https://orcid.org/0000-0002-0950-4390
Abstract
Cryptic species, defined as those that are morphologically indistinguishable but phylogenetically distinct, are common in bats and correspond to the majority of newly described species. Such seems to be the case of Glossophaga soricina, a generalist, glossophagine bat that is broadly distributed throughout the Neotropics. Several studies have found high levels of molecular variation within G. soricina, suggesting that it could be a cryptic species complex. Here, we explore data derived from two-dimensional geometric morphometric analyses of cranial characters and their covariation with environmental variables, aiming to test the existence of more than one species grouped within it and to contribute to the knowledge of its variation and evolutionary history. Variation in shape and size of skull and mandible supports the two main mitochondrial lineages reported in previous studies, one corresponding to the east side of the Andes (subspecies G. s. soricina) and the other one corresponding to populations from Mesoamerica and the west side of the Andes, in turn composed of at least three monophyletic and morphologically differentiated taxa. Environmental variables correlate with shape variation and might be responsible for convergences in shape and size between the species with the smallest distributions. Based on the evidence we present in this work we elevate four subspecies to the taxonomic level of species. The correct names for the species of the analyzed complex are: G. soricina (Pallas 1766), G. mutica Merriam 1898, G. antillarum Rehn 1902, and G. valens Miller 1913.
Las especies crípticas, definidas como aquellas morfológicamente indistinguibles pero filogenéticamente separadas, son particularmente comunes en murciélagos y corresponden a la mayoría de las especies descritas recientemente. Tal parece ser el caso de Glossophaga soricina (Chiroptera: Phyllostomidae), un murciélago glosofagino generalista de amplia distribución en el Neotrópico. Numerosos estudios han encontrado altos niveles de variación molecular en G. soricina, lo cual sugiere que se podría tratar de un complejo de especies crípticas. Exploramos datos derivados de análisis de morfometría geométrica en dos dimensiones de caracteres craneales y su covariación con variables ambientales, con el objetivo de explorar la existencia de más de una especie agrupada dentro de G. soricina, y contribuir al conocimiento de su variación e historia evolutiva. La variación en tamaño y forma apoya los principales linajes mitocondriales reportados previamente, uno correspondiente al Este de los Andes (subespecie G. s. soricina), y el otro correspondiente a las poblaciones mesoamericanas y del Oeste de los Andes, el cual a su vez se compone de al menos tres taxones monofiléticos y morfológicamente diferenciados. Las variables ambientales se correlacionan con la variación morfológica, y podrían ser responsables de las convergencias en tamaño y forma en las especies de distribución más reducida. Sugerimos que la evidencia presentada es suficiente para someter a G. soricina a una revisión taxonómica y elevar cuatro grupos a nivel de especie. Los nombres correctos de las especies en el complejo taxonómico analizado son: G. soricina (Pallas 1766), G. mutica Merriam 1898, G. antillarum Rehn 1902, y G. valens Miller 1913.
Historically, morphological features have been the first criterion under which biological entities are classified based in their resemblances, differences, and discontinuities (Evin et al. 2008), and these features have driven the questions about the processes behind such diversity (Darwin 1859). This proclivity of use has, however, also led to misclassifications, because phenotypic characters are prone to convergence (Cole et al. 2002; Cardini and Elton 2008) and subtle morphological differences easily can go unnoticed even after extensive examination, resulting in the phenomenon that phylogenetically distinct individuals that seem indistinguishable morphologically to the human observer are grouped together as the same taxonomic unit, giving origin to cryptic species complexes. The development of molecular techniques and integrative analyses have revealed phylogenetic patterns that are in discordance with the original taxonomic classifications (Ruedi and Mayer 2001; Bickford et al. 2007; Dávalos et al. 2012; Velazco et al. 2018), making morphologically cryptic species the majority of newly described species within bats, as well as mammals in general (i.e., Taylor et al. 2012; Molinari et al. 2017; Morales and Carstens 2018).
Bats are a notably diverse, speciose, and well-studied, order. Nonetheless, they also constitute a group that is particularly prone to the existence of overlooked species given their elusive nature: they are nocturnal, relatively small, flying animals whose morphological differences can be very subtle even between phylogenetically separated groups (Jones 1997; von Helversen et al. 2001; Mayer and von Helversen 2001; Dávalos et al. 2012; Clare et al. 2013; Morales and Carstens 2018).
There is evidence to consider that such is the case in the nominal species Glossophaga soricina (Pallas, 1766). However, its taxonomic situation has yet to be revisited in this respect. The species has been described as a widely distributed omnivorous phyllostomid bat that can be found in a variety of ecosystems, having no particular association with any of them (Álvarez et al. 1991; Clare et al. 2014), and whose only known dispersal restrictions are elevations above 3,000 m above sea level (Webster 1983). Glossophaga soricina has been divided into five subspecies corresponding to geographic distribution (Webster and Jones 1980; Webster 1983; Fig. 1): G. s. antillarum (Rehn 1902) in Jamaica; G. s. handleyi (Webster and Jones 1980; synonymized with G. mutica by Gardner 2008) from northern Mexico to Central America and western Colombia; G. s. mutica (Merriam 1898; synonymized with G. leachii by Hall 1981) in Tres Marías Islands of Mexico; G. s. soricina (Pallas 1766) in South America at the east side of the Andes; and G. s. valens (Miller 1913) in the west side of the Andes in Ecuador and Peru.
Map showing sites sampled for morphometric data within the distribution of the five subspecies of Glossophaga soricina: 1. G. s. antillarum; 2. G. s. handleyi; 3. G. s. mutica; 4. G. s. soricina; 5. G. s. valens. The dots indicate the sampling localities from which morphological data were obtained.
High levels of intraspecific molecular variation within G. soricina suggest that it might comprise at least two cryptic species: one corresponding to G. s. soricina and the other corresponding to the remaining subspecies. These two clades are monophyletic and the genetic distance between them is as large as those found between sister species (Ditchfield 2000; Hoffmann and Baker 2001; Clare 2011; Lim and Arcila-Hernández 2016; Dias et al. 2017; Hoffmann et al. 2019). This divergence is attributed to a vicariant event caused by the most recent Andes uplift, which would have imposed a physical barrier to their gene flow (Dias et al. 2017). When it comes to external differences, aside from the insular populations G. s. mutica and G. s. antillarum along with G. s. valens all being larger in size, and G. s. soricina being the smallest form and having a darker pelage, no evident phenotypic differences are remarked upon in their taxonomic description (Webster 1983; Miller 1913). Authors who have worked with this group (Hoffmann and Baker 2001; Clare 2011; Dias et al. 2017; Hoffmann et al. 2019) agree that, to solve the taxonomic status of the taxa within this complex, there is a need for more studies using larger sample sizes and more molecular markers, as well as the analysis of morphological characters.
Integrating information obtained from the analyses of different traits is useful for inferring the evolution of a species because each of them will provide independent information about the taxonomic support of the groups (de Queiroz 2007; Raxworthy et al. 2007; Rissler and Apodaca 2007; Gager et al. 2016). During the evolutionary process, traits of an organism are subject to different rates of diversification, so that they will not necessarily reflect the evolutionary history of the species in a congruent fashion (i.e., Ruedi and Mayer 2001; Evin et al. 2008). In fact, phyllostomid bats have shown incongruence in their morphologically and genetically derived phylogenetic hypotheses (Dávalos et al. 2012). The integration of distinct types of data therefore will decrease the chance of coming to incomplete or spurious conclusions about the phylogeny of a group of organisms. In particular, in the face of a potential instance of cryptic species being present, it is desirable to collate data that will sum evidence to the molecular findings (Clare 2011).
