INTRODUCTION
The rodent family Heteromyidae is exclusive to the Americas and includes five genera and 57 species of kangaroo rats (Dipodomys), kangaroo mice (Microdipodops), and pocket mice (Chaetodipus, Heteromys, and Perognathus; Williams et al. 1993; Patton 2005; Anderson et al. 2006; Hafner et al. 2007). They originated during the Oligocene (30 Myr) in North America (Hafner et al. 2007).
The subfamily Heteromyinae diversified during the Miocene (22.3 to 21.8 Myr; Hafner et al. 2007), principally due to climatic changes that led to the flooding of the lowlands in southern Mexico and Central America (Coates & Obando 1996; Baumgarten & Williamson 2007; Barber & Klicka 2010; Ordóñez-Garza et al. 2010; Vázquez- Domínguez & Arita 2010). During that period, the Panama land bridge connected the highlands of North and South America (Maldonado-Koerdell 1964; Coates & Obando 1996; Coates et al. 2004; Marshall 2007; Almendra & Rogers 2012). This land connection allowed the genus Heteromys to spread from North America to Central America and the northern part of South America, during the Great American Biotic Interchange (Pliocene, 5 to 3 Myr; Simpson 1980; Rogers & Vance 2005; Hafner et al. 2007). In the drier and colder conditions during the Pleistocene ( 2.5 Myr), Heteromys was probably restricted to refugia (zones in which changes in climate and vegetation were not so drastic) in southern Mexico and Central America (Toledo 1982; Alexander & Riddle 2005). Such refugia would have provided the isolation necessary for morphological, ecological, physiological, and ethological diversification of the genus (Anderson & Jarrín 2002; Anderson 2003; Alexander & Riddle 2005; Patton 2005; Anderson & Timm 2006; Hafner et al. 2007; Anderson & Gutiérrez 2009; Espinoza et al. 2011; Ramírez-Pulido et al. 2014), giving rise to the currently recognized species.
Morphological traits, such as overall cranial size, morphology of the interpterygoid vacuities, and length of temporal crests, have been traditionally used for species identification (Álvarez-Castañeda et al. 2015). Molecular analyses of allozymes and mitochondrial (cytochrome b, cytochrome c oxidase subunit I, and 12S and 16S rRNA) and nuclear (MYH6 and EN2) genes have confirmed monophyly in the basal relationships of the Heteromyidae, recovered Liomys as paraphyletic to Heteromys, and proposed synonymy of Liomys with Heteromys (Anderson et al. 2006; Hafner et al. 2007; Rogers & González 2010). Also, Heteromys irroratus, H. pictus, and H. salvini are paraphyletic (Rogers & Vance 2005; Rogers & González 2010), and H. desmarestianus, also paraphyletic, includes 4 subspecies and four candidate species with varying haplotypes whose genetic distances for cytochrome b range from 8.7% to 17% (Rogers & González 2010). Therefore, the genus Heteromys consists of 17 species clustered in six clades: 1) the H. adspersus group (including H. adspersus and H. salvini, previously identified as genus Schaeferia, Lehmann & Schaefer 1979); 2) the H. irroratus group (H. irroratus, H. pictus, and H. spectabilis); 3) subgenus Xylomys (H. nelsoni); 4) the H. anomalus group (H. anomalus, H. australis, H. catopterius, H. oasicus, and H. teleus; 5) the H. gaumeri group (H. gaumeri); and 6) the H. desmarestianus group (H. goldmani, H. oresterus, H. nubicolens, H. desmarestianus, and H. temporalis) (Anderson et al. 2006; Rogers & González 2010; Ramírez-Pulido et al. 2014) (Fig. 1).
Consequently, the taxonomy and biogeographic history of Heteromys are confusing, as the genus includes a set of species that are externally similar in morphology, yet vary considerably in their karyotypes, allozymes, and cranial morphology (Rogers 1990; Anderson et al. 2006). Taxonomic differences in the genus have not yet been fully clarified which has led to difficulties in estimating the boundaries between species. This is an important issue in sympatric species with similar colorations, cranial sizes, and morphology of the interpterygoid vacuities (H. irroratus, H. pictus, and H. spectabilis), and parapatric species that are similar in color but vary in the length of the temporal crest (H. desmarestianus, H. goldmani, and H. temporalis). In addition, the phylogenetic relationship between polytypic species with broad distributions (H. desmarestianus, H. irroratus, H. salvini, H. pictus, and H. adspersus) is not yet clear; nor is it clear if these species belong to groups with concordant phenotypic and genotypic traits. Furthermore, specimens of Heteromys exist as marginal records within the distribution of some species (Ramírez-Pulido & Sánchez-Hernández 1969; Hall 1981) and show unique morphological traits that differentiate them from other taxa. Therefore, they are possible candidates for new species, but further studies are necessary to identify them (Rogers & González 2010).
