Systematics of the broad-nosed bats, Platyrrhinus umbratus (Lyon, 1902) and P. nigellus (Gardner and Carter, 1972) (Chiroptera: Phyllostomidae), based on genetic, morphometric, and ecological niche analyses

ABSTRACT Platyrrhinus is a genus of leaf-nosed frugivorous bats that are endemic to the Neotropics. P. umbratus occurs in the Andean and costal mountain systems of Venezuela and Colombia. P. nigellus occurs along the Andes from Venezuela to Bolivia. Both species are medium-sized members of the genus possessing confusing taxonomic histories that have never intersected. Four of the 21 recognized species of Platyrrhinus, among them P. umbratus, do not have their taxonomic identification confirmed by molecular analyses. We provide the first genetic data (Cyt-b and ND2 sequences) for the species. Phylogenetic analyses including the new genetic data lead to the conclusion that P. umbratus and P. nigellus are conspecific. Through the use of Principal Components Analysis (PCA) and Ecological Niche Modeling (ENM), we confirm that P. umbratus and P. nigellus share high morphometric and environmental similarities. Based on such integrative approach, we regard P. nigellus as a junior synonym of P. umbratus. We provide an emended diagnosis of P. umbratus (subsuming P. nigellus) and draw morphological comparisons with other species of the genus with which it is sympatric. The conservation status of P. umbratus needs to be determined. The high rate of habitat destruction in the tropical Andes may soon cause P. umbratus to be reassigned to the Near-Threatened (NT) category of the International Union for Conservation of Nature (IUCN).

During the past decade, studies using morphometric, morphological, and molecular analyses have improved our knowledge of the systematics and taxonomy of Platyrrhinus [1][2][3][4][5]. Of the 21 recognized species of the genus, 17 have their taxonomic identification confirmed by molecular analyses [1,2,5]. The species for which this type of identification has not been performed are P. umbratus (Lyon, 1902), P. aquilus (Handley & Ferris, 1972), P. chocoensis (Alberico & Velasco, 1991), and P. nitelinea (Velazco & Gardner, 2009). In the case of P. aquilus and P. nitelinea, good quality tissues have not been available because all the known specimens were collected decades ago. In the case of P. umbratus and P. chocoensis, the laws regarding access to the genetic resources in most of the countries, in which these species occur, were not fully implemented until about a decade ago. This limited access to tissues needed to carry out molecular analyses [11,12].

Methods
Our assessment of the systematics of Platyrrhinus umbratus and P. nigellus was based on analyses of sequence variation of two molecular markers, followed by an assessment of museum specimens to search for morphometric and morphological congruence with the molecular results. The acronyms of the specimens and tissues included in this study are:

Molecular analysis
For five Venezuelan specimens of Platyrrhinus umbratus, we sequenced two mitochondrial genes, Cyt-b and ND2, following the protocols described by Velazco and Patterson [5]. We analyzed these sequences together with those used by Velazco and Lim [1], Velazco et al. [2], and Velazco and Patterson [5]. All sequences produced in this study were uploaded to the GenBank, with the accession numbers MH496512-MH496523 (Appendices a1). We aligned visually the sequences using CodonCode Aligner 7.0.1 (CodonCode Corporation, Dedham, MA). The Cyt-b and ND2 data sets contained, respectively, 1140 and 1044 nucleotide characters, for a total of 2184 characters. To select the models of nucleotide substitution, we used the Corrected Akaike Information Criterion (AICc), as determined by jModelTest 2.1.10 [28]: Cyt-b (GTR+ I + Γ); ND2 (TVM + I + Γ); and Cyt-b+ ND2 (TIM2 + I + Γ). We used PAUP* 4.0a (build 155) to calculate the intraspecific and interspecific Cyt-b uncorrected sequence divergence ("p") [29]. We conducted a maximum likelihood (ML) analysis using RAxML 8.2.10 in the CIPRES Science Gateway 3.3 [30,31]. We also used this software to assess support for three nodes by means of a Bootstrap resampling set to stop automatically (RAxML options: -s infile.txt -n result -p 12,345 -m GTRGAMMAI -x 12,345 -N 1000 -k -f a). Additionally, we used a Bayesian Inference (BI) analysis to estimate a phylogeny employing different models of molecular evolution for each locus. This analysis was conducted using Mr Bayes v. 3.2.6 in the CIPRES Science Gateway 3.3 [31,32]. Four (one cold and three heated) simultaneous Markov chains were run for 20 million generations. Trees were sampled every 1000 generations, with the first 25% of the generations discarded as "burn-in."