Geometric morphometrics are considered better suited than linear morphometric analyses for the study of subtle differences in the shapes of biological structures, and their power to separate cryptic taxa and variation in species level has been demonstrated in certain cases (i.e., Mayer and von Helversen 2001; Evin et al. 2008). Moreover, the use of environmental data can give an ecological context to evolutionary explanations (Rissler and Apodaca 2007; Hernández-Canchola and León-Paniagua 2017). It is known that environmental variables can influence bats’ morphology as a response to the different selective pressures that a species might encounter throughout its distribution (Evin et al. 2008; Richards et al. 2012; Hernández-Canchola and León-Paniagua 2017; Morales et al. 2018).
Our goal in this study is to explore the patterns of morphological variation to provide new data to test the hypothesis that G. soricina is a single, species level, taxonomic entity. A two-dimensional geometric morphometrics protocol was implemented to describe cranial shape and size variation, quantify morphological differences, and estimate morphological characters in the putative groups forming the G. soricina complex. We also evaluated whether climatic variables correlate with said morphological variation.
Materials and Methods
Group delimitation
To delimit the groups for which morphological and environmental variation was explored, we used the subspecies classification within the nominal species G. soricina, which corresponds to the five recognized subspecies (Webster 1983) based on molecular evidence and geographic separation: G. s. antillarum, G. s. handleyi, G. s. mutica, G. s. soricina, and G. s. valens (Fig. 1). The South American G. s. soricina initially was divided into two groups based on molecular findings (Hoffmann and Baker 2001; Dias et al. 2017; Hoffmann et al. 2019) that revealed it contains two separate clades: one in the north formed by populations in Trinidad, Venezuela, Suriname, and Guyana, and one in the south formed by populations from Brazil, Paraguay, Bolivia, and eastern Peru. However, our morphometric analyses showed that there are no significant differences in shape or size between the two genetic clades (P > 0.05), so G. s. soricina was therefore analyzed as a single unit.
Geometric morphometrics
A total of 231 specimens were photographed (Appendix I) from three mammal collections: Museo de Zoología “Alfonso. L. Herrera” at the Universidad Nacional Autónoma de México (Mexico City, Mexico); the National Museum of Natural History Mammal Collection (Washington, DC, United States); and the American Museum of Natural History (New York City, New York, United States). Our sampling contained individuals from the five subspecies throughout their distribution range (Fig. 1). Sample sizes are uneven, reflecting the differences in distribution areas among subspecies. However, we aimed to have a minimum of 10 specimens of each sex per subspecies, with an even number of males and females.
Digital photographs of lateral and ventral views of skulls, and lateral views of mandibles, were acquired using a tripod mounted Nikon D3200 reflex camera (Nikon Corporation, Tokyo, Japan) with an AF-S Micro Nikkor 60 mm lens (Nikon Corporation, Tokyo, Japan), keeping the skulls and mandibles always in the same position and at the same distance from the camera lens. Two-dimensional landmark and semilandmark coordinates from images were marked on the digital images using the program TpsDig 2.16 (Rohlf 2010). Semilandmarks were placed using guides drawn with the program Make Fan (Sheets 2014). Skull and mandible shapes were registered in four landmark configurations. The lateral view of the skulls was divided in two modules corresponding to separate developmental regions based on studies that have demonstrated separate rostral and basicranial modules for all mammals including bats (Marroig et al. 2009; Porto et al. 2009). One module corresponds to the fronto-maxillary region and was registered with a configuration of seven landmarks and six semilandmarks (Fig. 2A). The other skull module corresponds to the parieto-occipital region and was registered with a configuration of five landmarks and 14 semilandmarks (Fig. 2B). The ventral view of the skull was registered with a configuration of 11 landmarks and five semilandmarks (Fig. 2C). The lateral view of the mandible was unable to be divided in modules due to the lack of homologous marks to define such modules; therefore, the lateral view of the mandible was registered as a whole with a configuration of 11 landmarks and 16 semilandmarks (Fig. 2D). For a detailed description on the location of the landmarks and semilandmarks, see Supplementary Data SD1.
Four configurations of landmarks registering the shape of two lateral skull modules (A: fronto-maxillary region; B: parieto-occipital region); one ventral module (C); and one lateral mandible module (D). Landmark positions are shown by circles with white filling; semilandmarks are depicted by circles with gray filling. See Supplementary Data SD1 for description of landmarks.
After each shape character was registered on the image, each was aligned separately and shape variables obtained via a Generalized Procrustes Analysis, which translates each configuration to the origin, scales, and rotates it using a least squares distance criterion to obtain the shape information once all other variance factors have been removed (Rohlf 1990).
Semilandmarks were declared using a protocol and aligned along the contour to minimize the bending energy necessary to produce the change in the target outline to the consensus shape (Mitteroecker et al. 2013). We obtained shape variables as coordinates, and a size estimator called centroid size (CS) as the square root of the sum of distances from each landmark to the centroid (Bookstein 1997). CS values were log-transformed to procure the normality of their distribution (Mitteroecker et al. 2013). The geometric morphometric procedures were carried out in the package Geomorph 3.1.2 (Adams et al. 2020) in R version 3.6.0 (R Core Team 2020). In addition, two linear inter-landmark distances were measured to compare linear size descriptors between described species: greatest length of skull (GLS), measured from the anteriormost alveolar border of the first incisor to the posteriormost border of the foramen magnum in the lateral view of the skull, and the braincase breadth (BCB), measured at the greatest breadth of the globular part of the braincase in the ventral view. Each specimen was measured two times using the same scale, in order to decrease measurement error.
To evaluate the module partitioning in the lateral view of the skull, the correlation between the two modules was tested with a Partial Least Squares (PLS) analysis, which calculated the correlations between the configuration matrices by the first PLS vectors from each matrix. Permutation tests with 1,000 random repetitions of specimen order were conducted to test the significance of the correlations (Zelditch et al. 2004) using package Geomorph 3.1.2 (Adams et al. 2020).
Statistical analyses of size and shape variation
Analyses were carried out to test for differences in CS and shape variables between sexes (sexual dimorphism) and also between subspecies, for each shape character. Differences in CS inside subspecies were explored with paired sex comparisons using a t-test on the log-transformed CS. The mean CS values for each subspecies were compared using an ANOVA model and the significance of pairwise comparisons among subspecies was assessed with the Tukey Honest Significant Differences Method (Miller 1981). These analyses were undertaken with the R package stats (R Core Team 2020).
A Procrustes ANOVA model was applied to test the effect of CS, subspecies, and sex nested in subspecies, on shape variance, which was calculated as the Procrustes distance variance respect to the mean shape of each factor (Adams and Otálora-Castillo 2013). Considering that we did not have equal sample sizes between levels of factors and groups, and the loss of independence between shape data (Klingenberg and Monteiro 2005), the significance of the F-statistic to each factor and variable was tested using a resample test with 1,000 replicates on the residuals of the model in the R package RRPP version 0.4.2 (Collyer and Adams 2019). Differences in shape between sexes in each subspecies were explored using paired comparations between group means, and their significance was tested by permutation testing, comparing the observed Procrustes distance (Pdis) to those obtained from random assignment of observation to groups in the R package Morpho (Schlager 2017).