Investigation of variation in the cranial morphology of Heteromys allows for broadening the knowledge of its systematics and taxonomy. Therefore, the objectives of this study are 1) to compare, through geometric morphometrics, intra-and interspecific cranial variation in the specimens of Heteromys recognized species and candidate species from Mexico and Central America; 2) to evaluate whether the genetic variation in Heteromys species is reflected in variation in the shape of the skull; and 3) to evaluate the existence of geographic variation within Heteromys species. The cranial morphotype of each of the Heteromys species analyzed is expected to be different, with evident and unique differences in specific skull structures related to the physiographic regions that the species inhabit. It is also expected that these cranial morphotypes can be categorized into groups similar to those obtained by molecular analysis.
MATERIALS AND METHODS
A total of 638 specimen skulls (Table 1) from several localities were reviewed. Specimens were obtained from mam- mals collections of El Colegio de la Frontera Sur (ECO-SC- M), San Cristóbal de las Casas, Mexico; the Royal Ontario Museum (ROM), Ontario, Canada; Centro Interdisciplinario de Investigación para el Desarrollo Integral Regional Unidad Oaxaca (OAXMA), Oaxaca, Mexico; Instituto de Investigaciones Biológicas de la Universidad Veracruzana (IIB-UV), Xalapa, Mexico; and Colección Nacional de Mamíferos (CNMA) of Universidad Nacional Autónoma de México, Ciudad de México, Mexico (Appendix 1; Fig. 1). Each specimen was identified to the species level based on their tags, and morphology. The species considered in this study are characterized each by an unique combination of morphological traits, such as the greater length of the temporal crest in H. temporalis, the U-shaped interpterygoid fossa in H. irroratus, H. salvini, and H. pictus and the V-shaped interpterygoid fossa in H. australis, H. desmarestianus, H. gaumeri, H. goldmani, H. nelsoni, H. temporalis, and Heteromys sp., and the lateral length of the parietal bone throughout the lambdoidal ridge in H. nelsoni (Álvarez-Castañeda et al. 2015). These characteristics were not considered for location of cranial landmarks in this study, as we wished to compare skull morphology without considering traits unique to these species.
The H. desmarestianus clades identified by molecular studies of Rogers & González (2010) were considered as a single taxon unit since most of the specimens included in this study are from the northern distribution (clades II-IV) and very few are from the southern distribution (clades V-VIII) of the species; in addition, the clade I has already been identified as H. temporalis by Ramírez-Pulido et al. (2014).
Complete skulls of adult specimens were analyzed and identified by the presence of permanent premolars with little or evident wear, the eruption of M3, the presence of anterior and posterior molar lophs separated by a median valley, the M1-M3 usually worn with the anterior and posterior lophs connected (or have an O-shape) and fusion of cranial sutures (Rogers & Schmidly 1982). In the case of Heteromys sp., broken skulls recovered from owl pellets collected in Guerrero, Mexico, were also studied (Ramírez-Pulido & Sánchez-Hernández 1969). Using a 20.4-megapixel camera (Sony Cybershot HD DSC-HX50V, all skulls were photographed in three positions (dorsal, ventral, and lateral); the mandible was photographed in the right lateral view. The photographs were stored in JPEG image format and digitized to locate landmarks on half of the skull, under the assumption that mammal skulls are bilaterally symmetrical (Klingenberg & Mcintyre 1998; Figs. 2A-D; Supplement 1). The X and Y coordinates of each landmark in the digitized photographs were recorded using tpsUtil v1.44 (Rohlf 2009) and tpsDig v2.12 (Rohlf 2008). Using Procrustes analysis of overlapping landmarks, the effects of size, position, and scale were eliminated (Klingenberg & Mcintyre 1998; Klingenberg 2002; Vázquez-García 2016) with CoordGen8 (Sheets 2014a) of the Integrated Morphometrics Package (IMP), ver. 7 (Rohlf & Sheets 2004).