Morphological analyses
We restricted morphological comparisons to P. umbratus and P. nigellus; to species closely related to P. umbratus and P. nigellus as indicated by the molecular analysis presented in this study (P. aurarius, P. dorsalis, P. ismaeli, P. masu); and to species lacking molecular data but similar in size to P. umbratus and P. nigellus (P. aquilus, P. chocoensis).
To ascertain the congruence of phenotypic and genetic characters, we grouped museum specimens (P. umbratus, P. nigellus) for morphological and morphometric analyses based on the results of the molecular analysis (Appendices a2). External and osteological characters were based on, but not restricted to, those defined by Velazco and Gardner [3], Velazco [4], and Velazco and Solari [10]. The terminology for the dental homology of premolars follows Velazco [4]: first upper premolar (P3), second upper premolar (P4), first lower premolar (p2), and second lower premolar (p4).

Morphometric analysis
Based on the presence of closed epiphyses in wing digits, we verified all the specimens to be adults (for a list of the specimens examined, refer to Appendices a2). We used a digital caliper with a 0.01 mm accuracy to record one external (forearm) and 21 craniodental measurements. Craniodental measurements are the same illustrated by Velazco and Gardner [3, Figure 1]. Their description and abbreviations are as follows: Greatest length of skull (GLS), distance from the posterior-most point of the occiput to the anteriormost point of the premaxilla (excluding incisors).
Condyloincisive length (CIL), distance between a line connecting the posterior-most margins of the occipital condyles and the anterior-most surface of the upper incisors.
Condylocanine length (CCL), distance between a line connecting the posterior-most margins of the occipital condyles and a line connecting the anterior-most surface of the upper canines.
Braincase breadth (BB), greatest breadth of the braincase, excluding the mastoid and paraoccipital processes.
Palatal width at canines (C-C), least width across palate between the cingula of the upper canines.
Palatal length (PL), distance from the posterior palatal notch to the anterior border of the incisive alveolus.
Maxillary toothrow length (MTRL), distance from the anterior-most edge of the upper canine crown to the posterior-most edge of the crown on M3.
Molariform toothrow length (MLTRL), posterior border of the M3 alveolus to the anterior border of P3.
Maxillary breadth (MXBR), least width across the maxilla, from the lingual sides of the two M2. Mandibulary toothrow length (MANDL), distance from the anterior-most surface of the lower canine to the posterior-most surface of m3.
Coronoid height (COH), perpendicular height from the ventral surface of the mandible to tip of the coronoid process.
Width at mandibular condyles (WMC), greatest width between the inner margins of the mandibular condyles.
Width of m1 (m1W), greatest width of crown. To achieve normalization for statistical analyses, all measurements were log-transformed. We evaluated morphometric differences between populations of Platyrrhinus nigellus and P. umbratus, and between sexes of both species, by means of a PCA based on the variance-covariance matrix. We retained components with eigenvalues greater than 1. To show the relationships between groups in the morphospace, we plotted the principal component (PC) scores. For these procedures, we used PAST v3.14 [33].

Data input
We used ENM [34] to explore the environmental similarity between P. umbratus and P. nigellus. We created a niche model for each nominal species separately, and a pooled model by including all occurrence records for both species into a single data set (Appendices a3). These analyses were based exclusively on data from voucher specimens on deposit in natural history museums, with their taxonomic identifications confirmed via morphology. We carefully geo-referenced the collection localities of specimens lacking field GPS readings in their tag data. To reduce sampling bias [35], we spatially thinned our original data set using the spthin package in R [36]. While retaining the greatest number of localities possible, thinning ensured that the distance between all pairs of localities exceeds 10 km.
As potential predictors of the species' climatic niches, we chose six bioclimatic variables from the WorldClim data set, with a resolution of ca. 1 km 2 at the equator. These variables reflect information pertaining to temperature and rainfall [37]. Taking into account the known elevational distribution of P. umbratus and P. nigellus, we chose six climatic predictors known to impose physiological constraints on montane mammals [38]: BIO 01 (annual mean temperature), BIO 05 (maximum temperature of warmest month), BIO 06 (minimum temperature of coldest month), BIO 12 (annual precipitation), BIO 13 (precipitation of wettest month), and BIO 14 (precipitation of driest month). We calculated Pearson correlation coefficients for every pairwise comparison of variables across the study area using ENMTools [39].