Differences in shape among subspecies were explored using ordination methods. Differences between mean shape configurations of subspecies were evaluated with Canonical Variate Analysis (CVA) on a previous Principal Components Analyses. We selected the first five PCs as shape variables to avoid bias caused by differences in sample size among subspecies. For each of the four shape characters (fronto-maxillary region, parieto-occipital region, ventral view of skull, and lateral view of mandibles), we obtained the Mahalanobis distances between mean shapes for each of the subspecies, as well as their P-value, assessed by a permutation test on the original data matrix, with 1,000 replicates. Deformation grids and vectors for each consensus configuration were obtained to examine shape variation graphically among subspecies. Because changes were small and therefore difficult to appreciate visually, deformation grids were exaggerated by a magnitude of three in order for them to be more evident. This was done with the R package Morpho version 2.7 (Schlager 2017).
Allometry analyses were undertaken to evaluate differences in the effect of CS on shape variation (allometric vectors) among subspecies. First, multivariate regressions of shape variables on the CS values were carried out within subspecies. We then tested for differences in the direction of the allometric trajectories among subspecies, using the angle between vectors as the comparison parameter and a resample of residuals of the allometric model (shape ~ CS * subspecies) with 1,000 replicates in the package RRPP version 0.4.2 (Collyer and Adams 2019). Differences in allometric vectors were visualized for each subspecies, using the predicted shape for the first PC, for each CS with the previously described allometric model.
Association between shape and environmental variables
We explored whether environmental factors are associated with shape variation in the skull and mandible of G. soricina. To select the environmental variables, we obtained environmental data from the WorldClim database (Fick and Hijmans 2017) at a resolution of 1 km for current conditions, for an occurrence database of G. soricina obtained from VertNet that was visually inspected, depurated of dubious specimens, and spatially thinned, leaving only registers at least 10 km apart using the spThin Package in R (Aiello-Lammens et al. 2014). A Spearman correlation test was carried out to detect and exclude highly correlated variables (r ≥ 0.7). The selected variables were BIO2 (Mean Diurnal Range), BIO3 (Isothermality), BIO8 (Mean Temperature of Wettest Quarter), BIO9 (Mean Temperature of Driest Quarter), BIO13 (Precipitation of Wettest Month), BIO14 (Precipitation of Driest Month), BIO15 (Precipitation Seasonality), BIO18 (Precipitation of Warmest Quarter), and BIO19 (Precipitation of Coldest Quarter; Supplementary Data SD2). The degree of association of environmental variation with skull and mandible shape variables was assessed with a Two-block partial least squares analysis for Procrustes shape variables using the R package Geomorph version 3.1.2 (Adams et al. 2020). Finally, we extracted said environmental variables from the coordinates of the localities from which the examined specimens were collected.
Comparison between molecular and morphological evidence
To analyze the congruence between molecular and morphological evidence, we compared the position of the subspecies in the topology of the most recently published molecular phylogeny and the most parsimonious morphological tree, which was obtained by morphometric phylogenetic analyses of the four geometric morphometric characters. The aligned coordinates of each subspecies mean shape for each structure were analyzed under parsimony following Catalano et al. (2010) in the program TNT 1.5 (Goloboff and Catalano 2016). This analysis was executed initially approximating landmark positions with a 6 × 6 grid, nesting Sankoff two times, using iterative-pass for landmark optimizations and realigning during Tree Bisection and Reconnection (TBR; Goloboff and Catalano 2016); we carried out 1,000 replicates, each replicate starting from a Wagner tree.
Results
Modularity
The correlation between the two modules in which the lateral view of the skull was divided (parieto-occipital region and fronto-maxillary region) was significant (r = 0.686; P = 0.001), meaning there is a degree of integration with which these two modules covary (Adams 2016), as revealed by the Bootstrap analyses of the data with 1,000 replicates. We decided to keep module partition in spite of this, given the separate ontogenetic origins of each (Marroig et al. 2009; Porto et al. 2009).
Sexual dimorphism
The mandible was the only one of the four modules where sexual dimorphism was present, where all the subspecies except for G. s. valens (P = 0.086) showed significant differences in shape. The largest distance between mean shapes of male and females was found in G. s. mutica (dist = 0.034; P = 0.003), while the smallest difference was found in G. s. soricina (dist = 0.018; P = 0.001). The average female shape for the mandible is narrower in vertical height than that of the male, as shown by the deformation grids (Fig. 3) where the largest vectors of displacement with respect to male shape are more evident in the dentary bone in the alveolar region of the mandible. For this reason, posterior analyses of mandible shape were carried out separately for males and females. As for CS, only the subspecies G. s. antillarum differed between male and female size (t = −2.316; P = 0.036), with males being larger than females (mean males = 3.209; mean females = 3.188). In the ventral region, a small but significant difference in shape was found only for subspecies G. s. handleyi (dist = 0.008; P = 0.008), but not in remaining subspecies. No differences in CS were present between sexes in the rest of the studied modules.
Deformation grid showing vector displacements of the landmarks of the configuration of the mandible from the male mean shape to the female mean shape in the subspecies with the largest distance between mean shapes: Glossophaga soricina mutica. Deformation grids were exaggerated by a factor of 2 to improve visualization.
Size variation among subspecies
All four shape configurations exhibited the same pattern, in which insular subspecies (G. s. antillarum, G. s. mutica) along with continental subspecies G. s. valens, do not differ in size between one another (P > 0.05), while continental subspecies G. s. soricina and G. s. handleyi differ with them and also between each other (P < 0.001; Fig. 4). Glossophaga s. soricina is the smallest of the groups, G. s. handleyi is intermediate, and the other three subspecies form a cluster with the highest CS values.
Boxplot showing average size and standard deviations for males and females in the five subspecies of Glossophaga soricina for (A) anterior-maxillary region; (B) parieto-occipital region; (C) ventral region; and (D) mandible.
Shape variance in relation to size, sex, and subspecies
Procrustes ANOVAs show a significant effect of CS and subspecies on morphological variation, but that effect is not significant when sex was nested in the model, except for the case of the mandible, where the interaction between subspecies and sex did show a significant effect on shape variation (Table 1).
Shape variance components for the Procrustes ANOVA for each of the four studied modules. CS = centroid size.