Analyses of intra-and interspecific variation were carried out separately for each of the four skull views. The first intraspecific variation analysis focused on comparisons between sexes by species using a Goodall’s F-test (Webster & Sheets 2010) using TwoGroup8 of the IMP package (Sheets 2014b).
For interspecific comparison, canonical variate analysis (CVA) of partial warp scores was carried out [using the option of principal component analysis (PCA)-based dimensionality reduction due to the small number of specimens of some species]. Differences in specific cranial structures were obtained through the sheet deformation technique or TPS (Thin Plate Spline; Bookstein 1997) using the program CVAGen8 (Sheets 2014c) and were edited with the software PAST v.2.17c (Hammer et al. 2001).
The interspecific statistical analyses of each of the cranial views were carried out using the CVA scores obtained from the PCA dimensionality reduction, and included 1) a multivariate analysis of variance (MANOVA) with the Wilks’ lambda test (Zelditch et al. 2012) to analyze the differences between the nine species of Heteromys and the candidate species; 2) an analysis of canonical variables (CVA), which allows discrimination between specimens of Heteromys species; 3) a posteriori validation of the groups by allocation of percentages with the Jackknife method; and 4) a cluster analysis with Procrustes distances with PCA dimensionality reduction of the dorsal view, in which images of the nine species and the specimens considered the species candidate were obtained. All these analyses were carried out using the software PAST version 2.17c (Hammer et al. 2001).
For the second intraspecific analysis, a geographical division was obtained by using the package ArcGIS 10.2.2 with the intersection of the geographic information layers of the physiographic (Cervantes-Zamora et al. 1990) and mammal biogeographic regions (Ramírez-Pulido & Castro-Campillo 1990) of Mexico; the physiographic division of Guatemala and the topography of Belize were added, while for El Salvador, Costa Rica, Panama, and Ecuador, their political boundaries were used as physiographic units due to the absence of geographical representation of Heteromys records used in the analyses. The specimens were mapped and the CVA scores for interspecific analysis were grouped using the regions obtained. Clusters were analyzed using MANOVA with the Wilks’ lambda test (Zelditch et al. 2012) with the software PAST v.2.17c (Hammer et al. 2001).
The phenogram of the cluster analysis of the dorsal view was compared to the topology of the cladogram, based on the mitochondrial cytochrome b gene, obtained by Rogers & González (2010) in order to discern whether morphometric changes are consistent with the changes identified using molecular evidence. We limited our comparisons to the dorsal view because it was the most complete and the only view that allowed to fully allocate the landmarks in the specimens of Heteromys sp.
RESULTS
Sexual variation
Heteromys desmarestianus was the only species showing sexual variation in all four views (dorsal: F =2.36, p<0.05; ventral: F =1.89, p<0.05; lateral: F =1.61, p<0.05; lateral mandible: F =1.95, p<0.05), while H. irroratus showed sexual variation in three of the four views (dorsal: F =2.63, p<0.05; ventral: F =3.35, p<0.05; lateral: F =1.77, p<0.05). The remaining species (H. gaumeri, H. goldmani, H. nelsoni, H. pictus, H. salvini, and H. temporalis) showed sexual variation in only one or none of the views. Due to this, in analyses of interspecific variation, males and females of H. desmarestianus and H. irroratus were considered as separate groups for the posterior interspecific analysis. Two species were excluded from this analysis: H. australis, which was represented by only one specimen, and Heteromys sp. because the sex of its specimens was unknown.
In the geographic analyses of the two species that showed sexual variation, H. irroratus and H. desmarestianus, both sexes of H. irroratus were grouped together in four regions (North of the Southern Coastal Range, Southern Mountain Range of Puebla, Lakes and Volcanoes of Anahuac, and Chiconquiaco), whereas the two sexes did not coincide in four other regions, with only females occurring in Mountains and Valleys of Oaxaca and South of the Southern Coastal Range and only males occurring in Plains of Ojuelos-Aguascalientes and Plains andLow Elevations. Both sexes of H. desmarestianus coincided in 11 regions (Northern Mountains of Chiapas, Altos of Chiapas, Southern Mountains of Chiapas, Mayan Mountains of Guatemala and Belize, Carso and Campeche Hills, Lacandon Mountain, Coastal Plains of Chiapas and Guatemala, Lowlands of Peten, Cristaline Highlands of Guatemala and Motagua Depression, El Salvador, and Costa Rica); the two sexes did not occur together in three regions, with females exclusive to Plains and Swamps of Tabasco and Lower Coast of Belize and males exclusive to Panama.