Model calibration and evaluation
Niche models were built using a maximum entropy method (Maxent [40]), which uses localities of known presence and a random sample of pixels from the study region (i.e. background points) to characterize the environments preferred by species [35]. To be used as the calibration area of niche models, we delimited the study area to a buffer of three degrees encompassing all records, excluding areas that are unlikely to be accessible for P. umbratus and P. nigellus owing to limits in their dispersal capabilities [41]. To acquire a relatively good representation of environments available for these species, we included 100,000 random pixels within the delimited study area.
We calibrated and projected the niche models across the study areas corresponding to the three target taxa (P. umbratus, P. nigellus, and P. umbratus + nigellus). Because this involved transferring the niche models into a different space from that used for model calibration, we performed a Multivariate Environmental Similarity Surfaces (MESS) Analysis to quantify the similarity between the calibration and transference regions [42].
To select model settings approximating optimal levels of complexity, we constructed models with a wide variety of different combinations of feature classes (FC: Linear; Quadratic; Linear and Quadratic; Hinge; Linear, Quadratic, and Hinge) and regularization multipliers (RM: 1.0-6.0). We evaluated the models in the ENMeval package in R [43]. We applied a spatial block approach to data partitioning to obtain model evaluation statistics. To select optimal settings, we inspected threshold-dependent (omission rate for testing points, or OR 10 , using a threshold set by the 10% training omission rate) and threshold-independent (AUC for testing points, or AUC TEST ) evaluation statistics.

Molecular analyses
ML and BI analyses of the combined mitochondrial markers (2184 bp) produced similar highly supported topologies ( Figure 1). Four of the five Venezuelan specimens of Platyrrhinus umbratus were recovered forming a clade sister to the clade containing the five Peruvian specimens of P. nigellus. However, the fifth Venezuelan specimen of P. umbratus nested within the Peruvian P. nigellus, indicating the presence of polyphyly in the P. nigellus clade ( Figure 1). As in Velazco and Patterson [3], P. nigellus was recovered as sister to the P. ismaeli + P. masu clade ( Figure 1). The average Cyt-b pairwise distance among the specimens of P. nigellus + P. umbratus is 0.70%. The average pairwise distance between P. nigellus + P. umbratus and the closely related species P. ismaeli and P. masu is, respectively, 4.7% and 4.3% (Table 1).

Morphometric analysis
We compared a sample (n = 42) of Platyrrhinus umbratus from Colombia and Venezuela with a sample of (n = 56) of P. nigellus from Colombia, Ecuador, and Peru. We combined both samples into a single one, and performed a PCA to compare morphometrically both nominal species, and males with females irrespective of nominal species. The first three PCs accounted for 74.2% (PC1, 50.4%; PC2, 15.0%) of the total variance. A plot of the scores on the two first PCs (Figure 2a) showed a high overlap between P. nigellus and P. umbratus, indicating both nominal species to be similar in size and shape (Table 2). Another plot of the scores on the two first PCs (Figure 2b) showed a high overlap between males and females, indicating secondary sexual dimorphism in size and shape to be absent or nearly absent in P. nigellus + P. umbratus (Table 2).