| d.f. | SS | MS | R2 | F | Z | Pr(>SS) | ||
|---|---|---|---|---|---|---|---|---|
| Fronto-maxillary | CS | 1 | 0.051 | 0.051 | 0.11 | 26.191 | 5.076 | 0.001** |
| Subspecies | 4 | 0.035 | 0.009 | 0.076 | 4.537 | 4.501 | 0.001** | |
| Subsp.:Sex | 5 | 0.012 | 0.002 | 0.027 | 1.266 | 1.024 | 0.149 | |
| Residuals | 188 | 0.363 | 0.002 | 0.788 | ||||
| Total | 198 | 0.461 | ||||||
| Parieto-occipital | CS | 1 | 0.018 | 0.018 | 0.074 | 18.162 | 4.347 | 0.001** |
| Subspecies | 4 | 0.034 | 0.008 | 0.126 | 8.391 | 6.111 | 0.001** | |
| Subsp.:Sex | 5 | 0.006 | 0.001 | 0.025 | 1.234 | 0.865 | 0.189 | |
| Residuals | 188 | 1.19 | 0.001 | 0.765 | ||||
| Total | 189 | 0.248 | ||||||
| Ventral | CS | 1 | 0.01 | 0.01 | 0.085 | 19.805 | 6.117 | 0.001** |
| Subspecies | 4 | 0.01 | 0.0020 | 0.082 | 5.8 | 6.357 | 0.001** | |
| Subsp.:Sex | 5 | 0.002 | 0.0004 | 0.02 | 0.915 | −0.202 | 0.582 | |
| Residuals | 190 | 0.098 | 0.0005 | 0.813 | ||||
| Total | 200 | 0.12 | ||||||
| Mandible | CS | 1 | 0.017 | 0.017 | 0.057 | 13.538 | 5.185 | 0.001** |
| Subspecies | 4 | 0.022 | 0.006 | 0.075 | 4.504 | 5.929 | 0.001** | |
| Subsp.:Sex | 5 | 0.028 | 0.005 | 0.084 | 3.987 | 6.035 | 0.001** | |
| Residuals | 187 | 0.232 | 0.001 | 0.784 | ||||
| Total | 197 | 0.296 |
| d.f. | SS | MS | R2 | F | Z | Pr(>SS) | ||
|---|---|---|---|---|---|---|---|---|
| Fronto-maxillary | CS | 1 | 0.051 | 0.051 | 0.11 | 26.191 | 5.076 | 0.001** |
| Subspecies | 4 | 0.035 | 0.009 | 0.076 | 4.537 | 4.501 | 0.001** | |
| Subsp.:Sex | 5 | 0.012 | 0.002 | 0.027 | 1.266 | 1.024 | 0.149 | |
| Residuals | 188 | 0.363 | 0.002 | 0.788 | ||||
| Total | 198 | 0.461 | ||||||
| Parieto-occipital | CS | 1 | 0.018 | 0.018 | 0.074 | 18.162 | 4.347 | 0.001** |
| Subspecies | 4 | 0.034 | 0.008 | 0.126 | 8.391 | 6.111 | 0.001** | |
| Subsp.:Sex | 5 | 0.006 | 0.001 | 0.025 | 1.234 | 0.865 | 0.189 | |
| Residuals | 188 | 1.19 | 0.001 | 0.765 | ||||
| Total | 189 | 0.248 | ||||||
| Ventral | CS | 1 | 0.01 | 0.01 | 0.085 | 19.805 | 6.117 | 0.001** |
| Subspecies | 4 | 0.01 | 0.0020 | 0.082 | 5.8 | 6.357 | 0.001** | |
| Subsp.:Sex | 5 | 0.002 | 0.0004 | 0.02 | 0.915 | −0.202 | 0.582 | |
| Residuals | 190 | 0.098 | 0.0005 | 0.813 | ||||
| Total | 200 | 0.12 | ||||||
| Mandible | CS | 1 | 0.017 | 0.017 | 0.057 | 13.538 | 5.185 | 0.001** |
| Subspecies | 4 | 0.022 | 0.006 | 0.075 | 4.504 | 5.929 | 0.001** | |
| Subsp.:Sex | 5 | 0.028 | 0.005 | 0.084 | 3.987 | 6.035 | 0.001** | |
| Residuals | 187 | 0.232 | 0.001 | 0.784 | ||||
| Total | 197 | 0.296 |
**Significant probabilities.
Shape variance components for the Procrustes ANOVA for each of the four studied modules. CS = centroid size.
| d.f. | SS | MS | R2 | F | Z | Pr(>SS) | ||
|---|---|---|---|---|---|---|---|---|
| Fronto-maxillary | CS | 1 | 0.051 | 0.051 | 0.11 | 26.191 | 5.076 | 0.001** |
| Subspecies | 4 | 0.035 | 0.009 | 0.076 | 4.537 | 4.501 | 0.001** | |
| Subsp.:Sex | 5 | 0.012 | 0.002 | 0.027 | 1.266 | 1.024 | 0.149 | |
| Residuals | 188 | 0.363 | 0.002 | 0.788 | ||||
| Total | 198 | 0.461 | ||||||
| Parieto-occipital | CS | 1 | 0.018 | 0.018 | 0.074 | 18.162 | 4.347 | 0.001** |
| Subspecies | 4 | 0.034 | 0.008 | 0.126 | 8.391 | 6.111 | 0.001** | |
| Subsp.:Sex | 5 | 0.006 | 0.001 | 0.025 | 1.234 | 0.865 | 0.189 | |
| Residuals | 188 | 1.19 | 0.001 | 0.765 | ||||
| Total | 189 | 0.248 | ||||||
| Ventral | CS | 1 | 0.01 | 0.01 | 0.085 | 19.805 | 6.117 | 0.001** |
| Subspecies | 4 | 0.01 | 0.0020 | 0.082 | 5.8 | 6.357 | 0.001** | |
| Subsp.:Sex | 5 | 0.002 | 0.0004 | 0.02 | 0.915 | −0.202 | 0.582 | |
| Residuals | 190 | 0.098 | 0.0005 | 0.813 | ||||
| Total | 200 | 0.12 | ||||||
| Mandible | CS | 1 | 0.017 | 0.017 | 0.057 | 13.538 | 5.185 | 0.001** |
| Subspecies | 4 | 0.022 | 0.006 | 0.075 | 4.504 | 5.929 | 0.001** | |
| Subsp.:Sex | 5 | 0.028 | 0.005 | 0.084 | 3.987 | 6.035 | 0.001** | |
| Residuals | 187 | 0.232 | 0.001 | 0.784 | ||||
| Total | 197 | 0.296 |
| d.f. | SS | MS | R2 | F | Z | Pr(>SS) | ||
|---|---|---|---|---|---|---|---|---|
| Fronto-maxillary | CS | 1 | 0.051 | 0.051 | 0.11 | 26.191 | 5.076 | 0.001** |
| Subspecies | 4 | 0.035 | 0.009 | 0.076 | 4.537 | 4.501 | 0.001** | |
| Subsp.:Sex | 5 | 0.012 | 0.002 | 0.027 | 1.266 | 1.024 | 0.149 | |
| Residuals | 188 | 0.363 | 0.002 | 0.788 | ||||
| Total | 198 | 0.461 | ||||||
| Parieto-occipital | CS | 1 | 0.018 | 0.018 | 0.074 | 18.162 | 4.347 | 0.001** |
| Subspecies | 4 | 0.034 | 0.008 | 0.126 | 8.391 | 6.111 | 0.001** | |
| Subsp.:Sex | 5 | 0.006 | 0.001 | 0.025 | 1.234 | 0.865 | 0.189 | |
| Residuals | 188 | 1.19 | 0.001 | 0.765 | ||||
| Total | 189 | 0.248 | ||||||
| Ventral | CS | 1 | 0.01 | 0.01 | 0.085 | 19.805 | 6.117 | 0.001** |
| Subspecies | 4 | 0.01 | 0.0020 | 0.082 | 5.8 | 6.357 | 0.001** | |
| Subsp.:Sex | 5 | 0.002 | 0.0004 | 0.02 | 0.915 | −0.202 | 0.582 | |
| Residuals | 190 | 0.098 | 0.0005 | 0.813 | ||||
| Total | 200 | 0.12 | ||||||
| Mandible | CS | 1 | 0.017 | 0.017 | 0.057 | 13.538 | 5.185 | 0.001** |
| Subspecies | 4 | 0.022 | 0.006 | 0.075 | 4.504 | 5.929 | 0.001** | |
| Subsp.:Sex | 5 | 0.028 | 0.005 | 0.084 | 3.987 | 6.035 | 0.001** | |
| Residuals | 187 | 0.232 | 0.001 | 0.784 | ||||
| Total | 197 | 0.296 |
**Significant probabilities.
Differences in shape between subspecies
Significant differences between subspecies were found in all of the studied shape characters. Glossophaga s. soricina was the only subspecies that differed from remaining groups at all times. It displayed the largest differences with the insular groups and the smallest ones with G. s. handleyi. The largest differences among mean shapes were found in the ventral region of the skull (Pdis < 3.754) and in the mandible (Pdis < 3.576). In the ventral view, the subspecies that most differed were G. s. antillarum and G. s. soricina (Pdis = 3.754), while in the mandible the two subspecies that differed the most are G. s. antillarum with both G. s. soricina and G. s. valens (Table 2).