Only females showed differences in cranial morphology between groups and regions. Heteromys desmarestianus females of the Southern Mountains of Chiapas physiographic region were different from those of El Salvador (dorsal: F =2.69, p<0.05; ventral: F =3.46, p<0.05), Lacandon Mountain (ventral: F =3.46, p<0.05), and Mayan Mountains of Guatemala and Belize (ventral: F =3.46, p<0.05). Heteromys irroratus females of the Southern Mountains of Puebla were different from those of the Mountains and Valleys of Oaxaca (dorsal: F =4.42, p<0.05), South of the Southern Coastal Range (dorsal: F =4.42, p<0.05), North of the Southern Coastal Range (ventral: F =5.02, p<0.05), Lakes and Volcanoes of Anahuac (dorsal: F =4.42, p<0.05), Chiconquiaco (dorsal: F =4.42, p<0.05; ventral: F =5.02, p<0.05), and Plains of Ojuelos-Aguascalientes (ventral: F =5.02, p<0.05).
Interspecific variation
MANOVA showed that significant differences existed between species in the three cranial views (Table 2). The exceptions were H. desmarestianus and H. irroratus whose male and female specimens did not differ between the two species (p>0.05) in three or all four views, and H. salvini and H. australis, which were not statistically different in any of the four views.
With respect to CVA, canonical variable 1 (CV1) for the three cranial views (dorsal, ventral, and lat eral; Figs. 3A, C, E) separated H. salvini, H. irroratus (both sexes), and H. pictus (three species previously considered to belong to the genus Liomys) into one group and H. australis, H. desmarestianus, H. gaumeri, H. goldmani, H. nelson, and H. temporalis into another group. Furthermore, H. gaumeri and H. nelsoni were not superimposed, but were grouped on opposite sides of canonical variable 2 (CV2). Heteromys australis, H. desmarestianus (both sexes), H. goldmani, and H. temporalis were found to overlap in the three cranial views, although the CVA graph of the lateral mandible (Fig. 3G) showed no separation of any of the species.
The a posteriori Jackknife test (Table 3) resulted in each species showing high assignment percentages (>75%) for the three cranial views (except H. pictus for the ventral view, which had a value of 68%), while the lateral mandibular view had values ranging from 47% to 100%. Both sexes of H. desmarestianus and H. irroratus had assignment percentages that differed between views, with 50% and 63% for dorsal, 47% and 86% for ventral, 43% and 64% for lateral, and 39% and 50% for lateral mandible views, respectively. In the deformation grids of the dorsal view, the most notable changes were observed in the zygomatic arch; fewer differences were found in the coronoid suture separating the frontal and parietal bones, and the union between the zygomatic process with the squamosal bone and the posteriormost point of the auditory bullae (Fig. 3B). The ventral view shows deformation principally in the anteriormost and posteriormost regions of the zygomatic arch, palatine bone, interpterygoid fossa, and the anteriormost point between nasal bones, with fewer differences between species in the morphology of the maxillary tooth row, the posteriormost point of foramen magnum, and auditory bulla (Fig. 3D). The lateral view shows significant shape variation in the auditory capsule and the occipital condyle, and less variation in the braincase curvature (Fig. 3F). The lateral mandibular view shows greater deformation in the mandibular tooth row, coronoid process, and the posterior most point of the angular process than in the lowest point of the curve of the diastema, the anterior most point of the angular process, and mandibular condyle process (Fig. 3H). Deformations in the dorsal and lateral views (Figs. 3A-B, E-F) occurred from the negative end of CV1, where H. irroratus (males and females), H. pictus, and H. salvini were grouped, toward the positive end, where H. australis, H. desmarestianus, H. gaumeri, H. goldmani, H. nelsoni, H. temporalis, and Heteromys sp. were located. In the ventral and mandibular views (Figs. 3C- D, G-H), changes occurred from the positive to the negative end of CV1.