Ecological niche modeling
In terms of overfitting and discrimination, the pool of optimal settings that we selected showed a high  (Table 3). Pearson correlation coefficients indicated that BIO 05 and BIO 12 were the variables most highly correlated with each other (> 0.9); however, by looking at the lambda values we noticed that these two variables were not included in the final MAXENT models. The final models applying these optimal settings yielded ecologically realistic current predictions that showed a tight association with the distribution of montane forest in the northern Andes and the mountains of northern Venezuela.
The predictions for current climatic conditions for the three models showed clear qualitative similarities in geographic space (Figure 3). For P. umbratus, the prediction was tight and less-diffuse for Colombia and Venezuela, but relatively widespread for the lowlands and coastal region from the southern part of the study area ( Figure 3b). For P. nigellus and P. nigellus + P. umbratus, the highest suitability scores of the models corresponded to mountain forests above 500 m (Figure 3a and c). For the three models, some montane areas in northeastern Colombia and northern Venezuela were clearly separated by less-suitable lowland areas, with environments representing potential barriers to the dispersal of P. umbratus + P. nigellus. The potential distributions of Platyrrhinus umbratus and P. nigellus in the geographic space were highly similar (D = 0.707), suggesting both nominal species to possess broadly overlapping climatic niches. Comparison of the potential distributions of P. umbratus and P. umbratus + P. nigellus (D = 0.669), and P. nigellus and P. umbratus + P. nigellus (D = 0.936), also suggested broadly overlapping climatic niches. The MESS analysis showed non-analogous conditions in a small region of the projection area for the P. umbratus model, indicating that the prediction for areas in the southernmost region of the potential distribution should be considered with caution.  234). Skin and skull in good condition, with the exception of the proximal portions of the forearms were removed during preparation. The exact location of the type locality of P. umbratus has been the matter of confusion owing to vague information on the provenance of the specimens [13,45]. Some researchers place it in the Magdalena Department [9,19]. The MCZ collections place it in the Santander Department. Others place it in the La Guajira Department [14]. Currently, it is accepted that the type locality of P. umbratus is in the La Guajira Department, as stated by Helgen and MacFadden [14] and followed by Velazco and Gardner [3].

Distribution
Platyrrhinus umbratus is found in northern Venezuela and northern Colombia, south through Ecuador and Peru to south-central Bolivia (Figure 3). The species can sometimes be found in lowland habitats, but is mainly found at intermediate and high elevations (between 400 and 3,150) in the tropical Andes, and in the coastal mountain systems of Venezuela.

Emended diagnosis
Platyrrhinus umbratus is a medium-sized bat (FA 40-48 mm, .0 mm, CCL 21.5-26.3 mm; Table 4, Figure 4). The species is most easily distinguished from P. aquilus, P. aurarius, and P. ismaeli by its shorter skull ( Table 4). Most of the measurements of P. umbratus overlap those of P. chocoensis, P. dorsalis, and P. masu Table 3. Evaluation metrics of ecological niche models generated for the broad-nosed bats, Platyrrhinus umbratus and P. nigellus, in northern South America. FC, feature classes; RM, regularization multiplier; AUC TEST , threshold-independent metric based on predicted values for the test localities; OR 10 , threshold-dependent metric indicating the proportion of test localities with suitability values lower than that excluding the 10% of training localities with the lowest predicted suitability; p, number of parameters in the model; EXVAR, variables excluded from the final model.   (Table 4). With respect to these species, P. umbratus can be diagnosed by the following combination of characters: 1) facial stripes well marked but dusky ( Figure 4); 2) dorsal stripe conspicuous but narrow ( Figure 4b
External characters distinguishing P. umbratus are: 1) long (> 8 mm) dorsal hair, as opposed to short (< 8 mm) hair in P. chocoensis and P. masu; 2) upper surface of the feet covered by long (> 2.5 mm) and dense hair, as opposed to short (< 2 mm) and moderately dense in P. chocoensis and P. dorsalis; 3) ventral hairs possessing three color bands, as opposed to two bands in P. chocoensis; 4) uropatagial fringe covered by long (> 4 mm) and dense hair, as opposed to short and sparse hair in P. chocoensis and P. dorsalis; 5) well-marked fold lines in the pinnae, as opposed to poorly marked in P. dorsalis; 6) metacarpal III shorter than metacarpal V, as opposed to subequal in P. chocoensis and P. dorsalis.
Dental characters distinguishing P. umbratus are: 1) lingual cingulid of p4 present, as opposed to absent in P. chocoensis; 2) well-developed protocone on M1, as opposed to small and blunt in P. dorsalis and P. chocoensis;   3) lack of a stylar cuspule on lingual face of M2, as opposed to present in P. dorsalis; 4) lingual cingulum of M2 metacone not extending beyond the metacone, as oposed to continuous to paracone in P. chocoensis and P. dorsalis.