Distance values from mean comparisons between mean shapes of each subspecies. Only significant distances (P < 0.05) are shown.
| G. s. antillarum | G. s. handleyi | G. s. mutica | G. s. valens | ||
|---|---|---|---|---|---|
| Fronto-maxillary | G. s. handleyi | 1.673 | |||
| G. s. mutica | 1.537 | 1.608 | |||
| G. s. valens | 1.556 | 1.544 | |||
| G. s. soricina | 2.498 | 1.804 | 3.168 | 2.060 | |
| Parieto-occipital | G. s. handleyi | 1.157 | |||
| G. s. mutica | |||||
| G. s. valens | 2.381 | 1.815 | 1.548 | ||
| G. s. soricina | 2.542 | 1.794 | 2.749 | 2.935 | |
| Ventral | G. s. handleyi | 2.495 | |||
| G. s. mutica | 1.747 | 1.905 | |||
| G. s. valens | 2.123 | 2.122 | 1.376 | ||
| G. s. soricina | 3.754 | 1.585 | 2.706 | 2.974 | |
| Mandible | G. s. handleyi | 3.159 | |||
| G. s. mutica | 1.735 | 1.482 | |||
| G. s. valens | 3.576 | 1.876 | |||
| G. s. soricina | 3.539 | 1.100 | 1.852 | 1.083 |
| G. s. antillarum | G. s. handleyi | G. s. mutica | G. s. valens | ||
|---|---|---|---|---|---|
| Fronto-maxillary | G. s. handleyi | 1.673 | |||
| G. s. mutica | 1.537 | 1.608 | |||
| G. s. valens | 1.556 | 1.544 | |||
| G. s. soricina | 2.498 | 1.804 | 3.168 | 2.060 | |
| Parieto-occipital | G. s. handleyi | 1.157 | |||
| G. s. mutica | |||||
| G. s. valens | 2.381 | 1.815 | 1.548 | ||
| G. s. soricina | 2.542 | 1.794 | 2.749 | 2.935 | |
| Ventral | G. s. handleyi | 2.495 | |||
| G. s. mutica | 1.747 | 1.905 | |||
| G. s. valens | 2.123 | 2.122 | 1.376 | ||
| G. s. soricina | 3.754 | 1.585 | 2.706 | 2.974 | |
| Mandible | G. s. handleyi | 3.159 | |||
| G. s. mutica | 1.735 | 1.482 | |||
| G. s. valens | 3.576 | 1.876 | |||
| G. s. soricina | 3.539 | 1.100 | 1.852 | 1.083 |
Distance values from mean comparisons between mean shapes of each subspecies. Only significant distances (P < 0.05) are shown.
| G. s. antillarum | G. s. handleyi | G. s. mutica | G. s. valens | ||
|---|---|---|---|---|---|
| Fronto-maxillary | G. s. handleyi | 1.673 | |||
| G. s. mutica | 1.537 | 1.608 | |||
| G. s. valens | 1.556 | 1.544 | |||
| G. s. soricina | 2.498 | 1.804 | 3.168 | 2.060 | |
| Parieto-occipital | G. s. handleyi | 1.157 | |||
| G. s. mutica | |||||
| G. s. valens | 2.381 | 1.815 | 1.548 | ||
| G. s. soricina | 2.542 | 1.794 | 2.749 | 2.935 | |
| Ventral | G. s. handleyi | 2.495 | |||
| G. s. mutica | 1.747 | 1.905 | |||
| G. s. valens | 2.123 | 2.122 | 1.376 | ||
| G. s. soricina | 3.754 | 1.585 | 2.706 | 2.974 | |
| Mandible | G. s. handleyi | 3.159 | |||
| G. s. mutica | 1.735 | 1.482 | |||
| G. s. valens | 3.576 | 1.876 | |||
| G. s. soricina | 3.539 | 1.100 | 1.852 | 1.083 |
| G. s. antillarum | G. s. handleyi | G. s. mutica | G. s. valens | ||
|---|---|---|---|---|---|
| Fronto-maxillary | G. s. handleyi | 1.673 | |||
| G. s. mutica | 1.537 | 1.608 | |||
| G. s. valens | 1.556 | 1.544 | |||
| G. s. soricina | 2.498 | 1.804 | 3.168 | 2.060 | |
| Parieto-occipital | G. s. handleyi | 1.157 | |||
| G. s. mutica | |||||
| G. s. valens | 2.381 | 1.815 | 1.548 | ||
| G. s. soricina | 2.542 | 1.794 | 2.749 | 2.935 | |
| Ventral | G. s. handleyi | 2.495 | |||
| G. s. mutica | 1.747 | 1.905 | |||
| G. s. valens | 2.123 | 2.122 | 1.376 | ||
| G. s. soricina | 3.754 | 1.585 | 2.706 | 2.974 | |
| Mandible | G. s. handleyi | 3.159 | |||
| G. s. mutica | 1.735 | 1.482 | |||
| G. s. valens | 3.576 | 1.876 | |||
| G. s. soricina | 3.539 | 1.100 | 1.852 | 1.083 |
Glossophaga s. soricina and G. s. antillarum were the groups that differed the most, because they had the largest distances between their mean shapes for the three skull modules, and the second largest in the mandible (dist = 3.539) after G. s. antillarum and G. s. valens (dist = 3.576). This pattern can be observed in the CVA ordination plots, where these two groups show a tendency to occupy the most distant regions in morphospace (Fig. 5). The most similar groups were G. s. handleyi and G. s. valens, with no significant differences between the mean shapes of their mandibles and their fronto-maxillary shape. The ventral region of the skull was the only configuration in which all the subspecies differed significantly among each other (Table 2), suggesting this might be the region of the skull that holds the most phylogenetic information.
Ordination plots for the first two canonical axes from the ventral region of the skull and the mandible, with deformation grids derived from each canonical axis.
Allometry
Fitted linear models found a significant effect of size on shape variation in all the shape characters. The highest morphological variation explained by size was found in the parieto-occipital region (22.13%) and lowest for the ventral region of the skull (18.58%), while for the fronto-maxillary region it was 20.84%. For the mandible, the effect of the size on shape was similar between males (19.89%) and females (19.66%). The pairwise comparisons of directions of allometric vectors indicate significant differences only between G. s. antillarum and G. s. mutica in the parieto-occipital region (angle = 129°, P = 0.036); these subspecies showed opposite trends of shape variation with the increment of CS (Supplementary Data SD3).
Environmental effect on shape variation
There is a significant effect of environmental variables upon shape variation. The module with the highest adjustment was the ventral view of the skull (r2 < 0.417, P = 0.001), followed by the fronto-maxillary region (r2 < 0.405, P = 0.001). The lowest correlation between shape and environmental variation was in the mandible (r2 < 0.314, P = 0.012). Subspecies G. s. antillarum, G. s. mutica, and G. s. valens cluster in the same area of the PLS plot, indicating that they share the same morphospace and environmental region, which differs to that displayed by subspecies G. s. handleyi and G. s. soricina.
