptic species await to be described.  Table. Specimens examined for environmental and ecological niche modeling analyses. "+" indicates a literature record from Duellman and Mendelson [21]. " Ã " means the specimen was photographed but not collected. (XLSX) S3 Table. Primers employed in this study for PCR and DNA sequencing. Table. Statistical support for partitioning schemes applied for phylogenetic analyses.

Supporting Information
Bayes factor comparison and statistical support for three proposed DNA sequence data partitioning schemes for phylogenetic analyses. Entries are twice the log of the Bayes factor in the comparison between models M0 and M1 [2ln(B 10 Table. Principal components and discriminant analyses on environmental data for Pristimantis acuminatus complex. Loadings and percentage of variance explained for normed principal components (PCI-II) and discriminant axes (FI-III) based upon environmental variables used from WorldClim data set [73,119]. Bold numbers indicate highest loadings; asterisks indicate variables used to generate and validate ENMs. (XLSX)