Discussion
By regarding Platyrrhinus nigellus as a junior synonym of P. umbratus, we reduce the number of recognized species of Platyrrhinus to 20. Despite their complex taxonomic histories, the synonymies of P. umbratus and P. nigellus have never intersected. For over half a century since its description by Lyon [13], P. umbratus was considered as valid species. This changed when Sanborn [15], in his review of the genus Vampyrops (= Platyrrhinus), regarded P. umbratus as a junior synonym of dorsalis, an arrangement that was followed in a review of the Peruvian species of the genus [16]. More than two decades after Sanborn [15], Handley [46] resurrected P. umbratus as a valid species. However, P. umbratus continued to be regarded as a synonym of P. dorsalis for three decades [4,9,17], Thereafter, Handley's [46] view gained acceptance, and P. umbratus was again recognized as a valid species [3,19,20]. Platyrrhinus nigellus was described in 1972 [23]. It was deemed to be a junior synonym of P. lineatus [19,26,[47][48][49][50][51] or a subspecies of P. lineatus [20,24,25,27]. More than three decades after its description, based on morphological and morphometric analyses, Velazco and Solari [10] and Velazco [4], showed P. nigellus to be a valid species distinct from P. lineatus.
Two morphological-phylogenetic analyses have included both Platyrrhinus umbratus and P. nigellus as study taxa. Owen [52] found P. nigellus to be related to three small species (P. brachycephalus, P. helleri, P. recifinus), and P. umbratus to two large species (P. aurarius, P. vittatus). Velazco and Gardner [3] found both P. umbratus and P. nigellus to belong to a clade composed of the medium to large species of Platyrrhinus: P. umbratus was recovered as sister to P. chocoensis and P. nigellus as sister to a clade composed of nine species (P. albericoi, P. chocoensis, P. dorsalis, P. infuscus, P. ismaeli, P. masu, P. nitelinea, P. umbratus, and P. vittatus).
Sequences of P. umbratus were not available for the only published molecular-phylogenetic analysis, based on four genetic markers (Cyt-b, ND2, D-loop, RAG2) [5]. This analysis found P. nigellus to be sister to a clade containing P. ismaeli + P. masu. Our addition of Cyt-b and ND2 sequences from P. umbratus to the Cyt-b and ND2 published sequences [5], led to a topology ( Figure 1) similar to that based on the four genetic markers [5]: P. umbratus + P. nigellus were united in a clade sister to the P. ismaeli + P. masu clade.
The fast-evolving nature of mitochondrial genes allows them to be widely used to infer relationships among phyllostomid bats and to reveal hidden diversity or the need of synonymyzation (e.g. [5,7,[53][54][55][56][57]). In contrast, the slow-evolving nature of nuclear genes hinders their use in a similar manner, as shown by the failure to recover clades and/or species supported by mitochondrial genes and morphology [5,7]. Based on genetic distance values found to distinguish other species of Platyrrhinus and species in other phyllostomid genera [1,2,5,7,53], the low divergence (0.70%) among the Cyt-b sequences of the P. umbratus + P. nigellus specimens suggests that both nominal species should be synonymized into a single species, whose valid name should be P. umbratus by application of the Principle of Priority of zoological nomenclature. This course of action is further supported by the finding that one specimen of P. umbratus was nested within the P. nigellus clade (Figure 1). This synonymization is also supported by our morphometric analysis: the PCA plot ( Figure 3) showed a major overlap in size and shape between specimens assigned to both species. Finally, the climatic niche of the two nominal species is highly similar, which suggests that there is no ecological differentiation.
Given that four of the five Venezuelan specimens are in a clade, and that the five Peruvian specimens are in another clade (Figure 1), some kind of geographic structure may be present in P. umbratus, as it might be expected for a species with such an ample geographic distribution.
As part of the most recent IUCN (International Union for Conservation of Nature) Global Mammal Assessment [58,59], P. umbratus was listed in the Data Deficient (DD) category [60] owing to the absence of up-to-date information on its extent of occurrence, threats, status, and ecological requirements; whereas, P. nigellus was listed in the Least Concern (LC) category [60] based on its wide distribution and presumed large population, and because it is unlikely to be declining at nearly the rate required to qualify for listing in a threatened category. The conservation status of P. umbratus (subsuming P. nigellus) needs to be determined. The high rate of habitat destruction throughout the tropical Andes may soon cause P. umbratus to be reassigned to the NT category [60].

Disclosure statement
No potential conflict of interest was reported by the authors. Appendices a1. Species, tissue collection number and GenBank accession numbers for the Platyrrhinus and outgroup samples used in this study.

Appendices a2. Specimens Examined
The following list includes all the specimens examined in this study, with their respective localities. Refer Materials and Methods for abbreviations. Individuals or series marked with an asterisk were used in the elaboration of Table 4 and in morphometric analyses.