Comparison between molecular and morphological evidence
Our morphological assessment of relationships matches the most recent published molecular phylogeny in most of its topology (Supplementary Data SD4). It coincides in placing G. s. soricina basal to remaining subspecies, and in obtaining G. s. handleyi as sister to G. s. valens. The difference is that our morphological phylogeny groups G. s. antillarum and G. s. mutica as sister taxa recently split groups, in contrast with the molecular phylogeny which groups G. s. mutica and G. s. handleyi in the same clade (Hoffman et al. 2019).
Glossophaga s. antillarum has the largest amount of change on its branch when it comes to morphology. Glossophaga s. antillarum represents a clade with maximal support divergent in nuclear and mitochondrial markers (Hoffmann et al. 2019).
Discussion
Morphologically cryptic species complexes are common among all mammal groups, and bats are particularly prone to this kind of misclassification given their small size, elusive habits, morphological convergences, and subtle differences (Jones 1997; Mayer and von Helversen 2001; Clare et al. 2013). Given that morphological characteristics traditionally are the first criterion under which species are classified and described, it is common that groups with very subtle external differences are grouped together despite being phylogenetically distinct. In this study, we found that the subspecies currently contained within G. soricina differ in cranial and mandibular shapes. Together with their known molecular differentiation and nonsympatric distributions, this constitutes enough evidence to state that G. soricina is a complex of at least four taxa that are following independent evolutionary trajectories. We recommend that, moving forward, these taxa to be considered independent species, and their trinomen subspecies name be conserved, as follows: G. antillarum, G. mutica, G. valens, and G. soricina. More detailed description of distinguishing characters is provided below.
The taxa analyzed herein presented the same patterns of sexual dimorphism, in which sexes differ in mandible shape, but not in skull shape. While there are no differences in size, the body of females’ mandibles is significantly thinner than males’ mandibles. This likely is due to pregnancy and lactation: for other bat species, it has been reported that during these periods, fetal and neonate bats obtain their calcium requirements from the mother’s skeletal reserves, which leads to smaller volume of adult females’ bones, including the dentary region (Kwiecinski et al. 1987; Booher and Hood 2010). The lack of differences in CS between sexes (except for G. antillarum mandible) contrasts with previous descriptions of sexual dimorphism in this species (Webster 1983; Louzada and Pessôa 2013), because the estimation of CS values is different than that of linear measures (Bookstein 1997). A similar pattern has been reported previously for Myotis species (Ospina-Garcés et al. 2016), where differences between sexes where found when analyzing shape variables but not with CS.
Shape variation best is predicted by different factors, depending on the module, as shown by the Procrustes ANOVA, but in all cases the adjustment was low (R2 < 0.126), suggesting that shape variation could be explained better by other factors not taken into account by this model. The significant effect of size on shape variation is equal in all of the groups analyzed, which do not differ in allometric trends although they do differ in size. Subspecies affiliation also has a significant effect in shape variation in all shape characters, suggesting phylogenetic divergence is reflected in shape differentiation.
Glossophaga soricina (Pallas 1766) is the most distinct taxon, among the taxa examined herein, because it is the only one that in all cases presented significant differences with respect to all the studied groups, in both size and shape, for the four shape modules. According to previous molecular studies, G. soricina represents a monophyletic clade whose genetic distance to the other groups is more than twice the distance found between the remaining subspecies, and as large as that found between sister species (Hoffmann and Baker 2001; Dias et al. 2017; Hoffmann et al. 2019). Similarly, in our morphological phylogeny, G. soricina is sister to remaining taxa in this complex, reflecting the fact that this species is differentiated both molecularly and morphologically.
Glossophaga soricina has the smallest CS values of the groups within this complex, which agrees with earlier reports (Webster 1983; Miller 1913). In terms of shape, we present statistical evidence that supports previous descriptions of this taxon having a domed braincase (rather than square, as in the other taxa of the species group), and short rostrum (Supplementary Data SD5). We did not, however, find evidence of the rostrum being narrower and with a more moderate rostral slope (Webster 1983): deformation grids showed that the rostrum is taller on the lateral view and that the slope actually is the most pronounced among the five taxa examined. We conclude that the diagnostic characteristics for this group would be a smaller size (GLS: = 20.5 mm, range, 19.5–21.7 mm; BCB: = 9.2 mm, 8.7–9.5 mm), a domed, round shape for the braincase, and a shorter rostrum, along with the previously reported darker pelage (Webster 1983). This species is widely distributed on the eastern side of the Andes, from Colombia to Paraguay and northern Argentina.
Another taxon that showed important differentiation was G. antillarum, which displayed the largest distances when compared to remaining taxa (Table 2). Specifically, the distance between the mandible shape of G. antillarum and that of G. valens is the largest we found (dist = 3.576, P = 0.001), where the largest vector displacements are the ones corresponding to the landmarks of the condyloid process, which is notably larger in G. antillarum than in remaining taxa examined (Supplementary Data SD5). Recent findings (Hoffmann et al. 2019) place G. antillarum in a monophyletic clade with very high support, based on genetic differentiation in the molecular markers Cytb and Fgb. We consider these factors, combined with its distribution restricted to Jamaica and the aforementioned morphological particularities, to be sufficient to conclude that G. antillarum is an independent species.
Glossophaga antillarumRehn 1902 is characterized by its large size (GLS: = 22.1 mm, 21.6–22.7 mm; BCB: = 9.5 mm, 9.1–9.7 mm), slender and flattened cranium, and noticeably large condyloid process (Supplementary Data SD5). Deformation grids showed that this species also has the largest occipital bone, as shown by the displacement on the landmark set on the cranial suture between the occipital and temporal bones (landmarks 12–15). This species’ distribution is restricted to Jamaica and is not known to occur on any other island in the Caribbean (Webster 1983).
Molecular evidence indicates that what currently is recognized as G. s. handleyi is a paraphyletic group, which should be treated in its own right as a species complex that warrants further attention (Hoffman and Baker 2001; Dias et al. 2017; Hoffman et al. 2019). Individuals from Mexico and northern populations of Central America (Guatemala and El Salvador) form a monophyletic clade that includes G. s. mutica, with which there is no molecular separation. The taxonomic description of G. s. mutica (Merriam 1898) precedes that of G. s. handleyi (Webster and Jones 1980), so this taxon should henceforth be named G. mutica. Populations from the remainder of Central America (Panama and Nicaragua) seem to form another independent linage, whose relation to the rest of the species and taxonomic situation should be addressed in future studies.
Glossophaga muticaMerriam 1898 continental specimens are intermediate in size and shape between G. soricina and the insular groups (GLS: = 20.1 mm, range 19.9–21.9 mm; BCB: = 9.4, 9.1–9.9 mm). This is evident in the CVA plots, where G. mutica (originally G. s. handleyi) occupies an intermediate region of the morphospace for all studied characters, with the insular species always clustering to one side and G. soricina to the other. Glossophaga mutica differs from the insular species and G. valens in having a rounder braincase and a wider rostrum with a moderate slope, but not as round and wide as those presented in G. soricina (Supplementary Data SD5). It is identical to G. valens in the mandible and the fronto-maxillary shape, but can be distinguished from G. valens by being smaller in cranial size. Glossophaga mutica is a broadly distributed taxon that occupies a variety of environments and displays high variance in size and shape, as well as at a molecular level. It already has been suggested that it could be composed by more than one lineage (Dias et al. 2017; Hoffmann et al. 2019).