The intersection of the physiographic and biogeographic regions generated a total of 31 regions (Fig. 1). The specimens were grouped according to species and the physiographic region where they were collected. Heteromys australis, H. nelsoni, and the Heteromys species candidate were each grouped into a single region (Ecuador, Southern Mountains of Chiapas, and North of the Southern Coastal Range, respectively) making it impossible to compare geo graphic groups. Heteromys salvini and H. goldmani specimens were grouped into two regions (Central American Volcanoes and Southern Mountains of Chiapas), H. temporalis specimens into three regions (Coastal Plains of Veracruz, Los Tuxtlas Mountains, and Chiconquiaco), H. gaumeri into five regions (Yucatec Carso, Lower Coast of Quintana Roo, Plains and Swamps of the Laguna de Terminos, Lower Coast of Belize, and Carso and Campeche Hills), and H. pictus into six regions (Coastal Plains of Veracruz, Plains of the Isthmus, Los Tuxtlas Mountains, North of the Southern Coastal Range, Lakes and Volcanoes of Anahuac, and Chiconquiaco). The cranial morphology of Heteromys species was different between the physiographic regions within which they were grouped.
Specimens of H. gaumeri from the Yucatec Carso region were different from those from the Lower Coast of Belize (dorsal: F =2.47, p<0.05; lateral: F =3.44, p<0.05), Carso and Campeche Hills (dorsal: F =2.47, p<0.05; lateral: F =3.44, p<0.05), and Plains and Swamps of the Laguna de Terminos (lateral: F =3.44, p<0.05), whereas specimens from Carso and Campeche Hills were different from those from the Lower Coast of Quintana Roo (dorsal: F =2.47, p<0.05), Lower Coast of Belize (dorsal: F =2.47, p<0.05; lateral: F =3.44, p<0.05), and the Plains and Swamps of the Laguna de Terminos (lateral: F =3.44, p<0.05). The cranial morphology of H. goldmani was different between the Southern Mountains of Chiapas and the Volcanoes of Central America (dorsal: F =1.94, p<0.05; ventral: F =2.18, p<0.05; lateral: F =3.19, p<0.05).
The specimens of H. pictus from the north of the Southern Coastal Range showed differences from specimens from the Coastal Plains of Veracruz (dorsal: F =3.02, p<0.05), Plains of the Isthmus (ventral: F =3.82, p<0.05), and Los Tuxtlas Mountains (ventral: F =3.82, p<0.05). In the case of H. salvini, cranial morphology differed between specimens from the Volcanoes of Central America and those from the Southern Mountains of Chiapas (dorsal: F =13.72, p<0.05; ventral: F =45.86, p<0.05). Similarly, H. temporalis specimens showed differences between the Los Tuxtlas Mountains and Chiconquiaco regions (dorsal: F =3.30, p<0.05; ventral: F =3.88, p<0.05).
Cluster analysis
The cluster analysis of the dorsal view divided the genus into two groups; the first included H. pictus, H. salvini, and both sexes of H. irroratus, while the second included H. australis, H. nelsoni, H. gaumeri, H. goldmani, H. temporalis, and both sexes of H. desmarestianus. Heteromys sp. was grouped as a separate species (Fig. 4B). The separation of Heteromys species was similar to that obtained with maximum likelihood trees based on mitochondrial genes by Rogers and González (2010; Fig. 4A).
DISCUSSION
The variation in cranial shape and size of Heteromys species is a reflection of their evolutionary history, as a possible response to climatic, physiographic, and plant cover changes in the southern region of Mexico and Central America. This response has generated differences in cranial morphology between males and females of some species, as well as in the cranial characters used for species identification (Goldman 1911; Hall 1981; Carter & Genoways 1978; Dowler & Genoways 1978; McGhee & Genoways 1978; Schmidt et al. 1989; Rogers & Rogers 1992).
Sexual variation
In agreement with previous studies, our results show that two species have sexual variation: H. irroratus shows differences between sexes in cranial length and width (Dowler & Genoways 1978); coincidently, the average body size of H. desmarestianus males is greater than that of females (Espinoza et al. 2011). Our study did not find that H. gaumeri, H. goldmani, H. nelsoni, H. pictus, and H. salvini have sexual variation, contrasting with previous studies (McGhee & Genoways 1978; Carter & Genoways 1978; Schmidt et al. 1989; Rogers & Rogers 1992). These differences in results may be due to the techniques used. Linear morphometry has been used to describe the variation in some species of Heteromys (Anderson & Jarrín 2002; Anderson & Timm 2006; Anderson & Gutiérrez 2009). However, comparisons between linear and geometric morphometrics have shown that results may vary (Breno et al. 2011), when adding or removing a landmark or semilandmark (Schmieder et al. 2015). Therefore, a comparative study of both techniques and the use of different configurations of landmarks and semilandmarks would allow detailed analysis of sexual variation in Heteromys.