Insular specimens of G. mutica from Tres Marias islands are clearly recognizable from its continental counterparts by their larger size (GLS: = 21.9 mm, 21.2–22.4 mm; BCB: = 9.5 mm, 9.3–9.7 mm), long rostrum, flattened braincase, and larger condyloid processes in the mandible. This morphological differentiation that insular G. mutica presents when compared to its continental counterparts likely is a product of the occupation of the islands, hence the morphological convergence with G. antillarum, with whom differences are visually imperceptible both to the naked eye and when visually comparing their mean shapes with the deformation grids, even when shown to be statistically different. We consider that this insular subspecies requires a new trinomial name.
Glossophaga valensMiller 1913 is a large species (GLS: = 22.2 mm, 21.5–22.8 mm; BCB: = 9.7 mm, 9.3–10.1 mm) whose cranial and mandibular shape resembles continental G. mutica, but they can be distinguished easily by size. Although some phylogenetic studies have found introgression with G. soricina in some molecular markers (Clare 2011; Hoffmann et al. 2019), this species is geographically isolated by the Andes (Dias et al. 2017). The geographic isolation, molecular divergence, and morphological differentiation, in concert indicate that G. valens likewise should be elevated to species level.
The most phylogenetically informative of the analyzed shape characters was the ventral region of the skull, as previously has been reported for other groups, which is attributed to its structural, functional, and developmental complexity (Caumul and Polly 2005). The module that showed the lowest variation was the parieto-occipital region, because it houses organs of great importance such as the brain and all the structures related to sight and hearing; thus, in this region strong selective pressures allow scant variation to happen (Marroig and Cheverud 2004; Caumul and Polly 2005; Monteiro and Nogueira 2011; Ospina-Garcés et al. 2016; Rossoni et al. 2017). In contrast, the anterior-maxillary region of the skull and the mandible proved to be more malleable structures (Freeman 2000; Drake and Klingenberg 2010) that can undergo rapid evolutionary changes related to individual life stories and recent ecological adaptations and hence could be useful to infer responses to dietary pressures (Freeman 2000; Caumul and Polly 2005). A good example of this are continental G. mutica and G. valens, differing significantly in the shape of their parieto-occipital region and in the ventral view of their skull and more phylogenetically informative (Caumul and Polly 2005), while being identical in their mandible and fronto-maxillary shapes, which might reflect that the two groups are using the same dietary niches throughout their continental distribution (Monteiro and Nogueira 2011).
In fact, our results suggest that environment plays a significant role on this shape variation. It is evident that similar environments are producing similar shapes on separate species: in the PLS plot, G. antillarum, G. valens, and insular G. mutica cluster in the same area not only in terms of environmental components but also in terms of shape, while continental G. mutica (former G. s. handleyi) and G. soricina occupy a different region (Fig. 6). This convergent pattern also is observed in CS measures: G. antillarum, insular G. mutica, and G. valens are not different in size, as opposed to continental groups G. mutica and G. soricina that differ significantly between each other and also with the insular subspecies. This agrees with Miller (1913) who, in his revision of this group, stated that insular specimens are easily distinguished when compared to the continental ones, but not between one another. Webster (1983) also found that insular species and G. valens cluster together in a multivariate analysis of cranial and postcranial measurements, pelage color, and qualitative cranial characters. These three groups are not closely related, nor interbreeding with each other given their separate geographic ranges, so this similarity is an evolutionary convergence that can be explained by the “Island Rule” (Foster 1964), which states that mammals that generally are small in the mainland tend to evolve toward a larger size after invading islands because of the change in the selective pressures ensuing the insular invasion, such as enemy and competitive release, that enables the species to undergo rapid evolutionary change as a response to drastic changes in the environment (Lomolino 2005; Millien 2006). Glossophaga valens is the largest continental species, which could indicate that it has been released from some competitive pressures in an analog manner to those of taxa occupying islands. This adds up to the evidence that G. valens and the insular groups share a morphological convergence driven by facing similar selective pressures.
Partial Least Squares (PLS) plot for the ventral view of the skull, with deformation grids derived from each canonical axis.
The small size of G. soricina also could be explained by its environment. Previous studies of this species found that bats in regions with high biodiversity have a smaller size than those from arid regions (Louzada and Pessôa 2013). Because of resource competition, smaller sizes are favored in regions of high species richness such as the one this species encounters throughout its distribution range (Heaney 1978; Aguirre et al. 2002). Moreover, the distinctive shape of this species also could be reflecting the strong competition it encounters. The elongation of the mandible and rostrum is a feature that makes the nectar-feeding tribe Glossophaginae (Baker et al. 1989) stand out from the rest of the phyllostomid bats (Monteiro and Nogueira 2011), and is associated with a better support of the tongue at the expense of a weaker bite force (Aguirre et al. 2002; Monteiro and Nogueira 2011). However, G. soricina is the least specialized of the glossophagines: it has an omnivorous diet (Sánchez-Casas and Álvarez 2000) and displays a retention of morphological features related to insectivory, such as having three molars with an ectoloph pattern and well-developed incisors (Howell 1976), and the rostrum is the least elongated of the genus. These omnivorous traits allow them to explore a broader diversity of feeding niches, which could have the advantage of reducing resource competition (Howell 1974).
We find patterns of morphological differentiation are consistent with the previously documented molecular divergence (Dias et al. 2017; Hoffmann et al. 2019) and also suggest convergence in size and shape of insular populations. These previously unnoticed morphological differences have now been explored by virtue of the capacity of geometric morphometric techniques to detect subtle morphological differences. We therefore conclude that there is enough morphological evidence that, along with the molecular findings, allows G. soricina to be separated in at least four species, three of which correspond to previously recognized subspecies (G. s. antillarum, G. s. valens, and G. s. soricina). Glossophaga handleyi appears to be a species complex that includes at least two lineages: a monophyletic one of populations from Mexico to north Central America, including specimens from Tres Marías islands, for which we propose the name G. mutica, and another one in the southern populations that requires further attention.
Supplementary Data
Supplementary data are available at Journal of Mammalogy online.
Supplementary Data SD1.—Description of landmark and semilandmark positions for each of the four studied shape configurations.
Supplementary Data SD2.—Spearman correlation matrix showing the degree of association between 19 environmental variables for Glossophaga soricina presence records.
Supplementary Data SD3.—Allometric trend plots (shape regression score on CS) for each of the four studied shape characters.
Supplementary Data SD4.—Topologies for most the recent molecular tree as published by Hoffmann et al. (2019; A) and for the tree we built based on morphological characters (B). The numbers on the nodes represent the amount of morphological change on each branch from an ancestral shape configuration reconstructed using parsimony. The dotted line represents the position of G. s. mutica and G. s. handleyi, which have no molecular distinction between each other.
Supplementary Data SD5.—Deformation grids illustrating the changes in the fronto-maxillary (A) and parieto-occipital (B) regions of the skull between the mean shapes of the three most distinctive species: G. s. antillarum (now G. antillarum), G. s. handleyi (now G. mutica), and G. s. soricina (now G. soricina). Light dots represent the landmarks and semilandmarks corresponding to the target mean shape of the subspecies in Y; dark dots represent the landmarks and semilandmarks corresponding to the mean shape of the subspecies in X. The arrows show the magnitude and direction of the displacement vectors from subspecies in X to target (subspecies in Y), while the deformation in the grid shows the bending energy. Deformation grids were exaggerated by a factor of 3 to improve visualization.