Interspecific variation
All Heteromys species analyzed demonstrate greater interspecific variation in the three cranial views than in the mandible, as indicated by the statistical results (MANOVA, CVA, and Jackknife assignment) and the number of areas with changes in the deformation grids. This means that the skull is more informative than the mandible for taxonomic identification of species (Dowler & Genoways 1978; McGhee & Genoways 1978; Carter & Genoways 1978; Schmidt et al. 1989; Rogers & Rogers 1992).
The MANOVA and the cluster analysis results support the hypothesis of Ramírez-Pulido & Sánchez-Hernández (1969) that Heteromys sp. may be a smaller form of H. desmarestianus. This would place Heteromys sp. in the desmarestianus group (H. desmarestianus, H. goldmani, H. temporalis, H. nubicolens, and H. oresterus). The cluster analysis separates Heteromys sp. from the other species of Heteromys, except for H. irroratus, H. pictus, and H. salvini. Based on the differences in cranial morphology between Heteromys sp. and the other Heteromys species studied, it may be considered a different species within the genus; it is important to conduct genetic studies to corroborate this finding.
Variation in cranial morphology of the genus Heteromys is observed in four specific cranial structures: (1) the rostrum, (2) the zygomatic arch (mainly in the anteriormost and posteriormost points), (3) the braincase profile, and (4) the occipital and the foramen magnum. In addition, variation is seen in three mandibular structures: the angular process, the condylar process, and the curvature of the diastema. The mandible, as well as the rostrum, zygomatic arch, and braincase, are related to phylogenetic pressures dictated by diet and physiological factors, such as pressures on the masticatory muscle at their insertion in the skull (Bowers & Brown 1982; Cox et al. 2012; Klingenberg 2013). Variations in the posterior portion of the skull (the occipital bone and foramen magnum) are attributed to muscular pressure (of the masseter muscle) and circulatory pressure (of the stapedial artery), which allow for development of certain brain lobes and generate physical pressure on this area of the skull (Brylski 1990). It is important to point out that these characteristics of cranial structures correspond to those described in earlier studies of some Heteromys species (H. goldmani, H. temporalis, H. irroratus, H. desmarestianus, which identified differences in the zygomatic arch, rostrum width, braincase morphology, interparietal and frontal bones (Goldman 1911).
Besides physical pressures, Heteromyids, including the genus Heteromys, are susceptible to climatic pressures and abiotic environmental factors (Brown 1975; Wolf et al. 2009; Baumgardner & Kennedy 1993). This susceptibility was reflected in the results of the analysis of physiographic variation in species. Some species showed differences between physiographic regions whose limits are related to geological history and the geological faults in their ranges. An example is H. desmarestianus, whose females showed differences between regions (El Salvador, Southern Mountains of Chiapas, Lacandon Mountains, and the Mayan Mountains of Guatemala and Belize). These regions were delimited by the Motagua–Polochic and Tuxtla–Malpaso fault systems (Durán-Calderón et al. 2014). Both systems resulted in the formation of the Motagua valley and the Altos of Chiapas, two geographical barriers with an important role in the evolutionary history of Central America (Durán-Calderón et al. 2014). Other species, such as H. goldmani and H. salvini, showed differences between regions (Central American Volcanoes and Southern Mountains of Chiapas) whose boundary occurs to the north of the city of Tapachula, a region under the influence of the Tacaná volcano and the Polochic and Coatán River faults (García-Palomo et al. 2006; Durán-Calderón et al. 2014).
Similarly, H. irroratus, H. pictus, and H. temporalis (Dowler & Genoways 1978; McGhee & Genoways 1978; Williams et al. 1993; Rogers & González 2010) showed differences between physiographic regions with a heterogeneous profile. These regions varied from highlands with altitudes ranging from 0 to 5000 m a.s.l., such as the Southern Mountains of Puebla, Chiconquiaco, north and south of the Southern Coastal Range, Los Tuxtlas Mountains, Plains of Ojuelos-Aguascalientes, to lowlands ranging from 0 to 350 m a.s.l. (the Coastal Plain of Veracruz and the Plains of the Isthmus; Geissert Kientz 1999; Gutiérrez-Herrera et al. 2003; CONABIO 2011; Ortega-Corona et al. 2015). The altitudinally different areas (Geissert Kientz 1999) became islands or refugia during changes in the climate and vegetation cover (Munger et al. 1983) that occurred in the course of glacial and interglacial cycles (Rull 2004a,b; Mastretta-Yanes et al. 2015).
Unlike the previously mentioned species, for which geological and physiographic factors played an important role, H. gaumeri is distributed in a region of generally low elevation (less than 200 m a.s.l.; Schmidt et al. 1989); even so, this region has undergone significant changes in vegetation and precipitation throughout its history (Carrillo-Bastos et al. 2010, 2013; Torrescano-Valle & Islebe 2015), which are related to the glacial and interglacial periods that allowed biological colonization of the Yucatan Peninsula, as well as being barriers to species migration (Vázquez-Domínguez & Arita 2010). Considering these changes and their influence on populations (Wolf et al. 2009; Vázquez-Domínguez & Arita 2010) and the differences between the physiographic regions comprising the Yucatan Peninsula (Yucatec Carso, Carso and Campeche Hills, Lower Coast of Belize, and Plains and Swamps of the Laguna de Terminos), paleoclimatic models of precipitation gradients and changes in vegetation cover show similar patterns in these areas (Carrillo-Bastos 2013). Therefore, the cranial differences observed in H. gaumeri may be related more to the changes in precipitation and vegetation cover than to physiographic accidents.
Taxonomic implications
Despite not considering specific cranial structures used to identify Heteromys species, our study shows differences between species in cranial morphology, principally in the rostrum, zygomatic arch, braincase, and the region of the occipital bone and foramen magnum, as well as in the angular process, condyle process, and diastema. Cranial differences also exist between sexes in H. desmarestianus and H. irroratus; however, they are only perceptible in intraspecific comparisons, because at interspecific level these differences are reduced by the effect of inclusion of other species and increases in the sample size.
The division between the species previously included in the genus Liomys (H. irroratus, H. pictus, and H. salvini) and the other Heteromys species (H. australis, H. desmarestianus, H. gaumeri, H. goldmani, H. nelsoni, and H. temporalis) remains in analyses of variation in shape, even without considering the cranial structures used to identify the species (the form of the interpterygoid fossa). The relationships between species in the latter group are seen in the cluster analysis, in which H. gaumeri and H. nelsoni are totally separated, whereas the species in the desmarestianus complex (H. desmarestianus, H. goldmani, and H. temporalis) are grouped together. These clusters coincide with clades identified in phylogenetic studies based on various mitochondrial and nuclear genes; therefore, the phylogenetic relationships of Heteromys species at the molecular level are matched by the relationships from analyses of cranial shape variation by geometric morphometrics.
The groups identified by cluster analysis maintain the division of the genus Heteromys proposed by Anderson et al. (2006) between those species previously identified as belonging to the genus Liomys (H. irroratus, H. pictus, and H. salvini) and Heteromys species (H. australis, H. desmarestianus, H. gaumeri, H. goldmani, H. nelsoni, and H. temporalis). This topology also coincides with clades proposed using molecular evidence (Rogers & González 2010). This concordance indicates that the groups identified by morphological data represent the recognized genetic differences in the genus Heteromys. The understanding that morphometric differences in the skull reflect genetic differences, and that these morphometric differences are correlated with alterations in environmental factors (climatic, vegetation, or geographic; Cardini et al. 2007; De Moura Bubadué et al. 2016; Morales et al. 2016) would increase the knowledge available about the phylogeography and history of the genus Heteromys in the southern region of Mexico and Central America.
Finally, the species candidate Heteromys sp. is different in its cranial morphology from the rest of the Heteromys species; therefore, we consider it to be a new Heteromys species. It is important to obtain complete specimens of this species (skin, skull, and tissue) and carry out different studies that corroborate the morphometric differences found in this study since we could only partially analyze some broken skulls recovered from owl pellets. However, we may deduce that Heteromys sp. originated as a result of adaptation to a unique environment (rocky and arid; Ramírez-Pulido & Sánchez-Hernández 1969). The zone in which the specimens were found was possibly geographically isolated during the formation of the Balsas River depression, elevation of the Sierra Madre del Sur, and configuration of the Mountains and Valleys of Guerrero, leading to cranial variation and differentiation from the rest of the Heteromys species.