Appendix I
Specimens examined.—List of 231 specimens in alphabetical order by subspecies grouping. Museum names were abbreviated as follows: AMNH = American Museum of Natural History, New York, United States; MZFC = Museo de Zoología “Alfonso L. Herrera,” Facultad de Ciencias, UNAM, México; USNM = National Museum of Natural History, Smithsonian Institute, Washington, DC, United States.
Glossophaga soricina antillarum.—AMNH: M-271585, Jamaica, Clarendon, Portland Cottage; M-271587, Jamaica, Saint Ann, Mosely Hall Cave; M-214129, M-271588, Jamaica, Saint James, Montego Bay; M-271586, Jamaica, Saint Mary, Port Maria. USNM: 511236, 511237, 511238, 511239, Jamaica, Clarendon, Mahoe Gardens; 545155, 545156, 545157, 545158, Jamaica, Trelawny, Good Hope Estate; 545150, 545151, 545152, 545153, 511235, Jamaica, Trelawny, Quick Step.
Glossophaga soricina handleyi.—AMNH: M-269457, Colombia, Valle del Cauca, Dagua; M-136108, M-136109, M-136127, M-136129, M-136130, M-136134, Costa Rica, San José, San José; M-144704, M-144706, M-144717, M-144718, M-144720, Guatemala, Petén, La Libertad; M-126470, M-126477, M-126481, Honduras, Francisco Morazán, Guaimaca; M-203607, M-203608, M-203609, M-203613, M-203614, M-203615, M-203616, México, Veracruz, Veracruz; M-254616, México, Veracruz, Coatepec; M-213217, Nicaragua. MZFC: 6839, 6863, 6864, 7113, 7197, México, Chiapas, Ocosingo; 9324, 9333, 9336, 9339, México, Campeche, Tenabo; 961, 962, 963, México, Guerrero, Atoyac de Álvarez; 10558, 10570, 10573, México, Guerrero, Arcelia; 5799, 5800, 5806, 5808, 5812, México, Hidalgo, El Cardonal; 10170, 10172, 10173, 10174, México, Michoacán, Arteaga; 13835, 13836; México, Nayarit, San Blás; 4999, 5214, 5215, México, Oaxaca, San Juan Bautista Valle Nacional; 6415, México, Oaxaca, San José Chiltepec; 9761, 9762, 9763, México, Quintana Roo, Solidaridad; 11990, 11991, 12084, 12088, México, San Luis Potosí, Rayón; 666, 1090, México, Veracruz, San Andrés Tuxtla; 11310, 11311, México, Veracruz, Martínez de la Torre. USNM: 583013, Belize, Stan Creek, Cockscomb Basin Wildlife Sanctuary; 583011, 583012, Belize, Toledo, Big Falls; 250282, 250306, 250307, Costa Rica, Fuentes; 502198, 502199, 502200, 502202, Guatemala, Santa Rosa, La Avellana; 599194, 599201, Honduras, Cayo Cochinos; 565459, 565460, Honduras, Copán, Santa Rosa de Copán; 599227, Honduras, Islas de la Bahía, Roatan; 565461, Honduras, Lempira, Gracias; 564705, 564706, Honduras, Valle del Cauca; 252682, Honduras; 96834, 96837, 96853, México, Durango, Chacala; 564563, México, Yucatán, Pibtuch; 579006, 579010, 579011, 579012, Panamá, Bocas del Toro, Isla Escudo de Veraguas.
Glossophaga soricina mutica.—USNM: 512268, 512269, 512270, 512271, México, Nayarit, María Cleofas; 89271, 512255, 512256, 512258, 512259, 512261, México, Nayarit, María Madre; 512262, 512263, 512266, 512267, México, Nayarit, María Magdalena; 512272, México, Nayarit, San Juanito.
Glossophaga soricina soricina.—AMNH: M-209357, Bolivia, Beni, Itenez; M-209354, M-209355, M-209356, Bolivia, Beni; M-97230, M-97233, M-97244, M-97246, M-97248, Brazil, Para, Cameta; M-97263, Brazil, Para, Mocajuba; M-75236, Colombia, La Guajira, Riohacha; M-15157, M-15158, M-15159, Colombia, Magdalena, Santa Marta; M-207927, Colombia, Tolima, Mariquita; M-207315, Guyana, Demerara; M-207316, M-48251, Guyana, Demerara, Georgetown; M-182722, M-182723, Guyana, West Demara, Leguan Island; M-42615, M-42617, M-42621, M-42623, M-42627, M-42629, M-42632, M-42634, M-42635, M-42636, M-42637, M-42638, Guyana, Mahaica-Berbice Region, Canje River; M-217528, M-217530, M-217532, M-217533, Paraguay, Guaira,Villarica; M-175816, Trinidad and Tobago, Trinidad, Caroni County; M-175806, M-207064, M-183857, Trinidad and Tobago, Trinidad, Saint Andrew County; M-175812, Trinidad and Tobago, Trinidad, Victoria County; M-30682, M-30685, M-30686, M-30687, M-30689, M-30690, M-30691, M-30692, M-30693, M-30694, Venezuela, Bolivar, El Callao; M-130657, M-130658, M-130660, M-130661, Venezuela, Bolivar, Gran Sabana; M-31502, Venezuela, Carabobo, Puerto Cabello. USNM: 562721, 562722, Brazil, Amazonas, Laurete; 393653, 393654, Brazil, Mato Grosso, Serra Do Roncador; 391040, 391041, 391042, 391043, Brazil, Minas Gerais, Sete Lagoas; 361545, 361546, 361547, 361552, Brazil, Para, Utinga; 582297, 582298, Guyana, Barima-Waini, Baramita; 582296, Guyana, Upper Demerara-Berbice, Dubulay Ranch; 407705, 407707, Venezuela, Amazonas, San Juan; 545336, 545338, Venezuela, Amazonas, Rio Orinoco.
Glossophaga soricina valens.—AMNH: M-61426, M-61427, M-61428, Ecuador, El Oro, Santa Rosa; M-62943, M-62944, M-62945, Ecuador, Guayas, Duran; M-62105, Ecuador, Guayas, Guayaquil; M-64560, M-64562, M-64563, Ecuador, Guayas, Santa Elena; M-36276, Ecuador, Pichincha, Duale River.
USNM: 498834, 498836, 498837, 498839, Ecuador, Guayas, Balao; 498845, 498846, 498847, 498848, Ecuador, Los Ríos, Puebloviejo; 531243, Perú, Piura, Rio Chira.
Acknowledgments
We wish to thank everyone involved in this project, the Posgrado en Ciencias Biológicas (Universidad Nacional Autónoma de México), Suazanne Peurach and Darrin Lunde at the National Museum of Natural History Mammal collection, and Eleanor Hoeger and Sara Ketelsen at the American Museum of Natural History Mammal collection for their help in granting access to specimens. We thank Miriam Zelditch and Efraín de Luna for their valuable comments to improve this work. Partial funding was provided through a doctoral scholarship from Consejo Nacional de Ciencia y Tecnología (CONACyT), México to AC-O (253088) and Programa de Apoyo a los Estudiantes de Posgrado (PAEP). SMO-G is supported by the postdoctoral fellowship program from Direccion General de Asuntos del Personal Académico (DGAPA, Universidad Nacional Autónoma de México). The authors declare that they have no conflicts of interest or competing interest.
